Duke University Dissertation Template

Home , Drosophila innubila, Hybrid inviability

Parental Conflict, Parent of Origin Effects, and the Evolution of Hybrid Seed Failure in

Mimulus

Jennifer M. Coughlan

Department of Biology Duke University

Date:______Approved:

______John Willis, PhD, Supervisor

______Kathleen Donohue, PhD

______Mark Rausher, PhD

______Mohamed Noor, PhD

______Thomas Mitchell-Olds

Dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biology in the Graduate School of Duke University

2018

ABSTRACT

Parental Conflict, Parent of Origin Effects, and the Evolution of Hybrid Seed Failure in

Mimulus

Jennifer M. Coughlan

Department of Biology Duke University

Date:______Approved:

______John Willis, PhD, Supervisor

______Kathleen Donohue, PhD

______Mark Rausher, PhD

______Mohamed Noor, PhD

______Thomas Mitchell-Olds

An abstract of a dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the Department of Biology in the Graduate School of Duke University

2018 i

Abstract

The earth is home to roughly 9 million eukaryotic species. The formation and maintenance of this diversity requires the accumulation of barriers to reproduction. One of the most common post-zygotic barriers in plants is hybrid seed inviability (HSI).

Despite its commonality, we know relatively little about the genetic mechanisms and evolutionary forces which are responsible for this barrier, particularly in naturally co- occurring species. Here I tested the role of parental conflict in HSI between co-occurring monkey flowers in the M. guttatus species complex. I assessed the strength and directionality of HSI within and between phenotypically described perennial variants of the M. guttatus species complex. I find substantial HSI between two morphologically described variants, M. guttatus and M. decorus, amounting to ~30-50% reproductive isolation (RI). Genetically distinct clades of M. decorus vary in their ability to cross with

M. guttatus in both magnitude and direction of RI. Intriguingly, northern and southern clades of M. decorus are also incompatible with one another, showing strong and symmetric hybrid seed inviability. In all cases, HSI is accompanied by parent of origin effects on F1 seed size, consistent with a role of resource allocation to hybrid inviability.

Parent of origin effects on reciprocal F1s were not limited to final size. I characterized the developmental trajectory of seeds between M. guttatus and both the northern and southern clades of M. decorus. Hybrid seeds show similar developmental trajectories

depending on their end fate (i.e. viable vs inviable), despite differing in the maternal and paternal contributors in these crosses. Inviable seeds were characterized by patterns of paternal excess; chaotic and malformed endosperm. Viable hybrid seeds are characterized by maternal excess; limited, and precocious endosperm development.

Relatively normal embryo development early on suggests that hybrid seed inviability stems, in part, from malformed endosperm. Lastly, I use a combination of modeling, next generation sequencing, and marker genotyping approaches to determine the genetic architecture and basis of HSI. I find that the genetic basis of HSI is complex, wherein several maternally and paternally inherited nuclear loci interact to cause HSI. In total, I show the presence of multiple perennial species in the M. guttatus species complex. These species show little phenotypic or obvious ecological differentiation, but are genetically unique and have substantial post-zygotic reproductive isolation, namely through the formation of inviable seeds. Patterns of seed development and final size are indicative of parent of origin effects on resource allocation to endosperm. Despite relatively low genetic differentiation for the group, the genetic basis of HSI is quite complex, involving multiple maternally and paternally interacting alleles. Future work will be needed to determine the identities of these genes, as well as patterns of repeatability across the complex.

Dedication

For my Grandmother, Helen Marie Coughlan, who introduced me to the importance of game theory at an early age with her uncanny card skills.

Contents

Abstract ...... iv

List of Tables ...... xiii

List of Figures ...... xiv

Acknowledgements ...... xvii

Chapter 1: Introduction ...... 1

1.1 Gene Duplications and Translocations ...... 3

1.2 Genetic Linkage and Hitchhiking ...... 5

1.3 Compensatory Mutations and Systems Drift ...... 7

1.4 Divergent Ecological Selection ...... 9

1.5 Mutation Order Speciation in Response to the Environment ...... 11

1.6 Auto-Immune system [Co]-Evolution ...... 12

1.7 Parental Conflict ...... 15

1.8 Sex Chromosome formation and speciation ...... 19

1.9 Selfish Genetic Elements ...... 21

1.10 Recombination-related processes ...... 27

1.11 Conclusions ...... 30

Chapter 2: The Mimulus guttatus species complex as a model system for the evolution of hybrid inviability and sterility ...... 32

2.1 Hybrid Sterility ...... 33

2.2 Hybrid Inviability (late stage) ...... 35

2.3 Hybrid Seed Inviability ...... 36

vii

2.4 Research Overview ...... 38

Chapter 3: Patterns of hybrid seed inviability in perennials of the Mimulus guttatus sp. complex ...... 42

3.1 Introduction ...... 42

3.2 Materials and Methods ...... 46

3.2.1 Phenotypic and genetic survey of M. decorus ...... 46

3.2.2 Crossing survey between M. guttatus and M. decorus ...... 50

3.2.3 Assessing evolutionary mechanisms for variation in hybrid seed inviability . 52

3.2.3.1 Determining ploidy and its effects on hybrid seed inviability ...... 52

3.2.3.1 Assessing potential correlates of hybrid seed inviability ...... 53

3.2.4 Crossing survey within species ...... 55

3.2.5 Statistical Analyses ...... 56

3.3 Results ...... 57

3.3.1 Phenotypic and genetic divergence in perennials of the Mimulus guttatus species complex ...... 57

3.3.2 Hybrid seed inviability between M. decorus and M. guttatus ...... 60

3.3.3 The role of whole genome duplication in hybrid seed inviability between M. decorus and M. guttatus ...... 64

3.3.4 Correlates of hybrid seed inviability ...... 69

3.3.5 Hybrid seed inviability with M. decorus and M. guttatus ...... 72

3.4 Discussion ...... 77

3.4.1 Phenotypic and genetic diversity with perennials of the M. guttatus species complex ...... 78

viii

3.4.2 Hybrid seed inviability associated with whole genome duplication ...... 81

3.4.3 Hybrid seed Inviability and parental conflict ...... 82

3.4.4 Hybrid seed inviability is rapidly evolving ...... 84

3.4.5 Conclusions ...... 85

Chapter 4: Developmental trajectories of hybrid seeds ...... 87

4.1 Introduction ...... 87

4.2 Methods ...... 90

4.2.1 Growth conditions and plant crossing ...... 90

4.2.2 Crossing Assays ...... 91

4.2.3 Microscopy and measuring growth ...... 91

4.2.4 Embryo Rescues ...... 92

4.2.5 Statistical Analyses ...... 93

4.3 Results ...... 93

4.3.1 Post-Mating Reproductive Isolation ...... 93

4.3.2 Developmental Trajectories ...... 95

4.3.3 Embryo Rescues ...... 99

4.4 Discussion ...... 99

Chapter 5: Uncovering the genetic basis of hybrid seed inviability between co-occurring sister species ...... 103

5.1 Introduction ...... 103

5.2 Materials and Methods ...... 106

5.2.1 Plant material, growing conditions, crosses ...... 106

5.2.2 Determination of HSI associated loci ...... 109

5.2.2 Calculating selection coefficients and multi-locus genotype probabilities ..... 113

5.3 Results ...... 116

5.3.1 Hybrid seed incompatibilities are not caused by cyto-nuclear interactions ... 116

5.3.2 The genetic architecture of hybrid seed incompatibilities between M. guttatus and M. decorus is complex ...... 118

5.4 Discussion ...... 124

5.4.1 Hybrid seed incompatibilities are not caused by cyto-nuclear interactions ... 124

5.4.2 The genetic architecture of hybrid seed inviability is genetically complex .... 125

5.4.2 The effectiveness of hybrid seed inviability as a natural barrier to gene flow127

5.4.3 Problems, pitfalls, and future directions ...... 129

5.5 Conclusions ...... 131

Chapter 6: Dissecting the role of a large chromosomal inversion on life history divergence throughout the Mimulus guttatus species complex ...... 132

6.1 Introduction ...... 132

6.2 Methods ...... 137

6.2.1 Study Genus ...... 137

6.2.2 Assessing life history variation between annual M. guttatus and perennial species in the M. guttatus species complex ...... 138

6.2.3 Survey of the LG8 inversion orientation ...... 139

6.2.4 Determining whether the LG8 inversion is still associated with life history divergence in a more distantly related perennial species ...... 142

6.2.5 A test of the recombination suppression hypothesis- determining whether the LG8 region is still associated with life history divergence in the absence of the inversion ...... 144

6.2.6 Genotype-phenotype associations ...... 147

6.3 Results ...... 148

6.3.1 Assessing life history variation in the M. guttatus species complex: ...... 148

6.3.2 Species survey of the inversion ...... 150

6.3.3 The LG8 inversion is associated with life history divergence in M. decorus ... 151

6.3.4 The LG8 inversion region is associated with life history divergence in a collinear species ...... 152

6.4 Discussion ...... 162

6.4.1 The evolutionary history of the LG8 inversion ...... 162

6.4.2 The LG8 inversion region is associated with life history traits in the absence of the inversion ...... 165

6.5 Conclusions ...... 169

Chapter 7: Conclusion ...... 171

Appendix A: Assessing other reproductive barriers between M. guttatus and M. decorus ...... 173

Methods ...... 173

Results ...... 174

Appendix B: The evolution of underground stolons ...... 178

Methods ...... 178

Results ...... 180

Population Survey ...... 180

Appendix C: Supplemental figures and tables (in chronological order) ...... 186

Supplemental Figures and Tables: Chapter 3- Crossing Survey ...... 186

Supplemental Figures: Chapter 4- Seed Development Survey ...... 201

Supplemental Tables and Figures: Chapter 5- Genetics of Hybrid Seed Lethality ... 202

Supplemental Tables: Chapter 6: Inversions and Life History Adaptation ...... 206

References ...... 210

Biography ...... 245

xii

List of Tables

Table 1: Mixed model for type III ANOVAs for seed cell size across developmental time ...... 98

Table 2: Marker genotypes to confirm parent of origin segregation distortion...... 122

Table 2: Selection coefficients for regions associated with parent of origin effects...... 123

Table 3: Effect of the inversion karyotype on life history traits within the M. guttatus species complex ...... 158

Table 4: Population Collections of M. decorus used throughout this dissertation...... 192

Table 5: Average viability and standard error of crosses between populations of M. decorus averaged across populations of M. guttatus ...... 193

Table 6: Average viability and standard error of crosses between focal individual Odell Creek and populations of M. decorus...... 195

Table 7: Average viability and standard error of crosses between focal individual IMP and populations of M. decorus ...... 197

Table 8: Average viability of crosses between focal individual 35-48 and populations of M. decorus ...... 199

Table 9: Average seed size and standard error of crosses between populations of M. decorus averaged across populations of M. guttatus ...... 200

Table 11: Multi-locus genotypes: observed and expected values ...... 206

Table 12: Inversion survey across species of the Mimulus guttatus species complex and outgroup ...... 206

Table 13: Differences between parental lines and the effect of the inversion or inversion region ...... 207

Table 14: Dissecting the effect of the inversion karyotype on life history trait divergence ...... 208

xiii

List of Figures

Figure 1: BDMI model schematic ...... 3

Figure 2: Distribution of hybrid seed inviability across the genus Mimulus...... 38

Figure 3: Phenotypic divergence of perennials in the M. guttatus species complex...... 58

Figure 4: Genotypic divergence of perennials in the M. guttatus species complex...... 60

Figure 5: Sampling locales and hybrid seed inviability overview...... 62

Figure 6: Parent of origin effects on hybrid seed inviability...... 64

Figure 7: Proportion of viable seed produced in crosses between tetraploids and diploids ...... 68

Figure 8: Proportion of viable seed produced in crosses between sympatric accessions of M. decorus and M. guttatus ...... 70

Figure 9: Seed sizes for reciprocal F1s and parents ...... 71

Figure 10: Seed viability for within species crosses...... 73

Figure 11: Seed sizes for within M. decorus crosses...... 76

Figure 12: Proportion viable seed ...... 94

Figure 13: Seed development of IM62 x IMP crosses ...... 96

Figure 14: Seed development of IM62 x Odell Creek crosses ...... 97

Figure 15: Cell width for each cross type ...... 98

Figure 16: Model predictions for patterns of seed viability in F1 backcrosses...... 117

Figure 17: TRD mapping of M. decorus maternal alleles involved in hybrid seed inviability...... 120

Figure 18: TRD mapping of M. guttatus paternal alleles involved in hybrid seed inviability...... 121

xiv

Figure 19: Inversion survey schematic...... 144

Figure 20: Population-level survey of life history traits across perennial species in the Mimulus guttatus species complex ...... 150

Figure 21: Phenotypic differences between M. decorus and annual M. guttatus ...... 152

Figure 22: Phenotype-inversion associations for non-recombinants between annual M. guttatus and M. tilingii...... 154

Figure 23: Schematic for the F3 fine mapping experiment...... 155

Figure 24: Average effect of genotype on a trait for each of 4 recombinant classes ...... 160

Figure 25: Genetics of hybrid dwarfism...... 176

Figure 26: Proportion of underground stolons across all species and time points...... 181

Figure 27: Differences between above- and below-ground stolons for a number of traits ...... 182

Figure 28: Genetic map of underground stolon production ...... 184

Figure 29: Phylogeny for M. guttatus and M. decorus ...... 186

Figure 30: Population averages for hybrid seed inviability averaged across M. guttatus populations ...... 187

Figure 31: Population averages for hybrid seed inviability averaged across M. decorus populations ...... 188

Figure 32: Averages seed set for inter- and intraspecific crosses ...... 189

Figure 33: Chromosome squashes of three diploid populations of M. decorus: ...... 189

Figure 34: Chromosome squashes of three tetraploid populations of M. decorus ...... 190

Figure 35: 2C DNA content for each population of M. decorus...... 191

Figure 36: Total seed counts for (A) IMP and (B) Odell Creek crosses with IM lines ..... 201

Figure 37: Correlation between the proportion of wrong homozygous calls versus the number of missing SNPs in an individual...... 202

Figure 38: Mimulus decorus maternal alleles associated with HSI ...... 203

Figure 39: Mimulus guttatus paternal alleles associated with HSI ...... 204

xvi

Acknowledgements

Like any good species, the formation of a dissertation takes time, substantial selection, and even perhaps a bit of drifting. However, unlike the formation of most species, a good dissertation requires substantial ‘gene flow’ in the form of ideas, thoughts, criticisms, and support. I am forever grateful to numerous folks who have helped me along the way.

Firstly, thank you to my ever-enthusiastic advisor, John Willis. You’ve been a shining beacon of optimism, particularly when I’ve been a bit gruff around the edges.

You’ve also been my most critical cheerleader- honing in my ideas, challenging me to challenge myself, and helping form my worldview of all things genetics and evolution. I am forever grateful for your input and mentorship. Thank you to my committee:

Mohamed Noor for your unwavering encouragement and endless thumbs up. Words cannot describe how thankful I am for your continuing support to me as a graduate student and also a colleague. Kathleen Donohue- thank you for letting me squat in your lab for my first semester here, your thoughtful comments and practical advice. You have taught me to embrace to nuances of the natural world and reveal in life’s complexities!

Mark Rausher, thank you for your helpful criticisms and clear train of thought. You have helped teach me the importance of clarity, logic, and brevity (although the later has been abandoned in this acknowledgement section). Lastly, Tom Mitchell-Olds- thank

xvii

you for your thoughtful input, and attention to detail. Your ability to integrate statistical and molecular genetics ceases to astound me and has been formative for me.

I could not have completed this work without the helpful technical assistance from Mike Windham, Mel Turner, Dick White, John Kelly, and Miguel Flores-Vergara.

Kuddos to many undergraduates that helped me over this journey, particularly Maya

Wilson Brown and Madison Zamora, but also Faye Goodwin, Elysia Su, Sarah Moore,

Dorothy Jones, Julia Bellantoni, Shivam Dave, and Jamie Barry. I am grateful to Andrea

Sweigart, Ben Blackman, Dave Lowry, Yaniv Brandvain, Bob Franks, Elen Oneal, Jessica

Selby, and Kathy Toll for helpful feedback on many of these projects.

Duke University has also been home to amazing colleagues that I am grateful to.

In particular, I would like to thank Jessica Selby, Elen Oneal, Katherine Toll, Annie

Jeong, Kate Ostevik, Kieran Samuk, Lindsay Leverett, Maggie Wagner, Catherine

Rushworth, Liana Burghardt, Carrie Olsen-Manning, Mariano Alverez, Joanna Rifkin,

Rong-Chien Lin, Irene Liao, Ryan Campbell, Lauren Carley, Katharine Korunes, Allan

Castillo, Ashley Troth, and Laryssa Baldridge for all of our science chats and critical input in my own projects.

A dissertation can be a lonely place without good friends. I am forever indebted to Liana Burghardt, Lauren Carley, Eleanor Caves, Amanda Gorton, Patrick Green,

Young-Wha Lee, Lindsay Leverett, Kate Ostevik, Aspen Reese, Catherine Rushworth,

Kieran Samuk, Katie Thomas, Maggie Wagner, Ali Wardlaw and Robert Williamson for

xviii

their friendship. A most special thank you to Emily Josephs, without whom I wouldn’t have made the adventurous leap into the US. You have been my most enthusiastic cheerleader, helped shape me as a scientist, and been the best friend anyone could ask for. I can’t wait to continue our science adventures together!

Lastly, I am grateful to my family for supporting my ambitions. To my dad for instilling a sense of hard work, responsibility, and a good sense of humor. To my mom for being role model of resilience, strength, kindness, and unconditional love. To my brother for his support and encouragement, and his wonderful wife and their two kiddos for their kindness and patience with me. Thank you to Clover the dog, who forces me to take deep breaths and go for walks. I am deeply grateful for Emily Yates’ presence in my life. You inspire me daily. Your ambition, thoughtfulness, and intelligence are unriveled in my heart. Thank you for taking this journey with me, and believing that there was light at the end, even when I couldn’t see it.

xix

Chapter 1: Introduction

The most recent estimates suggest that the earth is home to ~9 million eukaryotic species (Mora et al. 2011). A major goal of evolutionary biologists is to understand how such a tremendous level of diversity can be both formed and maintained through time.

In particular, understanding the reproductive barriers that minimize or halt gene flow, and allow divergence to accumulate between incipient species, has been of long standing interest. Much recent research in speciation has focused on determining the role of natural and sexual selection in the formation of extrinsic barriers, particularly those that occur pre-zygotically (e.g. Seehausen & van Alphen 1999; Ramsey &

Schemske 2003; Rieseberg et al. 2003 McKinnon et al. 2004; Nosil 2007; Martin et al. 2008; van Rijssel et al. 2018), yet the evolutionary forces causing intrinsic post-zygotic reproductive isolation remain relatively less explored. Post-zygotic barriers are important in the speciation process, in part because once these barriers are complete, they are irreversible (unlike pre-zygotic barriers; Price 2007), and in part because almost all species which are considered ‘good’ (i.e. functionally no gene flow among them) possess both pre- and post-zygotic reproductive isolation (Coyne and Orr, 2004).

The evolution of hybrid sterility and inviability were deeply concerning for

Darwin and his contemporaries. This is because hybrids that have intrinsically low fitness, represent a fitness valley between two optima (i.e. the fertile and viable parents).

We now know that intrinsic post-zygotic reproductive isolation need not evolve in the 1

face of natural selection, and could simply evolve a byproduct of population divergence at two or more loci (i.e. BDMIs; Figure 1; Bateson 1909; Dobzhansky 1937; Muller 1942).

While other genetic models also contribute to post-zygotic reproductive isolation (i.e. changes in ploidy, structural genomic changes, or differences in endosymbionts; reviewed in Coyne and Orr 2004), there is substantial evidence that many hybrid incompatibilities evolve through these negative genic interactions, and genetic incompatibilities are common across biological kingdoms (reviewed in Coyne and Orr

2004; Rieseberg & Blackman 2010; Johnson 2010; Presgraves 2010; Blackman 2016).

Theoretical and empirical advances have shown that the number of loci involved in these interactions is predicted to depend on the level of divergence between the species pairs and may involve many interacting loci. These interactions may snowball through time, (Orr 1995; Matute et al. 2010; Moyle and Nakazato 2010) or show diminishing returns (Guerrero et al. 2017). They may stem from nuclear-nuclear or nuclear- cytoplasmic interactions (reviewed in Coyne and Orr 2004). However, the evolutionary causes of these incompatibilities, their incidence and importance in nature, and commonalities across kingdoms remain relatively unsolved.

Figure 1: BDMI model schematic as first proposed by Bateson (1909), Dobzhansky (1937), & Muller (1942).

Here, we aim to briefly synthesize the incidence and importance of intrinsic post- zygotic reproductive isolation in natural populations in the context of the evolutionary processes that are thought to cause them. When possible, we highlight the molecular genetic mechanism of these incompatibilities and argue that knowledge of the molecular genetic mechanism can give insight into the ultimate evolutionary causes. Lastly, we urge future studies to integrate these scales of study to more fully understand how intrinsic reproductive isolation evolves and contributes to the speciation process.

1.1 Gene Duplications and Translocations

Gene duplication is common in natural populations, and many examples of genes causing adaptation are the result of duplicated genes (Lynch & Connery 2001; 3

Zhang 2003; Des Marais & Rausher 2008), however the ubiquity of gene duplications in natural populations also paves the way for the evolution of hybrid incompatibilities

(Werth & Windham, 1991; Lynch & Force 2000; Moyle et al. 2010). Gene duplications can cause hybrid incompatibilities in a number of ways. Incompatibilities can arise if the ancestor to a pair of sister species undergoes a gene duplication event, but these duplicate copies are differentially lost in the subsequent sister species (Lynch & Force

2000), or differentially silenced (Durand et al. 2012). Incompatibilities may also arise between sister taxa if only one species underwent a gene duplication event, but subsequently looses the ancestral copy (i.e. Zuellig & Sweigart 2018). In all of these cases, an F1 hybrid will have the genotype A-B- (where ‘-‘denote a non-functional or silenced gene copy, and the A and B locus represent the duplicates of a given gene). This

A-B- individual will give rise to ¼ ‘--‘ gametes, and if selfed or inbred, 1/16 ‘----‘ individuals. Thus, depending on the function of the duplicated genes involved, either ¼ of the gametes or 1/16 of the F2 offspring will be inviable (resulting in F1 sterility or F2 inviability, respectively). The same patterns of hybrid incompatibility will arise if a single gene is translocated within the genome, so as to create a 2-locus system where species differ in which locus carries the functional and null alleles (Masley et al. 2006; reviewed also in Maheshwari & Barbash, 2011).

There are some examples of species pairs that have evolved reproductive isolation due to differential duplicate loss/gene translocation in plants (Oka 1974; Bikard

et al. 2009; Mizuta et al. 2010; Yamagata et al. 2010; Zuellig & Sweigart 2018), fungi

(Scannell et al. 2006), and animals (Masley et al. 2006). Maheshwari and Barbash (2011) suggest that plants might be expected to exhibit more gene-duplication related incompatibilities, because plants are more likely to undergo whole genome duplication than animals (Mable 2004). By this reasoning, animal lineages which have experienced more ancient whole genome duplication should have higher rates of reciprocal gene loss and reproductive isolation, but in teleost fishes this does not appear to be the case

(Kassahn et al. 2009). Further, Muir and Hahn (2015) show that many incidences of ancient whole-genome duplication events do not resolve as differential gene copy losses between sister species and therefore suggest that this process likely does not contribute substantially to post-zygotic reproductive isolation. Lastly, the examples of post-zygotic reproductive isolation due to gene duplication in plants (e.g. hybrid sterility in rice-

Mizuta et al. 2010; embryo viability in Arabidopsis thaliana-Bikard et al. 2009; and hybrid chlorosis in Mimulus- Zuellig & Sweigart 2018), all resulted from single gene duplication events (not ancient whole genome duplications). Thus, while there are a handful of convincing examples of the role of gene duplications in the evolution of post-zygotic reproductive isolation, but it’s overall role in speciation may be minimal.

1.2 Genetic Linkage and Hitchhiking

Alleles that cause reproductive isolation may fix between populations, because they are linked to positively selected alleles. While linked selection provides a neutral 5

explanation for the evolution of reproductive isolation, it is challenging to disentangle the effects of positive versus linked selection, and a thorough interrogation of the genetics of reproductive isolation is needed to distinguish the two.

Multiple examples of reproductive isolation evolving through linked selection have been uncovered, in both plants and animals. In Mimulus guttatus, hybrid necrosis between plants growing on copper mine tailings and non-tolerant plants is thought to have evolved, in part, because of linked selection between an allele involved in hybrid necrosis and physically proximate copper tolerant allele (MacNair & Christie 1983,

Wright et al. 2013). In Xiphophorus fishes, advanced hybrid melanoma phenotypes arise when X. helleri- which posses no tail spotting- and X. maculatus- which is polymorphic for tail spots- are mated (reviewed in Orr & Presgraves 2000; Shartl 2008). It was thought that this incompatibility may have evolved via differential sexual selection between these fish species, as tail spotting is a sexually selected trait in many fishes (Meyer et al.

1994; Fernandez & Morris 2008). However, genetic characterization of these melanomas reveal a trio of tightly linked genes: (1) the macro-melanophore determining locus (Mdl) which is responsible for tail spotting, and a pair of duplicated genes: Xiphophorus melanoma receptor kinase 1 and 2 (Xmrk-1 and Xmrk-2, respectively), the second of which has been identified as one of the melanoma-causing locus (Shartl 2008). Lastly, hybrid sterility loci between Mimulus lewisii and M. cardinalis occur inside of chromosomal re-arrangements that also house loci involved in floral evolution (Fishman

et al. 2013), suggesting that linked selection between floral characteristics and sterility loci could explain the presence of hybrid sterility in this system.

1.3 Compensatory Mutations and Systems Drift

It is well understood that there are many different genetic routes to get to a particular phenotypic destination. However, this degree of genetic redundancy can have implications for reproductive isolation between species. Systems Drift is the phenomenon by which a phenotypic optimum is conserved, despite evolution of the underlying genetic architecture (True & Haag 2001). Reproductive isolation manifests when subtle shifts in the genetic pathways involved in gamete formation, growth, or other standard development diverge such that hybrids between these populations then lack the appropriate genetic machinery to achieve the optimal phenotype.

While the name Systems Drift would suggest that only neutral processes are involved in changes to the genetic architecture of a trait, True and Haag (2001) emphasize that systems drift could occur through any number of evolutionary mechanisms. These include purely neutral processes, selection, or their combination.

Shiffman and Ralph (2017) show that incompatibilities involving systems drift can evolve quite easily under neutrality alone. Systems drift could also manifest through the evolution of compensatory mutations, which could evolve in response to slightly deleterious mutations, or to a pleiotropic gene which experiences positively selected mutations for some gene function, but for which a compensatory mutation evolves to 7

conserve a second gene function (Mack & Nachman 2016). Populations that then differ in the presence or type of compensatory mutation could then produce offspring with lower fitness than either parent.

Systems drift has primarily been applied to the divergence in the genetic systems underlying gene expression. This is because gene expression is a polygenic trait which is controlled by both cis and trans acting factors, which is also often under stabilizing selection (Ludwig et al. 2000; Lemos et al. 2013; Khatri & Goldstein. 2015). Despite the conservation of expression levels, regulatory divergence is common between closely related species across all kingdoms of life (reviewed in Mack & Nachman, 2016), allowing for incompatibilities to evolve. Compensatory evolution in gene expression has been reported in diverse groups, including flies, mice, yeasts, and nematodes (Wittkopp et al. 2004; Kuo et al. 2010; Barrière et al. 2012; Goncalves et al. 2012), and aberrant gene expression has been shown to be quite common across the genome in a number of hybrids, some of which also exhibit some form or reproductive isolation (Michalak &

Noor 2003; Landry et al. 2005; Haerty & Singh, 2006; Good et al. 2010; Meiklejohn et al.

2014; Brekke et al. 2016; Larson et al. 2016). However, few studies have been able to disentangle compensatory evolution that causes a reduction in hybrid fitness from gene mis-expression which is a consequence of being in a hybrid background. Recently, Mack and Nachman (2016) showed that many mis-expressed genes in sterile hybrid mice show patterns of compensatory evolution, relative to mis-expression in fertile hybrid

mice. However, we note that hybrid male sterility in this system has a complex genetic basis, involving multiple interacting loci (e.g. Good et al. 2008; Good et al. 2010; Larson et al. 2016), and thus it remains unclear how important this compensatory evolution was in the actual formation of species versus its effect since speciation occurred. Systems drift can theoretically contribute to the initial evolution and subsequent accumulation of intrinsic post-zygotic reproductive isolation, but evidence for a direct role of systems drift in speciation in naturally occurring systems is currently lacking.

1.4 Divergent Ecological Selection

Reproductive barriers can arise if loci responsible for traits under divergent ecological selection also cause RI (Schluter 2009). While the evolution of intrinsic post- zygotic reproductive isolation via ecological divergent selection is theoretically possible

(Agrawal et al. 2011), there are very few examples of genes causing both adaptation and reproductive isolation. Nonetheless, there is some evidence that ecologically divergent selection can result in the evolution of reproductive isolation.

This evidence falls into three general categories: (1) Ecologically divergent species that exhibit post-zygotic reproductive isolation (e.g. reduced hybrid viability between Lake Whitefish ecotypes; Lu & Bernatchez 1998; local adaptation conferred by the cytoplasm in the presence of cyto-nuclear incompatibilities in sunflowers; Sambatti et al. 2008; or hybrid sterility between serpentine-adapted and non-adapted ecotypes of

Collinsea; Moyle et al. 2012). (2) Macroevolutionary patterns of increase intrinsic 9

reproductive isolation with increasing ecotypic differentiation (e.g. with body size in fishes; Bolnick 2006; or generally across plants, invertebrates, and vertebrates; Funk et al.

2006). (3) Experimental evolution studies (e.g. ‘reproductive inferiority’ between flies adapted to foods of different pH levels; de Oliveira & Cordeiro, 1980, or lower growth rate in locally adapted yeast populations; Dettman et al. 2007).

In the first two types of studies there is no evidence that the loci that cause ecotypic differentiation are those that cause hybrid inviability or sterility, and thus are not direct evidence of ecologically divergent selection causing intrinsic post-zygotic reproductive isolation. The most convincing line of evidence for a direct link between ecological divergence and intrinsic post-zygotic reproductive isolation are the studies of

Dettman et al. (2007) and Anderson et al. (2010), which found partial inviability and sterility between yeast colonies which were differentially adapted to high salinity or low glucose habitats. Genetic studies show that mutations conferring higher salinity tolerance and low glucose tolerance (PMA1 and MKT1, respectively) interact to cause lower growth rate (Anderson et al. 2010), showing that intrinsic post-zygotic reproductive isolation can empirically evolve via ecological divergence.

The evolution of intrinsic reproductive isolation via divergent ecological selection has been a tantalizing hypothesis for many researchers. While it is theoretically possible, and can evolve in laboratory settings, concrete examples from natural populations are lacking. Knowing the identity of more genes that confer reproductive

isolation will help in determining how common ecological divergence in the evolution of post-zygotic RI is in nature.

1.5 Mutation Order Speciation in Response to the Environment

Mutation order speciation is the process by which reproductive isolation evolves as a byproduct of independent adaptations to the same (or a similar) selective pressure

(Schluter 2009). While each mutation might be advantageous in both populations, by chance, these different mutations conferring a similar selective advantage fix in different populations, and cause reproductive isolation when found in the same hybrid background. While Schluter (2009) proposes that Mutation-Order speciation also describes reproductive isolation conferred through conflict (such as those outlined in

Nosil & Crespi 2012), here we use a more restricted sense of the term to mean reproductive isolation which is conferred through convergent adaptation to similar ecological pressures through different genetic means. We restrict our discussion of mutation-order speciation to ecological adaptations, because a rich and more specific body of literature exists to explain the evolution of conflict-driven RI.

The conditions that allow for ecologically related Mutation-Order speciation to occur are extremely restrictive, and as such very few examples of mutation-order speciation are found in the literature. Mutation-order speciation essentially requires that the two convergently adapting populations experience no gene flow, because as soon as

gene flow occurs, selection will simply favor the most adaptive mutation (or in cases where the mutations are equally favorable, the selective benefit of each mutation will depend on their relative frequencies; Nosil & Flaxman, 2011). Despite these restrictions, mutation order speciation is theoretically possible. Ono et al. (2017) show that

Saccharomyces cereviseae populations that have been independently adapted to grow on the fungicide nystatin also evolve reproductive isolation from one another. In this case, the mutations which cause nystatin resistance are those which also interact to cause lower growth. In addition, Dettman et al. (2007) note that hybrids between parallel- adapted, replicate populations of yeasts (which were adapted to either high salinity or low glucose environments) showed some reduction in fertility relative to the parental populations. However, to our knowledge, there are no examples of ecologically related mutation-order speciation from wild populations.

1.6 Auto-Immune system [Co]-Evolution

In many systems, hybrid inviability has been linked to dysfunctional autoimmune systems. The general hypothesis is that host genomes are constantly evolving in the face of microbial communities (usually thought to be pathogenic-

Bomblies & Weigel, 2007, although a potential role for symbiotic microbes has been suggested- i.e. Brucker & Bordenstein 2013). Positive selection for novel microbial- response phenotypes will lead to divergence between populations that differ in their microbial communities. Hybrids between these populations may thus experience 12

incomplete or irregular innate immune responses, leading to hybrid inviability. This inviability is often manifest as stunted growth, necrotic tissue, and eventually lethality.

In plants, hybrid necrosis and dwarfism are extremely common barriers between species (i.e. Bomblies & Weigel, 2007; Bomblies 2009), and segregate within species

(Bomblies et al. 2007), even between neighboring stands (Świadek et al. 2017). Hybrid necrosis in these cases is caused by spontaneous auto-immune responses which result in inappropriate programmed cell death of healthy cells, causing necrosis and/or stunted growth (Bomblies et al. 2007; Alcazar et al. 2009). Pathogen-response has also been implicated in fungal hybrid dysfunction, particularly in filamentous fungi in which hybrid inviability is a relatively common reproductive isolating mechanism (Glass &

Kaneko 2003). In animals, incompatibilities involving the immune system are seemingly uncommon, although the MHC has been implicated in pre-zygotic barriers and rates of diversification (Eizaguirre et al. 2009; Karvonen & Seehausen 2012; Malmstrøm et al.

2016). One potential example is that of hybrid inviability associated with immune failure in Nasonia wasps (Brucker & Bordenstein 2013). Thus, general hybrid weakness caused by inappropriate innate immune development is at least possible in plants, animals, and fungi, though certainly this mechanism of RI appears most prevalent in plants.

Further evidence of pathogen response being causally related to hybrid necrosis comes from systems in which the genetic basis of hybrid necrosis has been linked to mismatches between epistatically interacting pathogen response genes (i.e. in

Arabidopsis, Capsella, tomatoes, and lettuce; Kruger et al. 2002; Bomblies et al. 2007; Jeuken et al. 2009; Sicard et al. 2015). One model of epistatically interacting immune genes is that of ‘R’ guard genes and their guardees. R proteins act as guards to specific guardee proteins which are commonly targeted for degradation by pathogens (van Der Biezen &

Jones 1998; Dangl & Jones 2001). An immune response is signaled when a change in the confirmation of the guardee protein is detected by the R guard protein (Dangl & Jones,

2001). Thus, mismatches between guard and guardee can result in constant autoimmune responses in hybrid plants. In fungi, hybrid inviability is associated with an incompatibility involving Het genes, which normally function to perform non-self recognition during vegetative hybridization (Glass & Kaneko 2003, Fedorova et al. 2005;

Kaneko et al. 2006; Paoletti & Saupe, 2009). It has been suggested that Het and R genes may play analogous roles which may result in their shared roles in hybrid inviabilities

(Bomblies & Weigel, 2007). In Nasonia wasps, the genetic basis of hybrid inviability remains unknown, although mis-expression of innate immune genes has been documented (Brucker & Bordenstein 2013).

While we have many examples wherein hybrid inviability is due to mismatches in autoimmune related genes, the exact drivers of their rapid evolution are mostly unknown. Scenarios such as host-pathogen co-evolution, positive selection induced by local adaptation to new pathogen environments, or balancing selection and diversification of response genes within and among populations have been

hypothesized. Both balancing and positive selection appears to be occurring in both plants and fungi defense genes. Many R genes exist as balanced polymorphisms (Stahl et al. 1999; Caicedo et al, 1999; Mauricio et al. 2003; Caicedo & Schaal. 2004; reviewed in

Tiffin & Moeller 2006), and positive selection has driven dramatic allelelic divergence between defense alleles in both plants (Mongradon-Palomino et al. 2002) and Het genes in fungi (Wu et al. 1998; Paoletti et al. 2007). A recent study in the genus Capsella indicates that balancing selection plays an important role in the maintenance of multiple alleles within the outcrosser Capsella grandiflora (Sicard et al. 2015). Interestingly, this polymorphism was divergently sorted among newly derived selfing species (i.e. C. rubella and C. orientalis) and has lead to complex reproductive isolation between these three species, dependent on the haplotype of C. grandiflora (Sicard et al. 2015).

Determining the relative frequencies of balancing selection, local adaptation, or host- pathogen arms races as a cause of immune system evolution is an important next step.

1.7 Parental Conflict

Parental conflict is thought to be an important evolutionary mechanism for the evolution of early-onset hybrid inviability in both plants and animals. In populations with any amount of sexual non-monogomy, optimal resource investment in offspring will differ between females and males. This is because while maternal parentage is assured, non-monogamy will result in variance in relatedness among offspring (Trivers,

1974; Haig & Westoby 1989). Selection should favor the evolution of paternally inherited 15

alleles that increase resource allocation to offspring, even at the expense of other offspring or the mother (Trivers 1974; Haig & Westoby 1989). However, selection should also favor maternally inherited alleles that prevent over-expenditure and promote equal distribution of resources among offspring (Trivers 1974; Haig & Westoby 1989). This difference in resource allocation optima can lead to an arms race between males and females. Incompatibilities can manifest if species diverge in the allelic target of this arms race, leading to a mismatch in maternally- or paternally- derived alleles in hybrids

(Brandvain & Haig 2005; Köhler et al. 2010).

Hybrid inviability caused by parental conflict is manifest through growth and developmental abnormalities in offspring, most often with asymmetric effects in reciprocal F1s (Scott et al. 1998; Vranna et al. 2000). Often, one direction of a cross will exhibit an over-growth of either the nutritive tissues or the developing zygote, while the opposite direction of the cross exhibits precocious and limited development. Patterns of reproductive isolation that are consistent with parental conflict are common in systems in which maternal investment is high and can be manipulated post-fertilization (e.g. placental mammals such as horses and donkeys; Mus, perymiscus, and dwarf hampsters;

Allen 1969; Vranna et al. 2000; Zechner et al. 2004; Berkke et al. 2016 and endosperm- bearing seed plants such as Arabidopsis, Capsella, Solanum, Mimulus, and corn; Lin 1984;

Scott et al. 1998; Rebernig et al. 2015; Wolff et al. 2015; Garner et al. 2016; Oneal et al. 2016;

Lafon-Placette et al. 2017; Roth et al. 2017, Lafon-Placette et al. 2018). In contrast to that

predicted by parental conflict, some systems which have lost post-fertilization resource partitioning still exhibit strong embryo inviability (e.g. orchids which have evolved a

‘dust seed’ phenotype lacking endosperm, but still display strong embryo inviability;

Scopece et al. 2007; 2008). In addition, placentation has also evolved independently in other non-mammalian animals, such as reptiles (Blackburn 2015) and fishes (Pollux et al.

2009), and parental conflict mediated hybrid inviability does not appear to play a major role in these systems (although there is some evidence for conflict; Schrader & Travis

2008; 2013; Pollux et al. 2009; Schrader et al. 2011; Blackburn 2015, Griffith et al. 2015).

Thus, parental conflict may be an important cause of RI in mammal and plant systems, but seems less important in other independent evolutions of maternal provisioning post fertilization.

Much progress has been made over the past decade to understand the genetic mechanisms of parental conflict driven RI. The most common hypothesis in both plants and animals is that incompatibilities result from the disruption of imprinted genes that are are necessary for embryo and nutritive tissue development (Haig & Westoby 1989;

Nosil & Crepsi, 2012). In mammals, support for this hypothesis is mixed. While Mus,

Perimyscus, and Phodopus (dwarf hamster) hybrids often exhibit disrupted imprinting patterns relative to either parent (Vrana et al. 2000; Zechner et al. 2004; Shi et al. 2005;

Brekke et al. 2016), only in Perimyscus do over-placentation phenotypes result from an incompatibility between a pair of normally imprinted loci (wherein hybrids exhibit loss

of imprinting of the maternal X and the paternal copy of the autosomal gene Peg3; Vrana et al. 2000). Despite very similar placental phenotypes, loss of imprinting at Peg3 cannot explain the hybrid abnormalities in either Mus or Phodopus (Zechner et al. 2004; Shi et al.

2005; Brekke et al. 2014), although both exhibit extensive bialleleic expression of normally imprinted genes (Shi et al. 2005, Brekke et al. 2016). In Donkey-horse hybrids

(i.e. hinnies and mules), perturbation of imprinted growth-related genes does not appear to explain asymmetric crossing barriers (Wang et al. 2013).

In contrast, disruption of imprinting status is well documented as the cause of abnormal hybrid seed development in interploidy plant crosses (Josefsson et al. 2006;

Cornejo et al. 2012; Wolff et al. 2015; Burkart-Waco et al. 2015; Florez-Rueda et al. 2016).

Imprinting in plants is caused by a complex interaction of DNA methylation and histone methylation by Fertilization Independent Seed- Polycomb Group Complex (FIS-PcG) and imprinting status is dosage dependent (reviewed in Köhler et al. 2012). In interploidy crosses of Arabidopsis thaliana, triploid block is caused by de-repression of maternal alleles of ADM, SUVH7, PEG2, and PEG9, which are normally only expressed paternally, and are involved in carbohydrate metabolism and endosperm cellularization

(Kradolfer et al. 2013; Wolff et al. 2015). More recent studies have investigated the genetic basis of parental-conflict driven reproductive isolation in diploid-diploid plant crosses, including Arabidopsis (Josefsson et al. 2006; Burkart-Waco et al. 2012), Capsella (Rebernig et al. 2015) and Mimulus (Garner et al. 2016). In the Arabidopsis thaliana x A. arenosa inter-

specific diploid crosses mimic interploidy crosses (Josefsson et al. 2006; Bukart-Waco et al. 2012), and hybrid incompatibility in this cross is associated with improper regulation of multiple normally imprinted maternal alleles, including MEDEA, MEIDOS, and casually linked with PHERES1 (Josefsson et al. 2006). Similarly, crosses between two diploid species of Capsella show elevated levels of normally imprinted genes, such as

AGL62, PHERES1, and ADM (although the later occurs too late in development to be causally associated with hybrid seed failure- Rebernig et al. 2015). Interestingly, both interspecific diploid crosses in Capsella and interploidy crosses in Solanum share overlap in the gene categories that were deregulated in inviable hybrid seeds (Cornejo et al. 2012;

Rebernig et al. 2015; Lafon-Placette & Köhler, 2016). Hybrid inviability between M. guttatus and M. tilingii, show several regions of the genome that have parent-of-origin effects on seed set (Garner et al. 2016). The genes responsible for hybrid incompatibilities- and whether hybrid dysfunction results from their improper imprinting- remain to be discovered.

1.8 Sex Chromosome formation and speciation

Sex chromosomes have long been known to play an important role in post- zygotic reproductive isolation (Haldane 1922). Indeed, the two rules of speciation both invoke sex chromosomes: (1) Haldane’s rule- that hybrid barriers first arise in the heterogametic sex (males in XY systems, and females in ZW systems; Haldane 1922;

Presgraves 2008; Brothers & Delph 2010), and (2) the large ‘X’ effect, whereby the X 19

appears to be significantly enriched for loci involved in post-zygotic reproductive isolation (Masley & Presgraves 2007). Several mutually non-exclusive hypotheses explaining both of these phenomenon are common in the literature, including fast evolution on the X chromosome (Charlesworth et al. 1987; Presgraves 2008); X regulation during spermatogenesis (Jablonka & Lamb 1991; Larsen et al. 2017), and meiotic drive and selfish evolution (reviewed in the selfish evolution section; Tao et al. 2001; Phadnis

& Orr 2009; Meiklejohn & Tao 2010). Here we focus on the link between the formation of sex chromosomes, sex chromosome turnover and speciation.

Conflict between sexes can arise due to the expression of characters that have different selective optima in males versus females (i.e. secondary sexual characters; Rice

1987; Zhou & Bachtrog 2012; Bachtrog 2013; Wright et al. 2017). Linkage between these loci and the sex determining locus on a non-recombining sex chromosomes can resolve this conflict (Wright et al. 2018). However, sex chromosomes are known to evolve rapidly in structure, gene content, and expression (Bachtrog et al. 2011). Differential specialization or degeneration of neo-x or y chromosomes among isolated populations could result in hybrids that lack the developmentally fundamental genes. Recently, studies in fish and beetles have proven instrumental in understanding how the formation of sex chromosomes impacts reproductive isolation. Sex chromosome turnover is incredibly common in fishes (Kitano & Pietchel 2012), and recent studies in stickleback map reproductive isolation to these neo sex chromosomes (although we note

that post-zygotic RI mapped to the ancestral X; Kitano et al. 2009). In the Mountain Pine

Beetle, East, West, and Southeast populations show almost no ecological or phenotypic differentiation, but suffer from substantial hybrid male sterility, and some inviability

(Bracewell et al. 2017). These populations all possess a neo Y chromosome, but differential neo Y specialization and degeneration between these populations is thought to be the primary driver of hybrid sterility (Bracewell et al. 2017).

1.9 Selfish Genetic Elements

Post-zygotic RI may also evolve through genomic host-parasite dynamics. Selfish genetic elements are genetic materials that are transmitted to the next generation more than would be expected by Mendelian inheritance. These elements experience strong selection, not because they benefit the organism as a whole, but because they need only benefit themselves (Werren et al. 1988). Instead, some selfish genetic elements actually impose a cost to the host, and selection favors the evolution of repressors. When populations which are divergent in the co-evolution of selfish genetic elements and their repressors are mated, reproductive isolation may manifest because cryptic selfish elements are unveiled. Most often, this results in hybrid sterility, although there are also cases of hybrid inviability (i.e. Ferree & Barbash 2009; Phadnis et al. 2015). As further evidence accumulates, selfish genetic elements may represent one of the most important evolutionary mechanisms by which post-zygotic reproductive isolation evolves (Johnson

2010). This selfish evolution can be further broken into three evolutionary mechanisms. 21

Firstly, selfish genetic evolution via meiotic drive can result in hybrid sterility

(Frank 1991; Hurst & Pomiankowski 1991; reviewed in Meiklejohn & Tao 2010). Drivers and their repressor systems are common in nature, and often associated with sex chromosomes (Taylor 1994; Phadnis & Orr 2009; Dyer et al. 2013; reviewed in Jaenike

2001), although not always (i.e. Herrmann et al. 1999; Fishman & Willis 2005; Dawe et al.

2018; Higgins et al. 2018). Drive associated hybrid sterility has been shown in multiple

Drosophila species pairs (Tao et al. 2001; Orr & Irving 2005; Phadnis & Orr 2009), yeasts

(Zanders et al. 2014), and stalk-eyed flies (Wilkens et al. 2014). Most of these examples of drive associated hybrid sterility are sex linked (i.e. Tao et al. 2001; Orr & Irving 2005;

Phadnis & Orr 2009; Wilkens et al. 2014). This bias may be due to the ease at which sex- linked drive is discovered (as X-linked drive results in an observable sex ratio distortion, whereas autosomal drive does not necessarily produce an obvious phenotypic effect).

Alternatively, drive may intrinsically be easier to evolve on sex chromosomes due to reduced recombination which can link the drive and responder locus (and avoid the formation of ‘suicide chromosomes’; Hartl 1974; Charlesworth & Hartl 1978; Burt &

Trivers 2006; although certainly autosomal drive in regions of low recombination is also quite possible, such as near centromeric regions; Fishman & Saunders 2008). The relative commonality of sex linked drive and sterility may be an important explanation for

Haldane’s rule (Frank 1991; Hurst & Pomiankowski 1991; Meiklejohn & Tao 2010).

However, more definitive tests of the role of sex chromosomes in the relationship

between meiotic drive and hybrid sterility will be needed when more examples of meiotic drive related hybrid sterility have been amassed.

Secondly, cytoplastic male sterility (CMS) is a common post-zygotic barrier in plants and one of the most well characterized hybrid incompatibilities (e.g. in Mimulus-

Fishman & Willis 2005; sunflowers- Sabar et al. 2003; maize- Wise et al. 1987; tobacco-

Gerstel et al. 1978; wild carrot- Linke et al. 2003; petunia- Bentolila & Hanson 2001;

Brassica- Brown 1999). CMS is thought to evolve via arms-race dynamics between cytoplasmic selfish elements that causes pollen sterility and restorer alleles that repress selfish cytoplasmic elements. Cytoplasmic male sterility is often caused by recombinational mishaps which result in chimeric gene sequences in mitochondrial open reading frames (reviewed in Hanson & Bentolila 2004; Chase 2007). Thus these selfish elements are transcribed with mitochondrial promoter regions or upstream mitochondrial genes. Almost all restorer of fertility (rf) alleles have been genetically mapped to PPR genes (polcentric peptide repeat genes), which are an exceptionally large gene family in plants responsible for policing organelle expression (Hanson & Bentolila

2004; Chase 2007). Although note that rf2 in Maize is not a PPR gene, but rather aldehyde dehydrogenase and is thought to inhibit pollen sterility by ameliorating excesses mitochondrial expression of the selfish element (Cui et al. 1996; Hanson &

Bentolila 2004). Some crosses have implicated multiple fertility restoring alleles (i.e. maize- Wen et al. 2003; rice-Kazama & Toriiyama 2003; Mimulus- Barr & Fishman 2010).

In animals, a special form of cytoplasmic male sterility/inviability is that exhibited by cytoplasmically inherited microbes, such as Wolbachia systems in Drosophila and other insects (reviewed in Coyne & Orr 2004; Turelli 1994; Bordenstein et al. 2001;

Dyer & Jaenike 2004; Smith et al. 2016; though see Cooper et al. 2017). The evolutionary dynamics of cytoplasmically inherited microbes is somewhat similar to that of cytoplasmic male sterility in plants, in that matrilineal inheritance of selfish elements can cause reproductive isolation. But here the similarity ends. Unlike CMS in plants,

Wolbachia infection does not have a restorer allele, and both theory and empirical work show that this incompatibility can be fleeting, as Wolbachia infected individuals enjoy higher reproductive success, and thus infection will spread to all reproducing members

(Turelli & Hoffmann 1995). Reproductive isolation may be maintained only under specific conditions in which two lineages which have competing Wolbachia strains which cause inviability in both directions of the cross (Keeling et al. 2003).

Thirdly, the evolution of transposable elements and repressor systems has been hypothesized to play a major role in speciation (Ginzburg et al 1984; Johnson 2010;

Castillo & Moyle 2012; Serrato-Capuchina & Matute 2018). The connection between speciation and TEs has come in many flavors through time, including hybrid dysgenesis caused by P, I, and hobo elements (Kidwell 1979; Bingham et al. 1982; Rubin et al. 1982), genome shock (Michalak 2009; Martienssen 2010; Castillo & Moyle 2012), and heterochromatic regions being associated with incompatibilities. Hybrid dysgenesis is

the set of correlated traits (including increased sterility, gonadal atrophy, and higher male recombination) that manifest in hybrids between different Drosophila strains

(Kidwell 1983). Hybrid dysgenesis occurs between parental strains that vary in TE content (P elements, I elements, hobo elements; or the TE family ‘Penelope’- Rubin et al.

1982; Bingham et al. 1982; Yannopoulos et al. 1987; Blumensteil & Hartl 2005), wherein dysgenesis manifests when a female lacking in TEs is mated to TE abundant males, but not the reverse cross. However, most of this work has been between lab and wild strains of D. melanogaster, and to our knowledge, there is no evidence of co-occurring species pairs in which hybrid dysgenesis is associated with reproductive isolation. Secondly, differential TE load or TE communities between parents may lead to misregulation of

TEs in hybrids, and cause reduced fitness (Michalak 2009; Martienssen 2010; Castillo &

Moyle 2012). While TE misregulation in hybrids is theoretically possible, there is limited evidence that this genomic shock contributes substantially to post-zygotic reproductive isolation in natural populations of either plants or animals.

The strongest evidence to date for the connection between transposable elements and reproductive isolation is that several genes which are known to be incompatibility loci interact with heterochromatic regions of the genome (such as hybrid male inviability in the Hmr/Lhr/gfzf system- Satyaki et al. 2014; Phadnis et al. 2015; and OdsH, which causes hybrid sterility between D. simulans and D. mauritiana- Ting et al. 1998), or are themselves heterochromatic regions of the genome (such as Zhr which has been linked

with hybrid female inviability between D. melanogaster and D. simulans- Ferree &

Barbash 2009). These genes have often undergone a long history if diversifying selection

(Ting et al. 1998; Barbash et al. 2003; Brideau et al. 2006), consistent with an arms race between TEs and their repressors (Johnson 2010). The association between post-zygotic incompatibilities and repetitive elements is much more tangential in plant systems, although there may be a role for TE regulation in hybrid seed inviability. Endosperm development is controlled by maternal or paternal specific expression of genes, and expression of these imprinted genes is associated with remnant TEs (Gehring et al. 2006;

Martienssen 2010). Hybrid seed incompatibilities between Arabidopsis thaliana and A. arenosa have been linked with expression of ATHILA- a retrotransposon (Joseffeson et al.

2006), and overall mismatches between TE load in pollen and egg and their respective repressors in the opposing gamete have been hypothesized to contribute to hybrid seed inviability (Marteinssen 2010; Castillo & Moyle 2012).

Many of the genes which have been found to cause post-zygotic intrinsic reproductive isolation are those which are in some way influenced by selfish genetic evolution, suggesting a strong role of selfish genetic elements in speciation. These systems occur in plants, animals, and fungi, although the evolution mechanisms responsible for selfish genetic elements vary between these three Kingdoms. Overall, selfish genetic element evolution and conflict may be a major contributor to post-zygotic reproductive isolation in the wild, but much more research is needed.

1.10 Recombination-related processes

Strong support for the relationship between recombination and reproductive isolation comes from two distinct sets of studies: (1) Studies of genes involved in recombination rates in hybrid male sterility, and (2) the role of mismatch repair systems in yeasts and prokaryotes sterility. Despite shared roles for these genes in reproductive isolation, the inferred mechanisms for their evolution vary between these two systems highlighting multiple ways in which selection for different aspects of recombination within groups can result in reproductive isolation between groups.

The only identified gene causing hybrid incompatibilities in mammals to date is

PRDM9 (Mihola et al. 2009), a gene that is known to control recombination hotspot locations across mammals (Berg et al. 2010; Parvanov et al. 2010; Baudat et al. 2010), and has undergone significant evolution across mammals (Oliver et al. 2009; Thomas et al.

2009; reviewed in Ponting 2011; Baker et al. 2016). PRDM9 is hypothesized to interact with Hstx2, a gene that inhibits repair of double stranded breaks. In hybrids, the wrong combination of PRDM9 and Hstx2 can result in disrupted chromosomal synapsis, asymmetric binding and meiotic arrest, and in turn F1 sterility (Mihola et al. 2009;

Gregorova et al. 2018). Interestingly, recent works have also implicated Meir1- a modifier of global recombination rate- in hybrid male sterility between the same two house mouse species, underling a potentially important link between the evolution of recombination landscapes and the evolution of incompatibilities more generally

(Balcova et al. 2016). However, despite a key understanding of the genetics underling hybrid male sterility in this system, we still lack knowledge of two important aspects of the its evolution: (1) The evolutionary causes of rapid evolution of PRDM9 and other recombination-related genes remain unknown, and (2) it is unclear how ubiquitous the relationship between recombination and speciation is across taxa.

Several hypotheses have been proposed to explain the rapid evolution of PRDM9 across mammals, some of which may apply to other genes involved in recombination evolution. One of the most popular ideas involves a special form of meiotic drive involved in gene conversion (reviewed in Ponting 2011). PRDM9 determines hotspots of recombination, with the strand possessing the hotspot motif becoming the invading strand (Myers et al. 2010; Ponting 2011). However, when this double strand break occurs, the leading strand experiences a gene conversion event, and the hotspot strand is thus written over by the coldspot strand (Myers et al. 2010; Ponting 2011). This has lead some researchers to believe that rapid evolution of PRDM9 is important for maintaining hotspots throughout the genome, particularly on the sex-chromosomes, to maintain appropriate chromosomal synapsis and meiosis. Davies et al. (2016) show that sterility in mice is correlated with binding asymmetry, and hypothesize that differential PRDM9 binding site erosion decreases binding symmetry in hybrids and causes sterility.

PRDM9 is one of the most rapidly evolving genes across diverse groups of mammals (i.e. primates, rodents, and equids; Oliver et al. 2009; Thomas et al. 2009;

Steiner & Ryder, 2013; Baker et al. 2016), and is known to control recombination hotspots in humans and mice (Baudat et al. 2010; Berg et al. 2010). While the link between hybrid sterility and recombination is quite strong in mice, PRDM9 is not associated with hybrid sterility in other systems with rapid PRDM9 evolution (such as equids; Steiner & Ryder

2013). Determining how ubiquitous PRDM9 (and other recombination-related genes) are in species formation is an important step in speciation research.

The second set of studies which link recombination to reproductive isolation come from that of mismatch repair and hybrid sterility in both yeasts and bacteria (i.e.

Hunter et al. 1996; Vulic et al. 1999; Hunter et al. 1996; Greig et al. 2003; Liti et al. 2006). In yeasts, sterility arises from hybrids being unable to produce viable spores (Greig et al.

2003). Increasing sterility between species of Saccharomyces, as well as between populations of the same species- are generally associated with sequence divergence

(except for one case of chromosomal sterility; Liti et al. 2006). By deleting genes involved in mismatch repair, Hunter et al. (1996) were able to restore hybrid fertility hybrid yeasts. In this case, sterility is thought to arise, because recombination is aborted between highly divergent homologous chromosomes (anti-recombination activity; reviewed in Greig 2009). As homologous recombination is necessary for proper meiotic segregation, hybrids which exhibit this anti-recombination activity are more likely to give rise to aneuploid spores (Greig 2009). Thus, hybrid sterility in this case is not caused by a specific genetic basis per se, but is related to general genomic divergence. As

in above, the generality of this phenomenon in plants and animals requires further study.

1.11 Conclusions

Determining the evolutionary mechanisms which cause hybrid incompatibilities is an ongoing goal of evolutionary biology. Although there are some examples in which reproductive isolating alleles evolve purely by neutrality, many genes causing hybrid sterility or inviability show a strong history of natural selection. From the examples to date, natural selection relating to ecological adaptations does not appear to contribute strongly to hybrid sterility or inviability, although it is theoretically possible. One potential exception is that of autoimmune related genes and hybrid necrosis in plants, which may evolve in response to variation in local microbes. Instead, conflict-driven processes, such as parental conflict and selfish genetic evolution appear to dominate. We caution that the number of incompatibilities in which the genic identity of the incompatibility alleles is known are very few, and most are in Drosophila, Arabidopsis or

Mus (and often in species pairs which do not frequency co-occur in nature). Thus, the evolutionary mechanisms responsible for reproductive isolation in naturally co- occurring species pairs remains an open question. We argue that a more holistic approach which incorporates determining the genes responsible for reproductive isolation, their effect in preventing introgression in nature, and the evolutionary mechanisms responsible is the next step in evolutionary genetics of speciation. 30

Chapter 2: The Mimulus guttatus species complex as a model system for the evolution of hybrid inviability and sterility

Mimulus is a diverse genus of wildflowers native to the Pacific Northwest and is a model system for ecology, evolution, and genetics due to its exceptional diversity (Wu et al. 2008, Twyford & Friedman 2015). Key insights have been made in mating system and life history transitions (Martin & Willis 2007, Fishman et al. 2002; Hall & Willis 2006;

Hall et al. 2010, Lowry & Willis 2010); floral evolution (Bradshaw & Schemske 2003;

Cooley et al. 2011; Stankowski et al. 2015; Fishman et al. 2015; Stankowski et al. 2017); structural variants and adaptation (Lowry & Willis 2010; Fishman et al. 2014; Flagel et al.

2014; Chapter 6); adaptation to edaphic conditions (McNair & Christie 1983; Ferris et al.

2017; Lowry et al. 2009; Aeschbacher et al. 2017; Lee & Coop 2018); species coexistence

(Kenney & Sweigart 2016; Ferris et al. 2018); the maintenance of genetic variation (Mojica et al. 2010, 2012; Lee et al. 2016); and speciation (Vickery 1964; Sweigart & Willis 2006;

Fishman & Willis 2006; Martin & Willis 2007; Lowry et al. 2008; Oneal et al. 2016; Garner et al. 2016l Kerwin & Sweiart 2017).

Despite the exceptional body of work which has accumulated for this genus, and in particular the M. guttatus species complex, much remains unknown about the many speices within this group. Work to date has primarily focused on describing annual species (e.g. Gardner & Macnair 2000; Martin & Willis 2007; Habecker 2012; Peterson et al. 2013; Kenney & Sweigart 2016; Oneal et al. 2016; Ferris & Willis 2018), or differences 32

between perennials and annuals (e.g. Hall & Willis 2006; Lowry et al. 2009; Peterson et al.

2016). Several morphological perennial variants have been described based on minimal phenotypic divergence (as described by herbarium vouchers; e.g. Nesom 2012), but much remains unknown about the ecological, phenotypic, and genetic diversity within perennials, as well as potential reproductive barriers which separate them. In this thesis,

I focus primarily on perennials within the M. guttatus species complex. Here, I outline the advances in intrinsic, post-zygotic reproductive isolation which have been achieved using Mimulus. I end by comparing these findings with summaries of the research presented in this dissertation.

2.1 Hybrid Sterility

By far, the best hybrid incompatibility studied in Mimulus is that of hybrid sterility. While crossing surveys suggest that many incidences of hybrid sterility exist

(Vickery 1964; Ferris et al. 2014), here we focus on three examples which have been worked on in depth.

Hybrid male sterility between M. guttatus and M. nasutus arises from two distinct sets of interactions. This first is Cytoplasmic Male Sterility (CMS; Fishman and Willis

2006), which is caused by interactions between a mitochondrial haplotype of M. guttatus which includes a selfish copy of nad6 (Case & Willis 2008), and two tightly linked nuclear inherited pentatricopeptide repeat genes (PPRs; labeled Rf1 and Rf2; Barr &

Fishman 2010). Mimulus nasutus lacks both the selfish and nuclear restorer alleles, and 33

thus when M. guttatus is crossed to M. nasutus as the maternal parent, one quarter of the

F2 offspring are sterile. Based on our current understanding, the CMS-Rf system is geographically restricted to Iron Mountain, OR (Case & Willis 2008). Evidence for an arms race between selfish mitochondrial alleles and fertility restoring nuclear alleles comes from population genetic studies which suggest that the Rf loci in M. guttatus from

Iron Mountain have undergone a recent selective sweep relative to nearby Cone Peak populations of M. guttatus which do not possess the selfish haplotype of nad6 (Case et al.

2016).

The second case of hybrid male sterility between M. guttatus and M. nasutus results from nuclear-nuclear interactions between a dominant M. guttatus allele at HMS1

(Hybrid Male Sterility-1) and a recessive M. nasutus allele at HMS2 (Sweigart & Willis

2006). This system is also associated with reductions in female fertility (Sweigart &

Willis 2006). While the M. nasutus contributing allele at HMS2 is fixed throughout its range, the M. guttatus allele at HMS1 is geographically restricted to the central Cascade

Mountains (Sweigart & Willis, 2007). Population genetic analyses suggest that HMS1 has rose to intermediate frequency through strong natural selection, although this allele exists within a nearly invariant 320 kb haplotype block, and so may have risen in frequency through linked selection (Sweigart & Flagel 2015). Although the exact mechanism of natural selection is unknown, Kerwin and Sweigart (2017) suggest that

selfish evolution may not be the cause, as patterns of transmission ratio distortion towards the M. guttatus allele at HMS-1 are unique to hybrid genetic backgrounds.

Lastly, hybrid male sterility has been explored in the M. lewisii-cardinalis complex. In this system, sterility is caused by both chromosomal translocations as well as genic interactions (Fishman et al. 2014; Stathos et al. 2014). Genetic mapping shows that sterility loci map to chromosomal inversions which also house loci responsible for pollinator-selected floral morphology (Fishman et al. 2014). Thus, hybrid sterility may have evolved via linked selection within the inversion (Fishman et al. 2014).

2.2 Hybrid Inviability (late stage)

Two reports of late-stage hybrid inviability have been documented in the

Mimulus guttatus species complex. In both of these examples, hybrid death results from inappropriate development or maintenance of vegetative tissues. The first is of hybrid necrosis between lines of M. guttatus which are adapted to grow in high copper soils, and M. guttatus which are copper intolerant (McNair & Christie 1983; Wright et al. 2013).

McNair & Christie (1983) found that the locus conferring copper tolerance is tightly linked or pleiotropic with that of one of the necrosis causing incompatibility loci. While this example was long touted as an example of ecologically divergent selection directly leading to intrinsic post-zygotic reproductive isolation, more recent genetic interrogation of these phenotypes finds they are in fact caused by tightly linked, but genetically separable loci (Wright et al. 2013). 35

The second example of later-onset hybrid inviability is that of hybrid chlorosis in crosses between M. guttatus and its close relative M. nasutus. When crossed, these individuals give rise to 1/16 achlorotic (white) seedlings in the F2 generation, consistent with a recessive nuclear-nuclear incompatibility (Zuellig & Sweigart 2018). Recent work has shown that this incompatibility arises from asymmetric retention of the duplicated gene pTAC14 (Zuellig & Sweigart 2018). In Mimulus guttatus this gene was duplicated, but the ancestral copy has pseudogenized, and is no longer expressed. In M. nasutus, pTAC14 either did not undergo a duplication event, or the duplicate version has been deleted entirely. Thus, in the F2 generation, 1/16th of the progeny receive a nonfunctional pTAC14 allele from M. guttatus and the null allele from M. nasutus. In Arabidopsis thaliana pTAC14 is a major regulator of chloroplast development and null alleles result in white seedlings (Gao et al. 2011). Interestingly, this incompatibility is not fixed throughout the range, perhaps due to differential introgression of the duplicated pTAC14 allele into M. nasutus populations (Zuellig & Sweigart 2018).

2.3 Hybrid Seed Inviability

Hybrid seed inviability is a relatively common outcome of crossing within

Mimulus, and may be a major barrier to gene flow in the genus (Vickery 1978; Figure 2).

While crossing surveys have been completed throughout the genus (Sobel unpublished), the most recent research has focused on members of the M. guttatus species complex.

Mimulus nudatus and M. guttatus are two highly outcrossing species which exhibit 36

strong, and symmetrical hybrid seed inviability (Oneal et al. 2016). Inviability is manifest by two distinct seed phenotypes; an early seed abortion phenotype, and seeds which exhibit inappropriate endosperm development between 5 and 9 Days After Pollinator

(DAP), and die shortly thereafter. These species are divergent in the timing and development of seeds. It remains to be tested whether divergence in seed development is associated with the ecological differences exhibited by M. nudatus and M. guttatus (e.g.

Toll unpublished), or parental conflict.

Mimulus tilingii and M. guttatus also exhibit strong and symmetric hybrid seed inviability (Garner et al. 2016). Genetic mapping in this system reveals a relatively complex genetic basis (Garner et al. 2016). There are eight QTL which affect seed lethality only when maternally inherited, three QTL that affect hybrid seed inviability only when paternally inherited, and two QTL which affect hybrid seed lethality through both parents (Garner et al. 2016). Interestingly, chromosomes 4 and 14 possesses QTLs with opposing effects- i.e. one QTL increased viable seed set when F2s were crossed to

M. guttatus as either the paternal or maternal donor, but decreases viable seed set when the same F2s are crossed to M. tilingii as the paternal or maternal donor. Further work will be needed to test whether these two QTL are indeed allelic, and indicate a role of opposing parent-of-origin effects on hybrid seed lethality.

Figure 2: Distribution of hybrid seed inviability across the genus Mimulus. Navy colored names indicate strong hybrid seed inviability (HSI), red indicates no evidence for HSI, grey indicates unknown status between sister species pairs. Check marks indicate range overlap between the sister species pairs. Tree is based on Beardsley et al. 2004. Crossing data are based on Sobel (dissertation), Garner et al. 2016, Oneal et al. 2016, and data presented in this dissertation. Color bars represent the sections of Mimulus that these species reside within. 2.4 Research Overview

Research of the evolution of hybrid inviability and sterility in the genus Mimulus indicates three major conclusions; 1) hybrid incompatibilities are common and can be quite strong, 2) despite their commonality, hybrid incompatibilities are often geographically restricted or variable, 3) while many of these incompatibilities display a

relatively simple genetic basis involving only two interacting loci, hybrid seed inviability appears to be genetically more complex. These three conclusions are also three of the main conclusions that I draw through my dissertation of the evolution of hybrid seed inviability (HSI) in the Mimulus guttatus species complex.

In Chapter 3, I assess the strength and geographic variability of HSI between M. guttatus and M. decorus. I find that hybrid seed inviability is strong (~50% reduction in survival in hybrids relative to parents), but the dynamics of this incompatibility vary throughout the range. Variation in hybrid seed inviability is caused both by differences in ploidy level and genetic differences between populations. Interestingly, M. decorus comprises at least two genetically distinct clades (referred to as north and south), and these clades exhibit asymmetric, but opposing, hybrid seed inviability with M. guttatus

(i.e. north populations produce inviable seeds when the maternal donor, and south populations produce inviable seeds when they are the paternal donor). Despite minimal phenotypic divergence, close geographic proximity (~30 miles), and being sister to one another, these north and south clades of M. decorus exhibit stronger reproductive isolation with each other than either does with the more genetically distant M. guttatus.

Evidence from reciprocal F1 seed size estimates suggests a role for parental conflict- both between M. decorus and M. guttatus, as well as between the north and south clades of M. decorus. Overall, the evolution of HSI in this group may be caused by rapidly evolving, parent-of-origin resource allocation alleles.

In Chapter 4, I assess the degree of developmental parallelism of reciprocal hybrid seeds between M. guttatus and both the north and south clades of M. decorus. I find substantial developmental parallelism between hybrid seeds made between M. guttatus and either a north or south M. decorus accession depending on the end fate of the seed (i.e. viable versus inviable). Inviable hybrid seeds display signs of paternal excess, with chaotic and mal-developed endosperm, while viable hybrid seeds appear precocious and underdeveloped. Overall, I find that hybrid seed inviability is due, at least in part, to inappropriate development of endosperm tissue, and inviable seeds display exceptional similarity independent of the parental genotypes involved.

In Chapter 5, I examine the genetic basis of hybrid seed inviability between a northern clade accession and M. guttatus. I determine a theoretical genetic model which best explains the patterns of hybrid seed inviability in F1 backcrossing data, and use next-generation sequencing and marker genotyping techniques to map loci associated with hybrid seed inviability. I find that HSI is likely caused by multiple interacting nuclear genes which have parent of origin effects. Each allele has a relatively small effect on the probability of death, and multiple maternally and paternally inherited alleles are needed to cause inviability.

Overall, I find that hybrid seed inviability is a strong, common and diverse barrier to reproduction in relatively closely related, co-occurring species within the M. guttatus species complex. Perennial taxa which exhibit significant hybrid seed inviability

are genetically distinct but exhibit no obvious ecological divergence. Patterns of hybrid seed inviability are instead associated with endosperm-defective development in both directions of the cross, wherein overgrowth phenotypes are more lethal than precocious development. Lastly, HSI between these closely related species is genetically complex relative to other incompatibilities described in this species complex.

Chapter 3: Patterns of hybrid seed inviability in perennials of the Mimulus guttatus sp. complex

3.1 Introduction

The origin and maintenance of species ultimately depends on the formation of barriers to reproduction between newly separated populations. Research to date has highlighted the role of ecological adaptations and sexual selection in the formation of many pre-zygotic barriers- such as pollinator isolation (i.e. Bradshaw and Schemske

2003; Ramsey and Schemske 2003), and habitat isolation (reviewed in Nosil et al. 2005), or extrinsic post-zygotic reproductive isolation (i.e. Arngard et al. 2014). However, relatively less is known about the evolutionary forces driving the formation of intrinsic post-zygotic reproductive isolation (i.e. hybrid sterility or inviability).

The formation of early-stage offspring inviability is one of the most common and powerful intrinsic reproductive barriers in nature (Tiffin et al. 2001; Lowry et al. 2008).

Early-stage offspring inviability commonly manifests in placental mammals and seed plants, such as over-growth related hybrid inviability in dwarf hampsters (Brekke et al.

2014, 2016) and Peromyscus mice (Vranna et al. 2000; Zechner et al. 2004), or hybrid inviability caused by endosperm malformation in Capsella (Rebernig et al. 2015), tomato

(Roth et al. 2017), Arabidopsis (Scott et al. 1998; Wolff et al. 2015), and Mimulus (Oneal et al.

2016; Garner et al. 2016). A common hypothesis to explain the evolution of early-stage offspring inviability is that of parental conflict (Trivers 1974; Haig and Westoby 1989).

Parental conflict arises because maternal and paternal optima for resource allocation to offspring differ (Trivers 1974; Haig and Westoby 1989). Since maternity is guaranteed, selection should favor equivalent resource allocation from mothers to all of her children.

However, in systems with multiple fathers, selection should favor the evolution of paternally derived alleles which enhance resource allocation to offspring- even at the expense of the mother or other offspring. Therefore, a co-evolutionary arms race can occur, wherein selection favors greedier alleles from paternal donors and more repressive alleles from maternal donors. Reproductive isolation can manifest if different populations have evolved different genetic responses to the conflict, resulting in a mismatch between maternal and paternal donors.

This hypothesis shows great promise, because it can explain why embryo inviability is so common in systems which partition resources after fertilization (e.g. seed plants and placental mammals; Vranna 2007; Lafon-Placette & Köhler 2016) relative to systems which do not (e.g. rates of early onset hybrid inviability in birds and frogs is substantially less than in mammals; Wilson et al. 1974; Prager & Wilson 1975; Fitzpatrick

2004), it is consistent with the observation that many of these crossing barriers are asymmetric (e.g. Vranna et al. 2000; Rebernig et al. 2015), and lastly, because most hybrid embryo inviability arises from issues with excessive or limited growth, it is consistent with the idea that incompatibilities in resource allocation are to blame (Vranna et al.

2000; Rebernig et al. 2015; Oneal et al. 2016). Yet, hybrid inviability may also evolve as a

byproduct of many other selective pressures (e.g. ecological reasons), or simply through neutral processes (like gene duplication; e.g. Lynch & Force 2000). Most studies of early- onset hybrid inviability are based on a single population pair, and as such represent a single evolutionary sample (e.g. Vranna et al. 2000; Rebernig et al. 2015; Wolff et al. 2015;

Brekke et al. 2016). Thus, we lack a comparative framework for understanding whether parental conflict can shape hybrid inviability, and what factors affect the magnitude and direction of inviability

Parental conflict can manifest in early onset hybrid inviability by way of two, mutually non-exclusive ways; (1) genetic changes between populations in the maternal vs paternal investment in offspring (as described above), or (2) changes in the copy number of offspring-investment genes between populations (most commonly through changes in overall ploidy; Köhler et al. 2010). The latter of these mechanisms is more relevant to plant systems than mammals (as mammals only very rarely undergo polyploidization; Mable 2004), and does not require a change in gene sequence to elicit a response in hybrid offspring phenotype. Rather hybrid seed inviability can simply be due to relative dosage of genes expressed in the endosperm (often referred to as ‘triploid block’- Comai 2005; Köhler et al. 2010). Indeed, the genetic mechanisms of hybrid seed inviability have been characterized in interploidy crosses in Arabidopsis (Kradolfer et al.

2013; Wolff et al. 2015). In this case, hybrid seed inviability manifests because of improper dosage of normally imprinted genes, resulting in an inappropriate balance of

maternally versus paternally expressed genes in the endosperm (Kradolfer et al. 2013;

Wolff et al. 2015). These relative differences in maternal and paternal inputs into seed development can be quantified by Endosperm Balance Numbers (EBN; Johnston et al.

1980; Birchler & Hart 1987; Lafon-Placette & Köhler 2016). Yet, despite the commonality of hybrid seed inviability as a reproductive barrier among diploid species (Tiffin et al.

2001; Lowry et al. 2008), it remains unknown whether the differences in conflict (or EBN) can explain hybrid seed inviability.

One potential way to test for the role of different factors in the evolution of reproductive isolation is to leverage natural variation in reproductive isolation that may manifest throughout a given taxonomic group (i.e. Case et al. 2016). In many species, the presence and strength of reproductive isolation varies between populations and geographic regions (Sweigart et al. 2007; Bomblies et al. 2007; Case & Willis, 2008; Martin

& Willis, 2010; Sicard et al. 2015; Barnard-Kubow & Galloway, 2017; Bracewell et al.

2017). Variation in reproductive isolation throughout the range of a pair of sister taxa may stem from differential hybridization rates, wherein populations which experience introgression will loose reproductive isolation or because reproductive isolation has evolved differentially throughout the range (potentially due to more complex patterns of allopatry within and between populations of sister species; e.g. Bracewell et al. 2017). By examining the variation in reproductive isolation within a group, we can begin to test specific hypotheses about the ultimate causes of these barriers.

Here we use perennials from the Mimulus guttatus species complex as a system for studying the evolution of hybrid seed inviability. We combine crossing surveys with common garden experiments and population genetics to determine the occurrence and strength of hybrid seed inviability, as well as assess the role of parental conflict in hybrid seed inviability in this group. We find exceptional cryptic diversity within perennials of the M. guttatus species complex which is driven, in part, by rapid evolution of hybrid seed inviability. Hybrid seed inviability is also strongly associated with parent of origin effects on growth, consistent with a role of parental conflict in the formation of reproductive barriers in this group.

3.2 Materials and Methods

3.2.1 Phenotypic and genetic survey of M. decorus

The Mimulus guttatus species complex is a morphologically, ecologically, and genetically diverse group which exhibit tremendous variation in life history and edaphic adaptations. There are many perennial variants within or sister to the M. guttatus species complex (Grant 1924; Nesom 2012). Indeed, perennial M. guttatus comprises several morphological variants, including two inland perennial variants (M. guttatus sensu stricto, and M. decorus), and a coastal variant (M. grandis), which are differentiated primarily on subtle differences in the length of the corolla and the types, densities and locations of stem trichomes (Nesom 2012). We refer to these as inland M. guttatus, M. decorus, and coastal M. guttatus, respectively. The most divergent perennial species is M.

tilingii, which represents a species complex in and of itself (Grant 1924; Nesom 2012;

Garner et al. 2016). Very little is known about the reproductive barriers which may separate perennials within the M. guttatus species complex as well as the genetic relationship between morphologically described perennial variants.

Since species delimitations are based on few and subtle differences, we first sought to assess the degree of phenotypic and genetic divergence between perennials of the M. guttatus complex. To determine the degree of phenotypic divergence between these perennial taxa, we grew six populations of annual M. guttatus, seven populations of coastal perennial M. guttatus, 16 populations of inland perennial M. guttatus, 18 populations of M. decorus and seven populations of M. tilingii in a common garden to assess genetic differences in several life history traits. Seeds of each maternal family were sprinkled onto moist Fafard 4P soil, cold stratified for 1 week, and seedlings were transplanted on the day of germination. For each population, we grew between 1-5 maternal families with 5 replicates per maternal family. Plants were grown under long days in the greenhouse (18h days, 21C days/18C nights).

On the day of first flower, we measured 13 morphological traits, including: days from germination to first flower, node of first flower, corolla length, tube length, corolla width, stem thickness, internode length between the cotyledons and first true leaves, and internode length between first and second true leaves, the length and width of the first true leaf, number of stolons, length of the longest stolon, width of the longest stolon,

number of side branches, length of the longest side branch. To determine how species varied in multivariate trait space, we then completed a PCA of all traits and taxa, with traits scaled to have unit variance. PCs 1 and 2 accounted for roughly 63% of the total trait variance, and according to a broken stick model, were the only PCs to explain more trait variance than expected by chance (PC1 accounted for 44.91% of trait variance, and

PC2 accounted for 18.31% of trait variance).

We also sought to determine the phylogenetic relationship among these taxa. To do this, we whole-genome re-sequenced 10 populations of M. decorus (1 individual per population). We then submitted whole genomic DNA extracted from young bud and leaf tissue to the Duke sequencing facility where 150bp, paired-end read, whole genome

Illumina libraries were created and run in 1 lane on the Illumina 4000 platform. In addition to the 10 M. decorus genomes, we sequenced 1 sympatric M. guttatus genome in the same manner as above. We combined these data with a previously sequenced line of

M. decorus from 11th population (IMP). To assess the relationship between a broader sample of M. guttatus populations and M. decorus, we used publically available GBS sequence data from a range wide collection of M. guttatus (from Twyford & Friedman,

2015; comprising 183 individuals: 70 populations of annual and perennial M. guttatus; 1

M. decorus population which was mis-identified as M. guttatus in the study, and 1 outgroup, M. moschatus var moniliformis). The whole-genome sequences are also part of a larger endeavor to more fully categorize the phylogenetic relationships between

members of the entire M. guttatus species complex (Monohan & Brandvain unpublished).

In total, we present results from 194 individuals; 12 populations of M. decorus, 70 populations of annual and perennial M. guttatus, and 1 accession of the outgroup M. moschatus car moniliformis.

For the whole-genome sequences, we first used Trim Galore! to trim adapter sequences and low-quality read ends from each pair of reads (Kreuger 2015). We then aligned the trimmed sequences to the M. guttatus Version 2.0 hard masked reference genome using BWA mem (Li & Durbin 2009; https://phytozome.jgi.doe.gov/). We used

Samtools to convert .sam files to .bam files, clean the reads, mark duplicates, and sort reads (Li et al. 2009). SNP calling was performed in GATK using haplotype caller for each sample independently (McKenna et al. 2010). We performed a similar set of analyses in the raw reads from Twyford and Friedman (2015), except we split individuals by barcode, discarded low quality reads, and trimmed barcodes using

STACKs before aligning using BWA mem (Catchen et al. 2013). After all samples had

SNPs called using GATK, we performed joingenocaller in GATK to merge all .g.vcf files and have greater confidence in SNP calls. We set the quality parameters so that only

SNPs with quality score of 30, minimum depth of coverage of 3x, a minor allele count of

>5%, and <25% missing data were kept. Indels were removed. We were left with 3840

SNP positions in 180 accessions of M. guttatus from across the range (including annual

and perennial life histories, accessions from 70 populations), 14 accessions of M. decorus

(from 12 populations), and one accession of M. moschatus.

We characterized the genetic composition of species by first performing a PCA using Tassel (Bradbury et al. 2007). PC 1 and 2 accounted for 16.5% of the total variance

(9.8% and 6.7%, respectively). We used the R package ape to build a neighbor-joining tree for M. guttatus and M. decorus, using M. moniliformis as an outgroup.

3.2.2 Crossing survey between M. guttatus and M. decorus

Based on the genomic analyses above, we determined that the perennial variant

M. decorus is a genetically unique clade, while other perennial variants belonging to M. guttatus are not genetically unique (see Results for details). To assess the extent of reproductive isolation between perennials in this group we crossed 19 populations of M. decorus to 1-4 populations of M. guttatus reciprocally, with an average of 13 replicate crosses (ranging from 2-37 replicate crosses; populations are outlined in Table 1). The four populations of M. guttatus that were used include one annual, two inland perennials, and one coastal perennial population, which span the phylogenetic and morphological diversity of M. guttatus (Twyford & Friedman 2015). We crossed individuals within populations to assess the standard levels of viable seed production for each population. For each fruit, seeds were counted and categorized as viable or malformed and appearing inviable, and the proportion of viable seeds was calculated.

To confirm that levels of seed inviability as assessed by morphology were related to the 50

ability of a seed to germinate, we also plated an average of 100 seeds per cross type onto

0.6% agar plates across 5 replicates (20 seeds per replicate plate. In crosses where less than 100 seeds were available, plates were still seeded with 20 seeds with fewer replicate plates). We did this for both directions of the cross, as well as within population crosses for each species. Plates were cold stratified for one week, then left to germinate under long days in the Duke growth chambers for three weeks. Final germination proportion was recorded. Hybrid seed viability as measured by morphological defects was correlated with actual germination (r2=0.66, p<<0.0001), but due to the low r2, so we use both morphologically defined inviability as well as germination ability to calculate total reproductive isolation (as in Ramsey & Schemske 2003):

Average hybrid seed viability in interspecific crosses (DxG and GxD) RI = Average seed viability in intraspecific crosses (DxD and GxG)

(where ‘G’ and ‘D’ refer to Mimulus guttatus and Mimulus decorus, respectively), and the symmetry of reproductive isolation, given as:

(proportion viable seeds, DxG) − (proportion viable seeds, GxD) symmetry of RI = (proportion viable seeds, DxG) + (proportion viable seeds, GxD)

Where values are bounded between -1 and 1. Values closer to -1 indicate an asymmetry where seeds fail when M. decorus is the maternal donor and values closer to

1 indicate an asymmetry where seeds fail when M. guttatus is the maternal donor, and values close to 0 indicate little asymmetry (but do not denote the severity of hybrid seed inviability).

3.2.3 Assessing evolutionary mechanisms for variation in hybrid seed inviability

Each population of M. guttatus exhibited a relatively consistent crossing phenotype when crossed to multiple populations of M. deorus (i.e. Supplemental Figure

31), but we found substantial variation between populations of M. decorus in both the magnitude and the directionality of hybrid seed inviability (Supplemental Figure 30).

We next sought to assess whether variance in the ability of M. decorus to cross to M. guttatus could be explained by variance in ploidy between populations of M. decorus.

3.2.3.1 Determining ploidy and its effects on hybrid seed inviability

We performed flow cytometry on 1-2 individuals per population of M. decorus to determine total genomic content using the method outlined in Galbraith et al. (1997).

Briefly, we chopped freshly collected, young leaf tissue in Galbraith buffer for each individual, as well as an internal standard, Arabidopsis thaliana (2n=2x, 2C=0.431pg;

Schmuths et al. 2004). Samples in buffer were filtered through a 40 μm, then 20 μm pore- sized nylon mesh. Shortly before analysis, we added 0.1mg/ml of 4’,6-diamidino-2- phenylindole (DAPI) and a small amount of the internal standard to each sample.

Sample nuclei stained for at least ten minutes before being run on a Becton Dickinson

LSRII flow cytometer. Total DNA content was calculated by the following equation:

푠푎푚푝푙푒 퐺1 푝푒푎푘 푚푒푎푛 2C DNA content (pg DNA) = 푥 푠푡푎푛푑푎푟푑 2퐶 퐷푁퐴 푐표푛푡푒푛푡 푠푡푎푛푑푎푟푑 퐺1 푝푒푎푘 푚푒푎푛

We discovered tremendous variation in total genomic content, ranging from 2C=

0.61- 3.06 pg (Supplemental Figure 35). To confirm that populations with larger genomic content were of higher ploidy, we also performed chromosomal squashes on select populations which spanned the range of genome size (0.67pg, 1.06pg, 1.22 pg, 1.43 pg, 2 pg, and 2.96 pg). The individuals with genome sizes smaller than 1.4pg were found to be diploid, while individuals with genomes between 1.43-2.96pg were found to be tetraploid. We thus inferred ploidy for the remaining populations based on these ranges.

For the remaining analyses we use only diploid accessions.

3.2.3.1 Assessing potential correlates of hybrid seed inviability

Variation in post-zygotic reproductive isolation may stem from two, mutually non-exclusive scenarios: (1) a post-zygotic reproductive barrier is being lost differentially throughout the range because of differential rates of hybridization or (2) post-zygotic reproductive isolation evolved differentially throughout the range.

We first assessed whether hybrid seed inviability between M. guttatus and M. decorus varies throughout the range of M. decorus because of variation in hybridization between these two species. We predict that if differential hybridization is responsible for variation in post-zygotic reproductive isolation throughout the range, then populations that occur sympatrically will exhibit the lowest levels of post-zygotic reproductive isolation. To test this prediction, we crossed sympatric collections of M. decorus and M. guttatus (N=2 pairs of populations) and assessed hybrid seed inviability. 53

In contrast, populations may vary in the extent or direction of reproductive isolation, because reproductive isolation has evolved differentially throughout the range. If parental conflict drives the evolution of hybrid seed inviability, then we predict that there should be parent-of-origin effects of resource allocation to offspring, manifesting as a parent-of-origin effect on hybrid seed size. Specifically, populations which have strong conflict should produce hybrid seeds which are large when they are the paternal donor and small when they are the maternal donor when crossed to a population with relatively less conflict. (2) The magnitude of growth defects should correspond to the magnitude of hybrid seed inviability. (3) Based on hybrid seed size and hybrid seed inviability, we should be able to assess the relative strength of parental conflict of each population, and these assessments should have predictive power in subsequent crosses.

To test these predictions, we measured total seed size by photographing seeds from each cross and measuring seed width in ImageJ (Scheider et al. 2012). We then assessed parent-of-origin effects on hybrid seed size by correlating the degree of asymmetry of hybrid seed inviability (in both morphological defects and germination ability) and the degree of symmetry in reciprocal F1 seed size (calculated in the same fashion as asymmetry in hybrid seed inviability). Lastly, we performed further crosses to determine if our conflict designations were predictive, outlined below.

3.2.4 Crossing survey within species

Given the full range of symmetries and magnitudes of hybrid seed inviability between diploid populations of M. decorus and M. guttatus, we infer that some populations of M. decorus have weaker conflict than M. guttatus (i.e. a smaller EBN), while others have stronger conflict (i.e. a larger EBN). Parental conflict theory predicts that these populations which show opposite crossing barriers with M. guttatus should display the strongest reproductive isolation, and this reproductive isolation should be accompanied by differences in reciprocal F1 seed size. To test this, we then performed a within species crossing survey in both M. decorus and M. guttatus. We crossed 12-16 populations of M. decorus reciprocally to three representative populations of M. decorus.

These representative populations of M. decorus varied in their crossing phenotypes with

M. guttatus, and also correspond to three genetically unique clades: 1) The population

IMP, which exhibits a strong and asymmetric crossing barrier when crossed to M. guttatus as the maternal donor (but none when the paternal donor); chromosome squash confirmed diploid, determined to belong to the ‘northern’ clade of M. decorus (discussed in the Results section), 2) Odell Creek, which exhibits a strong and asymmetric crossing barrier with M. guttatus as the paternal donor (but limited incompatibility when the maternal donor); confirmed diploid, determined to belong to the ‘southern’ clade of M. decorus (discussed in the Results section), and 3) Jct 35-48, which exhibits limited to no crossing barrier with M. guttatus and is an inferred tetraploid. We also crossed the four

populations of M. guttatus used in the original survey reciprocally. Each of these crosses was processed as above to score average seed viability based on morphology, average germination rate, and average size.

3.2.5 Statistical Analyses

To test for the presence of hybrid seed inviability, as well as differences in seed size between reciprocal F1s and parents, we performed two types of analyses. We firstly performed ANOVAs for each population with cross type (as a factor, of which the levels were DxD, DxG, GxD, GxG) as the predictor variable (i.e. Seed size ~ Cross Type; Table 6,

Table 7,

Table 8, Table 9). Based on genetic cluster, we also performed analyses in which populations were grouped into clades. In these analyses, we performed a linear mixed model with cross type as a fixed effect and population as a random effect using the lme4 and a type III ANOVA on the resultant model using car packages in the statistical interface R (i.e. Seed Size ~ Cross Type + (1|Population); Bates et al. 2015; Fox & Weisberg

2011). In both cases, we performed a post hoc pairwise T-test with Holm correction for multiple testing to determine which crosses differed significantly.

3.3 Results

3.3.1 Phenotypic and genetic divergence in perennials of the Mimulus guttatus species complex

Based on our characterization of life history traits, we find that M. decorus displays limited phenotypic divergence from other perennial populations of M. guttatus, particularly coastal perennials, although both perennial M. guttatus and M. decorus are phenotypically distinct from perennial M. tilingii and annual M. guttatus (Figure 3). We note that M. decorus does exhibit unique traits, such as the presence of underground stolons, slightly different calyx features (such as the angle at which the end of the calyx meets the peduncle), a layer of evenly spaced, dense, short trichomes coating the stems, and the presence of a deep purple or pink corolla tip. However, populations of M. decorus vary in whether they possess some of these traits, and some populations of M. guttatus also exhibit these features.

Figure 3: Phenotypic divergence of perennials in the M. guttatus species complex. Phenotypic distributions for the first 2 PCs of a PCA involving 13 morphological characteristics for perennials in the M. guttatus species complex, plus annual M. guttatus and M. tilingii. Trait loadings as follows: FT- days to first flower, FN- node of first flower, ST-stem thickness, CW- corolla width, TL- tube length, CL- corolla length, LL- leaf length, LW- leaf width, IL1- first internode length, IL2- second internode length, NS- number of stolons, SL- stolon length, SW- stolon width, SBL- side branch length, NSB- number of side branches.

In contrast, population genetic analyses suggest that M. decorus is a genetically unique group (Figure 4). Samples of M. decorus form a genetic cluster which is distinct from M. guttatus, while populations of M. guttatus which represent morphological variants do not show any genetic separation (e.g. annual and perennial forms of M. guttatus; Figure 4; also see Twyford & Friedman 2015). We note that M. decorus can be categorized into three clusters, from the lower most right-hand corner of Figure 4B: a

northern-diploid group which occur primarily in the central Cascade Mountains of

Oregon, a cluster which incorporates all tetraploid samples from throughout the range, and a cluster which includes the two southern-most diploid samples from the Southern

Cascade Mountains, OR, and is most proximal to the M. guttatus samples. These results are supported by phylogenetic analyses using NJ methods (Supplemental Figure 29).

Preliminary analyses based on whole genome sequences suggest that divergence between northern M. decorus and southern M. decorus, as well as between M. guttatus and each clade of M. decorus is approximately 3% (M. guttatus x Southern M. decorus: π:

0.03, se= 0.000018; M. guttatus x Northern M. decorus: π: 0.03, se= 0.000019; Northern M. decorus x Southern M. decorus π: 0.03, se= 0.000019). For comparison, mean divergence between M. guttatus and M. nasutus is approximately 3-4% (Brandvain et al. 2014;

Brandvain unpublished) and divergence between M. guttatus and M. tilingii is approximately 7% (for all sites; Garner et al. 2016). Thus, both northern and southern clades of M. decorus may represent a relatively recent split, and/or exceptionally high introgression post secondary contact.

Figure 4: Genotypic divergence of perennials in the M. guttatus species complex. A) Geographic distribution of samples included in sequencing from Twyford & Friedman (2015) as well as those generated here. MD= Mimulus decorus, MGA= Mimulus guttatus, annual, MGB= Mimulus guttatus, both annual and perennial forms existed at this site, MGP= Mimulus guttatus, perennial, MM= Mimulus moschatus. Color codes are the same as in B) PCA of genotypic space for all sequenced individuals.

3.3.2 Hybrid seed inviability between M. decorus and M. guttatus

There is significant hybrid seed inviability between M. guttatus and M. decorus, when measured as the proportion of viable seeds determined by morphological assessment (hybrids: mean= 57.29% viable, se= 3.19%; M. decorus within population crosses: 87.47%, se=1.855%; M. guttatus within population: 85.49%, se=6.52%) or germination proclivity (hybrids: mean=52.73% viable, +/- 3.155%, M. decorus within population crosses: 90.22%, se=1.61%; M. guttatus within population crosses: 79%,

se=6.48%). This translates into an average reproductive isolation of 0.37 (se=0.051) and

0.32 (se=0.055) for morphologically assessed hybrid seed inviability and germination proclivity, respectively. Despite a relatively large reduction in the number of viable hybrid seeds produced, there is no evidence of a general reduction in seed set in interspecific crosses relative to intraspecific crosses (i.e. although within-population M. guttatus crosses did produce more seeds than all other cross types; F=4.457, p=0.004, df=1016; Supplemental Figure 32). Thus, while there is a limited role for reproductive isolation conferred by pollen pistil interactions, hybrid seed inviability imposes a substantial reproductive barrier. We next sought to determine whether populations of

M. decorus, M. guttatus, or both varied in the level of reproductive isolation.

Figure 5: Sampling locales and hybrid seed inviability overview. Crossing patterns for four populations of M. decorus, representing the 4 main crossing phenotypes observed: A) no hybrid seed inviability, B) Strong, asymmetric inviability when M. decorus is the maternal parent, C) intermediate, and relatively symmetric hybrid seed inviability, D) strongly asymmetric inviability when M. guttatus is the maternal donor. E) Sampling locales of the 19 populations of M. decorus crossed (grey) and the 4 populations of M. guttatus crossed (red). F) a viable hybrid seed, G) an inviable hybrid seed.

Populations of M. guttatus did not drastically differ in their average hybrid seed viability or average symmetry of hybrid seed inviability when crossed with M. decorus

(Supplement Figure 31). Populations of M. decorus, however, differed substantially in both average hybrid seed inviability and average symmetry (for both morphologically assessed viability and germination; Supplemental Figure 30; Table 6). Northern, southern, and polyploid clades of M. decorus differed significantly from each other in the degree and directionality of hybrid seed lethality with M. guttatus (X2=42.1, df=3, p<0.001), as well as a genetic clade*cross type interaction (X2=152.9, df=5, p<0.0001), suggesting that the direction of hybrid seed inviability differs between these genetic clades.

In general, populations of M. decorus which belong to the northern clade tend to produce inviable seeds when they are the maternal donors in crosses with M. guttatus

(i.e. Figure 5B), although we note one population which did not exhibit significant hybrid seed inviability when crossed to M. guttatus (i.e. LL; Figure 5A). Populations from the southern clade of M. decorus either produce symmetrically inviable seeds or inviable seeds only when they are the paternal donor in crosses with M. guttatus (Figure

5C,D). Tetraploid M. decorus tend to produce intermediate levels of inviable seeds in both directions of the cross or stronger hybrid seed inviability when the tetraploid was the paternal donor (Figure 5C,D), although we note two populations which exhibit little to no hybrid seed inviability (Jct 35-48 & Hwy26; Figure 5A).

Parent of origin effects are displayed in Figure 6, where the average hybrid seed inviability as the maternal vs paternal donor are plotted. Populations which fall along

the 1:1 line denote populations with limited parent-of-origin effects on hybrid seed inviability, whereas points below the 1:1 line denote populations with a stronger hybrid seed inviability when M. decorus is the paternal parent and points above the 1:1 line denote populations with strong seed inviability when M. decorus is the maternal parent.

Figure 6: Parent of origin effects on hybrid seed inviability. Average hybrid seed viability as dad vs average hybrid seed viability as mom, plotted for both (a) morphologically assessed viability and (b) germination. Points are color coded based on inferred ploidy.

3.3.3 The role of whole genome duplication in hybrid seed inviability between M. decorus and M. guttatus

To assess the role of whole genome-duplication in hybrid seed inviability, we completed a combination of flow-cytometry and chromosome squashes to determine ploidy level of the M. decorus populations sampled. We find substantial variation in total

2C genome content (0.61- 3.06 pg; Supplemental Figure 35). Using chromosomal squashes, we find the populations of M. decorus sampled here comprise both diploids

and tetraploids (7 diploid populations, 10 tetraploid populations, and 2 populations for which tissue was not available for flow cytometry; Supplemental Figure 33;

Supplemental Figure 34). Based on chromosomal pairing during anaphase, we find preliminary evidence that some of the tetraploids in this group are of hybrid origin

(allotetraploids; homeologous chromosomes do not pair during metaphase, so all chromosome pairs are readily visible), while others are not (autotetraploids; all four copies of a homologous chromosome pair together and are pulled apart in sets of two, largely resembling diploids). However, further genetic and cytogenetic studies are needed to confirm this inference.

Most tetraploid populations of M. decorus display intermediate to strong hybrid seed inviability, with stronger hybrid seed inviability manifesting when the tetraploid accession was the paternal donor in crosses to diploid M. guttatus (i.e. Figure 5C). We note, however, that at least two populations of tetraploid M. decorus display almost no hybrid seed inviability when crossed to M. guttatus (Hwy 26 and Junction 35-48; Figure 5A; Table 6), suggesting that changes in genome copy are not sufficient to cause hybrid seed inviability in this group. Further support for this notion comes from our focal crosses within M. decorus. We see no hybrid seed inviability when tetraploid accessions are crossed to southern diploid accessions of M. decorus (i.e. Odell crossed to 4x individuals: F= 0.11, df=3, p=0.95; and Jct 35-48

crossed to southern populations: F=2.6, df=3, p=0.07;

Figure 7A,C; Table 7; Table 9 ), but significant hybrid seed inviability when tetraploid M. decorus were crossed with the northern diploid M. decorus (IMP crossed to 4x individuals: F= 78.75, df=3, p<0.0001; and Jct 35-48 crossed to nouthern

populations: F=213.1, df=3, p<0.0001;

Figure 7B,D;

Table 8; Table 9). It is important to note that while levels of hybrid seed inviability differ between these different diploid clades and tetraploid M. decorus, all triploid progeny grown to date are highly sterile, and so reproductive isolation is quite strong in all cases.

Figure 7: Proportion of viable seed produced in crosses between tetraploids and diploids within the M. decorus species complex using three focal populations crossed to 15 populations. (A) Focal plant: 4x individual from Junction 35-48 crossed to all southern diploid M. decorus (B) Focal plant: 4x individual from Junction 35-48 crossed to all northern diploid M. decorus (C) Focal plant: 2x individual from Odell Creek (southern clade) crossed to several tetaploid populations, (D) Focal plant: 2x individual from IMP (northern clade) crossed to several tetraploid populations. Letter groups denote that crosses different significantly from one another. NS indicates that crosses do not differ significantly from one another.

We conclude that polyploidization has likely influenced the evolution of hybrid seed inviability, but genome duplication alone cannot explain the variance in hybrid seed inviability between M. decorus and M. guttatus. Indeed, the diploid populations in our sample cover almost the full range of hybrid seed inviability phenotypes as is seen in the full dataset (Figure 5; Figure 6; Table 6). Therefore, there must be genic differences between populations in the alleles which cause differences in the extent and directionality of hybrid seed inviability. We restrict all further analyses to diploid populations to avoid any confounding effects of inter-ploidy crosses.

3.3.4 Correlates of hybrid seed inviability

To assess whether potential hybridization influences hybrid seed inviability, we crossed sympatric strains of M. decorus and M. guttatus. We find strong and asymmetric hybrid seed inviability in these crosses (F=75.04, df=3, p<0.0001; Figure 8), suggesting that differential hybridization is likely not the cause of differential hybrid seed inviability between M. decorus and M. guttatus.

Figure 8: Proportion of viable seed produced in crosses between sympatric accessions of M. decorus and M. guttatus. D= M. decorus, G= M. guttatus, maternal parent in cross is listed first.

To determine if hybrid seed inviability is associated with parent-of-origin effects on growth, we measured reciprocal F1 seed size. Reciprocal F1 hybrid seeds often varied in size (all populations of M. decorus which exhibit hybrid seed inviability when crossed to M. guttatus show significant or nearly significant size differences in reciprocal F1s;

Table 10). Overall, the diploid northern M. decorus produced F1s with larger seeds when they were the maternal donor, and the diploid southern M. decorus produced F1s with larger seeds when they were the paternal donor in crosses with M. guttatus (northern M. decorus crosses: F=87.4, df=3, p<0.0001; southern M. decorus crosses: F=95.55, df=3, p<0.0001; Table 10). The level of asymmetry in seed size was strongly correlated with the

level of asymmetry in viability- as measured by both morphology and germination

(Germination: r2=-0.53, df=22, p=0.008; Morphological viability: r2=-0.48, df=22, p=0.017), such that the reciprocal F1 seed which is larger tends to be the seed which is inviable.

Figure 9: Seed sizes for reciprocal F1s and parents for (A) northern clade populations, and (B) southern clade populations. Letters denote groups which were significantly different from one another. Correlations between seed size asymmetry and seed viability symmetry for (C) germination and (D) morphologically assessed viability. Blue lines indicate the linear regression line of best fit, and grey denotes the 95% confidence intervals of the linear regression.

Based on the levels of hybrid seed inviability and differences in reciprocal F1 seed size, we can assign populations of diploid M. decorus different levels of parental conflict/ EBNs relative to M. guttatus. In this case, northern populations of M. decorus 71

likely have less conflict and lower EBN than M. guttatus, as they are unable to prevent

M. guttatus paternal alleles from exploiting maternal resources. In contrast, southern populations of M. decorus must have stronger conflict and higher EBN than M. guttatus.

We next sought to determine whether these EBNs had predictive power in future crosses, was also true in this group.

3.3.5 Hybrid seed inviability with M. decorus and M. guttatus

We crossed populations of M. decorus which differed in their inferred EBN. As performing all pairwise crosses was impractical, we instead took a targeted approach, using focal accessions. To do this, we used diploid accessions from the north (IMP) and the south (Odell Creek). We crossed these lines to seven other diploid accessions and each other. These nine populations were then split into two groups: less conflict/lower

EBN than M. guttatus (seven populations) and more conflict/ higher EBN than M. guttatus (two populations). We also crossed the four populations of M. guttatus used above in a pairwise manner to determine if within-species hybrid seed inviability alleles segregate throughout the range of M. guttatus- which exhibited much less variance in

EBN.

Figure 10: Seed viability for within species crosses. Proportion viable seed between (A) IMP and other Northern/ low conflict populations, (B) Odell Creek and northern/low conflict populations, (C) IMP and southern/high conflict populations, (D) Odell Creek and other southern/high conflict populations, (E) between all populations of M. guttatus.

We find nearly complete reproductive isolation between northern and southern clades of M. decorus, but no reproductive isolation within groups (Figure 10;

Morphological inviability: IMP crossed to southern M. decorus: F=3599, df=3, p<0.0001;

IMP crossed to northern M. decorus: F= 0.47, df=3, p=0.626;

Table 8; Odell crossed to northern M. decorus: F=864, df=3, p<0.0001; Odell crossed to southern M. decorus: F=2.8, df=3, p=0.07; Table 7. Germination proclivity: IMP crossed to southern M. decorus: F=201.8, df=3, p<0.0001; IMP crossed to northern M. decorus: F=

0.17.27, df=3, p<0.0001 (due to low germination of within population crosses for Northern accessions),

Table 8; Odell crossed to northern M. decorus: F=42.4, df=3, p<0.0001; Odell crossed to southern M. decorus: F=15.79, df=3, p=0.0002 (due to low germination rate in the Odell within population crosses); Table 7). In addition, we note that geographically and phenotypically distinct populations of M. guttatus do not exhibit any signs of hybrid seed inviability (Figure 10; F=0.772, df=1, p=0.39).

Reciprocal F1s between northern and southern M. decorus crosses vary in seed size, in the direction predicted by parental conflict (Figure 11; IMP crossed to southern

M. decorus: F=39.45, df=3, p<0.0001; Odell crossed to northern M. decorus: F=129.3, df=3, p<0.0001). That is to say, Odell Creek produced larger seeds as the paternal donor and smaller seeds when the maternal donor when crossed to populations with less conflict, and IMP produced larger seeds when it was the maternal donor, and smaller seeds when it was the paternal donor when crossed to populations with more conflict. These differences in reciprocal F1 seed size do not occur in crosses within clades (Figure 11;

IMP crossed to northern M. decorus: F= 0.10.86, df=3, p<0.0001 (due to small seed size for

Northern accessions); Odell crossed to southern M. decorus: F=0.71, df=3, p=0.55;). We note that, although these seeds exhibited parent of origin associated size differences, all seeds in the crosses between northern and southern clades of M. decorus are smaller than either of the parental seeds.

Figure 11: Seed sizes for within M. decorus crosses. (A) IMP crossed to other northern clade populations of M. decorus. (B) IMP crossed to southern clade populations of M. decorus. (C) Odell creek crossed to northern clade populations of M. decorus, and (D) Odell creek crossed to southern clade populations of M. decorus.

Overall, this suggests that the morphological variant M. decorus likely comprises at least three biological species: a northern diploid, a southern diploid, and at least one tetraploid taxon. These species are 1) phenotypically nearly indistinguishable, 2) represent each others closest extant taxa, 3) inhabit populations which are a mere ~30 miles apart (Euclidean distance), 4) but display more reproductive isolation with one another than either does to M. guttatus.

3.4 Discussion

Here we explore the evolution of hybrid seed inviability in perennials of the

Mimulus guttatus species complex. Despite exceptional phenotypic overlap, we find that one morphological variant, M. decorus, is a genetically distinct and monophyletic taxon.

We surveyed 19 populations of M. decorus and 4 populations of M. guttatus for the presence of hybrid seed inviability, and find that hybrid seed inviability between these species is often quite strong, but the magnitude and directionality of reproductive isolation varies throughout the range of M. decorus. While M. decorus is composed of both diploid and tetraploid populations, we find that differences in ploidy cannot fully explain the differences in hybrid seed inviability between M. guttatus and different populations of M. decorus. Using only diploid accessions, we find that the strength of asymmetry in hybrid seed inviability is correlated with asymmetric growth defects between reciprocal F1s, suggesting a parent of origin effect on growth related to hybrid seed inviability. Surprisingly, populations of M. decorus which vary in their ability to cross with M. guttatus were also found to be incompatible with one another. Hybrid seeds in this set of crosses also display parent-of-origin effects on growth, in the direction predicted by parental conflict. Overall, we conclude that hybrid seed inviability in this group appears to be a substantial and rapidly evolving reproductive barrier, and the inviability of hybrid seeds is strongly linked with parent-of-origin effects on growth, consistent with parental conflict. Overall, we suggest that differential

parental conflict may play an important role in creating variation in hybrid seed inviability between perennial members of the M. guttatus species complex.

3.4.1 Phenotypic and genetic diversity with perennials of the M. guttatus species complex

Despite its morphological similarity to M. guttatus, M. decorus is both genetically unique from and exhibits strong reproductive isolation with M. guttatus. Mimulus decorus may not represent a cryptic species per se (as there are some distinguishing traits which can be used to differentiate the two), but the M. decorus group does represent a substantial amount of cryptic diversity within the M. guttatus species complex. Much work in the Mimulus guttatus species complex has focused on species diversity of annual forms (for example, M. nasutus, M. nudatus, M. laciniatus or M. glaucescens; Gardner &

Macnair 2000; Martin & Willis 2007; Habecker 2012; Peterson et al. 2013; Kenney &

Sweigart 2016; Ferris & Willis 2018), or differences between annuals and perennials (e.g.

Hall & Willis 2006; Lowry et al. 2009; Peterson et al. 2016), but relatively less has been done to establish the presence of multiple perennial species within the complex.

Intriguingly, although genetically and phenotypically very similar, the entity which is referred to as M. decorus in this work is likely three or more biological species in nature; a northern clade which produces inviable seeds when the maternal donor when crossed to M. guttatus, a southern clade which produces inviable seeds when the paternal donor when crossed to M. guttatus, and at least one tetraploid taxon. Similarly, work on a

montane perennial variant of M. guttatus (M. corallinus) suggests limited ecological divergence among perennials in the M. guttatus species complex which are located in

California (Peterson et al. 2016), but preliminary crossing studies suggest substantial hybrid seed inviability and F1 sterility between these montane perennial variants and inland M. guttatus (Coughlan & Peterson, personal observation). This work highlights a potential major difference between annuals and perennials in the M. guttatus species complex: annual species generally exhibit distinctive phenotypic divergence, usually linked to obvious ecological shifts, but very few exhibit strong post-zygotic reproductive isolation with M. guttatus (with the exception of M. nudatus). However, many of the perennial species we discover here are phenotypically very similar to M. guttatus and reside in essentially the same habitat, but exhibit relatively strong post-zygotic reproductive isolation. In a perhaps similar vein, in deceptive orchids of the

Mediterranean plants do not offer rewards to pollinators, but attract them either by mimicking either female pollinators (sexually deceptive), or other plants which do offer rewards (food mimics; reviewed in Cozzolino & Widmer 2005). Sexual mimics exhibit strong pollinator isolation, but little post-zygotic reproductive isolation, while food mimics exhibit limited pollinator isolation, but substantial post-mating reproductive isolation (Scopece et al. 2007; 2008). In both Mediterranean deceptive orchids and the

Mimulus guttatus species complex, major transitions in natural history (i.e. life history in the case of Mimulus and sexually deceptive strategies in the case of orchids) are

accompanied by either strong pre- or post-zygotic reproductive isolation, though rarely both. These systems provide interesting study systems for comparing the importance of pre- vs post-zygotic reproductive isolation on hybridization and introgression in nature.

In addition, as there is likely no biological constraint between the evolution of pre- vs post-zygotic barriers as many systems display both (e.g. Ramsey & Schemske 2003;

Christie & Strauss 2018), determining why these systems exhibit such striking dichotomies could be a fruitful avenue of research.

Compared to other post-zygotic reproductive barriers, hybrid seed inviability appears to be a common and relatively strong barrier to reproduction between M. guttatus and M. decorus. Unlike previous works on hybrid sterility or hybrid chlorosis

(Sweigart et al. 2007; Case & Willis, 2008; Case et al. 2016; Zuellig & Sweigart 2018), hybrid seed inviability manifest in the majority of populations surveyed (albeit with 3 of

21 populations showing very limited evidence for hybrid seed inviability). This contrasts with studied of Cytoplasmic Male Sterility or hybrid sterility caused by nuclear-nuclear interactions, in which the barrier loci are restricted to essentially a single population

(Sweigart et al. 2007; Case & Willis 2008). Because of its commonality, strength, and the fact that it stops gene flow in the first generation of hybridization, hybrid seed inviability may represent an important post-zygotic barrier in nature in this group.

Intriguingly, M. decorus exhibits striking variation throughout the range in the extent

and direction of hybrid seed inviability, which may allow us to dissect, at least in part, some of the evolutionary drivers of hybrid seed inviability.

3.4.2 Hybrid seed inviability associated with whole genome duplication

We find that a number of populations of M. decorus are tetraploid. While we have yet to determine the evolutionary origins of these tetraploids (i.e. auto- vs allotetraploids), patterns of chromosomal pairing observed in our chromosomal squashes suggests the presence of both auto- and allotetraploids. Further sequencing will be required to assess the degree of auto vs allotetraploids throughout the range of

M. decorus, as well as the number of origins for each type of tetraploid.

While most polyploid accessions of M. decorus exhibited hybrid seed inviability when crossed to M. guttatus in the direction expected by theory (i.e. larger, inviable seeds when the polyploid accession is the paternal donor; Haig & Westoby 1989; Köhler et al. 2010), two populations show no hybrid seed inviability when crossed to M. guttatus

(HWY 26 and Jct 35-48), suggesting that polyploidization is not always sufficient to create hybrid seed inviability in this group. Hybrid seed inviability in interploidy crosses has been studied extensively (Lin 1984; Scott et al. 1998; Kradolfer et al. 2013;

Lafon-Placette & Köhler 2016; Lafon-Placette et al. 2017), and can serve as an immediate reproductive barrier following whole genome duplication (referred to as triploid block;

Comai 2005, Köhler et al. 2010). In many of these interploidy crosses, fertility can be

restored via genome doubling of the diploid parent, suggesting that changes in the relative dosage between maternally and paternally inherited genes cause hybrid seed inviability (Scott et al. 1998; Josefsson et al. 2006; Lafon-Placette et al. 2017). Variance in the genetic composition of these tetraploid accessions may be responsible for the variance in hybrid seed inviability in interploidy crosses in this group. This may stem from tetraploids being of allo- and auto origin, as well as variance in the potential parental species in the case of allotetraploids. Future work will be needed to determine the causes of hybrid seed inviability in polyploids of the M. guttatus species complex.

Intriguingly, we also find that while most tetraploid accessions exhibit reproductive isolation with M. guttatus, and all tetraploid accessions exhibit nearly complete reproductive isolation with northern M. decorus populations, there is no reproductive isolation conferred from hybrid seed inviability between southern M. decorus and tetraploids. This may relate to the higher EBN inferred from these southern diploids relative to either M. guttatus or northern M. decorus. Ploidy manipulation studies would be useful in testing these hypotheses.

3.4.3 Hybrid seed Inviability and parental conflict

The evolutionary drivers of intrinsic post-zygotic reproductive isolation are still relatively unknown. Here, we test for a role of parental conflict in the evolution of hybrid seed inviability within perennials of the M. gutttus species complex. We find that reciprocal F1s vary in seed size, suggesting a parent of origin effect on growth. In 82

addition, variance in the degree of this asymmetry in hybrid seed size is strongly correlated with hybrid seed inviability. Overall, this suggests that parent of origin effects on growth are linked with hybrid seed failure, consistent with parental conflict.

Parental conflict has been invoked in a handful of studies of hybrid seed inviability between diploid species pairs in nature. Of note is that of hybrid seed inviability which occurs between Capsella rubella and C. grandiflora which produce slightly larger seeds when C. rubella is the maternal donor and C. grandiflora is the paternal donor, and seeds much reduced in size in the opposite direction (Rebernig et al.

2015). In addition, hybrid seed lethality in Capsella manifests in the larger of the reciprocal F1s, as it does in the Mimulus studied here. In diploid accessions of Arabidopsis lyrata and A. arenosa, as well as in wild tomato species, hybrid seeds display exceptional differences in F1 seed size which have been shown to be caused by improper endosperm proliferation (Lafon-Placette et al. 2017, Roth et al. 2017). We note that in both cases, reciprocal F1 hybrids are smaller than either parent. Similarly, in Mimulus we find that in crosses between the northern and southern clade of M. decorus, reciprocal F1s show a parent of origin effect on growth, but reciprocal F1s are smaller than both parents.

Interestingly, in these Arabidopsis, Solanum and Mimulus examples, severe inviability is exhibited in both directions of the cross. This could suggest that in cases of stronger hybrid seed inviability, seeds are aborted earlier, before achieving their final size.

Intriguingly, not all cases of hybrid seed inviability are coupled with asymmetric seed

sizes. In another Mimulus species pair for which seed size has been quantified (M. guttatus and M. nudatus), there is no parent of origin effect on growth or total seed size, although there is substantial, and symmetric seed inviability (Oneal et al. 2016).

3.4.4 Hybrid seed inviability is rapidly evolving

We also find that the morphological variant M. decorus is in fact two distinct diploid species which show extensive hybrid seed inviability- more than either does with M. guttatus. Interestingly, these two diploid species are morphologically almost indistinguishable, inhabit extremely similar habitats, are quite geographically proximate, and are each other’s most close relative. This finding highlights both the exceptional and understated diversity in the Mimulus guttatus species complex, as well as the rapidity at which hybrid seed inviability seems to have evolved in this group.

The fact that the taxa with the most reproductive isolation conferred by hybrid seed inviability are also each other’s closest relative suggests that alleles involved in hybrid seed inviability are evolving rapidly in this group. The accumulation of post- zygotic reproductive isolation has often been assumed to occur in a clock-like manner

(Coyne & Orr 1989; Orr 1995; Orr & Turelli 2001), but as evidence for conflict-driven evolution of reproductive isolation increases, this assumption may prove faulty. While parental conflict would predict rapid evolution of maternal and paternal resource allocation alleles (Trivers 1974; Haig & Westoby 1989), the variance in rates of conflict- driven evolution between northern and southern clades of M. decorus is curious. Much 84

work on the rate of evolution due to parental conflict focuses on the role of mating system or ploidy (i.e. Brandvain and Haig 2005; Rebernig et al 2015; Lafon-Placette &

Köhler 2016; Lafon-Placette et al. 2018 and Scott et al. 1998; Pennington et al. 2008; Lafon-

Placette et al. 2017, respectively). Both the southern and northern clades of M. decorus are diploid, and from phenotypic surveys of these populations all exhibit a floral morphology which is highly outcrossing (i.e. large corollas and significant stigma/anther separation). Yet, southern M. decorus exhibits crossing barriers with M. guttatus that suggest more rapid evolution of these alleles, while northern M. decorus exhibits very little conflict relative to either M. guttatus or southern M. decorus. One possibility is that differences in rates of conflict-driven evolution may be due to differences in clonal growth between these clades. The levels of clonal reproduction would affect the of variance in paternity each clade experiences (wherein clonal populations exhibit less variance in paternity than more sexual populations, and thus may have reduced conflict). However, this hypothesis remains uninvestigated as a source of variance in parental conflict.

3.4.5 Conclusions

Here we investigate reproductive isolation between perennials of the M. guttatus species complex. We find that despite minimal phenotypic divergence between perennials in this group, significant hybrid seed inviability between M. guttatus and the morphologically described variant, M. decorus. We find exceptional variation in the 85

magnitude and direction of hybrid seed inviability that is predominantly due to the population of M. decorus used. Variation in hybrid seed inviability between M. guttatus and M. decorus is related to both differences in genome copy number, as well as genetic differences between diploid populations. Hybrid seed inviability is associated with a parent of origin effect on reciprocal F1 seed size, consistent with parental conflict. We also find that populations of M. decorus that exhibit opposite patterns of hybrid seed inviability when crossed with M. guttatus exhibit the strongest hybrid seed inviability, and again reciprocal F1s differ in seed size. Together, this is consistent with the idea that hybrid seed inviability is associated with parental conflict. Lastly, we use reduced representation sequencing to discover that M. decorus is a genetically unique and monophyletic group, distinct from M. guttatus, despite substantial phenotypic overlap.

Given the levels of hybrid seed inviability between different populations of the morphologically described M. decorus, hybrid seed inviability appears to have rapidly evolved in this group. This work also highlights the potential for cryptic diversity within perennials of the M. guttatus species complex, driven, in part, by the rapid evolution of hybrid seed inviability.

Chapter 4: Developmental trajectories of hybrid seeds

4.1 Introduction

The repeatability of speciation is a contentious topic in evolutionary biology.

Much pre-zygotic reproductive isolation can be attributed to adaptations to the abiotic or biotic environment (Coyne & Orr 2004), and the repeatability of extrinsic reproductive isolation may be directly related to the repeatability of adaptation (i.e.

Rundle et al. 2000; Rolán-Alvarez et al. 2004; Stankowski 2013; Soria-Carrasco et al. 2014;

Stuart et al. 2017). Yet, intrinsic post-zygotic reproductive isolation also often manifests as canonical hybrid phenotypes (i.e. abdominal ablation in hybrid Drosophila species pairs- Orr et al. 1997; Gavin-Smyth & Matute 2013; CMS hybrid sterility phenotypes in plants- reviewed in Hanson & Bentolila 2004; Chase 2007; Rieseberg & Blackman 2010; hybrid necrosis in multiple plant and fungi species pairs- reviewed in Bomblies &

Weigel 2007), and much less is known about the evolutionary mechanisms that cause this repeatability at the phenotypic level. Moreover, in many systems it is unclear whether these common and repeated incompatibility phenotypes are developmentally similar, and whether this can inform our understanding of the drivers of hybrid incompatibilities. A classic hybrid incompatibility phenotype in plants is that of hybrid seed incompatibilities.

Proper seed development is achieved through the effective coordination of three tissues with differing genetic compositions: the maternally derived seed coat, the

developing embryo, and a nutritive tissue that sustains the embryo until germination- the endosperm. In diploid species, seed development begins with a double fertilization event; one in which the ovule is fertilized and the developing embryo is formed, while the other involves the fertilization of a diploid central cell that gives rise to a triploid endosperm tissue. The endosperm is thus comprised of 2:1 maternal:paternal genetic composition. This ratio is essential for appropriate seed development, because many genes exhibit parent of origin specific expression in the endosperm (i.e. imprinting;

Köhler & Weinhofer-Molisch 2010). In Arabidopsis, this is caused primarily by DNA methylation and histone methylation (Köhler et al. 2012), the later of which are controlled by the Fertilization Independent Seed- Polycomb Group Complex (FIS-PcG;

Köhler et al. 2003; Köhler et al. 2005; Gehring et al. 2006; Makarevich et al. 2006; Köhler &

Makarevic 2006; Gehring et al. 2009; Hsieh et al. 2009; Wolff et al. 2011; Waters et al. 2011;

Derkacheva et al. 2016). Indeed, imprinting in plants is thought to have evolved as a mechanism to decrease conflict over paternal and maternal interests in nutritive tissue growth (i.e. parental conflict or Kinship theory; Haig & Westoby 1989).

In Arabidopsis hybrid seed inviability in interploidy crosses can arise because of de-repression of normally imprinted genes, particularly that of ADM, SUVH7, PEG2, and PEG9, which are all associated with carbohydrate sequestration and cellularization in normally developing seeds (Kradofler et al. 2013; Wolff et al. 2015). Overexpression of these alleles causes excessive growth and improper cellularization of the endosperm

associated with paternal excess, while maternal excess is often associated with precocious development in Arabidopsis (Scott et al. 1998; reviewed in Lafon-Placette &

Köhler 2016). These defects in the endosperm are causative to hybrid inviability; both embryo rescues and dosage rescue experiments strongly suggest that in these interploidy crosses, hybrid seed inviability is due specifically to inappropriate development and growth of the endosperm, rather than incompatibilities arising in the embryo itself (Bushell et al. 2003; Josefsson et al. 2006; Dilkes et al. 2008; Erilova et al.

2009; Kradolfer et al. 2013).

While hybrid seed inviability is quite common in diploid species pairs, far fewer studies have determined the mechanistic underpinnings of hybrid seed inviability among diploids. In plants with nuclear endosperm development, such as Arabidopsis and

Capsella, hybrid seed inviability mimic interploidy crosses with a delayed or premature endosperm cellularization, depending on the maternal and paternal donor in the cross, and endosperm defects are intrinsically linked with hybrid seed inviability (Joseffson et al. 2006; Burkart-Waco et al. 2012; Rebernig et al. 2015; Lafon-Placette et al. 2017). In plants with cellular endosperm development (i.e. Solanum and Mimulus), hybrid seed inviability is much less explored, although there is some evidence to suggest that inappropriate replication of endosperm is at least partially causative of hybrid seed inviability in both M. nudatus and M. guttatus inviable hybrid seeds (Oneal et al. 2016), and in wild tomato varieties (Lafon-Placette et al. 2016; Roth et al. 2017).

We previously reported a wide diversity of hybrid seed barriers between M. guttatus and its morphological variant, M. decorus which confer relatively strong reproductive isolation (Chapter 3). We found that M. decorus is in fact multiple biological species; one which exhibits strong asymmetric hybrid seed inviability when it is the maternal donor in crosses to M. guttatus, one which exhibits strong asymmetric hybrid seed inviability when it is the paternal donor in crosses to M. guttatus, and several tetraploid accessions which exhibit variable levels of reproductive isolation via hybrid seed inviability. Here we test whether hybrid seed inviability between different diploid accessions of M. decorus and M. guttatus are associated with similar developmental defects, whether these defects are endosperm related, and whether they are consistent with paternal excess/maternal restriction, hypothesized by parental conflict. In addition, we assess the level of developmental parallelism of inviable and viable hybrid seeds between species pairs which have strongly asymmetric hybrid seed inviability, but in the opposite direction.

4.2 Methods

4.2.1 Growth conditions and plant crossing

Based on previous results, Mimulus decorus exhibits two distinct crossing patterns: the northern clade of M. decorus produces inviable seeds when they are the maternal plant in crosses with M. guttatus, while for the southern clade produces inviable seeds when they are the paternal plant in crosses with M. guttatus. To determine

the extent of developmental parallelism, we chose a representative diploid individual from each crossing type of M. decorus (northern clade: IMP; southern clade: Odell Creek) to cross with an annual population of M. guttatus- Iron Mountain (IM). For each population, we planted seed from 3-4 maternal families or inbred lines. For all plants, seeds were sprinkled onto moist Fafard 4-P soil, left to stratify at 4° Celcius for 1 week, then moved to the Duke greenhouses. Plants were transferred within the first 3 days after germination to one plant per 2” pot, then were grown in long days (18h days, 21C day/ 18C nights).

4.2.2 Crossing Assays

Plants were emasculated and crossed reciprocally, as well as self pollinated to yield an average of 6 crosses for each pairwise combination of maternal families between

IM and Odell Creek, and IM and IMP (IMx Odell range: 2-13 replicate crosses, IMxIMP:

2-10 replicate crosses for a total of 298 crosses). Fruits were allowed to develop until ripe but before dehiscence, then seeds for each cross were counted and categorized as inviable or viable based on morphology for each fruit.

4.2.3 Microscopy and measuring growth

We sought to determine the extent of developmental parallelism across hybrid seeds which show opposite patterns of inviability between these two populations of M. decorus. We next crossed plants, as in above, but collected fruits at 4,6,8,10 and 14 Days

After Pollination (DAP). Fruits were collected and immediately placed in FAA fixative. 91

Plants were processed in a similar manner to White & Turner (2012). In brief, after at least 24 hours in fixative, fruits were gradually dehydrated in a Tetrabutyl-Alcohol dehydration series, then mounted in paraffin wax with ~5% gum elemi resin to increase hardness. Paraffin mounted specimens were then sliced to 8 micron ribbons and mounted onto slides. We performed a staining series using Safranin-O and Fast-Green, which stain for nucleic acids and carbohydrates, respectively. Cover slips were then adhesed for permanent storage.

We visualized and photographed at least three developing seeds from a single fruit for each cross type and time point combination (40 unique combinations, 120 seeds measured). At least 25 endosperm cell widths were measured per seed using ImageJ

(Scheider et al. 2012).

4.2.4 Embryo Rescues

To assess whether hybrid seed inviability was due largely to inappropriate development of the endosperm tissue, we also performed embryo rescue assays on both

IMxOdell seeds and IMPxIM seeds, which would normally be inviable if left to develop fully in the seed pod. For this, we crossed IM62 reciprocally to both IMP and Odell creek, as well as performed self fertilizations for IMP, Odell Creek, and IM62. Fruits were left to mature for 7 or 9 days. Based on our developmental time series, embryos appear to die between 10-14 DAP for both sets of crosses. We plated whole developing seeds for each cross type on a sucrose rich growth media (1/2 MS salts, 0.7% agar, 4% 92

sucrose, adjusted to pH 5.9) at 7 and 9 DAP. Plates were cold stratified for 2 days, then placed in the Duke Greenhouses (18h days, 21C day/ 18C nights).

4.2.5 Statistical Analyses

Cross incompatibilities were determined with a linear mixed model using the package lme4 and a type III ANOVA using car (Bates et al. 2015; Fox & Weisberg 2011) in the statistical interface R. All analyses were completed separately for IMxOdell and

IMxIMP crosses. For both total seed number, and proportion viable seed, we treated cross type as a fixed factor and replicate maternal and paternal lines as random effects

(i.e. Proportion viable seed ~ Cross Type + (1|maternal replicate) + (1|paternal replicate) or total seed count ~ Cross Type + (1|maternal replicate) + (1|paternal replicate)). Pairwise T-tests with Holm correction for multiple testing were performed to determine which cross types varied significantly from one another. Differences in developmental trajectories were also assessed using a linear mixed model, as above. In this case, cross type, DAP, and their interaction were the fixed effects and the seed replicate was treated as a random effect.

4.3 Results

4.3.1 Post-Mating Reproductive Isolation

IMP and Odell Creek show consistent patterns of strong, and asymmetric hybrid seed inviability when crossed with individuals from Iron Mountain, but in opposing directions (Figure 12; Odell Creek: X2= 4598, df=3, p<0.0001; IMP: X2= 2969, df=3, 93

p<0.0001). IMP produces entirely inviable seeds when the maternal donor in these crosses, while Odell Creek produces entirely inviable seeds when the paternal donor

(and some inviable seeds when the maternal donor; Figure 12). As in the previous chapter, inviable seeds appear larger, disc-like, and shriveled.

There is some evidence of pollen-pistil incompatibilities between annual M. guttatus and Odell Creek (Odell Creek: X2= 49.33, df=3, p<0.0001; Supplemental Figure

36), wherein IM lines set less seed when pollinated by Odell Creek than any other cross type. We find no evidence of pollen-pistil interactions between IM and IMP, although we note that IMP has substantially larger ovaries and produces more seeds, regardless of the pollen donor (X2= 23.7, df=3, p<0.0001; Supplemental Figure 36).

Figure 12: Proportion viable seed for (A) IM x IMP crosses and (B) IM x Odell Creek crosses. Maternal parent is listed first.

4.3.2 Developmental Trajectories

The three parental species surveyed here display similar seed development. In all three parents, endosperm is tightly packed around the embryo and continues to grow until around 8-10 DAP. At this point, the endosperm is depleted, presumably to feed the growing embryo. Embryo development among the three parents is also quite similar, with IM displaying a slightly accelerated growth, relative to either Odell or IMP (i.e. they reach a heart-shaped embryo around 10 DAP, while this is not achieved in Odell or

IMP until around 14 DAP; Figure 13; Figure 14). By 14 DAP the embryo comprises most of the embryo sac, with a thin layer of endosperm remaining.

Hybrid seeds from both the IMxIMP and IMxOdell crosses show distinct developmental trajectories based on the maternal versus paternal donor of the cross, but both hybrid crosses display similar seed trajectories, depending on the fate of the seed

(i.e. whether it will remain viable or not). In both cross types, seeds which remain viable

(e.g. IMxIMP and OdellxIM) display a precocious development, with several small endosperm cells that are quickly degraded by the embryo (Figure 13; Figure 14) . By 14

DAP, the only remaining endosperm is a thin layer which surrounds the embryo (Figure

13; Figure 14). The overall seed size is also much smaller than either parent or the reciprocal hybrid. In contrast, the hybrid seeds which will eventually become inviable

(i.e. IMPxIM and IMxOdell) exhibit a chaotic endosperm growth, producing fewer, but larger and more diffuse endosperm cells (Figure 13; Figure 14). Embryos in this direction

of the cross remain small, relative to both parents and the reciprocal hybrid (Figure 13;

Figure 14). By 14 DAP, most embryos were entirely degraded.

Figure 13: Seed development of IM62 x IMP crosses at 4, 6, 8, 10, and 14 Days After Pollination (DAP). Maternal parent is listed first.

These patterns are recapitulated when looking simply at endosperm cell width.

In both crosses, the type of cross significantly effected the endosperm cell width (Table

1; Figure 15). Reciprocal F1 hybrids display the most extreme endosperm cell sizes, with inviable seeds exhibiting the largest endosperm cell size, and viable seeds displaying the smallest endosperm cells. The relationship between cell size and cross type was

dependent on Days After Pollination (DAP; significant interaction term: Table 1; Figure

15).

Figure 14: Seed development of IM62 x Odell Creek crosses at 4, 6, 8, 10, and 14 Days After Pollination (DAP). Maternal parent is listed first.

While the two different crosses display quite similar overall patterns of development, there are some distinctions between hybrid seeds formed between

IMxIMP, and IMxOdell Creek. For example, in IMPxIM crosses, the embryo remains comparatively small, and does not appear to grow past the 6 or 8 DAP. In contrast, seeds

formed from a IMxOdell cross often exhibit a developing embryo up until 10 DAP, suggesting that the seed is aborted slightly later in this cross.

Figure 15: Cell width for each cross type at 4, 6, 8, 10, and 14 Days After Pollination (DAP). I= IM62, N= Northern clade (IMP), S= Southern clade (Odell Creek). Maternal parent is listed first.

Overall, we see striking parallelism between two distinct seed incompatibilities with reciprocal F1 seeds displaying either excessive and chaotic growth patterns or precocious and stunted patterns, depending on whether the seeds are eventually aborted or not. We suggest that hybrid seed inviability in both of these cases is likely a result, at least in part, of defective endosperm development which manifests as chaotic and excessive growth.

Table 1: Mixed model for type III ANOVAs for seed cell size across developmental time for IMP and Odell Creek crossed reciprocally to IM62. Cross Type= inter- vs

intra-specific (DxD, DxG, GxD, GxG; treated as a fixed effect), DAP= Days After Pollination (fixed effect). Seed replicate was treated as a random effect.

IMP Odell Creek

X2 DF p X2 DF p Intercept 295.04 1 <0.0001 123.74 1 <0.0001 Cross Type 119.848 3 <0.0001 14.92 3 0.0018 DAP 36.033 1 <0.0001 0.0029 1 0.96 Cross Type*DAP 42.583 3 <0.0001 32.95 3 <0.0001

4.3.3 Embryo Rescues

Embryo rescues were generally unsuccessful in these crosses. No seeds germinated when collected 7 DAP. Only a small fraction of IM62 and hybrids which were viable (i.e. OdellxIM and IMxIMP) germinated in the 9 DAP collections (i.e. 5-10%).

Overall, this suggests that 7 DAP may be too premature for proper endosperm development of viable and inviable seeds to occur outside of the maternal environment, while only rapidly developing seeds are able survive transplants at 9 DAP.

Unfortunately, determining whether hybrid seed inviability is due solely to endosperm malformation is yet to be determined in this group.

4.4 Discussion

As previously reported, the northern and southern clade of M. decorus exhibit strong, and asymmetric reproductive isolation with M. guttatus, but in opposite directions of the cross. This observation is true for all maternal family/line combinations tested here (as well as multiple populations tested in Chapter 3), suggesting that HSI is

not geographically restricted, as other intrinsic post-zygotic barriers have been found to be (Sweigart et al. 2007; Case & Willis 2008; Martin & Willis 2010; Sicard et al. 2015;

Barnard-Kubow & Galloway 2017). Total reproductive isolation in these crosses was

~52% for both both species pairs, and thus may represent a significant barrier to reproduction. Hybrid seed inviability has been shown to be a formidable barrier to reproduction in other systems (i.e. Tiffin et al. 2001; Lowry et al. 2008), but the developmental causes of this incompatibility have really only been quantified for a small number of diploid species pairs.

Here we show that hybrid seed inviability is strongly associated with developmental defects in endosperm deposition. In inviable directions of the cross, endosperm is chaotically distributed, and the cells are few and large, suggesting a failure to proliferate. In the viable directions of the cross, seeds develop precociously, and endosperm deposition is limited, as witnessed by the general lack of endosperm remaining when seeds are fully developed, as well as the generally smaller size of both the cells and final seeds (shown in Chapter 3). Overall, these developmental patterns fit with a paternal excess/maternal repression model of hybrid inviability, wherein paternal excess is more likely to lead to seed abortion. Unfortunately, given the generally low success rate of the embryo rescues, it remains unclear whether hybrid seed inviability is due strictly to malformation of the endosperm, or whether genetic incompatibilities in the embryo also contribute.

100

These patterns have been shown in wild species pairs of both Arabidopsis and

Capsella, which have a nuclear-type endosperm development (Rebernig et al. 2015;

Lafon-Placette et al. 2017). In these crosses, as well as interploidy crosses, endosperm over-proliferation is accompanied by a failure of endosperm cells to cellularize (Scott et al. 1998; Bushell et al. 2003; Rebernig et al. 2015; Lafon-Placette et al. 2017), and at least in

Capsella, paternal excess is associated with inviability, but maternal excess still produces viable seed (Rebernig et al. 2015). Initial findings from species with cellular-type endosperm development (e.g. wild tomatoes and other Mimulus species pairs) also show endosperm defects as a major cause of hybrid inviability (Roth et al. 2017; Oneal et al.

2016). However, these defects to not represent defects in rates of cellularization, but rather rates of proliferation, cell size and distribution, and optical density (Oneal et al.

2016; Roth et al. 2017).

The developmental trajectories of viable vs non-viable hybrid seeds are strikingly similar between OdellxIM, and IMPxIM crosses, despite the fact that inviability occurs in the opposite cross direction. This suggests that similar evolutionary pressures may indeed be causing hybrid seed inviability within this complex, namely that of parental conflict. Despite considerable developmental parallelism, it remains to be seen how constrained the genetic targets of selection are between these species pairs. In other post- zygotic reproductive barriers, canonical hybrid incompatibilities often share similar genetic bases. For example, hybrid necrosis is often caused by interactions involving R

101

genes (reviewed in Bomblies & Weigel 2007), and CMS almost always involves PPR genes (polcentric peptide repeat genes; reviewed in Rieseberg & Blackman 2010). Yet, whether this remains true for hybrid seed inviability remains to be seen. Indeed, one may predict that the genomic target size for influencing resource provisioning to seeds is presumably larger than the number of genes responsible for auoto-immune function or managing organelle expression, as is the case with hybrid necrosis and CMS, respectively. However, this prediction remains to be tested.

102

Chapter 5: Uncovering the genetic basis of hybrid seed inviability between co-occurring sister species

5.1 Introduction

Genic incompatibilities are a common driver of speciation across the tree of life.

As divergence accumulates between species, so do pairs of alleles which interact to cause a reduction in hybrid fertility or viability. Curiously, within species these

‘incompatibility’ alleles are often essential for normal development, and do not cause a reduction in fitness. This model of speciation, first formalized by Bateson (1909),

Dobzhansky (1937), and Muller (1942) has gleaned much theoretical and empirical support, particularly in the era of modern genetics and genomics (e.g. Brideau et al. 2006;

Bomblies et al. 2007; Phadnis & Orr 2009; Mihola et al. 2009; Zuellig & Sweigart 2018, reviewed in Maheshwari & Barbash 2011, and Blackman 2016). Yet, we arguably know very little about the dynamics of incompatibilities in nature, because many well characterized examples of incompatibility loci are restricted to model systems, such as house mouse, Arabidopsis and Drosophila. Far fewer examples exist for species pairs which co-occur in nature.

A prominent barrier to reproduction in plants and mammals is the formation of embryos which die during development. In seed plants, hybrid seed inviability is quite common, frequently strong, and often asymmetric (Tiffin et al. 2001; Lowry et al. 2008), making it a potentially formidable barrier to gene flow. Hybrid seed inviability is often 103

associated with the malformation of endosperm; a nutritive tissue which is essential for proper embryo development (reviewed in Lafon-Placette & Köhler 2016). Endosperm is a triploid tissue, resulting from the fertilization of the diploid central cell of the ovule by a second sperm cell of the male gametophyte. Early crossing studies determined the importance of this 2:1 maternal:paternal ratio in endosperm development by manipulating the maternal and paternal dosage through interploidy crosses (Johnson et al. 1980; Birchler & Hart 1987). Interploidy crosses often have canonical features: when there is paternal excess, endosperm deposition is chaotic, slower, and often larger, while maternal excess results in precocious and smaller endosperm development (Birchler

1993; Scott et al. 1998; Bushell et al. 2003; Lafon-Placette et al. 2017). Intriguingly, many interspecific crosses mimic these interploidy crosses developmentally (Joseffson et al.

2006; Rebernig et al. 2015; Lafon-Placette et al. 2017).

Asymmetries in hybrid seed inviability have been interpreted as resulting from cyto-nuclear incompatibilities (e.g. Tiffin et al. 2001), yet we now know that many genes within the endosperm are imprinted and imprinting status is essential for normal endosperm development. In particular, both DNA methylation and histone methylation modified by the Fertilization Independent Seed- Polycomb Group Complex (FIS-PcG) are responsible for mediating imprinting in endosperm (reviewed in Köhler &

Makarevic 2006; Köhler et al. 2012). In interploidy crosses in Arabidopsis, endosperm

104

malformation in paternal excess crosses has been shown to result from improper imprinted expression of several paternally expressed genes (Kradolfer et al. 2013; Wolff et al. 2015). Much less work has been done to map the genetic basis of interspecific hybrid seed inviability between diploids, although the evidence to date suggests that interrupted imprinting may play a fundamental role. For example, hybrid seed inviability is associated with a general loss of imprinting status of endosperm expressed genes in some species pairs (Rebernig et al. 2015; Burkart-Waco et al. 2015; Florez-Rueda et al. 2016; Lafon-Placette et al. 2018). In addition, candidate gene approaches in

Arabidopsis, show that expression of maternal PHERES1 in endosperm, which is normally silenced, is causative to inviability between A. thaliana and A. arenosa (Joseffson et al. 2006). However, the limited studies which have assessed the genetic architecture of hybrid seed inviability indicate a relatively complex basis. In both Capsella and

Arabidopsis, modelling suggests 1-3 maternal loci that interact with 2-3 paternal loci, depending on the direction of the cross (Rebernig et a. 2015; Lafon-Placette et al. 2017).

Only one study has mapped loci controlling maternal and paternal contributions to hybrid seed inviability de novo. Garner et al. (2016) map hybrid seed inviability between two quite divergent Mimulus species- a coastal form of M. guttatus and M. tilingii, which produce almost no viable seeds in either direction of the cross. In this case, the genetic basis is quite complex, involving 18 Quantitative Trait Loci (QTLs) which contributed to

105

hybrid seed inviability (Garner et al. 2016). Yet, this species pair is quite divergent

(π=~7%, Garnder et al. 2016), and do not co-occur in nature. Thus, we lack an understanding of the genetic basis of hybrid seed inviability, particularly in young, co- occurring species pairs.

Here we use the more recently diverged M. decorus and its progenitor, M. guttatus to map hybrid seed inviability. These two species co-occur on fine spatial scales throughout the Cascade Mountains of Oregon, Washington, and into British Columbia, and display strong and asymmetric hybrid seed inviability. Mimulus decorus is itself comprised of multiple, reproductively isolated, biological species. In this study we focus on the northern clade M. decorus individuals and co-occurring populations of M. guttatus from Iron Mountain, OR.

5.2 Materials and Methods

5.2.1 Plant material, growing conditions, crosses

To determine the genetic model of hybrid seed inviability between northern clade M. decorus (hereafter IMPO) and M. guttatus (hereafter IM62), we performed F1 backcrosses in which the F1 was backcrossed to both parents in both directions. All F1s resulted from a cross between a maternal IM62 donor and a paternal IMPO donor. We calculated the proportion of viable seeds for each of these four crosses, as well as the proportion viable seed produced when the F1 was self fertilized. We compared these

106

rates of seed inviability to those predicted by a cyto-nuclear incompatibility, as well as models involving incompatibilities between imprinted nuclear loci.

For the models involving imprinted nuclear loci, we used the different backcrosses to determine the maternal vs paternal components of hybrid seed lethality.

The backcrosses towards IM62 are informative of the IMP maternal alleles which contribute to hybrid seed inviability, while the backcrosses towards IMP are informative of the paternal IM62 alleles which contribute to hybrid seed lethality. As hybrid seeds are only incompatible when the wrong combination of maternal IMPO and paternal

IM62 loci interact (i.e. not maternal IM62 and paternal IMPO), we treat the F1xIMP and

IMxF1 crosses are control crosses, and the F1xIM and IMPxF1 crosses as experimental crosses. Under a model in which one maternal and one paternal locus interact to produce aborted seeds, the expected proportion of inviable seeds is 50% for both

IMPxF1 and F1xIM crosses (and roughly 0 for the control crosses). However, both of these backcrosses differed substantially from this expectation (X2=30.3, df=3, p<0.001;

Figure 16), but in opposing directions (IMPxF1 backcross seeds had significantly higher levels of hybrid seed inviability, while F1xIM seeds had much lower rates of hybrid seed inviability). This suggests that the number of maternal and paternal loci interacting to form inviable seeds was greater than two and that the interaction among maternally

107

inherited incompatibility alleles differed from that of paternally inherited incompatibility alleles.

The observation that IMPxF1 backcrosses produced more inviable seed than expected suggests that IM62 paternal loci which interact to form inviable seeds do so redundantly (i.e. while there are many loci which contribute to hybrid seed lethality, only one or few is needed for seeds to abort), while the deficit of inviable hybrid seeds in the F1xIM backcross suggest that IMPO alleles at more than one loci are needed to invoke seed abortion. We integrate these observations into the model in a simplistic way; in the IMPxF1 model, we assume that inheritance of any paternal IM62 allele involved in hybrid seed inviability will result in an aborted seed. Thus the proportion of viable seeds is simply equal to the probability of inheriting multiple, independently assorting IMPO paternal alleles:

푝푟표푝표푟푡𝑖표푛 표푓 푣𝑖푎푏푙푒 푠푒푒푑푠 =0.5n

Where n is the number of loci. In the F1xIM model, we assume that all IMPO alleles involved in hybrid seed inviability must be inherited for a seed to be aborted. Thus, the proportion of viable seeds produced is equal to the probability of inheriting a maternal

IM62 allele at at least one locus:

푝푟표푝표푟푡𝑖표푛 표푓 푣𝑖푎푏푙푒 푠푒푒푑푠 = 1 −0.5n

108

To map the loci involved in hybrid seed inviability, and assess whether these loci have a parent of origin effect, we then grew all four backcross populations (i.e. F1xIM,

IMxF1, IMPxF1, F1xIMP) under long, warm days in the Duke Greenhouses. All seeds were sprinkled onto moist Fafard 4P soil, cold stratified for one week, then transplanted as early germinants. In total, we collected tissue for: 384 individuals for F1xIM, and 300 individuals for IMPxF1, and 96 individuals for each the IMxF1 and F1xIMP crosses in

96-well plates (although note that libraries were made for all IMPxF1 and F1xIM individuals, but only 64 F1xIMP and 68 IMxF1 individuals). Samples were flash frozen in liquid nitrogen and pulverized in a Geno/Grinder 2000.

5.2.2 Determination of HSI associated loci

We extracted high quality DNA using a modified CTAB extraction protocol

(Kelly & Willis, 1998), but included a brief RNAseA incubation after the initial chloroform wash (0.28 mg of RNaseA to each sample, followed by a 30-minute incubation at 37°C). An additional chloroform isoamyl wash was performed to remove residual RNAseA. DNA was precipitated using 5M NaCL and 95% EtOH for 20 minutes, pellets washed with 70% EtOH twice, then left to dry overnight and re- suspended in autoclaved distilled water. Total DNA content was measured for each sample using a Quant-iT PicoGreen dsDNA Assay kit (Invitrogen Life Technologies), and fluorescence was measured using a microplate reader.

109

To prepare genomic libraries for GBS sequencing, we used a modified

Multiplexed Shotgun Genotyping approach with the enzyme Csp6I (MSG; Andolfatto et al. 2011). Briefly, 25ng of each sample was digested, a unique barcode was ligated, then individual samples were pooled into groups of 48 (the maximum number of unique barcodes available). Pooled libraries were subject to a gel size selection using both gel electrophoresis and Ampure XP beads, and DNA fragments between 250-300bp were kept. Pooled libraries were then amplified for 16 cycles of PCR, during which the

Illumina adapters were annealed, so that each library possessed a unique Illumina adapter and each individual had a unique adapter/ barcode combination.

Libraries were run using the Illumina 4000 platform at the Duke Sequencing facility. Samples were split such that 3-5 libraries (144-240) individuals were sequenced per lane (yield= 1.7-5.9 billion bp per library). We combined these with previously sequenced parental DNA (both reduced representation libraries and whole genome sequences).

We split each library by barcode, then removed the barcode and cleaned the sample using Stacks (Catchen et al. 2013). Reads for each individual were then aligned to the M. guttatus hardmasked V2.0 genome using BWA mem (Li & Durbin 2009). We called

SNPs using GATK per individual, then joint genotyped all individuals, plus both GBS data and whole genome sequence data for each parent, again using GATK (McKenna et

110

al. 2010). We then performed the protocol outlined Flagel et al. (2018) to process the resulting VCF file. In brief, SNPs were filtered based on minimum and maximum coverage (i.e. 3-10 reads per individual), minor and maximum allele frequency (0.1-0.9, with individuals from all four crosses combined), and thinned so that only 1 SNP per read tag was selected. This resulted in 12,969 SNPs, for which the average coverage per individual was 6.9 (se=0.013), and at each SNP an average of 187 individuals had a SNP called (se=0.569). SNPs were then polarized by parental genotype so that only alternate homozygous loci remained (Resulting in 5452 SNPs kept). Genotypes for the parents and F1 backcrosses were then assigned based on 100kb windows (resulting in 1583 genotypes calls). We filtered the dataset to remove individuals which showed a high proportion of the wrong homozygous genotype (i.e. >15% of genotypes assigned were of homozygous for the non-backcross parent). These proportions roughly correlated with the amount of missing data (Supplemental Figure 37). We were left with 507 individuals

(F1xIM cross= 295 individuals, IMxF1 cross=46 individuals, IMPxF1=121 individuals,

F1xIM=41 individuals). We then ran custom scripts to determine the frequency of each genotype at each position for each cross. We subsequently filtered the dataset so that only positions in which 20 individuals had genotype calls were kept in the experimental crosses and at least 3 individuals had genotypes called in the control crosses (Resulting in 705 sites in the backcross towards IMP set of crosses and 858 in the backcross towards

111

IM set of crosses. Resulting average number of individuals per genotype call (with standard deviation in brackets) was: IMPxF1: 71 (15.96), F1xIMP= 16 (7.61), IMxF1=11

(5.3), F1xIM= 162 (48)). We performed a Chi-square test at each SNP to determine regions of the genome with significant transmission ratio distortion (TRD).

Unfortunately, the crosses in which we expect little to no TRD (i.e. IMxF1 and F1xIMP crosses) had relatively low number of individuals per genotype window, and thus were not used as a control in this analyses. Instead, we determined whether the counts of each genotype differed from the proportion expected for each cross type. In the IMPxF1 cross, this corresponded to a 1:1 ratio for homozygous IMPO and heterozygous genotypes. In the F1xIM crosses, the genome wide average for the frequency of IM62 homozygotes was higher than expected (i.e. 59%), so we used the expected probabilities of 0.6 for

IM62 homozgous genotypes and 0.4 heterozygous genotypes. We used a significance threshold of 0.01.

We find many regions of the genome which exhibit significant TRD. To confirm

TRD, and assess whether the TRD had a parent-of-origin effect, we also determined genotypes for these regions using fragment length-polymorphism based markers

(Fishman et al. 2014). To this end, we amplified between 1-8 markers for each scaffold for each experimental cross (Table 2). Markers which were significantly distorted in the experimental cross were then amplified in the control cross to determine if TRD had a

112

parent of origin effect. PCR products were pooled based on fragment size, and run on an

ABI 3730x1 machine. Fragment length genotypes were scored by hand in

GENEMARKER (https://softgenetics.com/). All TRD was determined using chi-square tests between the observed and expected ratios of genotypes (i.e. 1:1 homozygous: heterozygous), and subsequent contingency tests between the observed genotypes in the experimental vs control cross to show parent-of-origin effects.

5.2.2 Calculating selection coefficients and multi-locus genotype probabilities

We calculated the selection coefficient against heterozygous genotypes for the most distorted markers for all chromosomes which showed parent of origin TRD. While these calculations are based on the lack of heterozygous genotypes at a given locus, we note that selection is not acting against heterozygotes per se, but rather acting at the level of incompatible allelic combinations of the gametes, i.e. against M. decorus maternal alleles causing HSI loci in the F1xIM cross, and paternal M. guttatus alleles causing HSI in the IMPxF1 cross. To do this, we simply calculated one minus the ratio of survival probabilities of each genotype, given as:

푝푟표푝표푟푡𝑖표푛 표푓 ℎ푒푡푒푟표푧푦푔표푢푠 𝑖푛푑𝑖푣𝑖푑푢푎푙푠 푠 = 1 − 푝푟표푝표푟푡𝑖표푛 표푓 ℎ표푚표푧푦푔표푢푠 𝑖푛푑𝑖푣𝑖푑푢푎푙푠

We performed these analyses for all four cross for each locus in which one cross showed

TRD (Table 3). The genotype frequencies used were based on PCR marker genotypes.

113

We also sought to calculate the probability that surviving individuals were more likely to be homozygous at multiple loci than would be expected, given the genotype frequencies at each locus. To do this, we used the PCR marker genotype dataset which was filtered to discard any individuals with missing data. We counted how many loci were heterozygous versus homozygous for each individual, then summed the number of individuals which were homozygous for 0-N loci, where N is the number of loci included. We compared these counts to the expected probabilities, which we calculated as the probability of inheriting certain sets of genotypes based on the observed genotype frequency. For example, the probability of being entirely homozygous for all loci sampled would be:

푝푟표푏푎푏𝑖푙𝑖푡푦 (ℎ표푚표푧푦푔표푢푠 푎푡 푎푙푙 푙표푐𝑖)

= 푝푟표푝(ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 1) ∗ 푝푟표푝(ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 2)

∗ 푝푟표푝(ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 푁)

Where N= the number of loci, and prop() represents the proportion of individuals which were homozygous at a given locus. This calculation is suitable for calculating the probability that an individual is entirely homozygous or heterozygous, but there are multiple combinations of genotypes which would result in an individual being homozygous for a given number loci and heterozygous for the rest. To calculate these probabilities, we simply summed the probabilities of all possible genotype combinations

114

which would result in a certain number of loci being homozygous versus heterozygous.

For example, in a situation involving 3 loci:

푝푟표푏푎푏𝑖푙𝑖푡푦 (ℎ표푚표푧푦푔표푢푠 푎푡 푡푤표 푙표푐𝑖, ℎ푒푡푒푟표푧푦푔표푢푠 푎푡 표푛푒)

= (푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 1

∗ 푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 2

∗ 푝푟표푝표푟푡𝑖표푛 ℎ푒푡푒푟표푧푦푔표푢푠 푎푡 푙표푐푢푠 3)

+ (푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 1

∗ 푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 3

∗ 푝푟표푝표푟푡𝑖표푛 ℎ푒푡푒푟표푧푦푔표푢푠 푎푡 푙표푐푢푠 2)

+ (푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 2

∗ 푝푟표푝표푟푡𝑖표푛 ℎ표푚표푧푦푔표푢푠 푎푡 푙표푐푢푠 3

∗ 푝푟표푝표푟푡𝑖표푛 ℎ푒푡푒푟표푧푦푔표푢푠 푎푡 푙표푐푢푠 1)

We then performed a chi square test between the observed counts and the expected ratios. We performed these analyses including only loci which were significantly disorted in one direction of the cross (i.e. Table 11AD) as well as including loci which were significantly and suggestively distorted in one direction of the cross

(Table 11B,C). We note that the number of possible genotypic combinations when including all 5 loci implicated in the IMPxF1 cross was relatively high, and so we instead simplified these calculations by using the average genotype frequency across all loci (0.6 homozygous, 0.4 heterozygous). In this case, the probabilities were:

115

푝푟표푏푎푏𝑖푙𝑖푡푦 (ℎ표푚표푧푦푔표푢푠 푎푡 푇 푙표푐𝑖)

= 푛푢푚푏푒푟 표푓 푢푛𝑖푞푢푒 푔푒푛표푡푦푝𝑖푐 푐표푚푏𝑖푛푎푡𝑖표푛푠 ∗ 0.6푇 ∗ 0.4푁−푇

Where N is the total number of loci.

5.3 Results

5.3.1 Hybrid seed incompatibilities are not caused by cyto-nuclear interactions

Patterns of hybrid seed inviability in the F1 backcrosses rule out cytonuclear incompatibilities as the causative genetic model of hybrid seed inviability (Figure 16).

Specifically, that F1xIM62 backcrosses show a significant proportion of inviable seed

(~29%), while under cyto-nuclear incompatibilities involving any number of nuclear loci, the expected frequency should be 0, as these F1s had the IM62 cytoplasm. Our data represent a significant departure from this expectation (X2=20.8, df=3, p<0.001). In addition, we find approximately 1/3 of F2 offspring derived from a self-fertilized F1 are also inviable, which, again, is inconsistent with a cyto-nuclear incompatibility.

116

Figure 16: Model predictions for patterns of seed viability in F1 backcrosses. (A) Predictions (red circles) for cyto-nuclear interaction and real data, (B) Predicted level of hybrid seed viability against the number of IM62 paternal loci (white circles) and the observed value of hybrid seed inviability in the IMPOxF1 cross. (C) Predicted level of hybrid seed viability against the number of IMPO maternal loci (white circles) and the observed value of hybrid seed inviability in the F1xIM cross.

These patterns of hybrid seed inviability from F1 backcrosses also suggested a more complex genetic model than a simple 1 maternal: 1 paternal allele incompatibility

(X2=30.3, df=3, p<0.001). Our simplistic set of models based on hybrid seed inviability exhibited in the F1 backcrosses suggest that the hybrid seed incompatibility between

IM62 and IMPO is likely caused by ~2 maternal loci in which both are needed to cause inviability and ~4 paternal loci which act redundantly to cause hybrid seed inviability

(e.g. any one of these loci is sufficient to cause seed abortion). Levels of hybrid seed

117

inviability in F2 seeds are consistent with this model (X2=2.749, df=1, p=0.097; although note that the F2 data are also consistent with other models).

5.3.2 The genetic architecture of hybrid seed incompatibilities between M. guttatus and M. decorus is complex

Genetic mapping of hybrid seed inviability indicates a complex genetic basis.

TRD analyses based purely on the GBS data resulted in several spurious associations

(Figure 38 and Figure 39). We found 11 QTL in the F1xIM cross and 5 in the IMPxF1 cross. Many of these significantly distorted regions were simply caused by distortion at a handful of positions (i.e. 13 of the 16 significant QTLs were caused by distortion at 1-3 genotype windows). In contrast, the LG2 QTL in the IMPxF1 cross, as well as the LG 4 and 8 QTLs in the F1xIM cross show a number of genotype windows which were significantly distorted. We then used PCR-based marker genotypes across all fourteen linkage groups to confirm these QTL, as well as other QTL associated with hybrid seed inviability.

Using additional marker genotyping, we found 2-3 M. decorus alleles which exhibit TRD when maternally inherited, but not when paternally inherited (Figure 17; Table 2). We also found 3-5 M. guttatus alleles which exhibit TRD when paternally inherited, but not when maternally inherited (Figure 18; Table 2). However, levels of distortion at each given marker are relatively low (i.e. 57-67% homozygous instead of 50%). Nonetheless, selection coefficients against heterozygotes ranged from 0.24-0.51 (Table 3). We also calculated whether combinations of genotypes were more common than expected based on the frequency of each genotype at each locus individually and find no significant differences (

118

Table 11).

119

Figure 17: TRD mapping of M. decorus maternal alleles involved in hybrid seed inviability. Frequency of the IM62 homozgous genotype across three chromosomes with suggestive or significant parent of origin TRD: (A) LG8 (significant), (B) LG8 (significant), and (C) LG4 (suggestive). Unfilled circles are the data from GBS determined genotypes. Black outlined circles = frequencies for the F1xIM cross (i.e. experimental cross), grey outlined circles = frequencies for the IMxF1 cross (i.e. control cross). Filled triangles are the genotype frequencies determined by PCR marker genotyping. Grey triangles = data from the IMxF1 cross (control cross). Black, red, pink triangles = data from the F1xIM (experimental) cross, where color signifies significance (black= not significantly different from the Menedlian expectation, red= significantly different from both the Mendelian expectation and the marker genotypes in the control cross, pink= suggestive of parent of origin TRD, but not significant different from Mendelian expectations and control cross genotypes.

120

Figure 18: TRD mapping of M. guttatus paternal alleles involved in hybrid seed inviability. Frequency of the IMPO homozygous genotype across five chromosomes with suggestive or significant parent of origin TRD: (A) LG3 (significant), (B) LG6 (significant), (C) LG13 (significant), (D) LG2 (suggestive), and (E) LG14 (suggestive). Unfilled circles are the data from GBS determined genotypes. Black outlined circles = frequencies for the IMPxF1 cross (i.e. experimental cross), grey outlined circles = frequencies for the F1xIMP cross (i.e. control cross). Filled triangles are the genotype frequencies determined by PCR marker genotyping. Grey triangles = data from the F1xIMP cross (control cross). Black, red, pink triangles = data from the IMPxF1 (experimental) cross, where color signifies significance (black= not significantly different from the Menedlian expectation, red= significantly different from both the Mendelian expectation and the marker genotypes in the control cross, pink= suggestive of parent of origin TRD, but not significant different from Mendelian expectations and control cross genotypes).

121

Table 2: Marker genotypes to confirm parent of origin segregation distortion. Experimental crosses are listed in black, control in grey. Experimental values that are significantly different from the control cross are bolded. Values that are not significantly different from the control cross, but are trending in that direction are bolded and italicized.

Scaffold IMPxF1 F1xIMP IMxF1 F1xIM

LG1 MgSTS47 0.5

MgSTS436 0.36 0.53

MgSTS287 0.37 0.53

LG2 MgSTS737 0.42 0.44

MgSTS787 0.53 0.47

MgSTS699 0.53 0.5

MgSTS616 0.58 0.47 0.43

MgSTS513 0.51 0.55 0.58

MgSTS530 0.5

MgSTS652 0.49 0.57 0.58

MgSTS774 0.43 0.54 LG3 MgSTS474 0.66 0.49 0.47 0.44

MgSTS214 0.6 0.51 0.52

MgSTS592 0.54 0.56

MgSTS802 0.52 0.56

LG4 MgSTS573 0.47 0.56 0.62

MgSTS695 0.48 0.57 0.64

MgSTS750 0.63 0.63

MgSTS362 0.59

MgSTS449 0.43 0.56 0.56 0.66

MgSTS455 0.51

LG5 MgSTS762 0.51 0.51

MgSTS682 0.52 0.38

MgSTS784 0.52 0.38

LG6 MgSTS508 0.53

MgSTS72 0.56

MgSTS25 0.45

MgSTS724 0.6 0.47 0.5

MgSTS105 0.55 0.52

122

LG7 MgSTS806 0.48 0.51 0.64

MgSTS441 0.61

MgSTS369 0.61

LG8 MgSTS675 0.44 0.57

MgSTS330 0.54 0.38 0.67

MgSTS666 0.67

MgSTS743 0.67

MgSTS76 0.5 0.42 0.55

LG9 MgSTS745 0.25 0.67 0.68

MgSTS443 0.49 0.51

MgSTS849 0.36 0.6 0.61

LG10 MgSTS846 0.5 0.4 0.47

MgSTS82 0.49 0.46

LG11 MgSTS843 0.52 0.58 0.56

LG12 MgSTS394 0.49 0.47

MgSTS113 0.56 0.43 0.51

LG13 MgSTS526 0.64 0.44 0.56

MgSTS236 0.57 0.59 0.56

MgSTS361 0.59 0.58 0.54

MgSTS345 0.64 0.54

MgSTS276 0.61 0.42

LG14 MgSTS368 0.52 0.51

MgSTS448 0.49 0.54

MgSTS798 0.57 0.52

MgSTS783 0.51 0.47 0.53

Table 3: Selection coefficients for regions associated with parent of origin effects. Black= experimental cross, grey= control cross. Regions which are significantly distorted are bolded.

Scaffold marker s_IMPxF1 s_F1xIMP s_IMxF1 s_F1xIM

LG2 MgSTS616 0.275862069 -0.127659574 -0.325581395 LG3 MgSTS474 0.484848485 -0.040816327 -0.127659574 -0.272727273 LG4 MgSTS449 -0.325581395 0.214285714 0.214285714 0.484848485 123

LG6 MgSTS724 0.333333333 -0.127659574 0

LG7 MgSTS806 -0.083333333 0.039215686 0.4375

LG8 MgSTS330 0.148148148 -0.631578947 0.507462687

LG13 MgSTS276 0.360655738 -0.380952381 0.039215686

LG14 MgSTS798 0.245614035 0.076923077 0.113207547

5.4 Discussion

5.4.1 Hybrid seed incompatibilities are not caused by cyto-nuclear interactions

Given the strong parent-of-origin effects on hybrid seed lethality, one might suspect that incompatibilities causing embryo death are cyto-nuclear. However, we show that patterns of hybrid seed inviability in F1 backcrosses, as well as F2s reject this hypothesis. Indeed, hybrid seed inviability appears to be caused by interactions between nuclear loci with parent of origin effects. This is perhaps not surprising, as many genes regulating endosperm development are imprinted (Köhler et al. 2003; Makarevich et al.

2006; Gehring et al. 2006; 2011; Wolff et al. 2011; Pignatta et al. 2014; Klosinska et al. 2016;

Zhang et al. 2014; Waters et al. 2013). Curiously, very few to no genes in the embryo are imprinted in plants (Gehring et al. 2011; Waters et al. 2013), suggesting that hybrid seed inviability stems from incompatibilities in the endosperm, rather than the actual hybrid embryo. In both interploidy and interspecific crosses, imbalance of normally imprinted loci has been associated with hybrid seed inviability (Wolff et al. 2015; Rebernig et al.

2015; Burkart-Waco et al. 2015; Florez-Rueda et al. 2016; Lafon-Placette et al. 2018). The

124

only known gene to influence hybrid seed inviability, PHERES1, is normally maternally silenced, but mis-expression of PHERES1 during endosperm development causes hybrid seed inviability between A. thaliana and A. aeronsa (Joseffson et al. 2006). Yet, we know very little about the allelic variants which cause hybrid seed inviability in natural populations. Here we determined a preliminary genetic map for hybrid seed inviability between M. guttatus and northern M. decorus.

5.4.2 The genetic architecture of hybrid seed inviability is genetically complex

We find parent of origin TRD is associated with several regions of the genome.

Specifically, we find an excess of individuals who are homozygous for the IMP genotype at 3-5 regions of the genome in our IMPxF1 cross, but not the F1xIMP cross, and an excess of individuals who are homozygous for the IM allele at 2-3 regions of the genome in the F1xIM cross, but not the IMxF1 cross. These represent paternal M. guttatus alleles and maternal M. decorus alleles that contribute to hybrid seed lethality, respectively.

Each of these loci is only mildly distorted (i.e. by about 10-20%), suggesting that each locus has a relatively small impact on the probability of survival of hybrid seeds.

In Capsella, hybrid seed inviability between the recently diverged C. grandiflora and C. rubella is predicted to be caused by three paternal C. grandiflora and 1 maternal C. rubella loci (Guo et al. 2009; Foxe et al. 2009; Rebernig et al. 2015). Between Arabidopsis thaliana and A. arenosa seven maternal M. thaliana loci have been found to interact with 125

an unknown number of paternal M. arenosa alleles to influence hybrid seed lethality

(Burkart-Waco et al. 2012). In the much more divergent species pair, A. lyrata and A. arenosa hybrids, seeds are inviable in both directions of the cross (Koch & Matschinger

2007; Lafon-Placette et al. 2017). In one direction of the cross, genetic models predict that one A. lyrata maternal locus interacts with many paternal A. arenosa loci to cause inviable seeds, and in the other direction of the cross, three maternal and three paternal loci are predicted to interact to cause hybrid seed inviability (Lafon-Placette et al. 2017). In

Mimulus, hybrid seed inviability between M. tilingii and M. guttatus is caused by interactions between a number of nuclear loci: in one direction of the cross, seed lethality is associated with five M. guttatus maternal loci and five paternal M. tilingii loci, and in the other direction, hybrid seed inviability is associated with a single M. guttatus paternal locus and seven M. tilingii maternal loci. While the divergence time is unknown for M. guttatus and M. tilingii, genome wide divergence is ~7% (Garner et al. 2016), much more than M. guttatus and M. nasutus, which diverged roughly 200-500KYA (and have

~5% divergence at synonymous sites only; Brandvain et al. 2014). Thus, in all species surveyed to date hybrid seed inviability is genetically complex, regardless of the divergence of the parental species involved. This contrasts with many other examples of intrinsic post-zygotic reproductive isolation between plants (e.g. Fishman & Willis 2006;

Sweigart et al. 2006; Bomblies et al. 2007; Bikard et al. 2009; Zuellig & Sweigart 2018)

126

The complexity of the genetic basis of hybrid seed inviability may be influenced by the evolutionary mechanism by which it evolves. Parental conflict has been suggested to play a major role in hybrid seed inviability (Brandvain & Haig 2005).

Under this scenario, species rapidly cycle through maternal versus paternal alleles affecting resource allocation and acquisition, respectively (Trivers 1974; Haig & Westoby

1989). Divergence between populations or species in these arms races could lead to hybrids with mismatches in maternal- versus paternally advantageous alleles, and thus improper growth and development. These arms race dynamics could cause a series of relatively small effect alleles, each of which manipulate resource allocation only slightly.

However, a more rigorous examination of the genetic basis of hybrid seed inviability is necessary to determine the genes involved, and whether the identity of these genes fits with that of parental conflict.

5.4.2 The effectiveness of hybrid seed inviability as a natural barrier to gene flow

While hybrid seed inviability is relatively strong between the northern clade of

M. decorus and M. guttatus (30-50%), its effectiveness at preventing gene flow in nature remains unknown. Here we determined that the barrier has a relatively complex genetic basis, and that each locus contributes relatively little to reproductive isolation. Indeed, each locus which was associated with parent of origin specific TRD was distorted by 7-

17%, resulting in selection coefficients of 0.24-0.51. While these are relatively strong 127

selection coefficients, theory predicts that unless post-zygotic reproductive isolation is complete, hybridization will result in the loss of RI alleles (Price 2007, see Bracewell et al.

2017 for a potential example). Thus, after the first generation, the prevention of gene flow in subsequent generations may be minimal in this species pair. Mimulus decorus and

M. guttatus are known to co-occur and hybridize in nature (Puzey et al. 2017). Puzey et al.

(2017) show that introgression occurs from M. decorus into M. guttatus (although were unable to test the reverse), which is also the direction in which seeds are viable.

Phenotypic analyses of hybrids show that F1s flower at approximately the same time as the IM parent (Chapter 6), and if gene flow occurs wherein IM is the maternal parent,

F1s would likely germinate in IM habitat. Therefore, we predict that introgression may be more likely from M. decorus into M. guttatus, however this prediction remains untested. This is in contrast to Arabidopsis lyrata and A. arenosa, which display strong

(but incomplete) reproductive isolation due to hybrid seed inviability (Lafon-Placette et al. 2017). In this system, introgression is limited largely to geographic regions where HSI disapates due to a tetraploidization event in A. lyrata (Lafon-Placette et al. 2017).

Determining the effectiveness of hybrid seed inviability as a barrier to gene flow in nature is a logical next step for this system.

128

5.4.3 Problems, pitfalls, and future directions

We performed an initial mapping of hybrid seed inviability between co- occurring M. guttatus and northern M. decorus. While this work represents a first pass, several questions remain unanswered. Firstly, there is discrepancy between our genetic models and the empirical results. Our genetic models suggest that the M. guttatus paternal alleles should act redundantly, so that the presence of an M. guttatus allele at any locus involved in hybrid seed inviability should result in hybrid seed death.

Therefore, the surviving individuals should all be homozygous for the M. decorus allele.

However, we find no region of the genome with such strong TRD, instead finding 3-5 regions with modest distortion. One potential explanation could be that F1 backcross seeds which looked inviable were able to germinate, thus diluting TRD. While the flat, shriveled seeds essentially never germinate in crosses between the parental lines, it is possible that some of the backcross seeds which manifested this phenotype were in fact viable. Unfortunately, seeds were not separated into morphological categories before planting. Future experiments to test the rate of germination in inviable appearing seeds of these backcrosses will be in order.

Secondly, several regions in the genome-wide scan appear to be significantly distorted, but are not distorted in marker genotype confirmations (e.g. Figure 38; Figure

39). These regions of distortion are normally driven by one or two spurious SNPs.

129

Methods to improve genotype calling and work to increase the coverage per individual is currently being done. Work to improve the coverage of the control crosses is also necessary.

Lastly, we find one region of the genome on LG9 which exhibites TRD towards

IM62 in both the IMxF1 and F1xIM cross, as well as the IMPOxF1 cross (as denoted by an excess of heterozygous individuals). While this region is likely not associated with hybrid seed inviability, it does represent a region of the genome in which IM62 alleles outcompete IMPO alleles. This may be due to several biological or technical issues.

Firstly, there could be technical problems resulting from the IM62 allele amplifying better than the IMPO allele, and thus individuals are more likely to be called IM62 homozygous than heterozygous (although this does not account for the increase in heterozygous individuals found in the IMPxF1 backcross). It is also possible that the

IM62 allele is favored in pollen viability, pollen germination or fertilization, resulting in an excess of IM62 alleles in the resulting progeny. Lastly, though less likely is the possibility of meiotic drive of the IM62 allele in this region. While drive has been described in this population of M. guttatus, it genetically maps to LG11 (Fishman &

Willis 2005).

130

5.5 Conclusions

Here we show that the genetic basis of hybrid seed inviability between northern

M. decorus and M. guttatus is likely caused by interactions between several nuclear inherited loci, and each locus may affect the probability of inviability only slightly.

Much more work will be needed to more accurately map hybrid seed inviability, as well as fine map the regions discovered here. Future work should also include determining the overall importance of hybrid seed inviability as a barrier to gene flow in nature.

131

Chapter 6: Dissecting the role of a large chromosomal inversion on life history divergence throughout the Mimulus guttatus species complex

Publication Note: This manuscript is accepted for publication at Molecular Ecology

6.1 Introduction

Closely related ecotypes often vary in the structure and arrangement of chromosomes and these structural changes have long been hypothesized to play an important role in adaptation (Dobzhansky 1948; Umina et al. 2005; Manoukis et al. 2008;

Jones et al. 2012; Fishman et al. 2013; Lee et al. 2017). In particular, a link between chromosomal inversions and adaptation has been shown in numerous systems

(reviewed in Hoffman & Rieseberg, 2008). Research to date has provided many examples highlighting the potentially important role of inversions in the process of local adaptation, including examples in which the frequency of an inversion karyotype is associated with environmental variables (Dobzhansky 1948; Ward et al. 1974; Umina et al. 2005; Manoukis et al. 2008; Wallberg et al. 2017), putatively adaptive traits map to an inversion (Feder et al. 2003; Jones et al. 2012; Fishman et al. 2013), or local fitness itself is associated with an inversion and its contents (Lowry & Willis, 2010). However, the mechanistic role in adaptation remains relatively unknown for most inversions.

132

The evolutionary mechanisms underlying the association between an inversion and adaptation fall into three broad hypotheses. (1) The inversion does not contribute to local adaptation and the association between an inversion and locally adaptive traits occurs by chance. Such a pattern could occur if an inversion simply hitchhiked along with a beneficial mutation (Kirkpatrick & Barton 2006), or was fixed by drift and later accumulated locally beneficial mutations. (2) The inversion could play a direct role in adaptation via disruption of gene function or expression (Kirkpatrick & Barton 2006).

This could include disruption of gene function at the inversion breakpoints (Kirkpatrick

& Barton, 2006), or other effects on global or regional gene expression (as suggested in

Lavington & Kern, 2017; Said et al. 2018). Lastly, (3) The inversion could be indirectly favored by natural selection by preventing recombination between sets of locally adaptive alleles (Kirkpatrick & Barton, 2006; Charlesworth & Barton, 2018). Here we refer to this as the recombination suppression hypothesis.

While the recombination suppressing effects of inversions have long been known

(Sturtevant 1921), it wasn’t until the modern synthesis that Dobzhansky hypothesized an adaptive role for recombination suppression, in what he referred to as co-adapted gene complexes (Dobzhansky 1948, 1970). This hypothesis has undergone many iterations (see Rieseberg 2001; Kirkpatrick & Barton 2006; Feder & Nosil, 2009; Yeaman et al. 2011, 2013), but in essence states that if local adaptation is conferred by beneficial

133

alleles across multiple loci, and linkage among locally beneficial alleles reduces the probability of creating maladaptive recombinant genotypes, then chromosomal configurations that increase linkage (i.e. suppress recombination) will be favored

(Kirkpatrick & Barton 2006). However, the very fact that chromosomal inversions suppress recombination has made testing this hypothesis extremely challenging due to the inability to genetically dissect them. Despite these challenges, we can define two testable predictions of the recombination suppression hypothesis: (1) the alleles causing adaptation must predate the inversion itself and (2) there must be at least two alleles involved in the adaptive phenotype or local fitness.

There have been relatively few empirical tests of the role of recombination suppression in the process of adaptation. Recent studies have shown a role for recombination suppression in local adaptation (for example, in sticklebacks; Samuk et al.

2017), and recombination suppression via inversions specifically (in Boechera; Lee et al.

2017, song birds; Hooper & Price, 2017; and Drosophila Fuller et al. 2017). Indeed, some of the best evidence for the recombination suppression hypothesis comes from Boechera in which a recent inversion (~2-8kya) has captured approximately three quantitative trait loci (QTL) involved in water regime adaptation. Furthermore, this inversion only exists in a region of secondary contact between two differentially adapted subspecies (Lee et al.

2017; although see Charlesworth & Barton 2018). Despite these exciting advances, we

134

lack an understanding of the history of and evolutionary mechanism by which many inversions contribute to adaptation, and more examples are needed in order to make general conclusions.

One of the best studied inversion polymorphisms in plants is a large paracentric inversion which distinguishes life history ecotypes of Mimulus guttatus (Lowry & Willis

2010; Oneal et al. 2014; Twyford & Friedman 2015; Gould et al. 2017). QTL mapping has shown that a significant proportion of the divergence in life history traits between annual and perennial forms of M. guttatus maps to an inversion on linkage group 8 (here after LG8; Hall et al. 2006; Lowry & Willis, 2010; Friedman et al. 2014), and that local fitness is highly associated with this region (Lowry & Willis, 2010). Further, population genomic analyses found that the LG8 inversion is one of the only highly differentiated regions between annual and perennial M. guttatus populations (Twyford et al. 2015;

Gould et al. 2017), suggesting high levels of gene flow between these ecotypes outside of the inversion. However, we do not know which orientation of the inversion is ancestral and how widely distributed it is among other taxa within the Mimulus guttatus complex.

In addition, it is unknown whether alleles controlling life history divergence between annuals and perennials pre-date the LG8 inversion as predicted by the recombination suppression hypothesis.

135

In this study, we examined the evolutionary history of the LG8 inversion and its phenotypic effects on other species in the Mimulus guttatus species complex to understand its role in life history adaptation. To determine which orientation of the inversion is ancestral we surveyed the LG8 inversion karyotype within the Mimulus guttatus species complex. We further map the contribution of the LG8 to life history divergence between co-occurring annual M. guttatus and another perennial species, M. decorus. Lastly, to genetically dissect the LG8 inversion region and test whether this region contains two or more loci associated with life history, we use a distantly related perennial species, M. tilingii which is collinear with annual M. guttatus. We find that this region is still associated with life history divergence between M. tilingii and annual M. guttatus, and that there are at least two QTL in this region, in support of the recombination suppression hypothesis. However, we also find the M. guttatus inversion maps life history traits that are not mapped in either M. decorus or M. tilingii, suggesting that it may house unique alleles. We conclude that the LG8 inversion likely suppresses recombination between multiple loci which confer life history divergence, but its total contribution to local adaptation via life history may be more complex.

136

6.2 Methods

6.2.1 Study Genus

The Mimulus guttatus species complex is a morphologically and genetically diverse group of wildflowers, native to Western North America (also referred to as

Erythranthe of the family Phrymaceae; Nesom 2012). Species in this complex have evolved a variety of life-history strategies from diminutive, highly selfing, rapidly lived annuals to large, clonally spreading perennials, however perenniality is thought to be the ancestral form (Friedman et al. 2014). Across the complex, annuals and perennials are distinguishable by many traits, including branching patterns, the physical and developmental timing of flowering, size of vegetative and reproductive structures, and resource allocation to different reproductive strategies (i.e. clonal reproductive via stolons versus sexual reproduction; Friedman et al. 2015). Multiple studies have shown that life history trait differences among species in the complex are under local selection

(for example, Hall et al. 2006, 2010; Lowry & Willis 2010; Peterson et al. 2013; Peterson et al. 2016; Ferris et al. 2017), and often these species are found in close proximity (i.e.

Twyford & Friedman 2015; Kenney & Sweigart 2016; Ferris et al. 2017). However, we lack an understanding of whether the LG8 inversion plays an important role throughout the complex in distinguishing co-occurring species which vary in life history.

137

In this paper, we discuss a number of members of the M. guttatus species complex (diagramed in Figure 19) but we focus on two purely perennial species- M. decorus, and M. tilingii- as well as M. guttatus, which exhibits both annual or perennial ecotypes. While recurrent hybridization, rapid speciation, or their combination make taxonomic resolution of this clade difficult, genetic studies have shown that perennial M. tilingii is the most divergent member of the species complex (Beardsley et al. 2004;

Garner et al. 2016). Mimulus decorus has been relatively under-sampled, but has been shown to be a distinct genetic lineage relative to nearby M. guttatus populations (Puzey et al. 2017) and forms a monophyletic group nested within M. guttatus (Figure 29).

6.2.2 Assessing life history variation between annual M. guttatus and perennial species in the M. guttatus species complex

To confirm that different perennial species were phenotypically distinguishable from annual M. guttatus along similar phenotypic axes and to determine how much phenotypic variation existed between perennial species we performed a population- level common garden experiment. Seeds from six annual M. guttatus populations, seven

M. tilingii populations, 18 M. decorus populations and 26 perennial M. guttatus populations were sprinkled onto moist Farfard 4P soil, cold stratified for 1 week, then transferred to a long day greenhouse (18h days, 21C day/ 18C nights). For each population, 1-5 maternal families were grown, with 5 replicates of each maternal family.

For each plant, we measured 13 life history traits (days from germination to first flower, 138

node of first flower, total corolla length, corolla tube length, corolla width, stem thickness, leaf length and width, internode length between the cotyledons and first true leaf, internode length between the first and second pair of true leaves, the number of stolons, length of the longest stolon, width of the longest stolon). To evaluate overall clustering of species in trait space we performed a PCA with variables scaled to have unit variance. Using a broken stick model we determined that only the first two PCs explained a higher than random proportion of the variance in the phenotypic data, and thus use these two PCs to visualize and discuss the phenotypic differences among and between life history types (PC1 accounted for 44.91% of the phenotypic variance, while

PC2 accounted for 18.31%). To determine if species varied from each other in each life history traits, we completed ANOVAs with species as a fixed effect, and population and maternal family as random effects for all phenotypic traits using the lme4 package in the statistical software R (Bates et al. 2015). Post-hoc Tukey Honest Significant Difference tests were used to determine which species were significantly different from one another.

6.2.3 Survey of the LG8 inversion orientation

To understand how the LG8 inversion has contributed to life history adaptation, it is first important to understand its evolutionary history. To assess which orientation of the inversion is ancestral we surveyed the inversion orientation throughout the M.

139

guttatus species complex and within the genus more broadly. Using a combination of F2 mapping, linkage map comparison, and previously published results we determined

LG8 inversion karyotype for seven species within the genus Mimulus, six species of which reside in the Mimulus guttatus species complex.

(1) F2 mapping approach: We created F2 populations between five species within the M. guttatus species complex and an annual M. guttatus accession (see Table 12 for a list of species and their life histories). For four species, F2 populations were created by crossing a representative individual of each species to the highly inbred accession of annual M. guttatus, IM62. For the fifth species, we used DNA from a previously created F2 population, derived from a cross between sympatric populations of M. nudatus and annual M. guttatus. For one species, M. decorus, we also created an additional F2 population between M. decorus and a coastal perennial M. guttatus line (OPB; outlined below). Each F2 population was generated by self fertilizing a single F1 hybrid. For each

F2 mapping population, we planted seeds in 4 inch pots filled with moist Fafard 4P potting soil, cold stratifed them for one week at 4°C, and, after emergence, transferred young germinants (approximately 1-3 days old) to cell packs also filled with moist

Fafard 4P soil. Plants were grown in long day greenhouse conditions (18h days, 21C day/ 18C nights).

140

During peak flower, we harvested young bud tissue for each F2 individually into

96-well plates and stored the tissue at -80°C. We extracted DNA from these F2s, as well as the parental lines, using a modified CTAB protocol (outlined in Kelly & Willis, 1998).

Genomic DNA was re-suspended in autoclaved water and stored at -20°C until use. The hallmark of whether two species share the same orientation of an inversion is the presence of recombination in advanced hybrids. To assess recombination, we determined genotypes at two distant fragment-length polymorphism markers which were polymorphic among each pair of parents (MgSTS Markers, described in Fishman et al. 2014; markers ranged from 8-29cM apart. Primer sequences are available at: http://www.mimulusevolution.org/). To confirm recombinant genotypes, we re- amplified each marker pair in all putative recombinant F2s. All PCR products were run on an ABI 3730x1 machine, and fragment lengths were scored by hand in

GENEMARKER (https://softgenetics.com/). We then determined the inversion orientation based on the presence of recombination in the cross. While we cannot detect the exact number of recombination events in an F2 population due to double cross- overs, and an inability to phase doubly heterozygous individuals, we estimate a rough approximation of recombination rate for each F2 generation (Table 12). For one species,

M. decorus, of which 2 populations were surveyed, we found an absence of recombination when individuals were crossed to annual M. guttatus. We then assessed

141

recombination in an F2 population between the M. decorus population IMP and a coastal perennial M. guttatus population (OPB), which is known to have the perennial orientation of the inversion.

(2) Comparative linkage map approach: For the most distantly related species M. lewisii, our

F2 style of surveying inversion karyotype is impossible, as these species are approximately 20 million years diverged (Fishman et al. 2014). Instead, we use previously published integrated linkage maps of M. lewisii and annual M. guttatus to compare marker order (markers described in Fishman et al. 2014), and determine if these species are collinear in the LG8 inversion.

(3) Literature review approach: The LG8 inversion orientation has been published for a number of M. guttatus populations (Lowry & Willis 2010), as well as for two populations of other species in the complex (M. tilingii- Garner et al. 2016, and M. nasutus, Lowry &

Willis 2010). We include these results in our final survey to give a more holistic view of the complex (see Table 12).

6.2.4 Determining whether the LG8 inversion is still associated with life history divergence in a more distantly related perennial species

The perennial species M. decorus was found to differ in LG8 inversion orientation from that of annual M. guttatus. As these two species co-occur on fine spatial scales throughout the cascade mountains (Puzey et al. 2017), we were interested in whether the

LG8 inversion was associated with life history divergence between these species. To 142

assess an association between inversion orientation and life history traits between M. decorus and annual M. guttatus, we grew 385 F2s, 59 F1s (created with IM62 as maternal donor; the other cross direction produces only inviable seeds), 60 IM62 individuals and

57 IMP individuals in 2.5-inch post filled with moist 4-P Fafard potting soil in long days in the Duke University Greenhouses (18h days, 21C day/ 18C nights). On the day of first flower, we phenotyped these individuals for twelve life history traits: the number of days between germination and flowering, the node of first flower, corolla length, tube length, corolla width, stem thickness, leaf length and width for the first true leaf, the internode length between the cotyledons and first true leaves, the number of stolons, the length of the longest stolon, and the average number of side branches per node before first flower. After phenotyping, we collected early bud or leaf tissue and extracted DNA as above. We genotyped individuals at a minimum of 4 markers within the inversion to confirm that no recombination had occurred and to have confidence in calling the inversion karyotype in each F2. We then performed genotype-phenotype associations

(outlined below) for each trait with inversion orientation as the genotype.

143

Figure 19: Inversion survey schematic. (A) Proposed phylogenetic relationship between species of the M. guttatus species complex, based off of Beardsley et al. 2004. (B) Survey method for determining LG8 inversion orientation within the M. guttatus species complex. For the highly divergent M. lewisii, we compared the order of a conserved set of markers in an integrated M. guttatus and integrated M. lewisii map (Fishman et al. 2014). Across the phylogeny, species names written in red exhibit an annual life history, species names written in blue exhibit a perennial life history. The same color code is used in brackets beside species name to depict the inversion orientation discovered in each species (A= annual, P= perennial).

6.2.5 A test of the recombination suppression hypothesis- determining whether the LG8 region is still associated with life history divergence in the absence of the inversion

The recombination suppression hypothesis posits that an inversion could contribute to adaptation by preventing recombination among locally adaptive sets of loci. This hypothesis predicts that the adaptive loci should predate the evolution of the inversion, and that there should be more than one adaptive loci within the inversion

144

region. To test whether loci which are associated with life history exist within the LG8 inversion region in the absence of the inversion, we completed two mapping experiments using M. tilingii, which possesses the annual orientation of the inversion.

We first crossed an inbred accession of M. tilingii (SOP7) to an inbred accession of annual M. guttatus (IM62), then created an F2 mapping population by self-fertilizing a single F1 hybrid. For this experiment, we grew 71 SOP7, 67 IM62, 39 IM62 x SOP F1s (the reciprocal cross results in inviable hybrid seed, and accordingly, none germinated), and

265 F2 progeny. Individuals were potted in 4” pots as above and randomized across flats and benches. Individuals were scored on the day of first flower for 9 life history traits: days from germination to flowering, node of first flower, corolla tube length, corolla width, stem thickness, length of the first leaf, internode length between cotyledons and first true leaves, internode length between first and second true leaves, and stolon number. After all plants had been phenotyped, tissue collection and DNA extraction were carried out as described above. Four markers within the inversion, and 2 on the periphery of the inversion (while the exact breakpoints of the inversion are unknown, these markers occur less than 10-15cM from the closest marker within the inversion- region in a collinear M. guttatus linkage map; Fishman et al. 2014) were used to construct haplotypes for each F2.

145

Statistically determining whether an association between traits and genotypes are caused by more than one locus can be challenging when the region of interest is relatively small. To generate hybrids that were segregating for different sections of the

LG8 inversion region we picked several recombinant F2 individuals and self fertilized them to create an F3 population. As this species pair exhibits substantial post-zygotic reproductive isolation- including both seed incompatibilities (Garner et al. 2016) and pollen fertility issues (JMC personal obs), we were left with 9 recombinant genotypes which had sufficient F3 seed. These 9 recombinant lines fall into 4 different recombinant classes (outlined in Figure 23). We grew between 51 and 288 F3 offspring per recombinant F2 plant (for a total of 1535 plants) using the methods outlined above. We phenotyped them for traits which were statistically associated with the LG8 inversion region in the previous F2 map (or trended that way; days from germination to flower, node of first flower, corolla tube length, corolla width, stem thickness, internode length between cotyledon and first true leaves, internode length between first and second leaves, stolon number). We then amplified these F3s for one marker within the section of the inversion region for which they were segregating. For one recombinant where the

F3s were segregating for the majority of the inversion region, we amplified 5 markers across the segregating region and compared the non-recombinant alternate homozygotes across this region.

146

6.2.6 Genotype-phenotype associations

For both F2 association studies, as well as the F3 fine mapping study we use the same statistical approach for genotype-phenotype mapping. In each case, we first sought to determine if parental lines or F1 hybrids differed for each other. To assess the relationship between trait and genotype, we first completed a MANOVA, wherein all life history traits were treated as continuous, dependent variables, and our predictor trait was genotype, which was treated as a factor (i.e. Parent 1, F1, Parent 2). If the

MANOVA was significant, we then completed univariate analyses for each trait to determine which traits differed between genotypes. We then used Tukey Honest

Significant Difference tests to determine which genotypes differed from each other for each trait.

To determine the association between the LG8 inversion or LG8 inversion region

(in the M. tilingii x annual M. guttatus crosses), we repeated our above analyses, using F2 individuals which were alternate homozygotes for the inversion in the case of the M. decorus x annual M. guttatus experiment, and only non-recombinant alternate homozygote F2s in the case of the M. tilingii x annual M. guttatus experiment (we exclude recombinant genotypes to assess the contribution of the entire region to life history in this species pair). For the fine mapping experiment, we complete the same analyses for each recombinant class, but include line as a random effect in our models.

147

For each experiment, we also estimated the effect size between alternative homozygotes and the proportion of variance among parental lines explained by the inversion genotype to compare with previous published results of life history divergence between annual and perennial forms of M. guttatus (Lowry & Willis 2010).

6.3 Results

6.3.1 Assessing life history variation in the M. guttatus species complex:

To assess the distribution of annuals and perennials in phenotypic space, we completed a common garden experiment involving annuals and perennials within the

M. guttatus species complex. We found that annual M. guttatus is divergent from all perennial species in overall in PC1-2 space (Figure 20). This overall divergence between annuals and perennials is caused by divergence in the timing of reproduction, general size, and investment in different reproductive strategies (Figure 20). For most traits surveyed, perennial M. guttatus and M. decorus are very phenotypically similar, except in the number of stolons, in which M. decorus produced more stolons than perennial M. guttatus (Figure 20). Perennial M. tilingii was smaller and flowered earlier than perennial

M. guttatus or M. decorus, but was still divergent from annual M. guttatus in a number of these traits (Figure 20).

148

149

Figure 20: Population-level survey of life history traits across perennial species in the Mimulus guttatus species complex relative to annual M. guttatus. Panels A-F are trait means for (A) days from germination to first flower, (B) Corolla tube length, (C) Stem thickness, (D), Leaf length, (E) Number of Stolons, (D) Internode length for annual M. guttatus, Coastal perennial M. guttatus (CP), Inland perennial M. guttatus (IP), perennial M. decorus, and perennial M. tilingii. Panel G is the first two principal components of a PCA of all 13 life history traits for each species, ellipses are the 95% confidence ellipses for each species.

6.3.2 Species survey of the inversion

We next sought to determine which orientation of the LG8 inversion is ancestral by surveying the LG8 inversion karyotype throughout the species complex (Figure 19).

Almost all species in the M. guttatus complex have the annual orientation of the inversion, regardless of life history (Table 12). We determined that four of six populations surveyed have the annual orientation of the inversion (M. tilingii, M. glaucescens, M. laciniatus, M. nudatus) given the moderate levels of recombinants found between each species and annual M. guttatus. In contrast, both populations of M. decorus which were surveyed exhibited no signs of recombination with annual M. guttatus in the

LG8 inversion region, but substantial recombination when crossed to a perennial M. guttatus which is known to carry the LG8 inversion. We thus conclude that it likely has the perennial orientation. In addition, annual M. guttatus and M. lewisii show conserved marker order across the LG8 inversion region (which comprises about 1/3 of LG8 in M. guttatus and almost the entire homologous chromosome- LG4- in M. lewisii; Fishman et al. 2014), and thus likely also share the same orientation of the LG8 inversion. Taken

150

together, this suggests that the annual orientation of the inversion is likely the ancestral state, even though perenniality is thought to be the ancestral phenotype (Friedman et al.

2014).

6.3.3 The LG8 inversion is associated with life history divergence in M. decorus

Annual M. guttatus and perennial M. decorus differed substantially in all life history traits measured, with M. decorus having a delayed phenology, being generally larger, and exhibiting a greater investment in vegetative growth (Figure 20; Figure 21).

There is a strong association between the orientation of the LG8 inversion and almost every life history trait measured in a perennial M. decorus x annual M. guttatus F2 cross, except for the number of stolons produced (Figure 21;

Table 13). For traits which have been measured between multiple studies, the inversion karyotype accounts for similar levels of life history divergence among parental lines in this interspecific cross as it does between ecotypes of M. guttatus for almost all traits, except for the number of stolons (Table 4).

151

Figure 21: Phenotypic differences between M. decorus and annual M. guttatus: parental lines (grey) and alternate homozygotes in an F2 population (black) for 6 life history traits: (A) Days from germination to first flower, (B) corolla tube length, (C) stem thickness, (D) length of the first true leaf, (E) number of stolons, (F) internode length between cotyledons and the first true leaves. AA= annual lines (either annual M. guttatus line IM62 in the case of parental lines, or an F2 individual homozygous for the IM62 genotype), PP=perennial lines (either perennial M. decorus line IMPO in the case of parental lines, or an F2 individual homozygous for the IMPO genotype). Asterisks denote significance: NS (not significant); + 0.1>p>0.05, * 0.05>p>0.01; ** 0.01>p>0.001; ***p<0.001.

6.3.4 The LG8 inversion region is associated with life history divergence in a collinear species

The recombination suppression hypothesis predicts that the alleles conferring adaptation should pre-date the inversion, and thus the inversion region should still be

152

associated with adaptive traits in a collinear cross. Mimulus tilingii and annual M. guttatus vary in a number of life history traits, though not as extremely as either perennial M. guttatus or M. decorus (Figure 20). Nonetheless, we found that the LG8 inversion region is associated with life history divergence between annual M. guttatus and M. tilingii. As these species are collinear along LG8, this suggests that the LG8 inversion region is associated with life history even in the absence of the LG8 inversion

(Figure 22). In particular, the LG8 inversion region is associated with flowering time, flower size, stem thickness and internode length (Figure 22). The percentage of the parental divergence explained by the LG8 region in this cross is similar to that of what the LG8 inversion explains in both annual x perennial M. guttatus crosses and an annual

M. guttatus x perennial M. decorus cross (Table 4), although the absolute effect is smaller for some traits. We again note that this region does not seem to be associated with differences in the number of stolons produced between annual M. guttatus and perennial

M. tilingii.

153

Figure 22: Phenotype-inversion associations for non-recombinants between annual M. guttatus and M. tilingii. Phenotypic differences between parental lines (grey) and alternate homozygotes in an F2 population (black) for 6 life history traits: (A) Days from germination to first flower, (B) corolla tube length, (C) stem thickness, (D) length of the first true leaf, (E) number of stolons, (F) internode length between cotyledons and the first true leaves. AA= annual lines (either annual M. guttatus line IM62 in the case of parental lines, or an F2 individual homozygous for the IM62 genotype), PP=perennial lines (either perennial M. tilingii line SOP7 in the case of parental lines, or an F2 individual homozygous for the SOP7 genotype). Asterisks denote significance: NS (not significant); + 0.1>p>0.05, * 0.05>p>0.01; ** 0.01>p>0.001; ***p<0.001.

The recombination suppression hypothesis also predicts that the inversion region should hold more than one locus involved in adaptation. To determine whether the LG8 inversion region has more than one life history QTL, we grew up a panel of F3 offspring from 9 recombinant F2 individuals. These 9 recombinant lines comprised 4 different

154

recombinant classes, which were segregating for different regions of the LG8 inversion region (Figure 23). For each recombinant class, we sought to determine if alternate homozygotes differed in life history traits by performing a MANOVA for all life history traits with genotype and line as fixed variables. For all tests, we found that line had a significant effect, indicating that other loci which differed between these families also had an effect on life history. This is not surprising given the polygenic nature of this set of complex traits (Hall et al. 2006, Friedman et al. 2014).

Figure 23: Schematic for the F3 fine mapping experiment. The first column refers to the recombinant class different lines have been subscribed to, the second column refers to the original genotype of the F2 individual (or set of individuals) which were selfed to create an F3 family, shown in the third column, the forth column

155

indicates the number of lines in each class, and the fifth column is the number of individuals in that class with either alternate homozygous genotypes (i.e. homozygous for either the annual M. guttatus or perennial M. tilingii allele). Colors indicate the genotype: red= homozygous annual M. guttatus, blue= homozygous perennial M. tilingii, purple= heterozygote, red box with blue outline indicates that for recombinant classes in which there were more than 1 line, the region which is not segregating (i.e. was homozygous in the original F2 mother) was either homozygous annual M. guttatus or perennial M. tilingii.

We find that the recombinant class which possesses a very small part of the beginning of the LG8 inversion (class ‘4’ in Figure 23 & Figure 24) is not associated with any life history trait (F=1.7437, p=0.1055, DF=1; Figure 24). For F3 offspring which segregate for most of the inversion region (class ‘3’ in Figure 23 & Figure 24), genotype within the inversion region is associated with life history traits (F=3.4041, DF=1, p=<0.00001; Figure 24). The remaining two classes segregate for opposing sections of the

LG8 region; class 1 segregates for the first half of the inversion region, while class 2 segregates for the second half (Figure 23). Both of these classes are significantly associated with life history traits, when all life history traits measured are considered

(class 1: F=2.6678, DF=1, p= 0.0067; class 2: F=3.18939, DF=1, p=0.014; Figure 24). The presence of an association between genotype and phenotype for these 2 recombinant classes which segregate for different halves of the inversion region strongly suggests that there are at least 2 separate QTL for life history in this region. This finding is

156

consistent with at least two loci involved in life history adaptation pre-dating the inversion in this region.

157

Table 4: Effect of the inversion karyotype on life history traits within the M. guttatus species complex: within M. guttatus (from Lowry & Willis 2010; crosses indicated in light gray), between annual M. guttatus and M. decorus (medium gray), and collinear annual M. guttatus and perennial M. tilingii

Stem Internode Flowering Time Corolla Length Corolla Width Stolon Number Thickness Length Cros 2a/dif 2a/dif 2a/dif 2a/dif 2a/dif 2a/dif 2a 2a 2a 2a 2a 2a s f f f f f f CAN 5.61*** 2.74*** 0.73*** x 3.93** 0.4 0.7* 0.13 -5.6 -0.23 0.32 0.26 0.32 * * * BCB LMC - 3.29*** x 0.45 0.55* 0.18 19.39** -0.33 2.04*** 0.2 0.81* 0.1 NA NA * SWB * RGR 4.31*** 2.65*** x 3.47* 0.36 0.65** 0.27 2.76 0.1 0.49 0.3 1.12** 0.21 * * OPB SAM x 3.43*** 0.21 0.65** 0.18 2.96 0.12 2.62** 0.56 1.43** 0.23 NA NA OS * W IM x 2.39*** 1.08** 0.18 0.43 9.59*** 0.65 4.99*** 0.22 2.21*** 0.18 0.36 0.032 IMP * * IM x 1.67+ 0.41 0.34+ 0.26 -5.38* 0.53 2.06*** 0.45 2.92* 0.56 0.55 0.085 SOP

158

To determine which traits were significantly different between alternate homozygotes of the segregating inversion region in each recombinant class, we performed an ANOVA for each trait, treating genotype and line as fixed effects. We found that the first half of the inversion is associated with floral size (corolla length;

Figure 24;

Table 14), while the second half of the inversion is associated with timing of flowering (both days to flower and internode length) and weakly with floral size (Figure

24;

Table 14). As expected, the recombinant class which incorporated most of the inversion region was strongly associated with both floral size and timing of flowering.

159

Figure 24: Average effect of genotype on a trait for each of 4 recombinant classes (panels labeled as 1,2,3 or 4). Cartoon chromosomes above panels also denote the genetic composition of those F3 genotypic classes, as denoted in Fig. 5. Traits: (A) Days from germination to first flower, (B) corolla tube length, (C) stem thickness, (D) internode length between the first and second pair of true leaves. Asterisks denote significance: no denotation (not significant); + 0.1>p>0.05, * 0.05>p>0.01; ** 0.01>p>0.001; ***p<0.001.

160

We note that the effect of the LG8 inversion region in M. tilingii likely depends on both the genetic background and environmental influences. For traits which are statistically associated with only one half of the inversion region, the effect sizes are not equivalent between recombinant classes. For example, for flowering time, individuals which segregated for most of the inversion region (class 3; Figure 23) experienced a difference of 0.92 days between alternate homozygotes, and for individuals which segregated for the second half of the inversion (class 2; Figure 23), this difference was

0.73 days. Similarly, for the class 3 recombinants, the difference between alternate homozygotes was 4.15mm for internode length between the first and second pair of leaves, while for recombinant class 2 (which segregated for the second half of the inversion region) the effect size is only 2.7mm. Discrepancies between classes in their effect size could be due to small effect loci in other regions of the LG8 inversion region which we fail to statistically deem significant, or differences in genetic background between lines. The later is perhaps not a surprising finding, and background effects have been reported for the LG8 inversion itself in M. guttatus (Lowry & Willis 2010; Friedman et al. 2014). Similarly, differences in effect size between the F2 and F3 mapping experiments likely represent differences in life history expression due to slight environmental differences (such as, Friedman & Willis, 2013). Thus, while the LG8 inversion region has a strong effect on life history divergence between M. tillingii and

161

annual M. guttatus, the effect of the LG8 inversion region is likely also somewhat dependent on genetic background and environmental conditions.

6.4 Discussion

Here we explored the evolutionary history of a chromosomal inversion which is known to be important in the maintenance of annual and perennial life histories within

M. guttatus. We use a combination of crossing surveys, comparative mapping, and published results to survey the evolutionary history of the LG8 inversion. We then completed an F2 mapping experiment between another perennial species, M. decorus, and annual M. guttatus and show that the LG8 inversion is present and associated with life history divergence in M. decorus. Lastly, we tested two predictions of the recombination suppression hypothesis by completing an F2 mapping experiment and F3 fine mapping experiment between collinear annual M. guttatus and a more distantly related perennial species, M. tilingii. We found that the LG8 region is still associated with life history divergence, despite lacking the LG8 inversion, and this effect can be separated into at least 2 distinct QTLs. This finding is consistent with the recombination suppression hypothesis.

6.4.1 The evolutionary history of the LG8 inversion

Our study shows that the annual orientation of the inversion is widespread across the M. guttatus species complex, regardless of life history, which is consistent with

162

the annual orientation of the inversion being the ancestral orientation. The presence of the perennial orientation in the perennial species M. decorus, suggests that the inversion arose before the split between perennial M. guttatus and M. decorus. Mimulus decorus and

M. guttatus have a divergence of ~3% (JMC unpublished), compared to M. guttatus and M. tilingii which have an average divergence of 7% (Garner et al. 2016), or M. guttatus and

M. nasutus in which divergence is approximately 5%, and for which the divergence time is roughly 200 kya (Brandvain et al. 2014). Population genetic analyses of the LG8 inversion suggest that this inversion likely experienced an older selective sweep.

Twyford and Friedman (2015) show that perennial M. guttatus exhibit a significant drop in pairwise nucleotide diversity within the inversion (but not elsewhere in the genome), while annual M. guttatus show relatively stable levels of diversity genome wide, suggesting that the inversion swept in perennials. However, the authors infer that this sweep is likely old, as diversity in the LG8 inversion in perennial M. guttatus still relatively high, suggesting that substantial time has passed for diversity to recover

(Twyford & Friedman 2015). Given this model of evolution, we would posit that the split between perennial and annual M. guttatus is also quite old, but that more recent hybridization and introgression has homogenized the genomes of these ecotypes outside of the LG8 inversion (and other non-collinear genomic regions). This is consistent with the finding that within M. guttatus, populations of annuals and perennials cluster much

163

more closely by geography than life-history genome-wide (except at the LG8 inversion, in which annuals and perennials appear to be two distinct genetic groups; Twyford and

Friedman, 2015), or the lack of fixed SNPs separating annual M. guttatus from perennial

M. guttatus (Gould et al. 2017). Overall, our working hypothesis is that the LG8 inversion arose after the split between annual M. guttatus and perennial M. guttatus, but before the split between perennial M. guttatus and perennial M. decorus. Subsequent allelic evolution in the LG8 inversion in the lineage which resulted in M. guttatus occurred, which resulted in unique traits mapping to this region in perennial M. guttatus, but not other perennials (i.e. stolon growth, discussed below). Lastly, ongoing hybridization and introgression between annual M. guttatus and perennial M. guttatus has eroded divergence throughout most of the genome, except in regions which are no longer collinear.

However, we cannot definitively exclude two alternative hypotheses. Firstly, an inversion in this region could be independently derived in perennial M. guttatus and M. decorus. While this scenario is theoretically possible, it seems unlikely as M. decorus and perennial M. guttatus appear to be collinear for much of the inversion. Even if an inversion had evolved twice in this region, it would be highly unlikely to be largely collinear. Secondly, the perennial orientation of the LG8 inversion may have introgressed between perennial M. guttatus and M. decorus. Introgression between

164

perennial M. guttatus and M. decorus is possible, as both species co-occur throughout the

Cascade Mountains in Oregon, Washington, and British Columbia, exhibit no significant floral differentiation, and both species overlap in flowering in the field (Nesom 2012;

JMC personal observation). If the LG8 inversion did introgress into M. decorus it would likely pose a selective advantage, as M. decorus also co-occurs with annual M. guttatus and they are known to hybridize (Puzey et al. 2017). However, this hypothesis might be less likely than shared ancestry, because perennial M. guttatus appears to exhibit unique phenotypic effects that are associated with the inversion. Thus, one would need to invoke a much older introgression event between perennial M. guttatus and M. decorus.

A more detailed knowledge of the breakpoints, paired with further population genetics studies of all perennials in this species complex will be needed to definitively differentiate these hypotheses.

6.4.2 The LG8 inversion region is associated with life history traits in the absence of the inversion

The F2 and F3 genetic maps revealed that the LG8 inversion region is associated with a number of life history traits between collinear annual M. guttatus and perennial

M. tilingii (i.e. days to flower, corolla tube length, stem thickness, and internode length).

Since the recombinant classes 1 and 2 are segregating for different sections of the LG8 inversion, we can localize independent effects of each half of the inversion region to life 165

history. The LG8 region appears to harbor at least 2 QTL for life history between annual

M. guttatus and M. tilingii. The first half of the inversion is association with floral size, while the second half of the inversion is associated with both timing of reproduction, and to a lesser extent floral size. The presence of at least two QTLs is consistent with the recombination suppression hypothesis, in which multiple, adaptive alleles predate the evolution of an inversion. While we cannot definitively identify why the LG8 inversion rose to fixation (particularly without a knowledge of the historical selection coefficient of each of these loci, as well as historical migration rates), we can conclude that the LG8 inversion likely plays a role in repressing recombination among these loci in at least two perennial species which co-occur on fine spatial scales with annual M. guttatus and are known to hybridize with them (i.e. perennial M. guttatus- Twyford & Friedman 2015,

Gould et al. 2017; and perennial M. decorus- Puzey et al. 2017).

The recombination suppression hypothesis has gleaned much theoretical support

(Kirkpatrick & Barton 2006; Yeaman & Whitlock, 2011; Yeaman 2013), but only recently have people begun to test for the role of recombination in adaptation. In sticklebacks, regions of low recombination are associated with adaptive loci to a greater extent when populations are experiencing both divergent selection and gene flow (Samuk et al. 2017).

In passerine birds, chromosomal inversions are more likely to occur between sister taxa which have overlapping ranges (suggesting a potential for gene-flow; Hooper & Price,

166

2017). In ruffs, two ancient segregating chromosomal configurations appear to harbor at least two loci associated with mating behavior and plumage phenotypes, but it is unclear whether these alleles predate the structural changes (Küpper et al. 2017;

Lamichhaney et al. 2017). In Drosophila pseudoobscura over 30 unique gene arrangements are segregating, and harbor arrangement-specific derived alleles, and no disruption of genes at the breakpoints (Fuller et al. 2017). Each of these studies is consistent with a role of recombination suppression in adaptation. However, to our knowledge, the only other study to genetically dissect an inversion using more distantly related, collinear populations has been in Boechera (Lee et al. 2017). Here, the authors show that a recent inversion is strongly associated with local fitness in a region of secondary contact, and that this region harbors three QTL associated with adaptation to differential water regimes in collinear populations, strongly suggesting that the alleles predate the inversion. However, these subspecies of Boechera are highly selfing, and it has been pointed out that the suppression of recombination would offer only a weak selective advantage in this system (Charlesworth & Barton, 2018). While the number of studies empirically testing for the mechanisms of chromosomal inversions in adaptation remains limited, those to date highlight a role of recombination suppression.

While our data support the recombination suppression hypothesis, we also cannot rule out that the LG8 inversion does not also contribute to adaptive

167

differentiation in other ways. The LG8 inversion could have direct effects, such as adaptively altering gene expression. Or, the inversion could have accumulated beneficial mutations after fixation. In both of these scenarios, the association between life history divergence and the LG8 inversion region would be intimately linked with the inverted haplotype, and would not be apparent in a collinear annual M. guttatus x M. tilingii cross. We find several lines of evidence to suggest that the inverted and non-inverted haplotypes found in perennials are not genetically equivalent. Firstly, while the LG8 inversion region in M. tilingii explains relatively the same proportion of the parental divergence, the absolute effect size of this region on many life history traits is much smaller in M. tilingii than in perennials which have the inversion. This either suggests that the LG8 inversion possess some alleles which are unique from the M. tilingii haplotype, or that M. tillingii, the ancestor to M. guttatus and M. decorus, or both have evolved modifier alleles which act in trans with the inversion-region/inversion to alter life history phenotypes. Determining the alleles which are unique to inverted perennials, and how they have been influenced by the LG8 inversion (i.e. either directly through the

LG8 inversion disrupting gene function, or indirectly by simply being harbored within the non-recombining inversion haplotype) will likely need to be resolved with a combination of population genetic, gene expression and targeted gene manipulation studies.

168

We also find evidence that the LG8 inversion may harbor alleles which are unique to perennial M. guttatus. In particular, Lowry and Willis (2010) have found a strong association with the inversion karyotype and stolon production (and this association has been recapitulated by Friedman et al. 2014). However, we fail to find an association between stolon production and inversion genotype for both a perennial species which posses the LG8 inversion (i.e. M. decorus) and a perennial species which lacks the LG8 inversion (i.e. M. tilingii). Interestingly, both M. decorus and M. tilingii produce stolons in slightly different fashions from perennial M. guttatus. In particular, coastal M. guttatus produces surface dwelling stolons which arise from the first or second node, remain aboveground, and eventually become reproductive themselves. In addition to these aboveground stolons, both M. tilingii and M. decorus produce stolons which burrow underground, become white, highly branched and easily fragmented, and many of which do not become reproductive (Appendix B). The developmental genetic basis of stolon production- including the role of the LG8 inversion- may thus vary between these perennial species.

6.5 Conclusions

Overall, we find that an important structural change for local adaptation and speciation within the M. guttatus species complex- LG8 inversion- likely arose and fixed in the perennial ancestor to perennial M. guttatus and perennial M. decorus. We find that

169

this inversion contributes substantially to trait variance between sympatric populations of annual M. guttatus and perennial M. decorus, and may be an important factor in maintaining species boundaries in the face of gene flow throughout the species complex.

In addition, we use a more distantly related perennial species, M. tilingii, which is collinear with annual M. guttatus to genetically dissect the LG8 inversion region. We find that, even in the absence of the inversion, this region is associated with a number of life history traits, and explains roughly the same proportion of parental divergence as the LG8 inversion between annual M. guttatus and perennials which have the inversion, although the absolute effect is generally smaller. We further show that the LG8 inversion region possesses at least 2 separate QTL in this collinear cross, which is consistent with a role for recombination suppression. While we find support for the recombination suppression hypothesis, we also find evidence to suggest that alleles which are unique to inverted perennial M. guttatus have evolved after the split between M. guttatus and M. decorus. Overall, we suggest that multiple factors may contribute to the adaptive significance of the LG8 inversion in this group.

170

Chapter 7: Conclusion

Here we explore diversity within perennials of the M. guttatus species complex.

We find one morphological variant- M. decorus- is a genetically unique entity, despite minimal phenotypic divergence from M. guttatus. Furthermore, we show that M. decorus is indeed several biological species; at minimum, a northern diploid taxon, a southern diploid taxon, and at least one tetraploid taxon. The diploids of this group are reproductively isolated from one another, as well as from M. guttatus by several post- zygotic barriers, the most substantial of which is hybrid seed inviability. Patterns of hybrid seed inviability between M. guttatus and both southern and northern diploid M. decorus, as well as between southern and northern M. decorus, show patterns of parent of origin effects on hybrid growth. Developmental surveys between M. guttatus and each northern and southern clade M. decorus suggest that viable seeds exhibit a maternal- excess phenotype (i.e. precocious endosperm proliferation), while inviable seeds show paternal excess (i.e. chaotic endosperm proliferation), despite differing in the direction of the cross. Lastly, we show that the genetic basis of hybrid seed inviability appears complex, involving multiple nuclear loci with parent of origin effects.

In addition to our work on hybrid seed inviability between members of the M. guttatus species complex, we also explore the role of a large chromosomal inversion in life history adaptation. We find that the LG8 inversion likely arose in perennials of the

171

M. guttatus species complex. The LG8 inversion also influences life history divergence in northern M. decorus, to a similar degree as it does in M. guttatus for most traits, except for the production of stolon. This region is still associated with life history divergence between annual M. guttatus and M. tilingii, despite these species being collinear along

LG8, and we have separated the overall phenotypic effects of this region into at least 2 separate QTL. Overall, we propose a model in which the LG8 inversion evolved in the perennial ancestor to M. guttatus and M. decorus. This inversion captured at least two loci that affected life history adaptation between annuals and perennials, and subsequent evolution in the lineage leading to perennial M. guttatus resulted in the evolution of novel, stolon-related phenotypes. We posit that the LG8 inversion is likely an important structural change for the maintenance of species in the complex, both between perennial and annual forms of M. guttatus and between annual M. guttatus and perennial M. decorus.

172

Appendix A: Assessing other reproductive barriers between M. guttatus and M. decorus

Methods

We discovered two other post-zygotic reproductive barriers between M. guttatus and M. decorus. The first is hybrid dwarfism, which was uncovered in an F2 population between IM62 (M. guttatus) and a broadly sympatric accession of M. decorus (IMP). We cold stratified F2, F1, and parental seed for 1 week at 4°C on moist Fafard 4P soil, then transferred to warm, long days in the Duke University Greenhouses. Germinants were transferred on the day of germination to individual pots of Fafard 4P soil.

Approximately 22% of F2s exhibited a dwarfism phenotype, which was categorized as extremely small size- both overall, and of individual vegetative traits. While some individuals were able to produce 1-2 flowers, the majority did not. We selfed the dwarf individuals which flowered and grew F3 families to quantify the proportion of dwarf offspring per family. Specifically, we grew 15-32 F3 individuals for each of 11 selfed dwarf mothers (with an average of 29 plants per dwarf mom).

We also collected tissue for all dwarf and non-dwarf individuals and created

GBS libraries using a modified Andolfatto approach with the enzyme Csp6I. We used the Tassel 3 GBS Pipeline to de-multiplex, clean, and align reads to the IM62 hard masked genome, as well as call SNPs and assign parental genotypes (Bradbury et al.

2007; Glaubitz et al. 2014). We imputed these data using Tassel to fill in missing calls, 173

when the flanking genotype calls were identical. We then used custom scripts in R to determine the frequency of each of each genotype in each of dwarfs and non-dwarfs phenotypic class along the genome for sites in which at least 100 dwarf and 40 non- dwarf individuals were assigned genotypes. Significant differences in genotype frequencies between dwarf and non-dwarf classes was assessed using a Chi-square test at each SNP and a significance threshold of 0.001. We found two regions of the genome which are significantly different between dwarf and non-dwarf classes, and confirmed these QTL using PCR markers (fragment-length polymorphisms, outlined in Fishman et al. 2014).

We also assessed F1 sterility between M. guttatus and northern and southern M. decorus, as well as between north and south M. decorus by staining anthers from F1 flowers in analine blue dye for at least 5-24 hours, and assessing pollen sterility by eye, for which viable and inviable pollen is readily distinguishable (Kelly et al. 2002).

Results

We note two other reproductive barriers between M. guttatus and the northern and southern clades of M. decorus. The first is between IM62 and IMP in which ~22% of the resultant F2 hybrids display a dwarf phenotype (Figure 25). We did not observe this phenotype in a similar F2 cross between another IM767 and a nearby M. decorus

174

population, BR, suggesting that this barrier is polymorphic throughout the range of one or both species.

Patterns of hybrid dwarfism in F3 families made from selfing F2 dwarf individuals rule out a cyto-nuclear incompatibility, as dwarf mothers gave rise to both dwarf and non-dwarf offspring (ranging from 25-100% dwarf offspring, with an average of 65% +/- 7.5%l Figure 25B). However, we note that some mothers produce fewer dwarf individuals than expected under a simple 2-locus nuclear-nuclear model. We find two

QTL conferring hybrid dwarfism; one locus on LG7 where the frequency of IMP homozygous individuals increases to 47.5% (IM homozygote frequency = 6.4%, as opposed to 25% each; Figure 25) and one on LG11 where the IMP homozygous genotype frequency plummets to 3.7% (IM62 homozygote frequency = 51%, as opposed to 25% each; Figure 25).

175

Figure 25: Genetics of hybrid dwarfism. (A) representative photos of non- dwarf (left) and dwarf (right) individuals. (B) histogram of the proportion of F3 dwarf offspring that selfed dwarf F2 parents gave rise to. (C) QTL mapping for loci which showed significant genotypic differences between dwarf and non-dwarf individuals. Red points indicate SNPs which were significantly different at a threshold of p=0.001. The dwarf individuals are more likely to be (D) heterozygous or homozygous IMPO at the LG7 QTL and (E) IM62 homozygous at the LG11 QTL than non-dwarf individuals (AA= homozygous IM62, AB= heterozygous, BB=homozygous IMPO. Panels D and E are based on PCR marker genotypes).

176

The second barrier of note is hybrid sterility between the southern clade of M. decorus and both M. guttatus (including IM62 and a coastal perennial, OPB) and northern clade

M. decorus (specifically IMP). Both Odell Creek and Diamond Lake display significant F1 sterility when they are the maternal parent (i.e. between 86-95% sterile pollen; note that

F1s in the opposite direction are those which are inviable as seeds and thus are untested). F1s were also sterile sterility between Odell creek and IMP, again when Odell creek was the maternal parent (~95% pollen sterility), although F1s made in the opposite direction of the cross appear to have very low pollen sterility (only 5-10%, which approximates the general levels of sterility found in the parents). While this strong parent of origin effect suggests a cyto-nuclear interaction (such as seen in Fishman &

Willis, 2006), much more work is needed in this system.

177

Appendix B: The evolution of underground stolons

Methods

Diversity within perennials of the M. guttatus species complex has been relatively underexplored, in part because many perennial taxa of this group appear strikingly phenotypically similar, and often inhabit quite similar habitats. Intriguingly, perennials within the M. guttatus group do vary significantly in the type and development of stolons produced. While all perennials make above ground stolons which stem from the base nodes of the plant, are surface dwelling, and will eventually become reproductive, some taxa also make underground stolons. These stolons become white, highly branched, produce embryonic leaves, and are quite fragile. We sought to assess the incidence, development, and genetic basis of underground stolons in perennials within and outside of the M. guttatus species complex. To this end, we performed two simultaneous grow out experiments- a species-level survey and an F2 mapping experiment.

For the species level survey, we grew an average of three siblings per maternal family and two maternal families per population. We surveyed nine coastal perennial M. guttatus populations, eight inland perennial M. guttatus populations, three M. tilingii populations, 17 M. decorus (including northern, southern, and polyploid clades), and two populations of the high elevation M. guttatus variant M. corallinus. Seeds were sprinkled

178

onto fafard 4P, cold stratified for one week, and germanants were transplanted on the day of germination into 4” pots in the Duke Greenhouses. We then assessed several stolon phenotypes along a developmental time series. As the underground stolons are quite fragile, measuring these traits can be destructive. To this end, we measured one maternal sibling from the three planted at each time point: three weeks after transplant, the day of first flower, and approximately three months after transplant. On the day of sampling, the plants were unpotted and the underground biomass was washed gently to remove all soil particles. We measured the presence of belowground stolons, as well as a number of traits for both above and below ground stolons: the number of stolons, length of the longest stolon, width of the longest stolon, leaf length, number of branches per node (hereafter branchiness), and the highest node at which a stolon was produced.

For the F2 mapping experiment, we created an F2 population by selfing a single

F1 between an underground stolon producing individual of M. decorus (IMP-IA) and a coastal perennial M. guttatus individual which did not make underground stolons

(OPB). We grew 429 F2 individuals, and 25 of each parent and F1s simultaneously with the population survey (i.e. 4” pots, long greenhouse days). Again, plants were measured at three developmental time points, but we simply measured the number of above and belowground stolons, and the highest node of stolon production to avoid plant destruction. At time point two (i.e. day of first flower), we also measured a number of

179

size and phenological traits, including days to first flower, node of first flower, corolla length and width, and leaf length and width. After all phenotyping was completed, we collected and froze tissue for each F2 individual. DNA was extracted using a modified

CTAB approach and individual libraries were made with the enzyme CSP6I for 384 of

429 F2 individuals using the approach described in Chapter 5. Samples were run across

2 lanes of the Illumina 4000 platform. We used Tassel 3 to demultiplex, clean, align, call

SNPs and impute genotypes. We ran custom scripts to calculate the proportion of genotype at each SNP position for individuals which did vs did not produce underground stolons.

Results

Population Survey

We find that only three of the five species surveyed produce underground stolons: M. corallinus, M. decorus, and M. tilingii (Figure 26). Intriguingly, all of these species are higher elevation dwelling taxa relative to inland or coastal perennial M. guttatus. Belowground stolons may be an important phenotype for over-winter survival, however its adaptive significance remains untested.

180

Figure 26: Proportion of underground stolons across all species and time points. Species key: COR: Mimulus corallinus, CPG= coastal perennial M. guttatus, DEC= M. decorus, IPG= inland perennial M. guttatus, TIL= M. tilingii.

Using only the phenotypic data collected from the latest survey point (i.e. 3 months), we also find a number of phenotypic differences between above and belowground stolons- aboveground stolons are on average longer, wider, produce larger leaves, occur at higher nodes, and are less branched than below-ground stolons (Figure

27). Based on observation, belowground stolons also lack chlorophyll, and are more fragile. The combination of increased branching and easily fragmenting ability of underground stolons may be important for dispersal along the stream environments that these perennials inhabit.

181

Figure 27: Differences between above- and below-ground stolons for a number of traits: (A) width, (B) highest node of emergence, (C) leaf length, (D) length, and (E) branches per node. Species key: COR: Mimulus corallinus, CPG= coastal perennial M. guttatus, DEC= M. decorus, IPG= inland perennial M. guttatus, TIL= M. tilingii.

Lastly, we also roughly mapped the genetic basis of the ability to produce underground stolons. In accordance with our population survey, we find that the coastal perennial accession OPB did not produce underground stolons, while all M. decorus accession IMPIA produced underground stolons, and approximately 43% of stolons in

IMPIA were underground. All but one F1 also made underground stolons, and approximately 25% of stolons were underground in F1s. This suggests that the ability to 182

produce underground stolons is dominant, while the proportion of underground stolons shows little dominance. In our F2 population, we uncovered three phenotypic classes- individuals which produced underground stolons in which the primary branch is oriented underground, individuals that did not produce primary branches which oriented underground, but that produced secondary side branches which oriented underground, and individuals that produce no underground stolons in any capacity.

Sixty-five percent of F2s produce underground stolons in which the primary branch had burrowed underground, 26% of F2s produced side branches which burrowed underground, but no primary stolons, and only 8% produced solely aboveground stolons.

183

Figure 28: Genetic map of underground stolon production. (A) representative M. decorus (DEC) and coastal perennial M. guttatus (CPG), and (B) a magnification of their stolons. (C-E) Regions of the genome with genotype differences between individuals with underground stolons present vs absent. Black dots= difference between individuals with underground stolons absent vs present in the frequency of the OPB homozygous genotype, grey dots: difference between individuals with underground stolons absent vs present in the frequency of the IMP homozygous genotype. LG4 and LG5 are in the direction expected (i.e. IMP homozygous genotype is associated with underground stolon presence), while LG14 shows opposite allelic effect.

184

We have performed a preliminary map to determine regions of the genome which are associated with underground stolon production (including the burrowing of secondary side branches). We find three QTLs associated with the ability to produce underground stolons: one on each LG4, LG5, and LG14. The QTL on LGs 4 and 5 are in the direction expected by the parental phenotypes (i.e. the presence of the OPB homozygous genotype increases in individuals which do not produce underground stolons, and decreases in individuals which do; Figure 28). However, the QTL on LG14 shows phenotypic effects that are opposite of the parental phenotype (Figure 28). Much future work is needed to map underground stolon production. Future analyses will include the calculation of linkage maps and marker confirmations.

185

Appendix C: Supplemental figures and tables (in chronological order)

Supplemental Figures and Tables: Chapter 3- Crossing Survey

Figure 29: Phylogeny for M. guttatus and M. decorus using NJ methods. Color bars represent the species (pink= M. guttatus, yellow= M. moschatus, grey= M. decorus). Grey box outlines all samples of M. decorus as a monophyletic group.

186

Figure 30: Population averages for hybrid seed inviability averaged across M. guttatus populations: of (top down) viability, symmetry of viability, germination, symmetry of germination for each line of M. decorus tested (averaged across the 1-4 lines of M. guttatus).

187

Figure 31: Population averages for hybrid seed inviability averaged across M. decorus populations: Averages of (top down, Left to right) viability, symmetry of viability, germination, symmetry of germination for each line of M. guttatus tested (averaged across the 19 lines of M. decorus).

188

Figure 32: Averages seed set for inter- and intraspecific crosses between M. decorus (D) and M. guttatus (G). Maternal parent in the cross is listed first.

Figure 33: Chromosome squashes of three diploid populations of M. decorus: IMP (Iron Mountain Perennial), Odell Creek, LL (Lumberlost Campground). Chromosomes/chromosome pairs are labeled.

189

Figure 34: Chromosome squashes of three tetraploid populations of M. decorus: BM (Big Meadow), ZIG (ZigZag Creek), HWY26 (Highway 26). Chromosomes/chromosome pairs are labeled.

190

Figure 35: 2C DNA content for each population of M. decorus, indexed by average 2C DNA content. The transition between diploids and tetrpaloids occurs around 1.4pg.

191

Table 5: Population Collections of M. decorus used throughout this dissertation. Population code = abbreviation for population name, X indicates population was used in crossing survey.

Population Population Latitude Longitude Included Included in Ploidy Genetic code in focal cluster original individual crossing survey? survey? Big Meadow BM 44.4985167 -121.98321 X X 4x 4x campground Browder BR 44.371 -122.104 X X 2x North Ridge Quarry CAR CAR 47.0344167 -122.03486 X X 4x CHR CHR 46.7806833 -121.77908 X X 4x Diamond Lake DL 43.1575333 -122.13343 X X 2x South Hacklemen's HACK 44.40278 -122.07556 X 2x North Creek HAM HAM 47.5651833 -123.03233 X X 4x 4x HJA HJA 44.2332 -122.1762 X X 2x North Hwy 26 near Hwy26 45.2921833 -121.73482 X X 4x Gov't Camp HWY15 Hwy15 44.39323475 -122.14897 X 2x North IMP IMP 44.3934 -122.149 X X 2x North Junction of 35 35-48 45.3054667 -121.66688 X X 4x + 48 Kink Creek KINK 44.3032 -121.99622 X X 2x North Lumberlost LL 44.1739333 -122.0542 X X 2x North campground Multinomah Mult 45.5762 -122.1158 X 4x Falls N. Santiam NS 44.5237667 -121.99788 X X 4x River Odell Creek Odell 43.54795 -121.96278 X X 2x South Silver Falls Silf 44.877 -122.6552 X 4x Trillium Lake Trill 45.2665833 -121.74163 X X 4x 4x Wildwood WW 45.3498667 -121.99295 X 4x Rec. Area ZigZag river Zig 45.31105 -121.8892 X X 4x

192

Table 6: Average viability and standard error of crosses between populations of M. decorus averaged across populations of M. guttatusby both (A) morphologically assessed hybrid seed inviability and (B) germination. Colors of the populations of M. decorus correspond to the genetic clade: yellow= Northern clade, blue=Southern clade, green=polyploid. Maternal donor listed first: D= M. decorus, G= M. guttatus. F= F statistic from an ANOVA, asterisks denote significance: 0.1>p>0.05= +, 0.05>p>0.01=*,0.01>p>0.001=**,p<0.001=***.

Crossing Survey between M. decorus and M. guttatus (A) Morphological

DxD DxG GxD GxG F BR 0.91(0.06) 0.005 (0) 0.90(0.02) 0.87(0.05) 94.58*** HJA 0.31(0.2) 0.05 (0) 0.82(0.04) 0.87(0.05) 148.6*** IMP 1(0.0) 0.0005 (0) 0.83(0.03) 0.87(0.05) 125.6*** KINK 0.98 (0.01) 0.0004 (0) 0.87(0.03) 0.87(0.05) 223*** LL 0.82(0.03) 0.48(0.06) 0.72(0.06) 0.87(0.05) 0.292- Odell 0.96 (0) 0.62(0.12) 0.07(0.06) 0.87(0.05) 7.126*** DL 0.922(0.03) 0.83(0.02) 0.70(0.04) 0.87(0.05) 1.967- BM 0.91 (0.01) 0.73(0.04) 0.31(0.07) 0.87(0.05) 3.134* CAR 0.72 (0.02) 0.79(0.09) 0.87(0.05) 3.837*

CHR 0.83(0.03) 0.59(0.08) 0.87(0.05) 7.168** HAM 0.92(0.2) 0.85(0.03) 0.22(0.06) 0.87(0.05) 13.56*** Hwy26 0.79 (0.02) 0.87(0.02) 0.86(0.01) 0.87(0.05) 1.027- 35-48 0.82 (0.03) 0.83 (0.03) 0.79(0.03) 0.87(0.05) 0.134- MultF 0.88(0.01) 0.86(0.01) 0.32(0.16) 0.87(0.05) 5.097** NS 0.90(0.02) 0.91(0.01) 0.45(0.07) 0.87(0.05) 9.208*** SILF 0.96 (0) 0.85(0.06) 0.017(0.01) 0.87(0.05) 27.24*** TRILL 0.82(0.06) 0.87(0.01) 0.59(0.08) 0.87(0.05) 7.059*** WW 0.911 (0.0) 0.91(0.02) 0.063(0.03) 0.87(0.05) 5.807** ZigZag 0.89 (0.04) 0.77(0.04) 0.21(0.07) 0.87(0.05) 25.64*** (B) Germination

DxD DxG GxD GxG F BR 0.87(0.05) 0.03 (0.016) 0.78 (0.06) 0.79 (0.06) 50.37*** HJA 0.95 (0.03) 0.17 (0.04) 0.68 (0.06) 0.79 (0.06) 35.18*** IMP 0.98 (0.02) 0.01 (0.007) 0.41 (0.08) 0.79 (0.06) 50.6*** KINK 0.97 (0.02) 0.07 (0.02) 0.77 (0.07) 0.79 (0.06) 54.8***

193

LL 0.99 (0.01) 0.86 (0.04) 0.95 (0.02) 0.79 (0.06) 3.472* Odell 0.96 (0.03) 0.34 (0.03) 0.07 (0.02) 0.79 (0.06) 60.6*** DL 1 (0) 0.38 (0.07) 0.56 (0.05) 0.79 (0.06) 15.93*** BM 0.98 (0.02) 0.62 (0.05) 0.58 (0.04) 0.79 (0.06) 8.447*** CAR 0.47 (0.069) 0.74 (0.07) 0.79 (0.06) 6.68**

CHR 0.47 (0.07) 0.43 (0.06) 0.79 (0.06) 7.536** HAM 1 (0) 0.83 (0.034) 0.13 (0.02) 0.79 (0.06) 70.53*** Hwy26 0.96 (0.04) 0.82 (0.03) 0.88 (0.03) 0.79 (0.06) 2.44+ 35-48 1 (0) 0.67 (0.08) 0.9 (0.03) 0.79 (0.06) 4.518** MultF 0.92 (0.08) 0.33 (0.07) 0.16 (0.05) 0.79 (0.06) 14.55*** NS 1 (0) 0.44 (0.05) 0.63 (0.07) 0.79 (0.06) 11.9*** SILF 1 (0) 1 (0) 0.33 (0.11) 0.79 (0.06) 11.18*** TRILL 0.95 (0.05) 0.82 (0.05) 0.43 (0.05) 0.79 (0.06) 16.1*** WW 0.35 (0.11) 0.04 (0.03) 0.63 (0.15) 0.79 (0.06) 10.74*** ZigZag 0.99 (0.01) 0.85 (0.04) 0.66 (0.05) 0.79 (0.06) 5.567**

194

Table 7: Average viability and standard error of crosses between focal individual Odell Creek and populations of M. decorusfor both (A) morphologically assessed hybrid seed inviability and (B) germination. Colors of the populations of M. decorus correspond to the genetic clade: yellow= Northern clade, blue=Southern clade, green=polyploid. Maternal donor listed first: D= M. decorus, G= M. guttatus. F= F statistic from an ANOVA, asterisks denote significance: 0.1>p>0.05= +, 0.05>p>0.01=*,0.01>p>0.001=**,p<0.001=***.

Crossing Survey focal M. decorus: Odell Creek (A) Morphological DxD DxG GxD GxG F HACK 0.95 (0.04) 0 (0) 0.003 (0.002) 0.83 (0.05) 123.3*** HJA 0.78 (0.035) 0.001 (0.001) 0.006 (0.004) 0.83 (0.05) 252.6*** HWY15 0.88(0.07) 0 (0) 0 (0) 0.83 (0.05) 52.99*** KINK 0.98 (0.006) 0 (0) 0.0008 (0.0008) 0.83 (0.05) 394.6*** LL 0.88 (0.03) 0 (0) 0 (0) 0.83 (0.05) 154.3*** DL 0.5 (0.31) 1 (0) 0.96 (0.01) 0.83 (0.05) 2.996+ Hwy26 0.56 (0.21) 0.94 (0.04) 0.58 (0.19) 0.83 (0.05) 1.901- 35-48 0.93 (0.013) 0.76 (0.08) 0.89 (0.04) 0.83 (0.05) 0.952- NS 0.79 (0.18) 0.97 (0.01) 0.96 (0.02) 0.83 (0.05) 1.152- TRILL 0.96 (0.028) 0.93 (0.05) 0.98 (0.007) 0.83 (0.05) 2.07- WW 0.89 (0.04) 0.55 (0.16) 0.95 (0.022) 0.83 (0.05) 1.474- ZigZag 0.89 (0.04) 0.7 (0.15) 0.93 (0.07) 0.83 (0.05) 1.655- (B) Germination DxD DxG GxD GxG F HACK 0.7 (0.05) 0 (0) 0 (0) 0.3 (0.13) 71.88*** HJA 0.42 (0.11) 0 (0) 0 (0) 0.3 (0.13) 16.64*** HWY15 0.95 (0.05) 0 (0) 0 (0) 0.3 (0.13) 127*** KINK 0.28 (0.18) 0 (0) 0 (0) 0.3 (0.13) 11.56*** LL 1 (0) 0 (0) 0 (0) 0.3 (0.13) 218.2*** DL 0.96 (0.02) 1 (0) 0.96 (0.02) 0.3 (0.13) 3.783* Hwy26 0.68 (0.08) 0.98 (0.02) 0.46 (0.14) 0.3 (0.13) 2.36- 35-48 0.87 (0.09) 0.86 (0.14) 0.05 (0.03) 0.3 (0.13) 7.729*** NS 1 (0) 0.77 (0.19) 1 (0) 0.3 (0.13) 1.859- TRILL 0.99 (0.01) 0.73 (0.17) 0.56 (0.14) 0.3 (0.13) 4.819* WW 0.35 (0.0.12) 0.66 (0.21) 0.53 (0.2) 0.3 (0.13) 2.936*

195

ZigZag 0.99 (0) 1 (0) 0.9 (0.06) 0.3 (0.13) 2.156-

196

Table 8: Average viability and standard error of crosses between focal individual IMP and populations of M. decorusfor both morphologically assessed hybrid seed inviability and germination. Colors of the populations of M. decorus correspond to the genetic clade: yellow= Northern clade, blue=Southern clade, green=polyploid. Maternal donor listed first: D= M. decorus, G= M. guttatus. F= F statistic from an ANOVA, asterisks denote significance: 0.1>p>0.05= +, 0.05>p>0.01=*,0.01>p>0.001=**,p<0.001=***.

Crossing Survey focal M. decorus: IMP (A) Morphological DxD DxG GxD GxG F BR 0.58 (0.03) 0.60 (0.11) 0.92 (0.02) 0.93 (0.02) 17.02*** HACK 0.96 (0.01) 0.93 (0.02) 0.94 (0.01) 0.93 (0.02) 1.8- HJA 0.71 (0) 0.91 (0.04) 0.98 (0) 0.93 (0.02) 24.9*** HWY15 0.89 (0.06) 0.98 (0.02) 0.74 (0.03) 0.93 (0.02) 13.26*** KINK 0.93 (0.01) 0.71 (0.21) 0.96 (0.01) 0.93 (0.02) 0.109- LL 0.90 (0) 0.84 (0.03) 0.95 (0.01) 0.93 (0.02) 7.353** Odell 0.89 (0.02) 0.0007 (0) 0.002 (0) 0.93 (0.02) 1885*** DL 0.93 (0.02) 0.0007 (0) 0.0006 (0) 0.93 (0.02) 2305*** BM 0.94 (0.01) 0 (0) 0 (0) 0.93 (0.02) 3990*** HAM 0.98 (0.01) 0.002 (0) 0 (0) 0.93 (0.02) 2151*** Hwy26 0.79 (0.02) 0 (0) 0 (0) 0.93 (0.02) 1145*** 35-48 0.93 (0.01) 0 (0) 0.002 (0) 0.93 (0.02) 3301*** NS 0.95 (0.01) 0.016 (0) 0.0003 (0) 0.93 (0.02) 3058*** TRILL 0.92 (0.04) 0.001 (0) 0.0005 (0) 0.93 (0.02) 802.8*** WW 0.89 (0.04) 0 (0) 0.003 (0) 0.93 (0.02) 5315*** ZigZag 0.93 (0.02) 0.004 (0) 0.017 (0.01) 0.93 (0.02) 2409*** (B) Germination DxD DxG GxD GxG F BR 0.62 (0.07) 0.93 (0.03) 1 (0) 0.94 (0.02) 23.89*** HACK 0.7 (0.03) 0.79 (0.1) 0.89 (0.04) 0.94 (0.02) 212.9*** HJA 0.75 (0.1) 0.87 (0.06) 0.98 (0.02) 0.94 (0.02) 4.193* HWY15 1 (0) 0.96 (0.03) 0.84 (0.02) 0.94 (0.02) 4.14* KINK 0.61 (0.09) 1 (0) 0.99 (0.01) 0.94 (0.02) 18.5*** LL 0.9 (0.03) 0.88 (0.03) 0.9 (0.02) 0.94 (0.02) 1.449- Odell 0.89 (0.05) 0 (0) 0 (0) 0.94 (0.02) 351.3***

197

DL 0.66 (0.1) 0 (0) 0 (0) 0.94 (0.02) 82.74*** BM 0.97 (0.02) 0 (0) 0 (0) 0.94 (0.02) 753.3*** HAM 0.86 (0.06) 0 (0) 0 (0) 0.94 (0.02) 250*** Hwy26 0.58 (0.07) 0 (0) 0 (0) 0.94 (0.02) 139.2*** 35-48 0.94 (0.02) 0 (0) 0 (0) 0.94 (0.02) 654.5*** NS 0.98 (0.01) 0 (0) 0 (0) 0.94 (0.02) 888.6*** TRILL 0.64 (0.11) 0 (0) 0 (0) 0.94 (0.02) 69.31*** WW 0.35 (0.12) 0 (0) 0.01 (0.01) 0.94 (0.02) 41.42*** ZigZag 0.87 (0.02) 0 (0) 0 (0) 0.94 (0.02) 594.2***

198

Table 9: Average viability of crosses between focal individual 35-48 and populations of M. decorusfor both morphologically assessed hybrid seed inviability and germination. Colors of the populations of M. decorus correspond to the genetic clade: yellow= Northern clade, blue=Southern clade, green=polyploid. Maternal donor listed first: D= M. decorus, G= M. guttatus. F= F statistic from an ANOVA, asterisks denote significance: 0.1>p>0.05= +, 0.05>p>0.01=*,0.01>p>0.001=**,p<0.001=***.

Crossing Survey focal M. decorus: Jct. 35-48 (A) Morphological DxD DxG GxD GxG F HACK 0.05 (0.04) 0 (0) 0 (0) 0.87 (0.03) 98.56*** HJA 0.78 (0.04) 0 (0) 0 (0) 0.87 (0.03) 113.9*** HWY15 0.83 (0.08) 0 (0) 0 (0) 0.87 (0.03) 20.47*** KINK 0.98 (0.01) 0 (0) 0 (0) 0.87 (0.03) 225.1*** LL 0.88 (0.03) 0 (0) 0 (0) 0.87 (0.03) 227.1*** Odell 0.3 (0.13) 0.89 (0.04) 0.75 (0.08) 0.87 (0.03) 1.012- DL 0.5 (0.31) 0.89 (0.03) 0.88 (0) 0.87 (0.03) 2.418+ Hwy26 0.56 (0.21) 0.81 (0.12) 0.5 (0.15) 0.87 (0.03) 5.766** NS 0.79 (0.18) 0.77 (0.09) 0.97 (0.03) 0.87 (0.03) 0.667- TRILL 0.96 (0.028) 0.86 (0.1) 0.93 (0.02) 0.87 (0.03) 0.348- WW 0.89 (0.04) 0.85 (0.07) 0.76 (0.1) 0.87 (0.03) 1.087- ZigZag 0.89 (0.04) 0.71 (0.10) 0.58 (0.11) 0.87 (0.03) 6.12** (B) Germination DxD DxG GxD GxG F HACK 0.7 (0.05) 0 (0) 0 (0) 0.87 (0.09) 71.88*** HJA 0.42 (0.11) 0 (0) 0 (0) 0.87 (0.09) 16.64*** HWY15 0.95 (0.05) 0 (0) 0 (0) 0.87 (0.09) 127*** KINK 0.28 (0.18) 0 (0) 0 (0) 0.87 (0.09) 11.56*** LL 1 (0) 0 (0) 0 (0) 0.87 (0.09) 218.2*** Odell 0.3 (0.13) 0.05 (0.03) 0.86 (0.14) 0.87 (0.09) 22.52*** DL 0.96 (0.02) 0.64 (0.12) 0.83 (0.12) 0.87 (0.09) 3.783* Hwy26 0.68 (0.08) 0.92 (0.08) 1 (0) 0.87 (0.09) 2.36- NS 1 (0) 1 (0) 0.94 (0.06) 0.87 (0.09) 1.859- TRILL 0.99 (0.01) 0.62 (0.17) 1 (0) 0.87 (0.09) 4.819* WW 0.355 (0.12) 0.42 (0.16) 0.61 (0.21) 0.87 (0.09) 2.936* ZigZag 0.99 (0.01) 1 (0) 0.99 (0.01) 0.87 (0.09) 2.156-

199

Table 10: Average seed size and standard error of crosses between populations of M. decorus averaged across populations of M. guttatus. Colors of the populations of M. decorus correspond to the genetic clade: yellow= Northern clade, blue=Southern clade, green=polyploid. Maternal donor listed first: D= M. decorus, G= M. guttatus. F= F statistic from an ANOVA, asterisks denote significance: 0.1>p>0.05= +, 0.05>p>0.01=*,0.01>p>0.001=**,p<0.001=***.

Seed Sizes DxD DxG GxD GxG F BR 0.020 (0.001) 0.019 (0.0009) 0.017 (0.0006) 0.02 (0.0007) 2.476+ HJA 0.0217 (0.0012) 0.025 (0.0012) 0.014 (0.0042) 0.02 (0.0007) 24.83*** IMP 0.022 (0.0009) 0.023 (0.0009) 0.015(0.0005) 0.02 (0.0007) 24.43*** KINK 0.0252 (0.001) 0.025 (0.0008) 0.014 (0.0005) 0.02 (0.0007) 42.43*** LL 0.019 (0.0009) 0.026 (0.0007) 0.017 (0.0006) 0.02 (0.0007) 22.66*** Odell 0.02 (0.0005) 0.015 (0.0004) 0.024(0.0007) 0.02 (0.0007) 38.22*** DL 0.0178 (0.0006) 0.0137 (0.0004) 0.025 (0.0007) 0.02 (0.0007) 55.66*** BM 0.0232 (0.0009) 0.014 (0.0004) 0.0235 (0.0006) 0.02 (0.0007) 52.02*** CAR 0.0216 (0.0009) 0.017 (0.0007) 0.0224(0.00122) 0.02 (0.0007) 4.03** CHR 0.0179 (0.0006) 0.0178 (0.0008) .0216 (0.001) 0.02 (0.0007) 2.78* HAM 0.026 (0.0011) 0.017(0.0008) 0.026 (0.0012) 0.02 (0.0007) 16.19*** Hwy26 0.017 (0.0007) 0.014 (0.0008) 0.0237 (0.0012) 0.02 (0.0007) 12.08*** 35-48 0.029 (0.001) 0.17 (0.0007) 0.18 (0.0004) 0.02 (0.0007) 23.77*** NS 0.021 (0.0008) 0.015 (0.0004) 0.026 (0.0009) 0.02 (0.0007) 41.7*** TRILL 0.022 (0.001) 0.017 (0.001) 0.024 (0.0001) 0.02 (0.0007) 4.769** ZigZag 0.023 (0.012) 0.02 (0.0009) 0.017 (0.0009) 0.02 (0.0007) 4.1**

200

Supplemental Figures: Chapter 4- Seed Development Survey

Figure 36: Total seed counts for (A) IMP and (B) Odell Creek crosses with IM lines. Cross types of different letters denote groups with significant differences for IMP and Odell creek separately.

201

Supplemental Tables and Figures: Chapter 5- Genetics of Hybrid Seed Lethality

Figure 37: Correlation between the proportion of wrong homozygous calls versus the number of missing SNPs in an individual. Wrong homozygous calls are defined as homozygous genotype for the parent that the F1 was crossed away from (i.e. IM62 in IMPxF1 and IMPO in the F1xIM cross).

202

Figure 38: Mimulus decorus maternal alleles associated with HSI as determined from GBS data of the F1xIM cross. Each point represents a window, red points = positions with genotype counts that differ significantly from the expectated.

203

Figure 39: Mimulus guttatus paternal alleles associated with HSIas determined from the GBS data of IMPxF1 cross. Each point represents a window, red points = positions with genotype counts that differ significantly from 50/50 Mendelian expectation.

204

205

Table 11: Multi-locus genotypes: observed and expected values

A BCIMP_3 loci

# loci homozygous 0 1 2 3

#individuals 3 21 30 20

%individuals 0.040540541 0.283783784 0.405405405 0.27027027

probability 0.046243986 0.248560286 0.443094839 0.262100889

expected # individuals 3.422054964 18.39346115 32.78901808 19.39546581

B BCIMP_5 loci # loci homozygous 0 1 2 3 4 5 #individuals 1 5 19 21 13 12 %individuals 0.014084507 0.070422535 0.267605634 0.295774648 0.183098592 0.169014085 probability (assuming ~0.6 0.01024 0.0768 0.2304 0.3456 0.2592 0.07776 TRD towards IMP) expected 0.72704 5.4528 16.3584 24.5376 18.4032 5.52096

C BCIM_3 loci

# loci homozygous 0 1 2 3

#individuals 3 14 37 21

%individuals 0.052631579 0.245614035 0.649122807 0.368421053

probability (assuming ~0.6 0.035437037 0.217777778 0.4448 0.301985185 TRD towards IMP) expected 2.657777778 16.33333333 33.36 22.64888889

D BCIM_2 loci

# loci homozygous 0 1 2

#individuals 10 42 41

%individuals 0.107526882 0.451612903 0.440860215

probability (assuming ~0.6 0.102222222 0.435555556 0.462222222 TRD towards IMP) expected 9.506666667 40.50666667 42.98666667

Supplemental Tables: Chapter 6: Inversions and Life History Adaptation

Table 12: Inversion survey across species of the Mimulus guttatus species complex and outgroup M. lewisii. 206

Evidence for Inversion Population Species Life History Inversion Reference Orientation Orientation Conserved Linkage marker order maps from LF M. lewisii P A in linkage Fishman et maps al. 2014 SOP: 20% recombination LVR: Garner SOP, LVR M. tilingii P A in F2s with et al. 2015 IM62

BR, IMP: 0% recombination in F2s with IM62. IMP: BR, IMP M. decorus P P ~5% recombination in F2s with OPB

22% recombination OLM M. laciniatus A A in F2 with IM62 14% M. recombination PMO A A glaucescens in F2 with IM62

14% recombination in F2 with REM M. nudatus A A sympatric, annual M. guttatus Lowry & SF M. nasutus A A Willis 2010

Table 13: Differences between parental lines and the effect of the inversion or inversion regionin an F2 population for several life history traits. Traits: FT= days from germination to first flower, FN= node of first flower, CL= corolla length, TL= corolla tube length, CW= corolla width, ST= stem thickness, LL= ength of the first true leaf, LW= leaf width, IL1= internode length between cotyledons and first true leaf, IL2= internode length between first and second true leaves, NS= number of stolons, SL= stolon length, #B/N= number of side branches per node before first flower. P-

207

values are in parentheses for non-significant comparisons, otherwise asterisks denote significant: +=p<0.1, *=p<0.05, **=p<0.01, ***=p<0.001, ****=p<0.0001.

Associations Associations Parental between LG8 Parental between LG8 difference, M. inversion difference, M. inversion region

decorus x M. karyotype and tilingii x M. and trait (M. guttatus trait (M. decorus guttatus tilingii x M. x M. guttatus) guttatus)

F statistic F Statistic F statistic F Statistic Trait (DF=2) (DF=3) (DF=2) (DF=3) FT 701.6**** 16.68**** 48.15*** 2.875+ FN 977.5**** 6.233*** 49.19*** 3.159+ CL 1080**** 39.79**** TL 1116**** 40.76**** 169.6*** 10.72** CW 288.6**** 31.95**** 89.82*** 6.266* ST 509.3**** 61.14**** 220.3*** 4.929* LL 203.1**** 41.43**** 19.46*** 0.206 LW 158.3**** 30.95**** IL1 180**** 11.92**** 185.9*** 1.42 IL2 392.5*** 23.82*** 44.73*** 3.24+ NS 831.3**** 1.7 (0.166) 341.3*** 0.536 SL 439.9**** 35.97**** #B/N 87.84**** 4.96**

Table 14: Dissecting the effect of the inversion karyotype on life history trait divergence between M. tilingii and annual M. guttatus using an F3 fine mapping approach Asterisks denote significance among alternative homozygotes: +=p<0.1;*=p<0.05, **=p<0.01, ***=p<0.001. O denotes the effect is in the direction opposite to expected, based on parental phenotypes. For class 4, ANOVAs were not completed, as the MANOVA including all traits was not significant.

Class 1 Class 2 Class 3 Class 4 ANOVA DF=1 DF=1 DF=2 traits 208

FT 0.455 3.8* 3.522**

FN 0.48 0.011 1.17

TL 4.42* 3.584+ 6.116**

CW 0.038 5.602* 0.55

ST 0.125 12.2***O 0.022

IL1 0.45 2.416 5.108*

IL2 0.142 4.933* 3.874*

SN 1.036 3.4 3.874*

2.6678**, 3.1839*, 3.4041***, 1.7547 MANOVA DF=9 DF=4 DF=18 ,DF=8

209

References

Aeschbacher, S., Selby, J. P., Willis, J. H., & Coop, G. (2017). Population-genomic inference of the strength and timing of selection against gene flow. Proceedings of the National Academy of Sciences, 114(27), 7061–7066.

Agrawal, A. F., Feder, J. L., & Nosil, P. (2011). Ecological divergence and the origins of intrinsic postmating isolation with gene flow. International Journal of Ecology, 2011.

Alcázar, R., García, A. V, Parker, J. E., & Reymond, M. (2009). Incremental steps toward incompatibility revealed by Arabidopsis epistatic interactions modulating salicylic acid pathway activation. Proceedings of the National Academy of Sciences of the United States of America, 106(1), 334–9.

Allen, W. R. (1969). Factors influencing pregnant mare serum gonadotrophin production. Nature, 223, 177–178.

Anderson, J. B., Funt, J., Thompson, D. A., Prabhu, S., Socha, A., Sirjusingh, C., … Kohn, L. M. (2010). Determinants of divergent adaptation and dobzhansky-muller interaction in experimental yeast populations. Current Biology, 20(15), 1383–1388.

Andolfatto, P., Davison, D., Erezyilmaz, D., Hu, T. T., Mast, J., Sunayama-morita, T., & Stern, D. L. (2011). Multiplexed shotgun genotyping for rapid and efficient genetic mapping Multiplexed shotgun genotyping for rapid and efficient genetic mapping. Genome Research, 21(4), 610–617.

Arnegard, M. E., McGee, M. D., Matthews, B., Marchinko, K. B., Conte, G. L., Kabir, S., Bedford, N., Bergek, S., Chan, Y. F. Jones, F. C., Kingsley, D. M., Peichel, C. L., Schluter, D. (2014). Genetics of ecological divergence during speciation. Nature, 511, 307-311.

Bachtrog, D. (2013). Y-chromosome evolution: Emerging insights into processes of Y- chromosome degeneration. Nature Reviews Genetics, 14(2), 113–124.

Bachtrog, D., Kirkpatrick, M., Mank, J. E., McDaniel, S. F., Pires, J. C., Rice, W. R., & Valenzuela, N. (2011). Are all sex chromosomes created equal? Trends in Genetics, 27(9), 350–357.

210

Baker, Z., Schumer, M., Haba, Y., Holland, C., Rosenthal, G. G., & Przeworski, M. (2016). Repeated losses of PRDM9-directed recombination despite the conservation of PRDM9 across vertebrates. eLife, 6, 1–58.

Balcova, M., Faltusova, B., Gergelits, V., Bhattacharyya, T., Mihola, O., Trachtulec, Z., Knopf, C., Fotopulosova, V., Chvatalova, I., Gregorova, S., Forejt, J. (2016). Hybrid Sterility Locus on Chromosome X Controls Meiotic Recombination Rate in Mouse. PLoS Genetics, 12(4), 1–16.

Barbash, D. A., Siino, D. F., Tarone, a. M., & Roote, J. (2003). A rapidly evolving MYB- related protein causes species isolation in Drosophila. Proceedings of the National Academy of Sciences, 100(9), 5302–5307.

Barnard-Kubow, K. B., & Galloway, L. F. (2017). Variation in reproductive isolation across a species range. Ecology and Evolution, (August), 1–11.

Barr, C. M., & Fishman, L. (2010). The nuclear component of a cytonuclear hybrid incompatibility in mimulus maps to a cluster of pentatricopeptide repeat genes. Genetics, 184(2), 455–465.

Barrière, A., Gordon, K. L., & Ruvinsky, I. (2012). Coevolution within and between Regulatory Loci Can Preserve Promoter Function Despite Evolutionary Rate Acceleration. PLoS Genetics, 8(9), 1–13.

Bates, D., Maechler M., Bolker B., Walker S. (2015). Fitting Linear Mixed-Effects Models Using lme4. Journal of Statistical Software, 67(1), 1-48.

Bateson, W., 1909 Heredity and variation in modern lights, pp. 85-101 in Danvin and Modmz Science, edited by A. C. SEWARD. Cambridge University Press, Cambridge.

Baudat, F., Buard, J., Grey, C., Fledel-Alon, A., Ober, C., Przeworski, M., Coop, G. M., de Massy, B. (2010). PRDM9 is a major determinant of meiotic recombination hotspots in humans and mice. Science, 327, 1225–1237.

Beardsley, P. M., Schoenig, S. E., Whittall, J. B., & Olmstead, R. G. (2004). Patterns of Evolution in Western North American Mimulus (Phrymaceae). American Journal of Botany, 91(3), 474–489.

211

Bentolila, S., & Hanson, M. R. (2001). Identification of a BIBAC clone that co-segregates with the petunia restorer of fertility (RF) gene. Molecular Genetics and Genomics, 266(2), 223–230.

Berg, I. L., Neumann, R., Lam, K. W. G., Sarbajna, S., Odenthal-Hesse, L., May, C. A., & Jeffreys, A. J. (2010). PRDM9 variation strongly influences recombination hot-spot activity and meiotic instability in humans. Nature Genetics, 42(10), 859–863.

Bikard, D., Patel, D., Le Mette, C., Giorgi, V., Camilleri, C., Bennett, M. J., & Loudet, O. (2009). Divergent Evolution of Duplicate Genes Leads to Genetic Incompatibiolities Within A. thaliana. Science, 323, 623–626.

Bingham, P. M., Kidwell, M. G., & Rubin, G. M. (1982). The molecular basis of P-M hybrid dysgenesis: The role of the P element, a P-strain-specific transposon family. Cell, 29(3), 995–1004.

Birchler, J. A. (1993). Dosage Analysis of Maize Endosperm Development. Annual Review of Genetics, 27(1), 181–204.

Birchler, J. A., & Hart, J. R. (1987). Interaction of endosperm size factors in maize. Genetics, 117(2), 309–17.

Blackburn, D. G. (2015). Viviparous placentotrophy in reptiles and the parent-offspring conflict. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 324(6), 532–548.

Blackman, B. K. (2016). Speciation Genes. Encyclopedia of Evolutionary Biology, 4, 166–175.

Blumenstiel, J. P., & Hartl, D. L. (2005). Evidence for maternally transmitted small interfering RNA in the repression of transposition in Drosophila virilis. Proceedings of the National Academy of Sciences, 102(44), 15965–15970.

Bolnick, D. I., Near, T. J., & Wainwright, P. C. (2006). Body size divergence promotes post-zygotic reproductive isolation in centrarchids. Evolutionary Ecology Research, 8(5), 903–913.

Bomblies, K. (2009). Too much of a good thing? Hybrid necrosis as a by-product of plant immune system diversification. Botany, 87(11), 1013–1022.

212

Bomblies, K., Lempe, J., Epple, P., Warthmann, N., Lanz, C., Dangl, J. L., & Weigel, D. (2007). Autoimmune response as a mechanism for a Dobzhansky-Muller-type incompatibility syndrome in plants. PLoS Biology, 5(9), 1962–1972.

Bomblies, K., & Weigel, D. (2007). Hybrid necrosis: autoimmunity as a potential gene- flow barrier in plant species. Nature Reviews. Genetics, 8(5), 382–93.

Bordenstein, S. R., O’Hara, F. P., & Werren, J. H. (2001). Wolbachia-induced incompatibility precedes other hybrid incompatibilities in Nasonia. Nature, 409(6821), 707–710.

Bracewell, R. R., Bentz, B. J., Sullivan, B. T., & Good, J. M. (2017). Rapid neo-sex chromosome evolution and incipient speciation in a major forest pest. Nature Communications, 8(1), 1593.

Bradbury, P. J., Zhang, Z., Kroon, D. E., Casstevens, T. M., Ramdoss, Y., & Buckler, E. S. (2007). TASSEL: Software for association mapping of complex traits in diverse samples. Bioinformatics, 23(19), 2633–2635.

Bradshaw, H., & Schemske, D. W. (2003). Allele substitution at a flower colour locus produces a pollinator shift in monkeyflowers. Nature, 44(3), 138–47.

Brandvain, Y., & Haig, D. (2005). Divergent mating systems and parental conflict as a barrier to hybridization in flowering plants. The American Naturalist, 166(3), 330–8.

Brandvain, Y., Kenney, A. M., Flagel, L., Coop, G., & Sweigart, A. L. (2014). Speciation and Introgression between Mimulus nasutus and Mimulus guttatus. PLoS Genetics, 10(6), e1004410.

Brekke, T. D., Henry, L. a., & Good, J. M. (2016). Genomic imprinting, disrupted placental expression, and speciation. Evolution, 3(Haig 1996), 1–14.

Brekke, Th. D., & Good, J. M. (2014). Parent-of-origin growth effects and teh evolution of hybrid inviability in dwarf hamsters. Evolution, 68(11), 3134–3148.

Brideau, N. J., Flores, H. a, Wang, J., Maheshwari, S., Wang, X., & Barbash, D. a. (2006). Lethality in Drosophila, 314, 1292–1295.

213

Brothers, A. N., & Delph, L. F. (2010). Haldane’s rule is extended to plants with sex chromosomes. Evolution, 64(12), 3643–3648.

Brown, G. G. (1999). Unique aspects of cytoplasmic male sterility and fertility restoration in Brassica napus. Journal of Heredity, 90(3), 351–356.

Brucker, R. M., & Bordenstein, S. R. (2013). The Hologenomic Basis of Speciation: Gut Bacteria Cause Hybrid Lethality in the Genus Nasonia. Science, 341(6146), 667–669.

Burkart-Waco, D., Josefsson, C., Dilkes, B., Kozloff, N., Torjek, O., Meyer, R., Altmann, T., Comai, L. (2012). Hybrid incompatibility in Arabidopsis is determined by a multiple-locus genetic network. Plant Physiology, 158(2), 801–12.

Burkart-Waco, D., Ngo, K., Lieberman, M., & Comai, L. (2015). Perturbation of parentally biased gene expression during interspecific hybridization. PLoS ONE, 10(2), 1–19.

Bushell, C., Spielman, M., & Scott, R. J. (2003). The basis of natural and artificial postzygotic hybridization barriers in Arabidopsis species. The Plant Cell, 15(6), 1430–42.

Burt, A. & Trivers, R. 2006. Genes in conflict: the biology of selfish genetic elements. Cambridge, Massachusetts: First Harvard university Press

Caicedo, a L., & Schaal, B. a. (2004). Heterogeneous evolutionary processes affect R gene diversity in natural populations of Solanum pimpinellifolium. Proceedings of the National Academy of Sciences, 101(50), 17444–17449.

Caicedo, a L., Schaal, B. a, & Kunkel, B. N. (1999). Diversity and molecular evolution of the RPS2 resistance gene in Arabidopsis thaliana. Proceedings of the National Academy of Sciences, 96(1), 302–6.

Case, A. L., Finseth, F. R., Barr, C. M., & Fishman, L. (2016). Selfish evolution of cytonuclear hybrid incompatibility in Mimulus. Proceedings of the Royal Society B, 283.

Case, A. L., & Willis, J. H. (2008). Hybrid male sterility in Mimulus (Phrymaceae) is associated with a geographically restricted mitochondrial rearrangement. Evolution, 62(5), 1026–1039. 214

Castillo, D. M., & Moyle, L. C. (2012). Evolutionary Implications of Mechanistic Models of TE-Mediated Hybrid Incompatibility. International Journal of Evolutionary Biology, 2012, 698198.

Catchen, J., Hohenlohe, P. a., Bassham, S., Amores, A., & Cresko, W. a. (2013). Stacks: An analysis tool set for population genomics. Molecular Ecology, 22(11), 3124–3140.

Charlesworth, B., & Barton, N. H. (2018). The spread of an inversion with migration and selection. Genetics, 208(1), 377–382.

Charlesworth, B., Coyne, J. A., & Barton, N. H. (1987). The Relative Rates of Evolution of Sex Chromosomes and Autosomes. American Naturalist, 130, 113–146.

Charlesworth, B., & Hartl, D. I. (1978). Population dynamics of the segregation distorter polymorphism of Drosophila melanogaster. Genetics, 89(1), 171–192.

Chase, C. D. (2007). Cytoplasmic male sterility: a window to the world of plant mitochondrial-nuclear interactions. Trends in Genetics, 23(2), 81–90.

Christie, K., & Strauss, S. Y. (2018). Along the speciation continuum: Quantifying intrinsic and extrinsic isolating barriers across five million years of evolutionary divergence in California jewelflowers. Evolution, 1–17.

Cooley, A. M., Modliszewski, J. L., Rommel, M. L., & Willis, J. H. (2011). Gene duplication in mimulus underlies parallel floral evolution via independent trans- regulatory changes. Current Biology, 21(8), 700–704.

Cooper, B. S., Ginsberg, P. S., Turelli, M., & Matute, D. R. (2017). Wolbachia in the Drosophila yakuba complex: Pervasive frequency variation and weak cytoplasmic incompatibility, but no apparent effect on reproductive isolation. Genetics, 205(1), 333–351.

Cornejo, P., Camadro, E. L., & Masuelli, R. W. (2012). Molecular bases of the postzygotic barriers in interspecific crosses between the wild potato species Solanum acaule and Solanum commersonii. Genome, 55(7620), 605–614.

Coyne, J. A., & Orr, H. A. (1989). Patterns of Speciation in Drosophila. Evolution, 43(2), 362–381.

215

Coyne, J. A., & Orr, H. A. (2004). Speciation. Sunderland, Massachusett: Sinauer.

Cozzolino, S., & Widmer, A. (2005). Orchid diversity: An evolutionary consequence of deception? Trends in Ecology and Evolution, 20(9), 487–494.

Cui, X., Wise, R. P., & Schnable, P. S. (1996). The rf2 nuclear restorer gene of male-sterile T-cytoplasm maize. Science, 272(5266), 1334–1336.

Dangl, J. L., & Jones, J. D. (2001). Plant pathogens and integrated defence responses to infection. Nature, 411(6839), 826–833.

Davies, B., Hatton, E., Altemose, N., Hussin, J. G., Pratto, F., Shang, G., Gupta-Hinch, A., Moralli, D., Biggs, D., Diaz, R., Preece, C., Li, R., Bitoun, E., Brick, K., Green, C. M., Camerini-Otero, R. D., Myers, S. R., Donnelly, P. (2016). Re-engineering the zinc fingers of PRDM9 reverses hybrid sterility in mi. Nature, 530, 171–176.

Dawe, R. K., Lowry, E. G., Gent, J. I., Stitzer, M. C., Swentowsky, K. W., Higgins, D. M., Ross-Ibarra, J., Wallace, J. G., Kanizay, L., Alababy, M., Qiu, W., Tseng, K-F., Wang, N., Gao, Z., Birchler, J. A., Harkess, A. E., Hodges, Amy L., Hiatt, E. N. (2018). A Kinesin-14 Motor Activates Neocentromeres to Promote Meiotic Drive in Maize. Cell, 173, 1–12.

De Oliveira, A. K., & Cordeiro, A. R. (1980). Adaptation of drosophila willistoni experimental populations to extreme ph medium I. Changes in viability and developmental rate. Heredity, 44(1), 111–122.

Derkacheva, M., Liu, S., Figueiredo, D. D., Gentry, M., Mozgova, I., Nanni, P., Tang, M., Mannervik, M., Köhler, C., Hennig, L. (2016). H2A deubiquitinases UBP12/13 are part of the Arabidopsis polycomb group protein system. Nature Plants, 2(9), 16126.

Des Marais, D. L., & Rausher, M. D. (2008). Escape from adaptive conflict after duplication in an anthocyanin pathway gene. Nature, 454(7205), 762–5.

Dettman, J. R., Sirjusingh, C., Kohn, L. M., & Anderson, J. B. (2007). Incipient speciation by divergent adaptation and antagonistic epistasis in yeast. Nature, 447(7144), 585– 588.

216

Dilkes, B. P., Spielman, M., Weizbauer, R., Watson, B., Burkart-Waco, D., Scott, R. J., & Comai, L. (2008). The Maternally Expressed WRKY Transcription Factor TTG2 Controls Lethality in Interploidy Crosses of Arabidopsis. PLoS Biology, 6(12), e308.

Dobzhansky, T. (1948). Genetics of Natural Populations. Xviii. Experiments on Chromosomes of Drosophila Pseudoobscura from Different Geographic Regions. Genetics, 33(6), 588–602.

Dobzhansky, T., & Dobzhansky, T. G. (1937). Genetics and the Origin of Species. New York, New York: Columbia University Press.

Dobzhansky, T., & Dobzhansky, T. G. (1970). Genetics of the Evolutionary Process. New York, New York: Columbia University Press.

Durand, S., Bouché, N., Perez Strand, E., Loudet, O., & Camilleri, C. (2012). Rapid establishment of genetic incompatibility through natural epigenetic variation. Current Biology, 22(4), 326–331.

Dyer, K. A., Bray, M. J., & Lopez, S. J. (2013). Genomic conflict drives patterns of X- linked population structure in Drosophila neotestacea. Molecular Ecology, 22, 157– 169.

Dyer, K. A., & Jaenike, J. (2004). Evolutionarily stable infection by a male-killing endosymbiont in Drosophila innubila: Molecular evidence from the host and parasite genomes. Genetics, 168(3), 1443–1455. http://doi.org/10.1534/genetics.104.027854

Eizaguirre, C., Lenz, T. L., Traulsen, A., & Milinski, M. (2009). Speciation accelerated and stabilized by pleiotropic major histocompatibility complex immunogenes. Ecology Letters, 12(1), 5–12.

Erilova, A., Brownfield, L., Exner, V., Rosa, M., Twell, D., Scheid, O. M., Hennig, L., Köhler, C. (2009). Imprinting of the Polycomb group gene MEDEA serves as a ploidy sensor in Arabidopsis. PLoS Genetics, 5(9).

Feder, J. L., Roethele, J. B., Filchak, K., Niedbalski, J., & Romero-Severson, J. (2003). Evidence for inversion polymorphism related to sympatric host race formation in the apple maggot fly, Rhagoletis pomonella. Genetics, 163(3), 939–53.

217

Fedorova, N. D., Badger, J. H., Robson, G. D., Wortman, J. R., & Nierman, W. C. (2005). Comparative analysis of programmed cell death pathways in filamentous fungi. BMC Genomics, 6, 177.

Ferree, P. M., & Barbash, D. a. (2009). Species-specific heterochromatin prevents mitotic chromosome segregation to cause hybrid lethality in Drosophila. PLoS Biology, 7(10), e1000234.

Ferris, K. G., Barnett, L. L., Blackman, B. K., & Willis, J. H. (2017). The Genetic Architecture of Local Adaptation and Reproductive Isolation in Sympatry within the Mimulus guttatus Species Complex. Molecular Ecology, 26, 208–224.

Ferris, K. G., Sexton, J. P., & Willis, J. H. (2014). Speciation on a local geographic scale: the evolution of a rare rock outcrop specialist in Mimulus. Philosophical Transactions of the Royal Society of London. Series B, Biological Sciences, 369(1648), 20140001.

Ferris, K. G., & Willis, J. H. (2018). Differential Adaptation to a Harsh Granite Outcrop Habitat between Sympatric. Evolution, 1–17.

Fishman, L., Kelly, A. J., & Willis, J. H. (2002). Minor Quantitative Trait Loci Underlie Floral Traits Associated with Mating System Divergence in Mimulus. Evolution, 56(11), 2138–2155.

Fishman, L., & Kelly, J. K. (2015). Centromere-associated meiotic drive and female fitness variation in Mimulus. Evolution, 69, 1208–1218.

Fishman, L., & Saunders, A. (2008). Centromere-associated female meiotic drive entails male fitness costs in monkeyflowers. Science, 322(5907), 1559–1562.

Fishman, L., Stathos, A., Beardsley, P. M., Williams, C. F., & Hill, J. P. (2013). Chromosomal rearrangements and the genetics of reproductive barriers in Mimulus (monkey flowers). Evolution, 67(9), 2547–60.

Fishman, L., & Willis, J. H. (2005). A novel meiotic drive locus almost completely distorts segregation in Mimulus (monkeyflower) hybrids. Genetics, 169(1), 347–353.

Fishman, L., & Willis, J. H. (2006). A Cytonuclear Incompatibility Causes Anther Sterility in Mimulus Hybrids. Evolution, 60(7), 1372–1381.

218

Fishman, L., Willis, J. H., Wu, C. A., & Lee, Y.-W. (2014). Comparative linkage maps suggest that fission, not polyploidy, underlies near-doubling of chromosome number within monkeyflowers (Mimulus; Phrymaceae). Heredity, 112(5), 562–8.

Flagel, L., Blackman, B. K., Fishman, L., & Monnahan, P. J. (2018). A synthesis of mapping experiments reveals extensive genomic structural diversity in the Mimulus guttatus species complex. BioRxiv.

Flagel, L. E., Willis, J. H., & Vision, T. J. (2014). The standing pool of genomic structural variation in a natural population of Mimulus guttatus. Genome Biology and Evolution, 6(1), 53–64.

Florez-Rueda, A. M., Paris, M., Schmidt, A., Widmer, A., Grossniklaus, U., & Städler, T. (2016). Genomic Imprinting in the Endosperm Is Systematically Perturbed in Abortive Hybrid Tomato Seeds. Molecular Biology and Evolution, 33(11), 2935–2946.

Fox J. & Weisberg S. (2011). An {R} Companion to Applied Regression, Second Edition. Thousand Oaks CA: Sage.

Foxe, J. P., Slotte, T., Stahl, E. a., Neuffer, B., Hurka, H., & Wright, S. I. (2009). Recent speciation associated with the evolution of selfing in Capsella. Proceedings of the National Academy of Sciences, 106(13), 5241–5245.

Frank, S. a. (1991). Divergence of meiotic drive-suppression systems as an explanation for sex- biased hybrid sterility and inviability. Evolution, 45(2), 262.

Friedman, J. (2014). Genetic Determinants and Epistasis for Life History Trait Differences in the Common Monkeyflower, Mimulus guttatus. Journal of Heredity, 1–12.

Friedman, J., Twyford, A. D., Willis, J. H., & Blackman, B. K. (2015). The extent and genetic basis of phenotypic divergence in life history traits in Mimulus guttatus. Molecular Ecology, 24(1), 111–122.

Friedman, J., & Willis, J. H. (2013). Major QTLs for critical photoperiod and vernalization underlie extensive variation in flowering in the Mimulus guttatus species complex. The New Phytologist, 199(2), 571–83.

219

Fuller, Z. L., Haynes, G., Richards, S., & Schaeffer, S. W. (2017). Genomics of Natural Populations: Evolutionary Forces that Establish and Maintain Gene Arrangements in. Molecular Ecology, (212), 1–24.

Funk, D. J., Nosil, P., & Etges, W. J. (2006). Ecological divergence exhibits consistently positive associations with reproductive isolation across disparate taxa. Proceedings of the National Academy of Sciences, 103(9), 3209–13.

Galbraith, D. W., Lambert, G. M., Macas, J., & Dolezel, J. (2001). Analysis of nuclear DNA content and ploidy in higher plants. Current Protocols in Cytometry, Chapter 7, Unit 7.6.

Gardner, M., & Macnair, M. (2000). Factors affecting the co-existence of the serpentine endemic Mimulus nudatus Curran and it presumed progenitor, Mimulus guttatus Fischer ex DC. Biological Journal of the Linnean Society, 69, 443–459.

Garner, A. G., Kenney, A. M., Fishman, L., & Sweigart, A. L. (2015). Genetic loci with parent of origin effects cause hybrid seed lethality between Mimulus species.

Gavin-Smyth, J., & Matute, D. R. (2013). Embryonic lethality leads to hybrid male inviability in hybrids between Drosophila melanogaster and D. santomea. Ecology and Evolution, 3(6), 1580–1589.

Gehring, M., Bubb, K. L., & Henikoff, S. (2009). Extensive Demethylation of Repetitive. Science, 324, 1447–1451.

Gehring, M., Huh, J. H., Hsieh, T. F., Penterman, J., Choi, Y., Harada, J. J., Goldberg, R. B., Fischer, R. L. (2006). DEMETER DNA glycosylase establishes MEDEA polycomb gene self-imprinting by allele-specific demethylation. Cell, 124(3), 495–506.

Gehring, M., Missirian, V., & Henikoff, S. (2011). Genomic analysis of parent-of-origin allelic expression in arabidopsis thaliana seeds. PLoS ONE, 6(8).

Gerstel, D. U., Burns, J. A., & Burk, L. G. (1978). Cytoplasmic male sterility in Nicotiana, restoration of fertility, and the nucleolus. Genetics, 89(5379), 157–169.

Ginzburg, L. R., Bingham, P. M., & Yoo, S. (1984). On the theory of speciation induced by transposable elements. Genetics, 107(2), 331–341.

220

Glass, N. L., & Kaneko, I. (2003). Fatal attraction: Nonself recognition and heterokaryon incompatibility in filamentous fungi. Eukaryotic Cell, 2(1), 1–8.

Glaubitz, J. C., Casstevens, T. M., Lu, F., Harriman, J., Elshire, R. J., Sun, Q., & Buckler, E. S. (2014). TASSEL-GBS: A high capacity genotyping by sequencing analysis pipeline. PLoS ONE, 9(2).

Goncalves, A., Leigh-Brown, S., Thybert, D., Stefflova, K., Turro, E., Flicek, P., Brazma, A., Odom, D. T., Marioni, J. C.(2012). Extensive compensatory cis-trans regulation in the evolution of mouse gene expression. Genome Research, 22(12), 2376–2384.

Good, J. M., Dean, M. D., & Nachman, M. W. (2008). A complex genetic basis to X-linked hybrid male sterility between two species of house mice. Genetics, 179(4), 2213–2228.

Good, J. M., Giger, T., Dean, M. D., & Nachman, M. W. (2010). Widespread Over- expression of the X chromosome in sterile F1 hybrid mice. PLoS Genetics, 6(9), 30– 32.

Gould, B. A., Chen, Y., & Lowry, D. B. (2017). Pooled Ecotype Sequencing Reveals Candidate Genetic Mechanisms for Adaptive Differentiation and Reproductive Isolation. Molecular Ecology, 26, 163–177.

Grant, A. L. (1924). A monograph of the genus Mimulus. Annals of the Missouri Botanical Garden, 11(2), 99–388.

Gregorova, S., Gergelits, V., Chvatalova, I., Bhattacharyya, T., Valiskova, B., Fotopulosova, V., Jansa, P., Wiatrowska, D., Forejt, J.(2018). Modulation of Prdm9- controlled meiotic chromosome asynapsis overrides hybrid sterility in mice. eLife, 7, e34282.

Greig, D. (2009). Reproductive isolation in Saccharomyces. Heredity, 102(1), 39–44.

Greig, D., Travisano, M., Louis, E. J., & Borts, R. H. (2003). A role for the mismatch repair system during incipient speciation in Saccharomyces. Journal of Evolutionary Biology, 16(3), 429–437.

Griffith, O. W., Blackburn, D. G., Brandley, M. C., Van Dyke, J. U., Whittington, C. M., & Thompson, M. B. (2015). Ancestral state reconstructions require biological evidence to test evolutionary hypotheses: A case study examining the evolution of 221

reproductive mode in squamate reptiles. Journal of Experimental Zoology Part B: Molecular and Developmental Evolution, 324(6), 493–503.

Guerrero, R. F., Muir, C. D., Josway, S., & Moyle, L. C. (2017). Pervasive antagonistic interactions among hybrid incompatibility loci. PLoS Genetics, 13(6), e1006817.

Guo, Y.-L., Bechsgaard, J. S., Slotte, T., Neuffer, B., Lascoux, M., Weigel, D., & Schierup, M. H. (2009). Recent speciation of Capsella rubella from Capsella grandiflora, associated with loss of self-incompatibility and an extreme bottleneck. Proceedings of the National Academy of Sciences, 106(13), 5246–5251.

Habecker, N. M. (2012). Ecogeographical and intrinsic postzygotic isolation between Mimulus glaucescens and Mimulus guttatus (Masters thesis). Retrieved from California State University Chico repository.

Haerty, W., & Singh, R. S. (2006). Gene regulation divergence is a major contributor to the evolution of Dobzhansky-Muller incompatibilities between species of Drosophila. Molecular Biology and Evolution, 23(9), 1707–1714.

Haig, D., & Mark Westoby. (1989). Parent-specific gene expression and the triploid endosperm. American Naturalist, 134(1), 147–155.

Haldane, J. B. S. (1922). Sex ratio and unisexual sterility in hybrid animals. Journal of Genetics, 12(2), 101–109.

Hall, M. C., Basten, C. J., & Willis, J. H. (2006). Pleiotropic quantitative trait loci contribute to population divergence in traits associated with life-history variation in Mimulus guttatus. Genetics, 172(3), 1829–44.

Hall, M. C., Lowry, D. B., & Willis, J. H. (2010). Is local adaptation in Mimulus guttatus caused by trade-offs at individual loci? Molecular Ecology, 19(13), 2739–53.

Hall, M. C., & Willis, J. H. (2006). Divergent selection on flowering time contributes to local adaptation in Mimulus guttatus populations. Evolution, 60(12), 2466–77.

Hanson, M. R., & Bentolila, S. (2004). Interactions of mitochondrial and nuclear genes that affect male gametophyte development. The Plant Cell Online, 16, 154–170.

222

Hartl, D. L. (1974). Genetic dissection of segregation distortion. I. Suicide combinations of SD genes. Genetics, 76(3), 477–486.

Herrmann, B. G., Koschorz, B., Wertz, K., McLaughlin, K. J., & Kispert, A. (1999). A protein kinase encoded by the t complex responder gene causes non- mendelian inheritance. Nature, 402(6758), 141–146.

Higgins, D. M., Lowry, E. G., Kanizay, L. B., Becraft, P. W., Hall, D. W., & Dawe, R. K. (2018). Fitness Costs and Variation in Transmission Distortion Associated with Abnormal Chromosome 10 Meiotic Drive System in Maize. Genetics, 208(January), 297–305.

Hoffmann, A. A., & Rieseberg, L. H. (2008). Revisiting the impact of inversions in evolution: from population genetic markers to drivers of adaptive shifts and speciation? Annual Review of Ecology, Evolution, and Systematics, 39, 21–42.

Hooper, D. M., & Price, T. D. (2017). Chromosomal inversion differences correlate with range overlap in passerine birds. Nature Ecology & Evolution, 1, 1526–1534.

Hsieh, T. (2009). Genome-Wide Demethylation of. Science, 1451(June), 1451–1455.

Hunter, N., Chambers, S. R., Louis, E. J., & Borts, R. H. (1996). The mismatch repair system contributes to meiotic sterility in an interspecific yeast hybrid. The EMBO Journal, 15 (7), 1726–1733.

Hurst, L. D., & Pomiankowski, a. (1991). Causes of sex ratio bias may account for unisexual sterility in hybrids: A new explanation of Haldane’s rule and related phenomena. Genetics, 128(4), 841–858.

Jablonka, E., & Lamb, M. J. (1991). Sex Chromosomes and Speciation. Proceedings of the Royal Society B: Biological Sciences, 243(1308), 203–208.

Jaenike, J. (2001). Sex Chromosome Meiotic Drive. Annual Review of Ecology and Systematics, 32, 25–49.

Jeuken, M J., Zhang, N W., McHale, L K., Pelgrom, K., den Boer, E., Lindhout, P., Michelmore, R W., Visser, R G., Niks, R E. (2009). Rin4 causes hybrid necrosis and race-specific resistance in an interspecific lettuce hybrid. Plant Cell, 21(10), 3368– 3378. 223

Johnson, N. A. (2010). Hybrid incompatibility genes: Remnants of a genomic battlefield? Trends in Genetics, 26(7), 317–325.

Johnston, S. a., den Nijs, T. P. M., Peloquin, S. J., & Hanneman, R. E. (1980). The significance of genic balance to endosperm development in interspecific crosses. Theoretical and Applied Genetics, 57(1), 5–9.

Jones, F. C., Grabherr, M. G., Chan, Y. F., Russell, P., Mauceli, E., Johnson, J., Swofford, R., Pirun, M., Zody, M. C., White, S., Birney, E., Searle, S., Schmutz, J., Grimwood, J., Dickson, M. C., Myers, R. M., Miller, C. T., Summers, B. R., Knecht, A. K., Brady, S. D., Zhang, H., Pollen, A. A., Howes, T., Amemiya, C., Baldwin, J., Bloom, T., Jaffe, D. B., Nicol, R., Wilkinson, J., Lander, E. S., Di Palma, F., Lindblad-Toh, K., Kingsley, D. M. (2012). The genomic basis of adaptive evolution in threespine sticklebacks. Nature, 484(7392), 55–61.

Josefsson, C., Dilkes, B., & Comai, L. (2006). Parent-Dependent Loss of Gene Silencing during Interspecies Hybridization. Current Biology, 16, 1322–1328.

Kaneko, I., Dementhon, K., Xiang, Q., & Glass, N. L. (2006). Nonallelic interactions between het-c and a polymorphic locus, pin-c, are essential for nonself recognition and programmed cell death in Neurospora crassa. Genetics, 172(3), 1545–1555.

Karvonen, A., & Seehausen, O. (2012). The role of parasitism in adaptive radiations- when might parasites promote and when might they constrain ecological speciation? International Journal of Ecology, 2012, 10.1155/2012/280169.

Kassahn, K. S., Dang, V. T., Wilkins, S. J., Perkins, A. C., & Ragan, M. a. (2009). Evolution of gene function and regulatory control after whole-genome duplication: comparative analyses in vertebrates. Genome Research, 19, 1404–1418.

Kazama, T., & Toriyama, K. (2003). A pentatricopeptide repeat-containing gene that promotes the processing of aberrant atp6 RNA of cytoplasmic male-sterile rice. FEBS Letters, 544(1-3), 99–102.

Keeling, M. J., Jiggins, F. M., & Read, J. M. (2003). The invasion and coexistence of competing Wolbachia strains. Heredity, 91(4), 382–388.

Kelly, A. J., & Willis, J. H. (1998). Polymorphic microsatellite loci in Mimulus guttatus and related species. Molecular Ecology, 7(6), 769–774. 224

Kelly, J. K., Rasch, A., & Kalisz, S. (2002). A Method to Estimate Pollen Viability from Pollen Size Variation. American Journal of Botany, 89(6), 1021–1023.

Kenney, A. M., Sweigart, A. L., & Kenney, A. M. (2016). Reproductive isolation and introgression between sympatric Mimulus species. Molecular Ecology, 25, 2499–2517.

Kerwin, R. E., & Sweigart, A. L. (2017). Mechanisms of Transmission Ratio Distortion at Hybrid Sterility Loci Within and Between Mimulus Species. G3, 7, g3.300148.2017.

Khatri, B. S., & Goldstein, R. a. (2015). Simple biophysical model predicts faster accumulation of hybrid incompatibilities in small populations under stabilizing selection. Genetics, 201(4), 1525–1537.

Kidwell, M. G. (1979). Hybrid dysgenesis in Drosophila melanogaster: The relationship between the P–M and I–R interaction systems. Genetical Research, 33(3), 205–217.

Kidwell, M. G. (1983). Evolution of hybrid dysgenesis determinants in Drosophila melanogaster. Proceedings of the National Academy of Sciences of the United States of America, 80(6), 1655–9.

Kirkpatrick, M., & Barton, N. H. (2006). Chromosome inversions, local adaptation and speciation. Genetics, 173(1), 419–34.

Kitano, J., & Peichel, C. L. (2012). Turnover of sex chromosomes and speciation in fishes. Environmental Biology of Fishes, 94(3), 549–558.

Kitano, J., Ross, J. A., Mori, S., Kume, M. Jones, F. C., Chan, Y. F., Absher, D. M., Grimwood, J., Schmutz, J., Myers, R. M., Kingsley, D. M., Peichel, C. L. (2009). A role for a neo-sex chromosome in stickleback speciation. Nature, 461(7267), 1079– 1083.

Klosinska, M., Picard, C. L., & Gehring, M. (2016). Conserved imprinting associated with unique epigenetic signatures in the Arabidopsis genus. Nature Plants, 2, 16145.

Ko, C., Köhler, C., & Weinhofer-Molisch, I. (2010). Mechanisms and evolution of genomic imprinting in plants. Heredity, 105(1), 57–63.

225

Koch, M. a, & Matschinger, M. (2007). Evolution and genetic differentiation among relatives of Arabidopsis thaliana. Proceedings of the National Academy of Sciences of the United States of America, 104(15), 6272–6277.

Köhler, C., Hennig, L., Spillane, C., Pien, S., Gruissem, W., & Grossniklaus, U. (2003). The Polycomb-group protein MEDEA regulates seed development by controlling expression of the MADS-box gene PHERES1. Genes and Development, 17(12), 1540– 1553.

Köhler, C., & Makarevich, G. (2006). Epigenetic mechanisms governing seed development in plants. EMBO Rep, 7(12), 1223–1227.

Köhler, C., Mittelsten Scheid, O., & Erilova, A. (2010). The impact of the triploid block on the origin and evolution of polyploid plants. Trends in Genetics, 26(3), 142–8.

Köhler, C., Page, D. R., Gagliardini, V., & Grossniklaus, U. (2005). The Arabidopsis thaliana MEDEA Polycomb group protein controls expression of PHERES1 by parental imprinting. Nature Genetics, 37(1), 28–30.

Köhler, C., Wolff, P., & Spillane, C. (2012). Epigenetic Mechanisms Underlying Genomic Imprinting in Plants. Annual Review of Plant Biology, 63(1), 331–352.

Kradolfer, D., Wolff, P., Jiang, H., Siretskiy, A., & Köhler, C. (2013). An imprinted gene underlies postzygotic reproductive isolation in arabidopsis thaliana. Developmental Cell, 26(5), 525–535.

Kreuger, F. Trim Galore! 0.3. 7 In.

Kruger, J. (2002). A Tomato Cysteine Protease Required for Cf-2-Dependent Disease Resistance and Suppression of Autonecrosis. Science, 296(5568), 744–747.

Kuo, D., Licon, K., Bandyopadhyay, S., Chuang, R., Luo, C., Catalana, J., Ravasi, T., Tan, K., Ideker, T. (2010). Coevolution within a transcriptional network by compensatory trans and cis mutations. Genome Research, 20(12), 1672–1678.

Küpper, C., Stocks, M., Risse, J. E., Remedios, N., Farrell, L. L., Mcrae, B., Morgan, T. C., Karlionova, N., Pinchuk, P., Verkuil, Y. I., Kitaysky, A. S., Wingfield, J. C., Piersma, T., Zeng, K., Slate, J., Blaxter, M., Lank, D. B, Burke, T. (2015). A supergene

226

determines highly divergent male reproductive morphs in the ruff. Nature Genetics, 48, 79–83.

Lafon-Placette, C., Johannessen, I. M., Hornslien, K. S., Ali, M. F., Bjerkan, K. N., Bramsiepe, J., Glöckle, B. M., Rebernig, C. A., Brysting, A. K., Grini, P. E., Köhler, C. (2017). Endosperm-based hybridization barriers explain the pattern of gene flow between Arabidopsis lyrata and Arabidopsis arenosa in Central Europe. Proceedings of the National Academy of Sciences, 1027-1035.

Lafon-Placette, C., & Köhler, C. (2016). Endosperm-based postzygotic hybridization barriers: Developmental mechanisms and evolutionary drivers. Molecular Ecology, 2620–2629.

Lafon-Placette, C., Hatorangan, M. R., Steige, K. A., Cornille, A., Lascoux, M., Slotte, T., & Köhler, C. (2018). Paternally expressed imprinted genes associate with hybridization barriers in Capsella. Nature Plants.

Lamichhaney, S., Fan, G., Widemo, F., Gunnarsson, U., Thalmann, D. S., Hoeppner, M. P., Kerje, S., Gustafson, U., Shi, C., Zhang, H., Chen, W., Liang, X., Huang, L., Wang, J., Liang, E., Wu, Q., Lee, S. M.-Y., Xu, X., Höglund, J., Liu, X., Andersson, L. (2015). Structural genomic changes underlie alternative reproductive strategies in the ruff (Philomachus pugnax). Nature Genetics, 48, 84–88.

Landry, C. R., Wittkopp, P. J., Taubes, C. H., Ranz, J. M., Clark, A. G., & Hartl, D. L. (2005). Compensatory cis-trans evolution and the dysregulation of gene expression in interspecific hybrids of drosophila. Genetics, 171(4), 1813–1822.

Larson, E. L., Keeble, S., Vanderpool, D., Dean, M. D., & Good, J. M. (2016). The composite regulatory basis of the large X-effect in mouse speciation. Molecular Biology and Evolution, (590), msw243.

Lavington, E., & Kern, A. D. (2017). The effect of common inversion polymorphisms In(2L)t and In(3R)Mo on patterns of transcriptional variation in Drosophila melanogaster. G3: Genes, Genomes, Genetics, 7(November), 1–35.

Lee, C. R., Wang, B., Mojica, J. P., Mandáková, T., Prasad, K. V. S. K., Goicoechea, J. L., Perera, N., Hellsten, U., Hundley, H. N., Johnson, J., Grimwood, J., Barry, K., Fairclough, S., Jenkins, J. W., Yu, Y., Kudrna, D., Zhang, J., Talag, J., Golser, W., Ghattas, K., Schranz, M. E., Wing, R., Lysak, M. A., Schmutz, J., Rokhsar, D. S., 227

Mitchell-Olds, T. (2017). Young inversion with multiple linked QTLs under selection in a hybrid zone. Nature Ecology & Evolution, 1, 0119.

Lee, K. M., & Coop, G. M. (2017). Distinguishing Among Modes of Convergent Adaptation Using Population Genomic Data. Genetics, 207, 1591–1619.

Lee, Y. W., Fishman, L., Kelly, J. K., & Willis, J. H. (2016). Fitness variation is generated by a segregating inversion in yellow monkeyflower (Mimulus guttatus ). Genetics, 1–38.

Lemos, B., Meiklejohn, C. D., Cáceres, M., Hartl, D. L., & Url, S. (2013). Rates of Divergence in Gene Expression Profiles of Primates, Mice, and Flies: Stabilizing Selection and Variability among Functional Categories. Evolution, 59(1), 126–137.

Li, H., & Durbin, R. (2009). Fast and accurate short read alignment with Burrows- Wheeler transform. Bioinformatics, 25(14), 1754–1760.

Li, H., Handsaker, B., Wysoker, A., Fennell, T., Ruan, J., Homer, N., Marth, G., Abecasis, G., Durbin, R. (2009). The Sequence Alignment/Map format and SAMtools. Bioinformatics, 25(16), 2078–2079.

Lin, B. Y. (1984). Ploidy barrier to endosperm development in maize. Genetics, 107(1), 103–15.

Linke, B., Nothnagel, T., & Börner, T. (2003). Flower development in carrot CMS plants: Mitochondria affect the expression of MADS box genes homologous to GLOBOSA and DEFICIENS. Plant Journal, 34(1), 27–37.

Liti, G., Barton, D. B. H., & Louis, E. J. (2006). Sequence diversity, reproductive isolation and species concepts in saccharomyces. Genetics, 174(2), 839–850.

Lowry, D. B., Hall, M. C., Salt, D. E., & Willis, J. H. (2009). Genetic and physiological basis of adaptive salt tolerance divergence between coastal and inland Mimulus guttatus. The New Phytologist, 183(3), 776–88.

Lowry, D. B., Modliszewski, J. L., Wright, K. M., Wu, C. A., & Willis, J. H. (2008). The strength and genetic basis of reproductive isolating barriers in flowering plants. Philosophical Transactions of the Royal Society B, 363(1506), 3009–21.

228

Lowry, D. B., Rockwood, R. C., & Willis, J. H. (2008). Ecological reproductive isolation of coast and inland races of Mimulus guttatus. Evolution, 62(9), 2196–214.

Lowry, D. B., & Willis, J. H. (2010). A Widespread Chromosomal Inversion Polymorphism Contributes to a Major Life-History Transition, Local Adaptation, and Reproductive Isolation. PLoS Biology, 8(9), 1–14.

Lu, G., & Bernatchez, L. (1998). Experimental evidence for reduced hybrid viability between dwarf and normal ecotypes of lake whitefish (Coregonus clupeaformis Mitchill). Proceedings of the Royal Society B: Biological Sciences, 265(1400), 1025–1030.

Ludwig, M. Z., Bergman, C., Patel, N. H., & Kreitman, M. (2000). Evidence for stablizing selection in a eukaryotic enhancer element. Nature, 403, 564–567.

Lynch, M., & Conery, J. S. (2001). The Evolutionary fate and Consequences of Duplicate Genes. Science, 290(5494), 1151–1155.

Lynch, M., & Force, A. G. (2000). The Origin of Interspecific Genomic Incompatibility via Gene Duplication. The American Naturalist, 156(6), 590–605.

Mack, K. L., & Nachman, M. W. (2016). Gene Regulation and Speciation. Trends in Genetics, 33(1), 1–13.

Mack, K. L., & Nachman, M. W. (2017). Gene Regulation and Speciation in house mouse. Trends in Genetics, 33(1), 68–80.

Macnair, M. R., & Christie, P. (1983). Reproductive isolation as a pleiotropic effect of copper tolerance in Mimulus guttatus? Heredity, 50(3), 295–302.

Maheshwari, S., & Barbash, D. A. (2011). The Genetics of Hybrid Incompatibilities. Annual Review of Genetics, 45(1), 331–355.

Makarevich, G., Leroy, O., Akinci, U., Schubert, D., Clarenz, O., Goodrich, J., Grossniklaus, U., Köhler, C. (2006). Different Polycomb group complexes regulate common target genes in Arabidopsis. EMBO Reports, 7(9), 947–952.

Malmstrøm, M., Matschiner, M., Tørresen, O. K., Star, B., Snipen, L. G., Hansen, T. F., Baalsrud, H. T., Nederbragt, A. J., Hanel, R., Salzburger, W., Stenseth, N. C.,

229

Jakobsen, K. S., Jentoft, S. (2016). Evolution of the immune system influences speciation rates in teleost fishes. Nature Genetics, 48(10), 1204–1210.

Manoukis, N. C., Powell, J. R., Touré, M. B., Sacko, A., Edillo, F. E., Coulibaly, M. B., Traoré, S. F., Taylor, C. E., Besansky, N. J.(2008). A test of the chromosomal theory of ecotypic speciation in Anopheles gambiae. Proceedings of the National Academy of Sciences 105(8), 2940–5.

Martienssen, R. A. (2010). Heterochromatin, small RNA and post-fertilization dysgenesis in allopolyploid and interploid hybrids of Arabidopsis. New Phytologist, 186(1), 46– 53.

Martin, N. H., Sapir, Y., & Arnold, M. L. (2008). The genetic architecture of reproductive isolation in Louisiana Irises: Pollination syndromes and pollinator preferences. Evolution, 62(4), 740–752.

Martin, N. H., & Willis, J. H. (2007). Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species. Evolution, 61(1), 68–82.

Martin, N. H., & Willis, J. H. (2010). Geographical variation in postzygotic isolation and its genetic basis within and between two Mimulus species. Philosophical Transactions of the Royal Society B: Biological Sciences, 365(1555), 3261–3261.

Masly, J. P., Jones, C. D., Noor, M. a F., Locke, J., & Orr, H. A. (2006). Gene transposition as a cause of hybrid sterility in Drosophila. Science, 313(5792), 1448–50.

Masly, J. P., & Presgraves, D. C. (2007). High-resolution genome-wide dissection of the two rules of speciation in Drosophila. PLoS Biology, 5(9), 1890–1898.

Matute, D. R., Butler, I. A., Turissini, D. A., & Coyne, J. A. (2010). A test of the snowball theory or the rate of evolution of hybrid incompatibilities. Science, 329, 1518–1521.

Mauricio, R., Stahl, E. a., Korves, T., Tian, D., Kreitman, M., & Bergelson, J. (2003). Natural selection for polymorphism in the disease resistance gene Rps2 of Arabidopsis thaliana. Genetics, 163(2), 735–746.

McKenna, A., Hanna, M., Banks, E., Sivachenko, A., Cibulskis, K., Kerntsky, A., Garimella, K., Altshuler, D., Gabriel, S., Daly, M., DePristo, M. A.(2010). The 230

Genome Analysis Toolkit: A MapReduce framework for analysing next-generation DNA sequencing data. Genome Research, 20, 1297–1303.

Mckinnon, J. S., Mori, S., Blackman, B. K., David, L., Kingsley, D. M., Jamieson, L., Chou, J., Schluter, D. (2004). Evidence for ecology’s role in speciation. Nature, 429, 294–298.

Meiklejohn, C. D., Coolon, J. D., Hartl, D. L., & Wittkopp, P. J. (2014). The roles of cis- and trans-regulation in the evolution of regulatory incompatibilities and sexually dimorphic gene expression. Genome Research, 24, 84–95.

Meiklejohn, C. D., & Tao, Y. (2010). Genetic conflict and sex chromosome evolution. Trends in Ecology and Evolution, 25(4), 215–223.

Meyer, A., Morrissey, J. M., & Shartl, M. (1994). Recurrent origin of a sexually selected trait in Xiphophorus fishes inferred from a molecular phylogeny. Nature, 368, 539– 542.

Michalak, P. (2009). Epigenetic, transposon and small RNA determinants of hybrid dysfunctions. Heredity, 102(1), 45–50.

Michalak, P., & Noor, M. a F. (2003). Genome-wide patterns of expression in Drosophila pure species and hybrid males. Molecular Biology and Evolution, 20(7), 1070–1076.

Mihola, O., Trachtulec, Z., Vlcek, C., Schimenti, J. C., & Forejt, J. (2009). A Mouse Speciation Gene Encodes a Meiotic Histone H3 Methyltransferase. Science, 323(5912), 373–375.

Mizuta, Y., Harushima, Y., & Kurata, N. (2010). Rice pollen hybrid incompatibility caused by reciprocal gene loss of duplicated genes. Proceedings of the National Academy of Sciences, 107(47), 20417–20422.

Mojica, J. P., & Kelly, J. K. (2010). Viability selection prior to trait expression is an essential component of natural selection. Proceedings of the Royal Society B: Biological Sciences, 277(1696), 2945–50.

Mojica, J. P., Lee, Y. W., Willis, J. H., & Kelly, J. K. (2012). Spatially and temporally varying selection on intra-population QTL for a life history tradeoff in Mimulus guttatus. Molecular Ecology, 21(15), 3718–3728.

231

Mondragón-palomino, M., Meyers, B. C., Michelmore, R. W., & Gaut, B. S. (2002). Patterns of Positive Selection in the Complete NBS-LRR Gene Family of Arabidopsis thaliana Patterns of Positive Selection in the Complete NBS-LRR Gene Family of Arabidopsis thaliana, (Ml), 1305–1315.

Mora, C., Tittensor, D. P., Adl, S., Simpson, A. G. B., & Worm, B. (2011). How many species are there on earth and in the ocean? PLoS Biology, 9(8), 1–8.

Moyle, L. C. (2010). Hybrid incompatibility snowballs between solanum species. Nature, 329, 1521–1523.

Moyle, L. C., Levine, M., Stanton, M. L., & Wright, J. W. (2012). Hybrid Sterility over Tens of Meters Between Ecotypes Adapted to Serpentine and Non-Serpentine Soils. Evolutionary Biology, 39(2), 207–218.

Moyle, L. C., Muir, C. D., Han, M. V., & Hahn, M. W. (2010). The contribution of gene movement to the “two rules of speciation.” Evolution, 64(6), 1541–1557.

Moyle, L. C., & Nakazato, T. (2010). Hybrid incompatibility “snowballs” between Solanum species. Science (New York, N.Y.), 329(5998), 1521–3.

Muir, C. D., & Hahn, M. W. (2015). The limited contribution of reciprocal gene loss to increased speciation rates following whole-genome duplication. The American Naturalist, 185(1), 70–86.

Muller, H. J., 1942 Isolating mechanisms, evolution, and tempera- ture. Biol. Symp. 6: 71- 125.

Myers, S., Bowden, R., Tumian, A., Bontrop, R. E., Freeman, C., Macfie, T. S., Mcvean, G., Donnelly, P. (2010). Gene in Meiotic Recombination. Science, 327, 876–879.

Nesom, G. L. (2012). Taxonomy of erythranthe sect. simiolus (phrymaceae) in the usa and mexico. Phytoneuron, (1978).

Nosil, P. (2007). Divergent Host Plant Adaptation and Reproductive Isolation between Ecotypes of Timema cristinae Walking Sticks. The American Naturalist, 169(2), 151– 162.

232

Nosil, P., & Crespi, B. J. (2012). Conflictual speciation: species formation via intragenomic conflict. Trends in Ecology & Evolution, 28(1), 48–57.

Nosil, P., & Flaxman, S. M. (2011). Conditions for mutation-order speciation. Proceedings of the Royal Society B: Biological Sciences, 278(1704), 399–407.

Nosil, P., Vines, T. H., & Funk, D. J. (2005). Reproductive isolation casued by natural selection against immigrants from divergent habitats. Evolution, 59, 705–719.

Oka, H. (1974). Analysis of genes controlling F1 sterility in rice by the use of isogenic lines. Genetics, 77, 521–534.

Oneal, E., Lowry, D. B., Wright, K. M., Zhu, Z., & Willis, J. H. (2014). Divergent population structure and climate associations of a chromosomal inversion polymorphism across the Mimulus guttatus species complex. Molecular Ecology, 23(11), 2844–60.

Oneal, E., Willis, J. H., & Franks, R. G. (2016). Disruption of endosperm development is a major cause of hybrid seed inviability between Mimulus guttatus and Mimulus nudatus. New Phytologist, 210, 1107–1120.

Ono, J., Gerstein, A. C., & Otto, S. P. (2017). Widespread Genetic Incompatibilities between First-Step Mutations during Parallel Adaptation of Saccharomyces cerevisiae to a Common Environment. PLoS Biology, 15(1), 1–26.

Orr, H. A. (1995). Population Genetics of Speciation: Genetics, 139, 1805–1813.

Orr, H. A., & Irving, S. (2005). Segregation distortion in hybrids between the Bogota and USA subspecies of Drosophila pseudoobscura. Genetics, 169(2), 671–682.

Orr, H. A., Madden, L. D., Coyne, J. A., Goodwin, R., & Hawley, R. S. (1997). The developmental genetics of hybrid inviability: A mitotic defect in Drosophila hybrids. Genetics, 145(4), 1031–1040.

Orr, H. A., & Presgraves, D. C. (2000). Speciation by postzygotic isolation: Forces, genes and molecules. BioEssays, 22(12), 1085–1094.

Orr, H. A., & Turelli, M. (2001). The Evolution of Postzygotic Isolation: Accumulating Dobzhansky-Muller Incompatibilities. Evolution, 55(6), 1085–1094. 233

Paoletti, M., & Saupe, S. J. (2009). Fungal incompatibility: Evolutionary origin in pathogen defense? BioEssays, 31(11), 1201–1210.

Paoletti, M., Saupe, S. J., & Calve, C. (2007). Genesis of a funal non-self recognition repertoire. PLoS ONE, 2(3), 1435–1439.

Parvanov, E. D., Petkov, P. M., & Paigen, K. (2010). Prdm9 controls activation of mammalian recombination hotspots. Science, 327(5967), 835.

Pennington, P. D., Costa, L. M., Gutierrez-Marcos, J. F., Greenland, A. J., & Dickinson, H. G. (2008). When genomes collide: Aberrant seed development following maize interploidy crosses. Annals of Botany, 101(6), 833–843.

Peterson, M. L., Kay, K. M., & Angert, A. L. (2016). The scale of local adaptation in Mimulus guttatus: comparing life history races, ecotypes, and populations. New Phytologist, 211, 345–356.

Peterson, M. L., Rice, K. J., & Sexton, J. P. (2013). Niche partitioning between close relatives suggests trade-offs between adaptation to local environments and competition. Ecology and Evolution, 3(3), 512–22.

Phadnis, N., Baker, E. P., Cooper, J. C., Frizzell, K. A., Hsieh, E., de la Cruz, A. F. A., Shendure, J., Kitzman, J. O., Malik, H. S. (2015). An essential cell cycle regulation gene causes hybrid inviability in Drosophila. Science, 350(6267), 1552–1555.

Phadnis, N., & Orr, H. A. (2009). A single gene causes both male sterility and segregation distortion in Drosophila hybrids. Science, 323(January), 376–379.

Pignatta, D., Erdmann, R. M., Scheer, E., Picard, C. L., Bell, G. W., & Gehring, M. (2014). Natural epigenetic polymorphisms lead to intraspecific variation in Arabidopsis gene imprinting. eLife, 3, 1–24.

Pollux, B. J. a., Pires, M. N., Banet, a. I., & Reznick, D. N. (2009). Evolution of Placentas in the Fish Family Poeciliidae: An Empirical Study of Macroevolution. Annual Review of Ecology, Evolution, and Systematics, 40(1), 271–289.

Ponting, C. P. (2011). What are the genomic drivers of the rapid evolution of PRDM9? Trends in Genetics, 27(5), 165–171.

234

Presgraves, D. C. (2008). Sex chromosomes and speciation in Drosophila. Trends in Genetics, 24(7), 336–343.

Presgraves, D. C. (2010). The molecular evolutionary basis of species formation. Nature Reviews. Genetics, 11(3), 175–80.

Price, T. (2008). Speciation in Birds. Greenwood Village, Colorado: Roberts and Company.

Puzey, J. R., Willis, J. H., & Kelly, J. K. (2017). Population structure and local selection yield high genomic variation in Mimulus guttatus. Molecular Ecology, 26, 519–535.

Ramsey, J., Bradshaw, H. D., & Schemske, D. W. (2003). Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M. cardinalis (Phrymaceae). Evolution, 57(7), 1520–34.

Rebernig, C. a., Lafon-Placette, C., Hatorangan, M. R., Slotte, T., & Köhler, C. (2015). Non-reciprocal Interspecies Hybridization Barriers in the Capsella Genus Are Established in the Endosperm. PLOS Genetics, 11(6), e1005295.

Rice, W. R. (1987). The accumulation of sexually antagonistic genes as a selective agent promoting the evolution of reduced recombination between primitive sex chromosomes. Evolution, 41(4), 911–914.

Rieseberg, L. H. (2001). Chromosomal rearrangements and speciation. Trends in Ecology & Evolution, 16(7), 351–358.

Rieseberg, L. H., & Blackman, B. K. (2010). Speciation genes in plants. Annals of Botany, 106(3), 439–55.

Rieseberg, L. H., Raymond, O., Rosenthal, D. M., Lai, Z., Livingstone, K., Nakazato, T., Durphy, J. L. Schwarzbach, A. E, Donovan, L. A, Lexer, C. (2003). Major ecological transtitions in wild sunflower facilitated by hybridization. Science, 301, 1211–1216.

Rolán-Alvarez, E., Carballo, M., Galindo, J., Morán, P., Fernández, B., Caballero, A., Cruz, R., Boulding, E. G., Johannesson, K.(2004). Nonallopatric and parallel origin of local reproductive barriers between two snail ecotypes. Molecular Ecology, 13(11), 3415–3424.

235

Roth, M., Florez-Rueda, A. M., Griesser, S., Paris, M., & Städler, T. (2017). Incidence and developmental timing of endosperm failure in postzygotic isolation between wild tomato lineages. Annals of Botany, (January), 1–12.

Rubin, G. M., Kidwell, M. G., & Bingham, P. M. (1982). The molecular basis of P-M hybrid dysgenesis: The nature of induced mutations. Cell, 29(3), 987–994.

Rundle, H. D., Nagel, L., Boughman, J. W., & Schluter, D. (2000). Natural Selection and Parellel Speciation in Sympatric Sticklebacks. Science, 287(5451), 306–308.

Sabar, M., Gagliardi, D., Balk, J., & Leaver, C. J. (2003). ORFB is a subunit of F1F0-ATP synthase: Insight into the basis of cytoplasmic male sterility in sunflower. EMBO Reports, 4(4), 381–386.

Said, I., Byrne, A., Serrano, V., Cardeno, C., & Vollmers, C. (2018). Linked genetic variation and not genome structure causes widespread differential expression associated with chromosomal inversions. Proceedings of the National Academy of Science, 1–6.

Sambatti, J. B. M., Ortiz-Barrientos, D., Baack, E. J., & Rieseberg, L. H. (2008). Ecological selection maintains cytonuclear incompatibilities in hybridizing sunflowers. Ecology Letters, 11(10), 1082–1091.

Samuk, K., Owens, G. L., Delmore, K. E., Miller, S., Rennison, D. J., & Schluter, D. (2017). Gene flow and selection interact to promote adaptive divergence in regions of low recombination. Molecular Ecology, 26, 4378–4390.

Satyaki, P. R. V., Cuykendall, T. N., Wei, K. H-C., Brideau, N. J., Kwak, H., Aruna, S., Ferree, P. M., Ji, S, Barbash, D. A. (2014). The Hmr and Lhr Hybrid Incompatibility Genes Suppress a Broad Range of Heterochromatic Repeats. PLoS Genetics, 10(3), e1004240.

Scannell, D. R., Byrne, K. P., Gordon, J. L., Wong, S., & Wolfe, K. H. (2006). Multiple rounds of speciation associated with reciprocal gene loss in polyploid yeasts. Nature, 440(7082), 341–345.

Schartl, M. (2008). Evolution of Xmrk: An oncogene, but also a speciation gene? BioEssays, 30(9), 822–832.

236

Schiffman, J. S., & Ralph, P. L. (2017). System drift and speciation. bioRxiv, 1–24.

Schluter, D. (2009). Evidence for ecological speciation and its alternate. Science, 323.

Schmuths, H., Meister, A., Horres, R., & Bachmann, K. (2004). Genome size variation among accessions of Arabidopsis thaliana. Annals of Botany, 93(3), 317–321.

Schneider, C. a., Rasband, W. S., & Eliceiri, K. W. (2012). NIH Image to ImageJ: 25 years of image analysis. Nature Methods, 9(7), 671–675.

Schrader, M., Fuller, R., & Travis, J. (2013). Differences in offspring size predict the direction of isolation asymmetry between populations of a placental fish. Biology Letters, (July), 9–13.

Schrader, M., & Travis, J. (2008). Testing the viviparity-driven-conflict hypothesis: parent-offspring conflict and the evolution of reproductive isolation in a poeciliid fish. The American Naturalist, 172(6), 806–817.

Schrader, M., Travis, J., & Fuller, R. C. (2011). Do density-driven mating system differences explain reproductive incompatibilities between populations of a placental fish? Molecular Ecology, 20(19), 4140–4151.

Scopece, G., Musacchio, A., Widmer, A., & Cozzolino, S. (2007). Patterns of reproductive isolation in Mediterranean deceptive orchids. Evolution, 61(11), 2623–2642.

Scopece, G., Widmer, A., & Cozzolino, S. (2008). Evolution of postzygotic reproductive isolation in a guild of deceptive orchids. The American Naturalist, 171(3), 315–326.

Scott, R. J., Spielman, M., Bailey, J., & Dickinson, H. G. (1998). Parent-of-origin effects on seed development in Arabidopsis thaliana. Development, 125, 3329–3341.

Seehausen, O., & Van Alphen, J. J. M. (1999). Can sympatric speciation by disruptive sexual selection explain rapid evolution of cichlid diversity in Lake Victoria? Ecology Letters, 2(4), 262–271.

Serrato-Capuchina, A., & Matute, D. (2018). The role of transposable elements in speciation. Genes, 9(5), 254.

237

Shi, W., Krella, A., Orth, A., Yu, Y., & Fundele, R. (2005). Widespread disruption of genomic imprinting in adult interspecies mouse (Mus) hybrids. Genesis, 43(3), 100– 108.

Sicard, A., Kappel, C., Josephs, E. B., Lee, Y. W., Marona, C., Stinchcombe, J. R, Wright, S. I, Lenhard, M. (2015). Divergent sorting of a balanced ancestral polymorphism underlies the establishment of gene-flow barriers in Capsella. Nature Communications, 6, 7960.

Smith, D.A. S., Gordon, I. J., Traut, W., Herren, J., Collins, S., Martins, D. J., Saitoti, K., Ireri, P., Ffrench-Constant, R. (2016). A neo-W chromosome in a tropical butterfly links colour pattern, male-killing, and speciation. Proceedings of the Royal Society B: Biological Sciences, 283(1835), 20160821.

Soria-Carrasco, V., Gompert, Z., Comeault, A.A., Farkas, T. E., Parchman, T. L., Johnson, S. J., Buerkle, C. A., Feder, J. L., Bast, J., Schwander, T., Egan, S. P., Crespi, B. J., Nosil, P. (2014). Stick insect genomes reveal natural selection’s role in parallel speciation. Science, 344, 1497–1502.

Stahl, E. A., Dwyer, G., Mauricio, R., Kreitman, M., & Bergelson, J. (1999). Dynamics of disease resistance polymorpism at the Rpm 1 locus of Arabidopis. Nature, 400(August), 667–671.

Stankowski, S. (2013). Ecological speciation in an island snail: evidence for the parallel evolution of a novel ecotype and maintenance by ecologically dependent postzygotic isolation. Animal Biology, 2726–2741.

Stankowski, S., Sobel, J. M., & Streisfeld, M. a. (2017). Geographic cline analysis as a tool for studying genome-wide variation: a case study of pollinator-mediated divergence in a monkeyflower. Molecular Ecology, 26(1), 107–122.

Stankowski, S., Streisfeld, M. A., & Streisfeld, M. A. (2015). Introgressive hybridization facilitates adaptive divergence in a recent radiation of monkeyflowers. Proceedings of the Royal Society B: Biological Sciences, 282, 20151666.

Stathos, A., & Fishman, L. (2014). Chromosomal rearrangements directly cause underdominant F1 pollen sterility in Mimulus lewisii-Mimulus cardinalis hybrids. Evolution, 68(11), 3109–3119.

238

Steiner, C. C., & Ryder, O. a. (2013). Characterization of Prdm9 in Equids and Sterility in Mules. PLoS ONE, 8(4), 1–7.

Stuart, Y. E., Veen, T., Weber, J. N., Hanson, D., Ravinet, M., Lohman, B. K., Thompson, C. J., Tasneem, T., Doggett, A., Izen, R., Ahmed, N., Barrett, R. D.H., Hendry, A. P., Peichel, C. L., Bolnick, D. I. (2017). Contrasting effects of environment and genetics generate a continuum of parallel evolution. Nature Ecology and Evolution, 1(6), 1–7.

Sturtevant, A. H. (1921). A Case of Rearrangement of Genes in Drosophila. Proceedings of the National Academy of Sciences, 7(8), 235–237.

Sweigart, A. L., Fishman, L., & Willis, J. H. (2006). A simple genetic incompatibility causes hybrid male sterility in mimulus. Genetics, 172(4), 2465–2479.

Sweigart, A. L., & Flagel, L. E. (2015). Evidence of Natural Selection Acting on a Polymorphic Hybrid Incompatibility Locus in Mimulus. Genetics, 199(2), 543–554.

Sweigart, A. L., Mason, A. R., & Willis, J. H. (2007). Natural variation for a hybrid incompatibility between two species of mimulus. Evolution, 61(1), 141–151.

Świadek, M., Proost, S., Sieh, D., Yu, J., Todesco, M., Jorzig, C., Rodriguez C., Andrés E., Plötner, B., Nikoloski, Z., Chae, E., Giavalisco, P., Fischer, A., Schröder, F., Kim, S., Weigel, D., Laitinen, R. E. (2017). Novel allelic variants in ACD6 cause hybrid necrosis in local collection of Arabidopsis thaliana. New Phytologist, 213, 900–915.

Tao, Y., Hartl, D. L., & Laurie, C. C. (2001). Sex-ratio segregation distortion associated with reproductive isolation in Drosophila. Proceedings of the National Academy of Sciences, 98(23), 13183–13188.

Taylor, D. R. (1994). The genetic basis of sex ratio in Silene alba (= S. latifolia). Genetics, 136(2), 641–651.

Thomas, J. H., Emerson, R. O., & Shendure, J. (2009). Extraordinary molecular evolution in the PRDM9 fertility gene. PLoS ONE, 4(12), e8505.

Tiffin, P., & Moeller, D. a. (2006). Molecular evolution of plant immune system genes. Trends in Genetics, 22(12), 662–670.

239

Tiffin, P., Olson, M. S., & Moyle, L. C. (2001). Asymmetrical crossing barriers in angiosperms. Proceedings of the Royal Society B: Biological Sciences, 268(1469), 861–7.

Ting, C. (1998). A Rapidly Evolving Homeobox at the Site of a Hybrid Sterility Gene. Science, 282(5393), 1501–1504.

Trivers, R. L. (1974). Parent-Offspring Conflict. American Zoology, 14(1), 249–264.

True, J. R., & Haag, E. S. (2001). Developmental system drift and flexibility in evolutionary trajectories. Evolution and Development, 3(2), 109–119.

Turelli, M. (1994). Evolution of Incompatibility-Inducing Microbes and Their Hosts. Evolution, 48(5), 1500.

Turelli, M., & Hoffmann, A. a. (1995). Cytoplasmic Incompatibility in Drosophila simulans: Dynamics and parameter estimates from natural populations. Genetics, 140, 1319–1338.

Tuttle, E. M., Bergland, A. O., Korody, M. L., Brewer, M. S., Newhouse, D. J., Minx, P., Stager, M., Betuel, A., Cheviron, Z. A., Warren, W. C., Gonser, R. A., Balakrishnan, C. N. (2016). Divergence and Functional Degradation of a Sex Chromosome-like Supergene. Current Biology, 26(3), 344–350.

Twyford, A. D., & Friedman, J. (2015). Adaptive divergence in the monkey flower Mimulus guttatus is maintained by a chromosomal inversion. Evolution, 69(9), 1476–1486.

Umina, P. A., Weeks, A. R., Kearney, M. R., McKechnie, S. W., & Hoffmann, A. A. (2005). A rapid shift in a classic clinal pattern in Drosophila reflecting climate change. Science, 308(5722), 691–3.

Van Der Biezen, E. a., & Jones, J. D. G. (1998). Plant disease-resistance proteins and the gene-for-gene concept. Trends in Biochemical Sciences, 23(12), 454–456.

Van Rijssel, J. C., Moser, F. N., Frei, D., & Seehausen, O. (2018). Prevalence of disruptive selection predicts extent of species differentiation in Lake Victoria cichlids. Proceedings of the Royal Society B: Biological Sciences, 285(1871), 20172630.

240

Vickery, R. (1964). Barriers to Gene Exchange Between Members of the Mimulus guttatus Complex (Scrophulariaceae). Evolution, 18(1), 52–69.

Vickery, R. (1978). Case studies in the evolution of species complexes in mimulus.

Vrana, P. B. (2007). Genomic Imprinting as a Mechanism of Reproductive Isolation in Mammals, 88(1), 5–23.

Vrana, P. B., Fossella, J. a, Matteson, P., del Rio, T., O’Neill, M. J., & Tilghman, S. M. (2000). Genetic and epigenetic incompatibilities underlie hybrid dysgenesis in Peromyscus. Nature Genetics, 25(1), 120–124.

Vulic, M., Lenski, R. E., & Radman, M. (1999). Mutation, recombination, and incipient speciation of bacteria in the laboratory. Proceedings of the National Academy of Sciences, 96(13), 7348–7351.

Wallberg, A., Schöning, C., Webster, M. T., & Hasselmann, M. (2017). Two extended haplotype blocks are associated with adaptation to high altitude habitats in East African honey bees. PLoS Genetics, 13(5), 1–30.

Wang, X., Miller, D. C., Harman, R., Antczak, D. F., & Clark, a. G. (2013). Paternally expressed genes predominate in the placenta. Proceedings of the National Academy of Sciences, 110(26), 10705–10710.

Ward, B. L., Starmer, W. T., Russell, J. S., & Heed, W. B. (1974). The Correlation of Climate and Host Plant Morphology with a Geographic Gradient of an Inversion Polymorphism in Drosophila pachea. Evolution, 28(4), 565–575.

Waters, A. J., Bilinski, P., Eichten, S. R., Vaughn, M. W., Ross-Ibarra, J., Gehring, M., & Springer, N. M. (2013). Comprehensive analysis of imprinted genes in maize reveals allelic variation for imprinting and limited conservation with other species. Proceedings of the National Academy of Sciences, 110(48), 19639–19644.

Waters, A. J., Makarevitch, I., Eichten, S. R., Swanson-Wagner, R. A., Yeh, C.-T., Xu, W., Schnable, P. S., Vaughn, M. W., Gehring, M., Springer, N. M. (2011). Parent-of- Origin Effects on Gene Expression and DNA Methylation in the Maize Endosperm. The Plant Cell, 23(12), 4221–4233.

241

Wen, L., Ruesch, K. L., Ortega, V. M., Kamps, T. L., Gabay-Laughnan, S., & Chase, C. D. (2003). A Nuclear restorer-of-fertility Mutation Disrupts Accumulation of Mitochondrial ATP Synthase Subunit ?? in Developing Pollen of S Male-Sterile Maize. Genetics, 165(2), 771–779.

Werren, J. H., Nur, U., & Wu, C. I. (1988). Selfish genetic elements. Trends in Ecology and Evolution, 3(11), 297–302.

Werth, C. R., & Windham, M. D. (1991). A Model for Divergent, Allopatric Speciation of Polyploid Pteridophytes Resulting from Silencing of Duplicate-Gene Expression. American Naturalist, 137(4), 515–526.

White, R. A., & Turner, M. D. (2012). The Anatomy and Occurrence of Foliar Nectaries in Cyathea (Cyatheaceae). American Fern Journal, 102(2), 91–113.

Wilkinson, G. S., Christianson, S. J., Brand, C. L., Ru, G., & Shell, W. (2014). Haldane’s rule is linked to extraordinary sex ratios and sperm length in stalk-eyed flies. Genetics, 198(3), 1167–1181.

Wise, R. P., Fliss, a. E., Pring, D. R., & Gengenbach, B. G. (1987). urf13-T of T cytoplasm maize mitochondria encodes a 13 kD polypeptide. Plant Molecular Biology, 9(2), 121– 126.

Wittkopp, P. J., Haerum, B. K., & Clark, A. G. (2004). Evolutionary changes in cis and trans gene regulation. Nature, 430(6995), 85–88.

Wolff, P., Jiang, H., Wang, G., Santos-González, J., & Köhler, C. (2015). Paternally expressed imprinted genes establish postzygotic hybridization barriers in Arabidopsis thaliana. eLife, 4, 1–14.

Wolff, P., Weinhofer, I., Seguin, J., Roszak, P., Beisel, C., Donoghue, M. T A., Spillane, C., Nordborg, M., Rehmsmeier, M., Köhler, C.(2011). High-Resolution analysis of parent-of-origin allelic expression in the Arabidopsis endosperm. PLoS Genetics, 7(6).

Wright, A. E., Darolti, I., Bloch, N. I., Oostra, V., Sandkam, B., Buechel, S. D., Kolm, N., Breden, F., Vicoso, B., Mank, J. E. (2017). Convergent recombination suppression suggests role of sexual selection in guppy sex chromosome formation. Nature Communications, 8, 14251. 242

Wright, A. E., Fumagalli, M., Cooney, C. R., Bloch, N. I., Vieira, F. G., Buechel, S. D., Kolm, N., Mank, J. E. (2018). Sex-biased gene expression resolves sexual conflict through the evolution of sex-specific genetic architecture. Evolution Letters, 2(2), 56– 61. http://doi.org/10.1101/176990

Wright, K. M., Lloyd, D., Lowry, D. B., Macnair, M. R., & Willis, J. H. (2013). Indirect evolution of hybrid lethality due to linkage with selected locus in Mimulus guttatus. PLoS Biology, 11(2), e1001497.

Wu, C. A., Lowry, D. B., Cooley, a M., Wright, K. M., Lee, Y. W., & Willis, J. H. (2008). Mimulus is an emerging model system for the integration of ecological and genomic studies. Heredity, 100(2), 220–30.

Wu, J., Saupe, S. J., & Glass, N. L. (1998). Evidence for balancing selection operating at the het-c heterokaryon incompatibility locus in a group of filamentous fungi. Proceedings of the National Academy of Sciences, 95(21), 12398–12403.

Yamagata, Y., Yamamoto, E., Aya, K., Win, K. T., Doi, K., Sobrizal, Ito, T., Kanamori, H., Wu, J., Matsumoto, T., Matsuoka, M., Ashikari, M., Yoshimura, A. (2010). Mitochondrial gene in the nuclear genome induces reproductive barrier in rice. Proceedings of the National Academy of Sciences, 107(4), 1494–1499.

Yannopoulos, G., Stamatis, N., Monastirioti, M., Hatzopoulos, P., & Louis, C. (1987). hobo is responsible for the induction of hybrid dysgenesis by strains of Drosophila melanogaster bearing the male recombination factor 23.5MRF. Cell, 49(4), 487–495.

Yeaman, S. (2013). Hybridization and the porous genome: patterns of isolation and introgression in manakins. Molecular Ecology, 22(12), 3195–3197.

Yeaman, S., & Whitlock, M. C. (2011). The genetic architecture of adaptation under migration-selection balance. Evolution, 65(7), 1897–911.

Zanders, S. E., Eickbush, M. T., Yu, J. S., Kang, J. W., Fowler, K. R., Smith, G. R., & Malik, H. S. ing. (2014). Genome rearrangements and pervasive meiotic drive cause hybrid infertility in fission yeast. eLife, 3, e02630.

Zechner, U., Shi, W., Hemberger, M., Himmelbauer, H., Otto, S., Orth, A., Kalscheuer, V., Fischer, U., Elango, R., Reis, A., Vogel, W., Ropers, H., Rüschendorf, F., Fundele,

243

R. (2004). Divergent genetic and epigenetic post-zygotic isolation mechanisms in Mus and Peromyscus. Journal of Evolutionary Biology, 17(2), 453–460.

Zhang, J. (2003). Evolution by gene duplication: An update. Trends in Ecology and Evolution, 18(6), 292–298.

Zhang, M., Xie, S., Dong, X., Zhao, X., Zeng, B., Chen, J., Li, H., Yang, W., Zhao, H., Wang, G., Chen, Z., Sun, S., Hauck, A., Jin, W., Lai, J. (2014). Genome-wide high resolution parental-specific DNA and histone methylation maps uncover patterns of imprinting regulation in maize. Genome Research, 24(1), 167–176.

Zhou, Q., & Bachtrog, D. (2012). Sex-specific adaptation drives early sex chromosome evolution in Drosophila. Science, 337, 341–346.

Zuellig, M. P., & Sweigart, A. L. (2018). Gene duplicates cause hybrid lethality between sympatric species of Mimulus. PLoS, 14(4), e1007130.

244

Biography

J.M. Coughlan was born and grew up in central Ontario with her parents; Shirley and Michael, older brother, Sean, and several farm pets. Jenn completed a BSc. at Trent

University, in Peterborough Ontario. There she worked concurrently on two honors theses projects on the phylogeography of prairie plants with Dr. Joanna Freeland, and the biogeochemical and physiological side-effects of urban storm water holding ponds on local cattail communities with Dr. Paul Frost. After graduation, Jenn traveled to the subarctic to work for Dr. Erica Noel in monitoring Semi-palmated Plover populations in

Churchill, MB. She then completed fieldwork in Northern BC collecting Crataegus in the lab of Tim Dickinson. Jenn completed an MSc. at the University of Toronto under the supervision of Dr. Tim Dickinson and Dr. Sasa Stefanovic. She published her master’s work in Journal of Biogeography and Molecular Ecology. Jenn then moved to Durham, NC to complete her PhD under the supervision of Dr. John Willis in the Biology Department at Duke University. After completing a rotation with Dr. Kathleen Donohue, Jenn published works in Plant Species Biology and Seed Science Research. One of her thesis chapters has been accepted for publication at Molecular Ecology. During her tenure at

Duke, Jenn was awarded the Maya and William Waldo Boone Fellowship. She belongs to the American Society of Naturalists, and Society for the Study of Evolution. She has received external funding from the National Science Foundation, and the Student

245

Research award for both the Society for the Study of Evolution and American Naturalist

Society. She has also been awarded the Biology Department Graduate Teaching award and was a finalist in the Hamilton Best Student Talk award at the Society for the Study of Evolution conference in Portland, OR, based on her dissertation works. Jenn is excited to begin a postdoc in the lab of Daniel Matute at UNC Chapel Hill next year.

246