Understanding the Breeding Systems of Cakile Edentula, Cakile Maritima

Understanding the breeding systems of Cakile edentula, Cakile maritima

(Brassicaceae) and their hybrids Chengjun Li

(ORCID: orcid.org/0000-0001-6127-7423)

Doctor of Philosophy

September 2017

School of Biosciences The University of Melbourne

This thesis is being submitted in the total fulfilment of the nominated degree and the research project is sponsored by the China Scholarship Council (CSC) (grant number: 201306240013).

Abstract

Invasive plant species and their influences have been a frequent topic of discussion in recent decades; however, most studies treat invasive species individually and their arrival and spread as independent events. In reality, species invade sequentially, building in number and interacting. When these species are closely related, they may hybridise.

This thesis explores a case study of two such invasive species, Cakile maritima and Cakile edentula in Australia and in particular the roles that the breeding system has played in hybridisation and invasion. The sea rockets, Cakile edentula and C. maritima (Brassicaceae), are closely related and hybrids can sometimes be found between the two. They have contrasting breeding systems: C. maritima is self-incompatible and therefore is an outbreeder, whereas C. edentula is self-compatible.

Seeds at different distances from the invasion front of the two species, Cakile maritima and Cakile edentula, were collected and used to produce hybrids, including F1s, F2s and backcrosses. Hybrids and the parental species were also used to assess their relative fitness in a common garden experiment. Results showed that hybrids were successfully produced in both reciprocal directions but differences were observed. Self-incompatibility was inherited in most hybrids but could be enhanced or reduced, inducing a flexible breeding system. Maternal effects were obvious in hybrids. Backcrosses with the self-incompatible species had higher fitness in all hybrids and even higher than both parental species, indicating the existence of heterosis.

This thesis has shown that hybridisation between Cakile species with contrasting breeding systems can be achieved in both directions and the interaction is more likely to occur between hybrids and the self-incompatible species after hybridisation due to the dominance of self- incompatibility and heterosis in hybrids. Therefore, unlike traditional mechanisms of plant invasion, the hybridisation with closely related species and the following interaction between hybrids and their parents can provide the newly arriving species with increased mate availability when it spreads into a new environment, easing the Allee effect and thus helping its establishment.

Keywords: Cakile edentula, Cakile maritima, breeding system, hybridisation, self-incompatibility, self-compatibility, fitness

Declaration

This is to declare that:

(i) The thesis comprises only my original work towards the PhD. (ii) Acknowledgement has been made in the text to all other material used. (iii) The thesis is fewer than the maximum word limit in length, exclusive of tables, maps, bibliographies and appendices.

Acknowledgements

First, I would like to acknowledge my supervisor, Professor Roger Cousens, whose patience with me and enthusiasm for my research seems boundless. I am honoured to have had him as my advisor. I also would like to acknowledge my co-supervisors, Dr Peter Ades and Dr Mohsen Mesgaran, for their insightful comments and supports with my research.

Meanwhile, I would like to express my great gratitude to all the members of my academic committee, Susan Hoebee, Ed Newbigin and Phillip Salisbury, who were always patient with my questions and provided helpful feedback. I could not make it without their valuable help and insightful suggestions.

I would like to thank the nursery and glasshouse technical officers at the University of Melbourne, especially Nicholas Osborne, who provided excellent care of my experiments at Burnley. Your help and kindness made my research enjoyable. In addition, I would like to acknowledge the laboratory technicians and staff at the Burnley Campus who allowed me to use the lab equipment and provided excellent support for all the lab related issues during my study.

I thank my fellow lab mates and PhD students, Sara Ohadi and Lynda M. Hanlon, for the stimulating discussion, support and more importantly for all the fun we have had together. I would like to thank all my friends in Melbourne, whose support is precious to my life and research.

Last but not the least, I would like to thank my family: my parents and my wife, Qi Sun, for supporting me throughout my PhD and in all aspects of my life. Thank you so much and I love you all!

iii Table of Contents

GENERAL INTRODUCTION ...... 1

1.1 INTRODUCTION ...... 1

1.2 HYBRIDISATION AND ITS CONSEQUENCES ...... 2

1.3 DETERMINANTS OF HYBRIDISATION ...... 4

1.4 MODEL PLANT SPECIES ...... 6

1.5 CONCLUSIONS ...... 8

LITERATURE REVIEW ...... 9

2.1 FLORAL DEVELOPMENT ...... 9 2.1.1 Early development ...... 9 2.1.2 Stigma receptivity and pollen ripeness ...... 11

2.2 PLANT BREEDING SYSTEM ...... 13 2.2.1 Pollen post-arrival processes ...... 13 2.2.2 Self-incompatibility ...... 14 2.2.3 Self-compatibility ...... 18 2.2.4 The SI × SC rule ...... 20 2.2.5 Partial and cryptic self-incompatibility ...... 21 2.2.6 Inheritance of breeding system ...... 23

2.3 PROGENY FITNESS ...... 25 2.3.1 Seed traits ...... 25 2.3.2 Progeny growth and survival ...... 27 2.3.3 Reproductive attractiveness and output ...... 29

2.4 CONCLUSIONS AND RESEARCH QUESTIONS ...... 30

GENERATION OF F1, F2 AND BACKCROSSED HYBRIDS ...... 33

3.1 INTRODUCTION ...... 33

3.2 PRODUCING F1 HYBRIDS ...... 34 3.2.1 Seed collections ...... 34 3.2.2 Seed treatments ...... 35 3.2.3 Crossing plan for F1s ...... 36 3.2.4 Methodology of pollination ...... 38

3.3 PRODUCING F2 AND BACKCROSSED HYBRIDS ...... 39 3.3.1 Seed collections ...... 40

iv 3.3.2 Seed treatments ...... 40 3.3.3 Crossing plan for BCs ...... 41 3.3.4 Crossing plan for F2s ...... 42

3.4 CONCLUSIONS ...... 43

THE ROLE OF FLORAL DEVELOPMENT IN HYBRIDISATION BETWEEN C. MARITIMA AND C. EDENTULA ...... 44

4.1 INTRODUCTION ...... 44

4.2 MATERIALS AND METHODS ...... 46 4.2.1 Bud development in Cakile ...... 46 4.2.2 Anther dehiscence ...... 47 4.2.3 Stigma receptivity ...... 47

4.3 RESULTS ...... 48 4.3.1 Bud development...... 48 4.3.2 Anther dehiscence ...... 51 4.3.3 Stigmatic receptivity ...... 52

4.4 DISCUSSION ...... 55 4.4.1 Anther dehiscence ...... 55 4.4.2 Distance between stamen and stigma ...... 55 4.4.3 Stigma receptivity ...... 56 4.4.4 Explanations for the occurrence of hybridisation ...... 56

THE INHERITANCE OF BREEDING SYSTEM IN HYBRIDS BETWEEN SI AND SC SPECIES ...... 60

5.1 INTRODUCTION ...... 60

5.2 MATERIALS AND METHODS ...... 62 5.2.1 Pollen germination and pollen tube growth...... 63 5.2.2 Fruit and seed set ...... 64 5.2.3 Pollen viability ...... 64 5.2.4 Determination of self-incompatibility ...... 65 5.2.5 Pollen-ovule ratio ...... 65

5.3 RESULTS ...... 66 5.3.1 Pollen germination and pollen tube growth...... 66 5.3.2 Fruit and seed set ...... 70 5.3.3 Pollen viability ...... 71

v 5.3.4 Determination of self-incompatibility ...... 72 5.3.5 Pollen-ovule ratio ...... 74

5.4 DISCUSSION ...... 76 5.4.1 Breeding system barriers upon hybridisation ...... 76 5.4.2 Inheritance of the breeding system ...... 77 5.4.3 Impact of the variation of breeding system on sex expression ...... 79 5.4.4 Ecological implications ...... 79

FITNESS COSTS OF HYBRIDISATION BETWEEN TWO INVASIVE PLANTS WITH CONTRASTING BREEDING SYSTEMS ...... 81

6.1 INTRODUCTION ...... 81

6.2 MATERIAL AND METHODS ...... 82 6.2.1 Seed mass and germination ...... 83 6.2.2 Seedling survival ...... 83 6.2.3 Plant development ...... 84 6.2.4 Floral display ...... 84 6.2.5 Plant projected area ...... 85 6.2.6 Reproductive output ...... 86 6.2.7 Cumulative fitness ...... 86

6.3 RESULTS ...... 87 6.3.1 Seed mass and germination ...... 87 6.3.2 Seeding survival ...... 88 6.3.3 Plant development ...... 89 6.3.4 Floral display ...... 90 6.3.5 Plant projected area ...... 93 6.3.6 Seed number ...... 93 6.3.7 Cumulative fitness ...... 94

6.4 DISCUSSION ...... 95 6.4.1 Fitness costs at early life stages ...... 95 6.4.2 Fitness costs at late life stages ...... 96 6.4.3 Implications for co-invasion of related species ...... 97

GENERAL CONCLUSIONS ...... 100

7.1 WINDOW OF OPPORTUNITY FOR HYBRIDISATION ...... 100

7.2 BREEDING SYSTEM UPON HYBRIDISATION ...... 102

vi 7.3 FITNESS COSTS OF HYBRIDISATION ...... 104

7.4 LIMITATIONS ...... 105

7.5 CONCLUSIONS ...... 106

APPENDICES ...... 107

APPENDIX TO CHAPTER 4 ...... 107

APPENDIX TO CHAPTER 5 ...... 112

APPENDIX TO CHAPTER 6 ...... 114

REFERENCES ...... 117

vii List of Tables

Table 2.1 Two main types of stigma, revised from (Richards 1997)...... 14 Table 2.2 Major differences between GSI and SSI, revised from De Nettancourt (1997)...... 15 Table 3.1 Population sources of individual used in the experiment...... 37 Table 3.2 Crossing plan to produce F1s and outcrosses within C. maritima and C. edentula. Left panel is the crossing plan for producing F1 hybrids where “×” indicates crossings without allocated labels; other crossings were labelled with corresponding labels from “F11-F115” (see detailed explanation in Section 3.3.1). “M1-M10” refer to the 10 C. maritima plants; “E1-E10” represents the 10 C. edentula plants. Each plant acted as both pollen donor and recipient. Right panel is the outcrossing plan for producing outcrossed progeny, which was numbered as shown in the right panel...... 38 Table 3.3 F1s used in producing F2s and their lineage. M2 and M3 related F1s were excluded as M2 and M3 were identified as hybrids. M9 died during the experiment and thus no seeds were producing between M9 and E4...... 40 Table 3.4 Crossing plan of producing backcrossed hybrids. Crossings listed in the table are all reciprocal...... 42 Table 3.5 Groups divided by F1s lineage. In a single group, F1s’ parents share the same origin, i.e., derived from the same population...... 42 Table 3.6 Crossing plan between F1 groups to produced F2s. “×” indicates crossing to produce F2s...... 43 Table 3.7 Detailed crossing plan to produce F2s...... 43 Table 4.1 Summary of the stages of floral development in Cakile and the landmark events used to define each stage...... 48 Table 5.1 Crossing directions and corresponding codes used in experiments. For details of seed collection and experiment design, please refer to Chapter 3...... 62 Table 5.2 Estimated odds ratios of (i.e., differences between) the degree of pollen tube growth when producing EE, MM, F1s, F2s and BCs. Estimated odd ratios were based on the higher value of pollen tube growth*. Significant comparisons are in bold (P < 0.05). Only parts of pollen tube growth when producing hybrids were sampled due to the limited availability of buds in each plant type. Therefore, pollen tube growth condition was not sampled when producing MH...... 70 Table 5.3 The proportion of viable pollen in parental species and all produced hybrids. Pollen viability was assessed using FDA. Significance groupings (lower case letters) were based on Tukey’s post hoc tests (pair-wise comparisons with a significance level of P < 0.05). Means with the same letter are not significantly different. Minimum pollen viability with a value of zero is in bold...... 72 Table 5.4 Proportion of fruit set in the artificial or natural selfing in C. edentula. All selected racemes were bagged to prevent potential pollinator’s visitation. Significance groupings (lower case letters) are based on Tukey’s post hoc tests (pair-wise comparisons with a significance level of P < 0.05). Means with the same letter are not significantly different...... 73

viii Table 5.5 Breeding system in the parental species and their progeny. The “selfing rate” column in the table indicates the selfing rate in those SC plants of each plant type. Selfing rate was estimated in the field station by artificial selfing on chosen racemes covered with pollination bags. “1: 0” and “0: 1” ratios (SI: SC) indicate that all plants in that row were self-incompatible (1: 0) or self-compatible (0: 1)...... 73 Table 5.6 Pollen grain number (back transformed) and pairwise comparisons between plant types. Significance groupings (lower case letters) are based on LSMEANS (pair-wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different...... 74 Table 5.7 Ovule number produced by each plant type. Significance groupings (lower case letters) are based on the differences of least squares means of the odds ratios in ovule number per fruit, estimated using PROC LOGISTIC. Groups with the same letters are not significantly different (P < 0.05)...... 75 In parental species, C. maritima had the largest pollen-ovule ratio (P < 0.001) (no significant difference between C. maritima and MM, P > 0.05) while C. edentula and EE had the smallest pollen-ovule ratio (Table 5.8)...... 75 Table 5.8 Pollen-ovule ratio (back transformed means). Means with the same letter are not significantly different...... 75 Table 6.1 Means of seed mass (mg) of all plant types. Lower case letters are based on Tukey’s post hoc test (pair-wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different. M, E and H indicate C. edentula, C. maritima and the F1 hybrids used in producing backcrosses (see details about crossing plan in Chapter 3), respectively. When the initials combine, the first letter indicates the pollen donor; for example, EM and ME indicate two reciprocal crossings used to produce F1s (the first with E as the pollen donor, and the second with M as the pollen donor). MM and EE are outcrossed seeds within natural C. maritima and C. edentula, respectively...... 87 Table 6.2 Average germination percentage in each plant type. Twenty seeds, if applicable, of each replicate within a plant type were used. The number of germinated, un-germinated and TTC viable seeds is the sum of all replicates in each plant type. Tetrazolium chloride (TTC) was used to test the viability of un-germinated seeds...... 88 Table 6.3 Means of the proportion of seedling survival in different plant types. Multiple comparisons between different plant types indicated that none of them were significantly different from any others. Numbers are combined across all replicates; survival was calculated separately for each replicate...... 88 Table 6.4 Test of the differences between plant types within each stage in phenology. Significant P values are in bold. Least Square Means were used to estimate the differences within stages. Stages are all measured as days since germination...... 89 Table 6.5 Time (days) required to reach the stages of bud emergence (stage 4) and first open flower (stage 5). Significance levels are indicated by lower case letters, based on Tukey’s post hoc test (pair-wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different...... 89

ix Table 6.6 Means of petal area per flower (mm2) in each plant type. Means with the same letters (lower case) are not significantly different. Significance level was P < 0.001...... 90 Table 6.7 Frequency distribution of petal number in all plant types. The column “total” presents the total number of samples collected in each plant type...... 91 Table 6.8 Means of daily flower number in each plant type. Means with the same letters are not significantly different (P > 0.05)...... 91 Table 6.9 Means of the total lifetime flower number in each plant type. Means with the same letters are not significantly different (P > 0.05)...... 92 Table 6.10 Means of the daily floral display area (mm2) in each plant type. Means with the same letters are not significantly different (P > 0.05). When analysing, data were square root transformed...... 92 Table 6.11 Means of plant projected in all plant types at maturity. Means with the same letters are not significantly different (P > 0.05)...... 93 Table 6.12 Means of seed number at maturity in different plant types. Seed numbers with the same grouping letters (lower case) are not significantly different (P > 0.05)...... 94 Table 6.13 Relative fitness at different stages. Within each column, all values for that fitness component were divided by the largest value. Thus, a fitness score of 1.00 was assigned to the highest value in the column (in bold), and a proportion to all other plant types. No statistical significance is implied by these numbers (Johnston et al. 2003)...... 94 Table 6.14 The main assumptions and parameters used in the model of Mesgaran et al. (2016). “-” means the assumption has not been tested or confirmed in this study...... 97

x List of Figures

Figure 1.1 Determinants of hybridisation and their interactions. As shown in the diagram, plant breeding system plays the key role and interacts with all other parts. Pollinator is the only external factor involved...... 5 Figure 1.2 Photographs of individual plants and their fruits of C. maritima (A, C) and C. edentula (B, D) (revised from Cousens 2013)...... 7 Figure 2.1 A conceptual model for the consequences of a GSI (left) and SSI (right) system on the viability of pollen and the possible outcome of progeny. Red represents failed pollen while green represents successful pollen; brown-coloured progeny indicates a homozygote. The dominance hierarchy of SSI in the figure is: S1 > S2 > S3 > S4...... 17 Figure 3.1 A schematic diagram of producing all lines in this experiment, where “M” and “E” are two parental species while “H” stands for the F1 hybrids. “MM” and “EE” represent intraspecific outcrossing of M and E respectively; “BC” stands for backcrossing between the F1s and parental species. F2s are produced using F1 hybrids...... 33 Figure 3.2 A schematic drawing of relative positions of sampling sites. Green and black arrows indicate the invasion fronts of C. maritima and C. edentula, respectively. Dashed line indicates a certain species is either replaced or not colonized. Narrowed solid lines indicate the sympatric zone of the two species...... 35 Figure 3.3 Seedlings in biodegradable pots...... 36 Figure 3.4 Seedlings transplanted into the plastic pots (left) and grew bigger (right)...... 36 Figure 3.5 Experimental layout of the two species. Left panel shows the arrangement of plants within the polyhouse and right panel shows an enlarged picture of a single plant in a pot with pollination cages. The white square represents the pollination cage covering the plant. The green region shows the area outside each pollination bag. The grey circle shows the position of the pot...... 37 Figure 3.6 Plants covered by cages in the Poly tunnel...... 37 Figure 3.7 Diagram of pollination processes...... 39 Figure 4.1 Pictures of stages 1-4 taken under Leica M205A dissecting microscope using its built- in Z-stack function, showing landmark events in C. edentula (upper panel) and C. maritima (lower panel). In C. maritima, scale bar = 1 mm at stages 1-4; in C. edentula, scale bar = 0.5 mm at stage 1, scale bar = 1 mm at stages 2-4...... 49 Figure 4.2 Pictures of stages 5-9 taken under dissecting microscope, showing landmark events in C. edentula (upper panel) and C. maritima (lower panel). In C. maritima, scale bar = 2 mm at stages 5-9; in C. edentula, scale bar = 1 mm at stages 5-6, scale bar = 2 mm at stages 7-9...... 49 Figure 4.3 Difference in length between the pistil and upper stamens in C. edentula (a) and C. maritima (b). Positive values indicate that the stigma is above the tallest anther...... 50 Figure 4.4 Difference in length between the pistil and lower stamens in C. edentula (a) and C. maritima (b). Positive values indicate that the stigma is above the tallest anther...... 50

xi Figure 4.5 Anther dehiscence along with the development of buds in C. edentula (a) and C. maritima (b). Scale bars in C. maritima, stages 2-3 = 1 mm, stages 4-6 and stage 7 = 2 mm. Scale bars in C. edentula, stages 2-3 = 1 mm, stages 4 and 6-7 = 0.5 mm, stage 5 = 1 mm...... 51 Figure 4.6 Analysis of anther dehiscence by fitting logistic regression. Figures show predicted probabilities for anther dehiscence in C. maritima (left panel) and C. edentula (right panel). Upper stamens in C. edentula fitted better (P < 0.001) compared with that in C. maritima (P < 0.05)...... 52 Figure 4.7 A close-up look of pollen grains (yellow arrows) that attached to the stigmatic papillae (black arrows) with pollen tubes (green arrows) elongating in C. edentula (a, stage 5) and C. maritima (b, stage 7) 2 h after pollination. Scale bar = 100 µm...... 52 Figure 4.8 Pollen germination and pollen tube elongation in C. edentula from stage 2 to 7 under a light microscope (before staining) and UV light (after staining). Yellow arrows indicate the pollen grains attached to the stigma surface; green arrows mark the pollen tubes growing in the style. Pictures from light microscopy were presented for a clear view of the embryos (green arrowhead) and vascular bundles (red arrows). Scale bar = 250 µm...... 53 Figure 4.9 Pollen germination and tube elongation in C. maritima from stage 2 to 7 under a light microscope (before staining) and UV light (after staining). Pollen grains (yellow arrows) germinated, inducing pollen tube (green arrows) growth and elongation into the stigma. Pictures from light microscopy were presented for a clear view of the embryos (green arrowhead) and vascular bundles (red arrows). Stigmas were harvested 2 h after pollination. Scale bar, 250 µm...... 54 Figure 4.10 Examples of pollen shedding onto the stigma at stage 7 in C. edentula. Pollen could completely cover the stigma surface (A) or a single area with complete clean stigma elsewhere (B), and even did not shed (C and D) despite fully flowering and anther dehiscence. Scale bar = 2 mm...... 58 Figure 4.11 Pollen adhesion positions of C. edentula in different samples. All samples in the upper panel had pollen grains adhered mostly on both sides of the stigma whereas pollen was all over the stigma surface of the samples in the lower panel. Naturally, after anthers dehisced, the pollen should shed atop the stigma evenly and cover all the stigma surface (as shown in lower panel) instead of over a single and certain area, especially on both sides of the stigma. Scale bars: A and F, 1 mm; B, C and E, 2 mm; C, D, E and H, 500 µm...... 58 Figure 5.1 Pollen germination and pollen tube growth under UV light 2 h after pollination when the pollen of C. edentula landed on the stigma of C. maritima (a and b) and vice versa (c and d). Figure a, b and c were under 5x magnification while d was under 10x magnification. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule...... 67 Figure 5.2 Examples of pollen tube growth in the intraspecific outcrossing of C. edentula (a and b) and C. maritima (c and d). Figure a and b were under 10x and 20x magnification, respectively; while c and d were under 5x magnification, respectively. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule; SP, stigmatic papillae...... 68 Figure 5.3 Pollen germination and pollen tube growth under UV light 2 h after pollination in the backcrossing of F1s with C. maritima (F1 was the pollen donor in a and b) and interspecific

xii crossing of F1s (F1s were the pollen donor as well as recipient in c and d). All figures were taken under 5x magnification. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule...... 69 Figure 5.4 Proportion of fruit set and seed set (± SE) of all reciprocal crossings conducted in this study. M, C. maritima; E, C. edentula; H, F1 hybrids. There were two and four different crossing directions when producing F1s and BCs, respectively, i.e., EM and ME in producing F1s, and F1 hybrids backcrossing C. edentula (EH and HE) and C. maritima (MH and HM) in producing BCs. Arrows in different colours indicate different crosses: red, reciprocal crossing producing F1s; blue, backcrossing between F1s and parental species; black, outcrossing within parental species and F1s...... 71 Figure 5.5 Examples of pollen viability of C. martima (a), C. edentula (b), F1s (c, HM), F2s (d) and BCs (e, EH; f, MH). Pollen grains were stained with FDA and observed under UV light (wavelength = 495 nm). Bright pollen grains were judged as viable while those greyed out were not. Scale bar = 100 µm...... 72 Figure 6.1 Common garden experiment. Germinated seeds were firstly transplanted into biodegradable pots (a) for a short period and then transferred to plastic pots (b). All pots then were randomly placed in the field station (c) at a spacing of around 0.5 m...... 84

xiii General introduction Invasive plant species and their influences have been a frequent topic of discussion in recent decades (Harrison and Darby 1955, Rodman 1980, Vallejo‐Marín and Hiscock 2016). Most studies treat invasive species individually and their arrival and spread as independent events. However, species invade sequentially, building in number and interacting. When these species are closely related, they may hybridise during invasion (Ellstrand and Schierenbeck 2000). This thesis will explore a case study of two such invasive species, Cakile maritima and Cakile edentula in Australia (Rodman 1986), and in particular the roles that breeding system has played in hybridisation and invasion. In this chapter, I will present the reasons for choosing hybridisation and breeding systems in invasive Cakile species as the topic and outline the main contents of this thesis.

1.1 Introduction Invasive plants can be defined as plants that are introduced to a new habitat with a potential to survive, establish and colonize, thus posing a variety of risks to biological diversity, crop production and human well-being. They may have both negative and positive impacts. Paini et al. (2016) gave an accurate and fast increasing invasion cost worldwide, in which China and the USA suffer most ($117.29 and $70.381 billion total invasion cost, respectively) while the cost in Australia is estimated as $7.815 billion, ranking 15th worldwide. Invasive plants can create novel habitats for disease vectors (Pejchar and Mooney 2009). For example, the invasion of dense stands of Lantana (Lantana camara) in East Africa has provided a new habitat for the tsetse fly (Glossina spp.), which carries sleeping sickness (Pejchar and Mooney 2009). Potential benefits of invasive plants may include their use as food, fibre and even fuel to local communities; they also have the ability of taking up heavy metals (Pejchar and Mooney 2009).

There are various explanations for why some species are particularly successful after invasion. Firstly, they may have escaped from predators and competitors (Mitchell and Power 2003) and thus could spread fast in some similar habitats. Secondly, some plants may have pre-adaptations. For example, they may have evolved higher plasticity, competitiveness, or tolerance to disturbance, and interfere with other plants chemically before their invasion (Hobbs and Huenneke 1992, Levine and D'Antonio 1999, Callaway and Ridenour 2004). Thirdly, successful invasion may be also related to the post-adaptations of invasive plants to a broad range of environments (Kriticos et al. 2003, Adger et al. 2005, Bedsworth and Hanak 2010) in which EICA (the Evolution of Increased Competitive Ability) theory is a good case. It predicts that invasive plants, having escaped from their natural enemies in their introduced range, evolve reduced resource allocation to herbivore defence and increase vegetative growth or reproductive effort (Bernd and Rolf 1995). Although EICA became popular among ecologists, there is evidence both for (Joshi and Vrieling 2005, Hull-Sanders et al. 2007) and against (Felker‐Quinn and Schweitzer 2013). Colautti et al. (2009) also showed how it may be confused with other adaptations.

As a matter of fact, from a genetic and evolutionary point of view, both pre- and post- adaptations are determined by genetic variation. Adaptation occurs because of inheritable variation in relevant traits and selective pressure. There are many processes that affect genetic variation, with some increasing variation and some decreasing it: (1) genetic mutations,

1 changing the nucleotide sequence of the genome (Eyre-Walker and Keightley 2007) ; (2) genetic bottleneck, largely restricting and reducing the number of alleles available (Allendorf 1986); (3) genetic drift, which is caused by random changes that result in some alleles becoming abundant while others reducing in frequency or even disappearing (Wright 1931, Fisher 1937, Masel 2011); (4) multiple introductions of the species (Ciosi et al. 2008, Cousens et al. 2013), which may also bring together combinations of genes that have never co-existed in their native range (Kolbe et al. 2007); (5) hybridisation with another species could potentially increase variation among genes and may produce new species that can survive in different habitats (Baack and Rieseberg 2007, Payseur 2010, Ludwig et al. 2013). Rather than producing new species, some genes from the hybrids may introgress into one of the species, thus increasing its genetic diversity (Snow et al. 2001, Ohadi et al. 2015).

Extensive theoretical studies have been conducted on the implications of genetic variation for adaptation (Kondrashov 1988, Hyten et al. 2006, Eyre-Walker and Keightley 2007, Masel 2011) but there have been relatively fewer studies specifically about implications of hybridisation (Goodwillie and Ness 2013, Mesgaran et al. 2016). Although many empirical studies have revealed that hybridisation is important for invasive species (Klinger and Ellstrand 1994, Ellstrand and Schierenbeck 2000, Snow et al. 2001) and their spread (Ridley and Ellstrand 2009), they mostly recorded the occurrence without considering the mechanisms and processes behind hybridisation. As suggested by Goodwillie and Ness (2013), despite extensive literature on both hybridisation and mating system evolution in plants, relatively little consideration has been given to the effect of mating system traits on rates of hybridisation and the role that hybridisation might play in shaping mating systems. Therefore, it is a neglected, under- researched but important area, which is worth further investigation.

1.2 Hybridisation and its consequences Hybridisation can be defined as the production of viable offspring from interspecific mating (Baack and Rieseberg 2007). Under natural conditions it is rare, although some researchers have argued that hybridisation is more common than thought (Mallet 2005, Baack and Rieseberg 2007).

There are many barriers preventing hybridisation from happening. These barriers may be either pre-zygotic (Ganders 1979, Grant 1994, Brandvain and Haig 2005) or post-zygotic (Campbell 2003). Pre-zygotic barriers consist of pre-pollination barriers and pollen-pistil interactions (also known as “post-pollination but pre-zygotic barriers”). Pre-pollination barriers are exclusively pre-zygotic (Carrió and Güemes 2014), and may arise as a result of ecological (e.g., geographical distributions and habitats), behavioural (e.g., pollinator behaviour) and temporal differences or divergent floral traits (such as flowering phenology). Pollen–pistil interactions are caused by plant breeding system, inducing corresponding hetero-specific pollen rejection, unsuccessful pollination and further seed abortion (Brandvain and Haig 2005, Goodwillie and Ness 2013) (see details of pollen-pistil interaction in Section 2.2). Post-zygotic barriers include inviability and sterility of hybrids caused by genetic incompatibilities or low fitness of progeny (see details in Section 2.3) (Brandvain and Haig 2005, Goodwillie and Ness 2013).

If barriers are removed, hybrids can be generated. There are many ways to overcome barriers. For example, if pre-pollination barriers exist between species, e.g., ecological isolation, they can

2 be broken by introducing one species to the habitats of another, thus eliminating geographical barriers (Rodman 1980, 1986). Meanwhile, if pollen-pistil interactions (i.e., self-incompatibility) occur when hybridising, some treatments, for example, bud pollination or treating stigmas with cycloheximide (Hiscock and Dickinson 1993) may disable such barriers and thus produce fertile hybrids. Moreover, although in some species hybrids may show inferiority to their parental species, evolution of breeding system can help reduce post-zygotic barriers (e.g., sterility and low viability) and thus increase the fitness of hybrids, which in turn reduces barriers and promotes hybridisation (De Nettancourt 1997, Richards 1997, Brennan et al. 2013).

After barriers are overcome and hybrids are generated, there are different possible outcomes. Hybrids may not persist and have no ecological consequences while they may also temporarily show impact on population process without necessarily having any long term genetic consequences. In contrast, hybrids may promote introgression between parental species, increasing their genetic variation and even forming new species.

Firstly, hybridisation may lead to speciation through divergent selection and corresponding reproductive isolation (Hatfield and Schluter 1999). Theories of reproductive isolation have been well-developed by Mayr (1942, 1970) from the viewpoint of a zoologist and then further explained by many other plant scientists (Hatfield and Schluter 1999, Campbell 2003, Provine 2004, Baack and Rieseberg 2007, Brennan et al. 2011a). Hybrids have a different genome than their parents, which can act as the “raw material” for speciation (Rieseberg et al. 1999). The source of divergent selection in flowering plants may be either animal pollinators or environmental features of habitats. This complex mixture of selection, mediated by pollinators and other resources, leads to physiological differentiation of hybrids from their parents, e.g., time of flowering and attractiveness to pollinators (see detailed review in Campbell 2003). Subsequently, these adaptations, often associated with spread of hybrids into alternative habitats, result in the re-establishment of reproductive barriers (mainly pre-zygotic at first) and thus contribute to speciation. Post-zygotic barriers may exist as well but are rather weak compared with pre-zygotic ones under such conditions (Lowry et al. 2008). Examples of well- studied speciation resulting from hybridisation include sunflowers (Ungerer et al. 1998, Rieseberg 2006) and the invasive Spartina anglica (Baumel et al. 2003).

Hybridisation may also enable gene flow through introgression (Baack and Rieseberg 2007), without resulting in new species. Introgression is defined by Rieseberg and Wendel (1993, p. 71)as the stable integration of genetic material from one species into another via repeated backcrossing. There is mounting evidence showing that hybridisation-derived introgression occurs in nature. For example, Rieseberg et al. (1999) and Yatabe et al. (2007) have found significant gene flow between two sunflower species, Helianthus annuus and H. petiolaris. Similarly, Snow et al. (2001) reported that wild radish (Raphanus raphanistrum) can hybridise with cultivated radish (R. sativus), producing hybrids that can transfer genes from the crop to the weed. Many studies by Ellstrand and his co-workers (Klinger and Ellstrand 1994, Arriola and Ellstrand 1997) have also demonstrated that hybridisation between crops and weeds has contributed to gene flow, and suggested the possibility of introgression and subsequent benefits to weed spread (Ellstrand and Schierenbeck 2000).

The previous outcomes from hybridisation, speciation and introgression are about selection and adaptation; however, previous and current ongoing studies by Cousens and co-workers (Cousens and Cousens 2011, Ohadi et al. 2012, Cousens et al. 2013, Ohadi et al. 2015) have led to a third possibility that the temporary formation of hybrids may enable species to overcome population constraints (Mesgaran et al. 2016). Populations, when they are small, can be pollen- limited due to limited mate availability, as plants fail to attract sufficient pollinators and thus there is less production of seeds (the “Allee effect”, see details in Section 2.2.3). This will not be a problem for an inbreeder (self-compatible, SC) but could be particularly severe for self- incompatible (SI) species. Usually when a single founder of an SI species arrives in a new site, it cannot reproduce and establish a population. However, if it can hybridise with another congeneric species that is already there (e.g., SC species), then its genes may establish. The resulting hybrids may give rise to a permanent hybrid population or they may backcross with further plants of the new invader that arrive later.

The population genetics of hybrids is an ongoing area of exciting research internationally while the interaction of breeding system in hybridisation has drawn little attention. Cousens and co- workers (Cousens et al. 2013, Mesgaran et al. 2016) indeed have done a lot of work on this under-researched area but more efforts are needed to further investigate and understand it.

1.3 Determinants of hybridisation To figure out the role of interaction of breeding systems in hybridisation, the priority is to understand what the determinants of hybridisation are. Figure 1.1 shows, in a scenario of hybridisation, factors that could have profound influences on hybridisation in which the plant breeding system interacts with all other factors. Basically, these factors can be divided into two categories: pollinator and intrinsic traits of plants, such as floral development and pollen-pistil interactions.

Pollinators and their behaviour, as shown in Figure 1.1, interact with plants and determine whether there will be any pollen and what kinds of pollen arrive on the stigma. Pollinators visit flowers for various reasons. A large proportion of pollinators just go to flowers for food, feeding on nectar and sometimes also on the pollen itself. Occasionally, flowers offer neither pollen nor nectar as a foodstuff to their visitors but yield other rewards instead; or they may offer these as “extras” in addition to some pollen (see details in Chapter 9, Willmer 2011). This is mainly affected by the traits of plants available to the pollinators at the time (e.g., attractiveness of species, frequency and abundance, inherited from parental plants). Thus, pollinator behaviour will be greatly affected by the traits of the target species and its neighbouring species (Seifan et al. 2014). Pollen competition can occur if a mixture of compatible pollen is carried and deposited onto a stigma by pollinators. The order of arrival of different pollen types determines which one gains preferential access, initiating pollen germination ahead of others. When the pollen load is a mixture of different pollen types arriving at the same time, there will also be competition between pollen tubes in the stigma such that the pollen type with faster pollen tube growth will have a greater opportunity to fertilise the ovules.

Figure 1.1 Determinants of hybridisation and their interactions. As shown in the diagram, plant breeding system plays the key role and interacts with all other parts. Pollinator is the only external factor involved. Besides pollinators, hybridisation is also closely related to intrinsic traits of plants, in which the breeding system plays a key role. As shown in Figure 1.1, the plant breeding system interacts with and largely determines floral development and fitness of progeny (see details in Chapter 2).

Firstly, when a flower emerges, traits (such as herkogamy and cleistogamy etc., determined by plant breeding system, see detail review in Chapter 2) inherited from parents will determine the processes of floral development, such as timing of stigma receptivity and pollen dehiscence. As mentioned above, pollinators are just one of the main pre-zygotic barriers; with the help of pollinators, hybrids may occur if their habitats are sympatric or at least parapatric. However, hybridisation cannot happen if flowering of the species does not overlap in time or space or simply because self-pollen has fertilised the ovules prior to the arrival of alien pollen. These plant traits, thus, are obstacles for hybridisation occurrence regardless of the presence of pollinator.

Secondly, after deposition of pollen, the plant breeding system directly affects the success or failure of hybridisation. The plant breeding system (or plant mating system) affects the reproduction processes from the moment pollen arrives onto the stigma to seed set and seed dispersal (Dafni 1992). It directly determines what kinds of pollen will successfully fertilise the ovule because of complicated mechanisms maintaining pre-zygotic and post-zygotic barriers. When pollen arrives on a stigma, compatible pollen will germinate, pollinate the ovule and thus complete fertilisation in which SI plays an important role. It is noteworthy that not only self- pollen but also incompatible alien pollen will be rejected if SI is functionally operating in the pollen recipient species (see details in Kitashiba and Nasrallah 2014, as well as in Chapter 2).

Last, but not least, the plant breeding system also has great impacts on the fitness of progeny. After hybridisation, the resulting seeds may be inviable or vary in fitness. Fitness of progeny could potentially be affected by genetically inherited traits as well as the environment. In the experiments conducted in this thesis, all plants were grown in the same environment, therefore the variation of fitness in progeny could only be due to genetically inherited traits. As competition occurs between individual progeny after seeds germinate, become seedlings and grow to maturity, those with more vigorous growth will have a greater ability to compete. It is common that selfing progeny of outcrossing species suffers from inbreeding depression and thus shows reduced fitness compared with outcrossing lines, termed as inbreeding depression while outcrossing progeny of inbreeders suffers from outbreeding depression (Fox 2005). Similarly, after hybridisation, hybrid vigour may vary greatly depending on the parents, i.e., some hybrids have negative heterosis (outbreeding depression) while others have positive heterosis (hybrid vigour) (see details of fitness in Chapter 2).

Therefore, the plant breeding system plays a key role in hybridisation but most studies in recent years failed to address this interaction, which justifies the subject of my thesis: understanding the breeding system of hybridising species and their hybrids and revealing exactly what the role breeding system plays and how it works in hybridisation. Although pollinators and their interactions with plants are important to the supply of pollen, these aspects alone would require another PhD thesis. Therefore, pollinators and their behaviour were excluded from this thesis. Meanwhile, the primary aim is to get a clear understanding of the role that the breeding system plays in hybridisation and thus the scenario of pollen mixture (competition) was not taken into consideration although it is very common in the wild (Ward et al. 2013, McCallum and Chang 2016, Pélabon et al. 2016).

Considering the interaction between hybridisation and breeding system, there are many questions remained to be answered; however, what I am interested in is the hybridisation between contrasting breeding systems. Hybridisation could occur between many closely related species that are mostly congeneric and even recently formed through speciation. These species may or may not have the same breeding strategy, for example, hybridisation of SC × SC, SI × SI, and SC × SI. Specifically, hybridisation of SC × SI would be the most exciting area as it involves different breeding systems and their interactions. It is intriguing to know: what happens when species with different breeding systems form hybrids? What breeding system could be inherited in hybrids? Is this inheritance partial or complete? Could this inheritance break down or be continuous in future generations? What are the fitness costs of hybridisation? To answer these questions, I chose two closely related introduced species in Australia as the model species and focused on the interaction of breeding system upon their hybridisation.

1.4 Model plant species The model system I will use in Australia for studying hybridisation and the interactions of breeding systems involves two introduced and invasive species: Cakile maritima and Cakile edentula, family Brassicaceae. Here I will present a brief introduction of these two species in Australia.

The sea rockets, Cakile edentula (Bigelow) Hook. and C. maritima Scop. (Brassicaceae), are closely related and hybrids can sometimes be found between the two species (Rodman 1980, Ohadi et al. 2012, Cousens et al. 2013). They have contrasting breeding systems: C. maritima is self-incompatible and therefore is an outbreeder, whereas C. edentula is self-compatible.

A B C D

Figure 1.2 Photographs of individual plants and their fruits of C. maritima (A, C) and C. edentula (B, D) (revised from Cousens 2013). These two species both invaded Australia in the late 19th century. C. edentula originated from sand beaches of eastern North America (first collected in Australia in the 1860’s near Melbourne). It was already widespread in Australia by the time that C. maritima arrived from the Mediterranean and Western Europe. C. maritima was first collected in Australia in the 1890’s near Perth (Rodman 1974, 1980). The species are mainly distributed on strand and fore dunes of coastal areas. This is a harsh environment for most species and the absence of competitors in this habitat perhaps contributes to their wide distribution all around the southern and eastern Australian coasts (Heyligers 1984). From herbarium records, it was originally inferred that after the introduction in Melbourne, C. edentula spread along the coast of Victoria to South Australia, south into Tasmania, and north into New South Wales and southern Queensland. C. maritima spread eastwards from Western Australian along the coast of southern Australia, then into New South Wales and Tasmania (Rodman 1974, 1976, 1980, 1986). However, recent morphological and genetic studies have revealed that there may have been multiple introductions of C. maritima rather than a single arrival starting in Perth; eastern populations of C. maritima almost certainly do not derive from those in the west (Ohadi et al. 2015).

C. maritima has completely replaced C. edentula in much of southern and eastern Australia (Rodman 1974, 1986) as it has in southern California (Boyd and Barbour 1993) and the North Island of New Zealand (Cousens and Cousens 2011). Field surveys have shown that the invasion fronts of C. maritima are currently in southern Tasmania, and in northern New South Wales and southern Queensland. Microsatellite studies showed that there have been multiple introductions of C. maritima into the eastern parts of Australia, while there is only one introduction into Western Australia. These two species can be found on the same beaches in these sympatric regions. Sequencing of chloroplast genes was used to develop species-specific markers, which indicated that there has been introgression from C. edentula to C. maritima (Ohadi et al. 2015).

All the evidence presented above implies that hybridisation happened between these two species and is still happening in their sympatric zones. Although hybrids have been reported in these two species, no report has studied it from the point of the interaction between breeding

7 systems and hybridisation, which makes the two species perfect model species. With Cakile as the model species, I want to ask the following main questions.

1. What is the window of opportunity for hybridisation between the two closely related species? This will mainly deal with the reproductive barriers, especially barriers caused by the differences in floral development between the two closely related species. 2. How are the breeding systems inherited in hybrids? This may include the following sub-questions: which crossing direction of hybridisation is preferred? What are the reasons behind the preference? Is SI inherited in hybrids? Is the inheritance continuous in the next few generations? 3. What are the fitness costs of hybrids compared with their parents? It is of great importance to check whether the hybrids show any inferiority or superiority compared with their parents. Meanwhile, if hybrids backcross with their parental species, what would happen? What are the implications of the hybridisation?

1.5 Conclusions Both hybridisation and breeding system have been extensively and thoroughly researched in which their origin, evolution and rationales have been fully explained. However, there is little study involving their interaction, which invokes the following questions: will hybridisation have any impact on breeding system? Will hybridisation, as products of breeding systems, in turn, shape and facilitate the evolution of the breeding system itself? In other words, what are the interactions between breeding systems and hybridisation?

To understand the interactions between breeding system and hybridisation, this thesis focused on the model species, C. maritima and C. edentula, and explored the following questions: how does the hybridisation occur between the two species? What is the inheritance of the plant breeding system in their hybrids? What are the implications? Each question may involve several sub-questions. For example, the hybridisation processes may involve differences in reproductive barriers and crossing direction of hybridisation. These specific research questions will be raised after literature review.

Literature review In chapter 1, three major research components were discussed: floral development, breeding system, and progeny fitness. In this chapter, I will present a detailed literature review of each part in order to help understand the roles of breeding system in hybridisation and how these processes interact with each other and help the establishment of invasive species. Firstly, I will identify the research gaps within each part and raise relevant research questions, and finally I will propose the research hypotheses to be examined in this thesis.

2.1 Floral development Floral development (flower development) has been studied over the last century. Early studies about floral development were mainly descriptive research of a flower (such as Darwin 1877, Raunkiaer 1934) and its relevant traits (e.g., distyly and tristyly) (see detailed review in Ganders 1979). By late 1950s, researchers had identified the biology of these traits in flowers, figuring out their relationships with breeding system, in which these traits were involved with the incompatibility of certain flower types (Baker 1955, Harrison and Darby 1955, Lewis and Crowe 1958, Ganders 1979). After that, new techniques and methodologies enabled further studies into cellular and molecular levels (Smyth et al. 1990, Lawrence 2000, Brandvain and Haig 2005, Eckert et al. 2009, Austen and Weis 2015).

In this section, instead of following the timeline of the progress achieved in this area, I present a literature review based on the view of developmental processes of a flower to explore how the organs form, when the stigma will be receptive and pollen will be ripe, and the implications of the research gaps on understanding pollination and hybridisation.

2.1.1 Early development In a flowering plant, after transition from vegetative growth to reproduction, the first sign is the emergence of the flower (bud). A flower, as its modern definition, is the bisexual reproductive shoot of an angiosperm, in which the reproductive organs are surrounded by sterile organs (Glover 2014, p. 8). The transition from vegetative growth to reproduction needs induction, mainly by environmental variables, such as day length, temperature, and, to a lesser extent, water stress (Glover 2014). Once induced, the plant is on its way to flowering and involves changes in gene expression in the apical meristem, which were extensively studied in the late 1990s (Bowman et al. 1991, Weigel et al. 1992, Yanofsky 1995, Mena et al. 1996, Krizek and Fletcher 2005), especially for Arabidopsis (see detailed review in Yanofsky 1995) and crops such as rice (Izawa et al. 2000, Izawa et al. 2002), wheat and barley (Glover 2014, p. 81).

During this period, the ABC model for floral development was proposed by Coen and Meyerowitz (1991), where region A comprise whorls 1 (sepal) and 2 (petal); region B, whorls 2 and 3 (stamen); region C, whorls 3 and 4 (carpel). The action of several genes in these overlapping regions could give each whorl a unique combination of functions (Coen and Meyerowitz 1991). In whorl 1, only A function genes are expressed; while in whorl 2 and 3, AB and BC are expressed, respectively. However, in whorl 4, C function genes are solely expressed again. There are two key assumptions for this model: a) organ identity is a function of the ABC genes expressed in developing primordia instead of the position of each whorl.; b) the boundaries of A and C function genes are set by mutual antagonism (summerized by Glover

2014). Specifically, for example, if B function genes are lost through mutation, then the whorls expressed in a flower will be sepal, sepal, carpel, and carpel; if C function genes are lost, then the expressed whorls will be sepal, petal, petal, and sepal (Coen and Meyerowitz 1991).

After the proposal of this model, many studies were conducted to test and amend details of how genes work in this model, not only in Arabidopsis and Antirrhinum but also in Solanales. For example, studies verified that there are two A function genes although both play other roles in plant development (see detailed review in Chapter 10, Glover 2014). Then, two additional classes of homeotic genes, D and E function genes, were included into this model, in which D function genes are described and defined as the necessary genes for ovule development (Favaro et al. 2003) and even seed development in Solanales (Colombo et al. 1997) while E function genes are involved in all stages of early floral development, with the interaction between the ABC and E based on physical contact and complex formation (Mandel and Yanofsky 1998, Pelaz et al. 2000, Ditta et al. 2004). Later on, this ABC model and its related revised models have always been an important topic of floral development in gymnosperms, angiosperms, monocots and grasses (detailed review can be found in Chapter 11, Glover 2014).

Although the genetic control of floral development in many flowering plants has been the focus for decades and studies have revealed the structure and development of flowers from anthesis to seed set (reviewed in Bowman et al. 1989, and Smyth et al. 1990), there have been few reports about visualization of the intervening early stages of flower development (from bud emergence to flower opening) until Smyth et al. (1990). In his paper, Smyth et al. (1990) described the structure of the model plant, Arabidopsis thaliana, thoroughly with the help of SEM, from its first appearance as a buttress on the apical meristem until it opens as a fully developed flower, dividing the floral development into 12 continuous stages with landmark events. These stages and corresponding landmarks events could be modified according to different species.

Through visualizing and dividing early floral development into different stages, possible to describe exactly what the status of the flower organs is. The relative position of the stamens and stigma surface can give an idea whether there is any chance of contact between the two sex organs, which is important for assessing possible contamination of self-pollen before flowering. In a self-compatible (see details in Section 2.2.3) species, the relative position between the two sex organs is a vital factor for assessing the pollination process. This is because if the stigma surface is always higher than the stamens, it is impossible, although species is self-compatible, for the plant to pollinate itself before flowering. As a result, it could lead to outcrossing and even hybridisation.

Such morphologic studies (for example Smyth et al. 1990) are extremely rare in recent decades as most researchers have been focusing on the genes and genetic pathways controlling plant growth instead of plant growth itself (Mena et al. 1996, Izawa et al. 2000, Pelaz et al. 2000, Izawa et al. 2002, Favaro et al. 2003, Ditta et al. 2004, Krizek and Fletcher 2005, Takebayashi et al. 2006, Luo and Widmer 2013). In recent years, some studies started focusing on this topic and presented a few case studies, such as Iwamoto et al. (2015) and Nores et al. (2015), in which visualization has helped greatly understand the phylogenetic position in species as well as present new models for plant development. All these studies indicate that studying floral

10 development through visualization is of great value because it can a) provide simple but robust information about floral development before flowering, b) provide new insights into phylogeny and floral development in species, and c) thus potentially provide tools for assessing chance of self-pollen vs. alien pollen for pollination, especially in the scenario of hybridisation.

2.1.2 Stigma receptivity and pollen ripeness For self-pollen contamination, there is another scenario that should be taken into consideration: even though self-pollen has successfully landed on the stigma surface, there are still two prerequisites for the occurrence of self-pollination, i.e., stigma receptivity and pollen ripeness (and viability) (see details in Section 2.2.1). If a stigma is yet to be receptive or pollen has not ripened before contact with it, pollen contamination would not be a problem at all. In the natural world, there are various ways to avoid self-pollen contamination, either spatially or temporally, involving the plant breeding system, such as the incompatibility between different types of pollen derived from different forms of flowers (heteromorphy) within the same species (see details in Chapter 7 of Richards 1997).

The flower I have discussed so far is a “perfect” flower (hermaphrodite), containing both female and male reproductive organs, i.e., pistil and stamen, and producing female (ovule) and male gametes (pollen). Separation of sex organs in plants will now be discussed.

In the natural world, hermaphroditic plants have various strategies to adjust the timing of their floral development, separating stigmas and stamens spatially and reducing self-pollen contamination. This mechanism is termed as herkogamy. In a herkogamous flower, its stigma and stamens are positioned such that self-pollen cannot be passively transferred to the stigma surface unless it is transferred by pollinators (Glover 2014). The stigma normally has a higher position compared with the stamens; therefore, when self-pollen is transferred by foraging pollinators, the pollinators is likely to carry pollen from other individuals, providing a mixed load of self- and alien pollen and thus promoting outcrossing. The distance between the stigma surface and the top of the stamen, which is correlated with the degree of outcrossing, can be used as an indicator of the degree of herkogamy in a species (Takebayashi et al. 2006). However, herkogamy does not always result in any significant effect on the rate of outcrossing, although the degree varies. This is because it can only reduce the chance of self-pollen in the same flower pollinating itself but is unable to prevent pollinators depositing pollen from a different flower on the same plant (Lysak et al. 2005). Moreover, it is also dependent on the pollinator not brushing by the stigma when it leaves the flower (Medrano et al. 2005). Nevertheless, recent studies have suggested that even in predominant selfers, such as Arabidopsis, herkogamy still occurs in natural populations to reduce selfing and presumably allow enhanced outcrossing (Luo and Widmer 2013, Glover 2014), indicating that herkogamy is still a useful way to reduce self- pollen contamination and selfing.

Some hermaphroditic species have also evolved strategies to avoid contact temporally, i.e., dichogamy, an extremely common phenomenon, with 87% of angiosperm species showing some degree of dichogamy (Bertin and Newman 1993). Dichogamy is a slightly more reliable method of preventing autogamous pollination than herkogamy (Glover 2014) as it separates the stigma and stamens by the timing of stigma receptivity and anther dehiscence, which can avoid the possibility that pollinators accidently pollinate flowers when leaving. However, it also suffers

11 from another problem that herkogamy has: if pollinators forage primarily within an inflorescence, then the pollen they transfer to receptive stigmas will still be the self-pollen from different flowers on the same plant. To tackle this dilemma, some species has evolved synchronous dichogamy, in which all male organs mature at one point and females mature at another (Glover 2014). According to the timing of the maturity of reproductive organs, dichogamy can be divided into two different types, i.e., protandry and protogyny. In protandrous flowers, the anthers dehisce and the pollen is shed several days before the stigma is receptive; in contrast, in protogynous flowers, the stigma becomes receptive before the pollen ripens, which is less common (Glover 2014).

Besides, there are two strategies that can largely prevent the occurrence of autogamy, i.e., monoecy and dioecy, in which male and female organs are separated into different flowers or different plants and thus the timing of stigma receptivity and pollen ripeness are not involved. Monoecy is defined as the state that plants are hermaphroditic but only one sex organ occurs in each flower, completely preventing self-pollination within the same flower. Dioecious plants, in contrast, have a single sex of flowers on the same plant and thus completely solve the problem of self-pollination. It is reported that around 6% of angiosperm species show complete separation of sexes (Barrett 2002) and are scattered throughout the angiosperm phylogenetic tree.

All above strategies consider the situation after or at flowering while little attention has been drawn to the stages before flowering in hermaphroditic flowers. Studies about the timing and position of the stigma and stamens mainly focus on the stages after or at flowering but stages before flowering are of great value as well. For example, cleistogamy, mainly occurs in peas and grasses, is a type of autogamy in certain species that can self-pollinate before flowering (as a matter of fact, the flowers never open). Another example is Arabidopsis: it has small flowers and may usually be self-fertilised before buds open (Richards 1997). Therefore, it is very important to consider these stages which can provide critical information about stigma receptivity and anther dehiscence before flowers open, vital for assessing the chance of outcrossing and hybridisation, especially for selfers.

Overall, most studies of floral development have either focused on genetic aspects or events after flowering while less attention has been drawn to early flower development, its visualization as well as the development of stamen and stigma. As a matter of fact, combining these areas can provide new insights into floral development. For example, in the early floral development of a selfer, it is not only crucial to estimate whether the stamens have any chance to contact the stigma surface before flowering but also important to know the timing of stigma receptivity and pollen ripeness. This is because, even though stamens could reach the stigma surface, neither non-ripe pollen on receptive stigma nor ripe pollen on non-receptive stigma could cause any pollination.

In Cakile, C. edentula, as a selfer, when hybridising with C. maritima, as an outbreeder, fits the above scenarios very well. When considering hybridisation between these two species, the priority is to determine whether the selfer would be self-fertilised before flowering (i.e., there is no window of opportunity for alien pollen). Thus, there are two priorities that should be taken into consideration: observing the development of stamen and stigma (relative position) in early

12 stages and the timing of stigma receptivity and pollen ripeness. However, such topics have barely been studied previously. Therefore, more research is needed to determine how hybridisation between the two species can occur in nature.

2.2 Plant breeding system The plant breeding system (or plant mating system) is a reproductive strategy that a plant has evolved, spanning the reproduction processes from pollen arriving on the stigma to seed set, seed dispersal and seed germination etc. It includes pollen recognition and directly determines the success of particular genotypes in seeds. In its broad sense, all aspects of sex expression in plants that determine the relative genetic contributions to the next generation of individuals within species are included (Dafni 1992). In plants, there are two main breeding systems: sexual reproduction and asexual reproduction. Asexual reproduction produces new individuals without producing gametes by meiosis and subsequent gamete fusion, genetically identical to the parental plants unless mutations occur. Sexual reproduction produces offspring by the fusion of gametes, resulting in offspring genetically different from their parents, and has significant genetic advantages. During sexual reproduction, any harmful recessive mutant which is phenotypically expressed in the gametophyte generation (the pollen tube and the embryo-sac in flowering plants) is screened out by failure or gametophytic competition and does not persist through to the dominant sporophyte generation. The breeding system is a key property of natural populations because the variation in the pattern of mating is expected to have significant ecological and evolutionary consequences (Eckert et al. 2009).

In flowering plants, sexual reproduction is the predominant mating system and thus hereafter the plant breeding system in this thesis is the sexual reproductive one unless otherwise stated. As mentioned in Chapter 1, the plant breeding system is a key determinant of hybridisation; therefore, in this section, the plant breeding system, mechanisms underlying as well as research gaps will be introduced.

2.2.1 Pollen post-arrival processes Normally, when landing onto the stigma, successful pollen with a particular genotype simply goes through hydration, germination, and pollen tube growth and then fertilises the ovule. However, there are two prerequisites: the pollen is viable and the stigma is receptive. Pollen longevity varies between different species, genus and orders (Dafni and Firmage 2000, Willmer 2011). In some extreme cases, pollen may even have no viability, e.g., hybrid pollen with particular genotypes (Runquist et al. 2014). It is likely, however, that there will be variation in viability and longevity amongst genotypes, even if the genotype is fertile. The period of stigma receptivity is also limited and varies from just a few minutes in some grasses to around three weeks in some orchids (Willmer 2011). Imported pollen must arrive before the stigma is no longer receptive. For allogamous pollen landing on the stigma of a selfing plant, it is possible that there is a narrow time window between flowering and selfing to fertilise the ovule and create a hybrid. However, little attention has been given to this time constraint in the formation and frequency of hybrids.

When these two prerequisites are satisfied, pollen can then absorb water from the stigma. The hydration processes vary between different types of stigma. There are two basic types of stigma recognised, termed as “wet” and “dry”, which can be differentiated as per Table 2.1 (Richards

1997, Willmer 2011). Pollen on a wet stigma goes through relatively simple phases. Wet stigmas have a gappy surface and form many papillae. They produce a stigmatic exudate which is relatively free of protein and not selective for pollen adhesion. Thus, arriving pollen can successfully hydrate. After hydration, the pollen tube grows down the outside of the stigma papillae and penetrates the stigmatic tissues at the base of the papillae, dissolving the cell-wall middle layer of pectate (Willmer 2011). In contrast, pollen on a dry stigma experiences more complicated processes. The surface of a dry stigma is continuous instead of gappy and has a thin layer of exudates (therefore “dry”). Unlike a wet stigma where pollen hydration happens on the surface (Willmer 2011), pollen on a dry stigma adheres to the stigma surface and then absorbs water from the stigmatic cells (Richards 1997). After hydration, the pollen tube penetrates the cuticle of the stigma and then grows through the cells (Richards 1997).

Table 2.1 Two main types of stigma, revised from (Richards 1997).

Stigma type Stigmatic exudate Pollen cuticle Pollen hydration Pollen entry Wet Present Gappy External Intercellular Dry Absent Continuous Internal Intracellular

Within the style is a “pollen tube transmitting tract”, whose length varies between taxa; the pollen tube continues to grow in the tract after its penetration. In most plants, the growth period of pollen in the tract lasts between 12 and 48 h from pollen germination to ovule fertilisation, not only because the tract length may vary but also due to the properties of particular species (Richards 1997, Willmer 2011). Upon the arrival of the pollen tube to the ovary, the sperm cell in the pollen is released, initiating a series of events that lead to fertilisation (Willmer 2011; Chapter 3).

If all goes well, the normal pollen germination will end up with a successful fertilisation; however, things may not go so smoothly under natural conditions. In wet stigmas, pollen can become hydrated when landing on the stigma surface. However, if the pollen is from a species of another genus or family, despite being able to absorb water and germinate, it may not grow through the transmitting tract simply because the sizes of the pollen tube and tract do not fit. In some cases, although the pollen tube can grow into the tract, it cannot reach the ovule. In dry stigmas, if the pollen landing on the stigma is not conspecific, it may not be retained as pollen hydration cannot occur. Through these mechanisms, plants can ensure that legitimate pollen germinates and fertilises the ovule, which contributes to the spread of their genes.

2.2.2 Self-incompatibility It is believed that the original flowering plants were all hermaphrodite, i.e., they had stamens (androecium) as well as pistils (gynoecium) (Richards 1997, Charlesworth 2006, Goodwillie et al. 2010). As a result, after pollination, it is possible for self-pollen to fertilise the ovule, i.e., self- fertilisation. As mentioned in the last section, once pollen, either self or alien, is deposited on the stigma surface, the normal fertilisation processes will be initiated. However, pollination and fertilisation processes could be far more complex than it was originally thought.

There is an interesting phenomenon that not all co-sexual plants can be self-fertilised. Many plants tend to recognise self-pollen as illegitimate and thus prevent self-fertilisation, termed as “self-incompatibility” (SI). SI is defined as a genetically determined post-mating but pre-zygotic

14 barrier to fertilisation by self- or self-related pollen that eliminates any risk of inbreeding and therefore optimizes the potential for outbreeding afforded by pollinators (Hiscock and McInnis 2003). SI has important significances for plant species. It prevents self-fertilisation, thus avoiding inbreeding depression. Meanwhile, through promoting outcrossing, it can increase genetic diversity and allow adaptation for changed conditions (Willmer 2011; Chapter 3). Therefore, it is of great value to understand SI and how it works.

2.2.2.1 SSI and GSI SI has been appreciated for a long time, since Darwin mentioned it as “illegitimate” and “legitimate” mating between different forms of flowers on the same species (Darwin 1877, cited in Hiscock and McInnis 2003). At least 100 flowering plant families have been reported to contain SI species. This may be a conservative estimate because of limited appropriate studies (Igic et al. 2008). There are two types of SI, i.e., heteromorphic SI, e.g. “pin flowers” and “thrum flowers” (described by Darwin 1877), and homomorphic SI where there is no such difference in flower morphology (see detailed review in Charlesworth 2006). Another major difference between these two systems, at the genetic level, is that homomorphic SI has single-locus but multi-allelic SI genes while heteromorphic SI has only diallelic SI genes (De Nettancourt 1977, 1997, Richards 1997, Hiscock and McInnis 2003). Although heteromorphic SI is scattered in 24 families, it does not occur in the Brassicaceae family (Ganders 1979, Barrett and Harder 1996, Hiscock and McInnis 2003, Charlesworth 2006). Therefore, homomorphic SI will be the focal topic in the following discussion and SI will be the homomorphic one if not specified.

Classical genetic studies have identified two distinct genetic forms of SI: gametophytic self- incompatibility (GSI) and sporophytic self-incompatibility (SSI). In the GSI system, the incompatibility phenotype of the pollen is determined by its own haploid (gametophyte) genome; in contrast, the incompatibility phenotype of the pollen in SSI is determined by the diploid genome of the plant (sporophyte) (Hiscock and McInnis 2003). Major differences between these two forms of SI are listed in Table 2.2 (De Nettancourt 1997, Hiscock and McInnis 2003).

Table 2.2 Major differences between GSI and SSI, revised from De Nettancourt (1997).

Form Pollen phenotype determined by Stigmatic type Site Dominance SSI Haploid genome Dry Stigma Co-dominance GSI Diploid genome Wet Style Complex dominance

One major difference between GSI and SSI is the rejection site of pollen. As mentioned in Table 2.1, pollen deposited on a stigma has different hydration strategies, depending on the stigmatic type. In SSI species, the secretion of factors for proper rehydration and germination is prevented. In some cases, even the adhesion of pollen could be inhibited, leading to the rejection of incompatible pollen. In contrast, pollen deposited on the wet stigma of GSI species, regardless of its compatibility, could absorb water and nutrients from the exudates and thus grow into style.

Another significant difference between GSI and SSI is the dominance of alleles. Most SI systems typically have simple genetic control based on a single multi-allelic locus (S-locus). The S-locus usually consists of one pistil-expressed S gene (female determinant) and one pollen-expressed

S gene (male determinant), forming a non-recombining S haplotype (S allele) (Brennan et al.

2011b). For example, in the GSI system, if the maternal plant’s genotype is S1S2, and the genotype of paternal plant is (Figure 2.1):

• a) S1S2, then no progeny can be produced;

• b) S1S3, then S3 pollen can fulfil fertilisation and produce the progeny with the genotypes

of S1S3 and S2S3;

• c) S3S4, then the genotypes of the progeny will be S1S2, S1S3, S2S3 and S2S4.

However, in an SSI system, the outcomes would be distinctive due to a complex dominance hierarchy. Here, the dominance of the four alleles is assumed as S1 > S2 > S3 > S4. If the genotype of the maternal plant is S1S2, then the corresponding results are:

• a) no progeny will be produced when the paternal genotype is S1S2;

• b) no progeny will be produced when the paternal genotype is S1S3;

• c) progeny will be produced with the genotypes of S1S3, S1S4, S2S3, and S2S4, when the

paternal genotype is S3S4.

If an extra paternal genotype S2S4 is considered, there will be another possibility:

• d) progeny will be produced with the genotypes of S1S4, S2S2 and S2S2 (see details in Figure 2.1).

It should be noted that it is possible to have homozygotes within populations of an SSI species, while it is theoretically impossible to observe this phenomenon in the GSI system (see possible outcomes of progeny of both SI systems in Figure 2.1). In the examples of the SSI system given above, the dominance hierarchy is very simple; however, it could be much more complicated in the real world, such as independent, co-dominant, dominant and combinations of different dominance hierarchies in the pollen and style. In contrast, in the GSI system there is no such a complicated dominance hierarchy (Richards 1997, Schierup et al. 1997, Billiard et al. 2007).

Although the dominance hierarchy plays an important role in the SSI system, relevant studies are very limited. Previous studies have revealed that two major dominance classes are in the Brassicaceae family, with class I alleles being dominant over class II alleles in the pollen and co- dominant in the stigma (Nasrallah et al. 1991). Although progress has been made, it is still obscure in other SI systems; even in the Brassicaceae, further research is still required (see details and review in Brennan et al. 2011b, Brennan et al. 2013). For example, in the genus Cakile, no single published study has tried to define the dominance of the S-alleles. In addition, in other families with the SSI system, only limited and scattered studies have focused on the dominance hierarchy of S-alleles (e.g., Stevens and Kay 1989, Kowyama et al. 1994).

GSI A B C SSI A B C D

POLLEN POLLEN PARENT S1S2 S1S3 S3S4 PARENT S1S2 S1S3 S3S4 S2S4

POLLEN S1 S2 S1 S3 S3 S4 POLLEN S1 S2 S1 S3 S3 S4 S2 S4

OVULE OVULE PARENT S1 S2 S1 S2 S1 S2 PARENT S1 S2 S1 S2 S1 S2 S1 S2

S1S2 S1S3 S1S2 S1S3 S1S4 S2S2 PROGENY PROGENY NONE S1S3 S2S3 NONE NONE S2S3 S2S4 S2S3 S2S4 S2S4

Figure 2.1 A conceptual model for the consequences of a GSI (left) and SSI (right) system on the viability of pollen and the possible outcome of progeny. Red represents failed pollen while green represents successful pollen; brown-coloured progeny indicates a homozygote. The dominance hierarchy of SSI in the figure is: S1 > S2 > S3 > S4. It can be inferred from Figure 2.1 that the less frequent the S allele is in a population, the higher the chance of successful mating would be. This is a negative frequency-dependent balancing selection in the SSI system (Brennan et al. 2013), termed as an S-Allee effect (Levin et al. 2009, Busch et al. 2014) (see details about Allee effect in Section 2.2.3). Therefore, SSI populations should be more vulnerable to extinction because they are more likely to experience an S-Allee effect, where the population growth rate may become negative (i.e. population size declines) when fewer alleles are available at the S-locus, limiting successful offspring production (Levin et al. 2009, Busch et al. 2014).

As for data on the number of S-alleles in a population, most estimates on the number of alleles present in natural populations are for the GSI system (Mable et al. 2003). Estimates of the number of alleles in species with SSI are few and less precise than those with GSI because of the complication of dominance of S-alleles. About 13 to 100 S-alleles have been reported in the GSI system. For the SSI system, the reported number of alleles ranges from 10 to 49 in local populations, while estimates and reports of the species-wide number of alleles range from 2 to 60 (Schierup et al. 1997, Lawrence 2000, Brennan et al. 2006, Busch et al. 2014). Recent studies revealed that populations with the SSI system should harbour less S-allele diversity than populations with the GSI system (Levin et al. 2009).

2.2.2.2 Genetic basis of GSI and SSI Within a given angiosperm family, homomorphic SI is always of one type (Hiscock and McInnis 2003), e.g. the Brassicaceae family is exclusively SSI (therefore, Cakile is SSI). Previous studies have revealed the molecular basis of SSI and GSI in two models, i.e., a) the SSI in the Brassicaceae family and b) the SRNase-mediate GSI, mainly in the Solanaceae but also in the Rosaceae.

Theoretically, in the SSI system, different genes regulate the expression of SI in the pollen and pistil. Therefore, a minimum of two genes are predicted to be present at an S-locus. However, in reality, only the SSI system in the Brassicaceae family has been clearly identified with both pollen and stigma S-specificity determinants so far (Hiscock and McInnis 2003). Therefore, the best model to use as an illustration is the SSI system.

Two stigma expressed genes have been identified at the S-locus, i.e., S-locus Glycoprotein (SLG) and S Receptor Kinase (SRK); while the pollen determinant of SSI is a small cysteine-rich peptide located in the pollen coat, named S Cysteine Rich protein (SCR) or SP11 (S Pollen 11). An interaction occurs between these three genes: SLG enhances the S-haplotype-specific rejection

(SRK) of pollen (SCR) but it is a non-essential component of the SI rejection process. To illustrate the whole process, a stigma genotype S1S2 is used as an example (S allele hierarchy S1 > S2; thus, pollen haplotype is S1). The pollen protein SCR1 is recognized by and binds to the extracellular receptor of SRK1, which induces the dimerization of SRK1 and autophosphorylation on serine and threonine residues in the kinase domain. Activation of the SRK initiates further intracellular signalling cascade within the stigma papilla cell and therefore prevents the germination of pollen. SLG is not involved in these processes and just acts as a reinforcement to them (see review in Hiscock and McInnis 2003, Hiscock and Allen 2008).

The GSI system, mainly in Solanaceae, Rosaceae and Scrophulariaceae, is the most common system in plant species. It is characterised by the female determinant of GSI, a secreted ribonuclease (S-RNase) that accumulates to a very high concentration in the transmitting tract of the stigma and style (McClure et al. 1989, cited in Hiscock and McInnis 2003). If the pollen arriving on the stigma is incompatible, although the pollen tube can grow into the style, the S- RNase in the style will enter the pollen tube and degrade ribosomal RNA and mRNA, causing the termination of pollen tube growth. If the pollen is compatible, then the S-RNase will not inhibit the growth of pollen tube, resulting in fertilising the ovule and generating seeds.

As an example of the GSI system, the following genotypes are used: S1S2 as the female and S1S3 as the male. The pollen haplotype will be either S1 or S3, while the secreted S-RNases will be S1 and S2. The GSI response occurs later than the SSI response and the pollen tube will grow into the style where the inhibition occurs. Here are the two situations:

• a) pollen S1 successfully germinates, enters the style and absorbs the S-RNases (S1 and

S2) but ends up with pollen tube growth failure. The S-RNase (S1) has the same genotype

of the pollen tube (S1) and therefore will bind together with an S-specific RNase

interactor to prevent the interaction between S-RNase (S1) and general RNase-inhibitor

secreted by S1. Thus S-RNases retain the enzymic activity to fulfil the prevention of further pollen tube growth (see details in Luu et al. 2000). It should be noted that in Papaver, unlike the S-RNase mediated GSI, pollen tube arrest occurs at the stigma surface and is mediated by the Ca2+-based signalling system, which is induced by the exposure of pollen to a stigmatic S-protein of the same haplotype (McCubbin and Kao 2000, cited in Hiscock and McInnis 2003);

• b) pollen S3 lands on the stigma, germinates and grows. The pollen tube will grow into

the style and the S-RNases (S1 and S2) will be taken up non-specifically by the pollen tube.

However, the S-RNases (S1 and S2) will be inhibited by the general RNase-inhibitor

secreted by S3 and thus the inhibition of pollen tube by S-RNases (S1 and S2) will not happen.

Recent studies have revealed that many specified SI systems exist in other species (see detailed review in Hiscock and McInnis 2003), such as the SSI in the Asteraceae which was thought to be the same as that of the Brassicaceae but turned out to be different (Allen et al. 2011). These systems need further investigations, especially in the aspect of their molecular mechanisms.

2.2.3 Self-compatibility Under natural conditions, many flowering plant species indeed can self-fertilise. Compared with self-incompatible species, self-compatible ones have many advantages as well as disadvantages.

Self-compatibility also has a close relationship to self-incompatibility (Donohue 1998, Hiscock and McInnis 2003, Willi and Fischer 2005, Charlesworth 2006, Cartwright 2009, Willi 2009, Dufay and Billard 2012, Murren and Dudash 2012, Dart and Eckert 2013). Given the existence of self- incompatibility, why does self-compatibility still occur? What are the advantages and disadvantages? In this section, these questions will be discussed.

Self-compatibility, the counterpart of self-incompatibility, refers to the ability in some plants to be self-fertilised by their own pollen without any imported pollen. It is quite common and sometimes the norm because of its great importance to many plants. The main advantages can be listed as:

• a) it increases the probability of seed production, especially in isolated habitats or at an invasion front by overcoming the limitation of pollinators, mates and consequences of the Allee effect. The Allee effect, introduced by W.C. Allee (Allee 1931, cited in Stephens et al. 1999), is defined as a relationship between any components of individual fitness and either the number or density of conspecifics (Levin et al. 2009) that results in the population growth rate becoming negative at low population abundance (see details in Stephens et al. 1999, Rodger et al. 2013). When populations are small, plants with self- incompatibility will experience an enhanced Allee effect since more plants share the same S-alleles and fewer of the plants will be able to fertilise each other (S-Allee effect) (Levin et al. 2009, Busch et al. 2014). However, self-compatible plants will not be limited by the availability of pollinators and mates (Cheptou 2004, Rodger et al. 2013), especially in patchy habitats or at an invasion front (Cheptou 2004, Wagenius et al. 2007, Levin et al. 2009, Rodger et al. 2013, Busch et al. 2014). This advantage can be best explained by Baker’s rule (Baker 1955, 1967) that species with self-compatibility will tend to be prevalent in successfully established species after long-distance dispersal. • b) it provides the reliability of a fall-back or fail-safe mechanism (often termed reproductive assurance) when crossing has not occurred, especially for annual plants with short lives and only one flowering season (Willmer 2011). • c) it is cheaper and quicker than self-incompatibility (Willmer 2011). Most habitually self-pollinating plants have flowers that are smaller and fewer in number than their close relatives that require cross-pollination. Selfing plants clearly do not need the expensive investment in flower colour, size, and number to put on an attractive floral display.

However, self-compatible plants will also suffer negatively from selfing because of:

• a) inbreeding depression (ID). Inbreeding depression is defined as the inferiority of inbred progeny compared to outbred progeny primarily due to the expression of deleterious recessive mutations in homozygous inbred progeny (Eckert et al. 2010). Inbreeding, on one hand, eases limitations of pollinators and mates; but on the other hand, it increases the possibility of homozygosity. This can produce inbreeding depression for two possible reasons (Dudash and Carr 1998, Charlesworth and Willis 2009). First, deleterious recessive or partially recessive alleles that are masked at heterozygous loci by dominant alleles become fully expressed in homozygotes; second, alleles may interact in an over-dominant manner, such that the fitness of either type of

homozygote is lower than that of heterozygotes. These two mechanisms will produce long-term effects in populations experiencing an increased level of inbreeding. Some studies have indicated that inbreeding depression can be reduced or even removed completely with increased self-fertilisation among populations (Dudash and Carr 1998, Charlesworth and Willis 2009, Dart and Eckert 2013); however, the inbreeding depression produce by overdominance will resist and cannot be removed in these populations without lowering the mean fitness of the target populations (Dudash and Carr 1998). It is a good explanation for the existence of inbreeding depression in some species after long-term inbreeding, e.g., Cakile edentula (Donohue 1998). • b) no variation that allows adaptation to changed conditions. The lack of variation can also, of course, be a major and potentially lethal disadvantage if the habitat changes. Self-compatible plants always have a fixed genotype and phenotype. This could presumably become an additional handicap over evolutionary time, with separate subpopulations eventually so different that they cannot interbreed and restore variation (Willmer 2011).

Nevertheless, selfing could also contribute significantly to the evolution of plant species due to the coadaptation of the gene complexes. The concept of coadaptation of genes was firstly proposed by Dobzhansky and defined as genes coadapted if high fitness depends upon specific interactions between them (Wallace 1968, cited by Ohta 1980). Such coadaptation of genes can contribute to the adaptation of populations to unique environments and rapid colonisation of those environments (Allard et al. 1972, Clegg et al. 1972), which can be well maintained by selfing. However, outcrossing could break down such interactions between the coadapted genes, leading to the collapse of the coadapted gene complexes (Matioli and Templeton 1999).

Although SC is quite different from SI, they have a close relationship. Studies on the phylogenetic analyses of S-RNases suggest that SI was the ancestral state in the majority of dicots and thus SC plants derived secondarily from SI species (Igic and Kohn 2001, Steinbachs and Holsinger 2002). Meanwhile, it is clear that the SI in species could break down and thus become SC (Vogler et al. 1999). For example, bud emasculation could also break down the SI, possibly due to the immaturity of the stigma (Hiscock and Dickinson 1993).Another good example is U’s triangle, where the breakdown of SI is due to allopolyploidy occurring during hybridisation (see details in Section 2.2.6), indicating that interspecific hybridisation could also disrupt SI.

Recent studies have indicated some novel mechanisms underlying the breakdown of SI (Brennan et al. 2011b, Dwyer et al. 2013). Intriguingly, instead of causing irreversible changes between SI and SC, these mechanisms allow species to regain SI after the breakdown under certain circumstances. As a consequence, they may act as a bridge between species, promoting interspecific hybridisation, introgression and even contributing to the survival against extreme conditions (see details in Nasrallah et al. 2007).

2.2.4 The SI × SC rule It is intriguing to consider what will happen when closely related SI species encounter SC species and co-exist in the same patch. Is it possible for the two species to hybridise and how does this happen? In this section, the possibility of hybridisation between such species will be discussed.

Obviously and inevitably, when two species co-occur in the same habitat, there will be pollen flow between the two species by pollinators. It is possible that hybridisation could occur and hybrids may form, if there are no reproductive barriers between the two species. However, there are always barriers in the real world, either pre-zygotic or post-zygotic or even both (Tiffin et al. 2001, Bushell et al. 2003, Brandvain and Haig 2005, Lowry et al. 2008). It has been suggested that a pre-zygotic barrier exists between the pollen of SC species and the pistil of their SI relatives, resulting in the rejection of the pollen from SC species on the pistil of its SI relatives (Lewis and Crowe 1958, reviewed in De Nettancourt 1997).

The SI × SC rule was firstly proposed by Harrison and Darby (1955), which illustrates that the pollen from SC species will be rejected by the stigma of SI species but not vice versa, and is a part of unilateral incompatibility (UI) (Lewis and Crowe 1958). Unilateral incompatibility occurs when the pollen of one species is rejected by the pistil of another. There appear to be many exceptions, e.g., in the Brassicaceae (Hiscock and Dickinson 1993), and thus the rule seems to be no more than a general tendency rather than a universal law. UI can be demonstrated based on three crosses between two close relatives: SI × SI, SI × SC and SC × SC. Specifically, SC× SC crosses are usually compatible; SI x SC crosses show unilateral incompatibility (usually incompatible when the SI species is the female and the SC species is the male), while SI x SI crosses are often incompatible or unilaterally incompatible. Studies have revealed that S-alleles may be involved in UI (reviewed in Hiscock and Dickinson 1993, De Nettancourt 1997).

The SI × SC rule concerns pollination in the current generation, but there is little information on the subsequent generations: in crosses between SI and SC species, which breeding system will be inherited (which is dominant)? Previous studies observed that in SI x SI crosses some hybrids can be produced (Hiscock and Dickinson 1993, Brennan et al. 2013) and inherit the SI system from their parents (Brennan et al. 2013). However, in SI × SC crosses, information on the subsequent generations (e.g., inheritance in hybrids) is still rare (Martin 1961, Nasrallah et al. 2007).

C. maritima (SI species) and C. edentula (SC species) are two closely related congeners, known to hybridise artificially (Rodman 1976) as well as in the field (Ohadi et al. 2015). They are thus good examples to observe (a) whether the SI x SC rule applies to Cakile species and (b) the inheritance of breeding systems in subsequent generations.

2.2.5 Partial and cryptic self-incompatibility When hybridisation occurs between two incipient species, there is another possibility that should be taken into consideration: will there be any discrimination (difference) of pollen performance between the two directions of crossing and other crossings with produced progeny? Here, I introduced two additional SI systems in this section: Cryptic Self Incompatibility (CSI) and partial SI (PSI), which are all involved in stylar discrimination and thus introduced in this section.

CSI was firstly introduced in Bateman (1956), defined as a phenomenon that self-pollination results in full seed set when only self-pollen is available but the success of self-pollen is strongly reduced when it competes with outcross pollen (Chapter 5, Korbecka 2004), and is based on the differences between pollen tube growth rates. It is a mechanism that results in subtle

21 discrimination between self- and outcross pollen, which is only evident after pollination with mixtures of different types of pollen (Cruzan and Barrett 2015).

Since the study of Bateman (1956), the concept of CSI, although not so intensively, has been studied for decades, and is proposed as a mechanism accounting for the low selfing rates in some fully self-compatible species (Galloway et al. 2003). CSI can also be regarded as a functionality analogous to delayed selfing, one of the mechanisms of reproductive assurance (see details in Busch and Delph 2012, Brys et al. 2016), as it not only allows plants to preferentially outcross, but also reproduce through selfing when pollinator service is low (Cruzan and Barrett 2015).

There are two ways to estimate the degree of CSI. Traditionally, CSI have always been tested by applying equal proportions of mixed pollen types onto the stigma surface, performing paternity analysis of the produced offspring, and then comparing results with those of single donor experiments (Bateman 1956, Eckert and Allen 1997). As suggested by the definition of CSI, pollen tube growth observation can also be used as a way of detecting CSI, which is more frequently used and combined with paternity analysis method in recent studies (Korbecka 2004, Figueroa- Castro and Holtsford 2009, Cruzan and Barrett 2015).

When the stylar discrimination between self and outcross pollen is based on pollen-tube growth rate, this bias towards outcross pollen has the potential to produce a ﬂexible mating system (i.e., CSI) that may be adaptive when the supply of outcross pollen is unpredictable (Cruzan and Barrett 2015), thus acting as a reproductive assurance. In other words, when outcross pollen is sufficient, outcrossing is the predominant mating strategy; in contrast, self-fertilisation is the primary breeding system when lacking alien pollen. CSI based on the differences of pollen tube growth would be of great advantage when plants are subjected to harsh or various environments where pollinator service or pollen is limited. As a result, CSI largely depends on the stigmatic pollen-load capacity as well as stylar-discrimination ability (Cruzan and Barrett 2015). The differences between these two factors in different plants may reflect a trade-off selection between increased outcrossing and greater reproductive assurance (Cruzan and Barrett 2015).

In contrast with CSI, there is another possibility when the stylar discrimination happens: instead of elongating at a slow rate, pollen tube growth is ceased in the style, i.e., PSI. Partial self- incompatibility is a stylar discrimination based on pollen tube attrition (i.e., failure of pollen tubes due to the cessation of growth before fertilisation (Cruzan 1989), resulting in partial or complete SI, which does not provide a ﬂexible mating system that favours outcross pollen when it is more abundant. Some researchers (e.g., Korbecka 2004) included pollen tube attrition based SI in CSI which is wrong in my opinion. This is because, according to its definition, CSI only occurs after mixture of different types of pollen are deposited, while PSI could occur when the pollen load is single or mixed. Thus, PSI should be an independent concept instead of part of CSI. As a matter of fact, most related studies of pollen tube attrition used comparisons of single pollen donor experiments of self and outcross pollen to assess degree of PSI (Cruzan 1989, Plitmann 1993). In recent decades (see detailed review in Swanson et al. 2016), numerous pollen tubes, seemly compatible, have been detected in the stigma but few reached the ovules, as mentioned in Plitmann (1993).

Partial SI is featured by the pollen tube attrition in style, which was firstly defined by Cruzan (1989); however, not all attrition is due to genetic-controlled incompatibility. Instead, various reasons could lead to it, including genetic and environmental factors (Cruzan 1989, Plitmann 1993). Cruzan (1989) in his study mentioned that the genetic mechanisms of attrition in some species may be the same as that of self-incompatibility, especially for partial- or pseudo- incompatibility. Plitmann (1993) also showed that species with the SSI system had much higher attrition rates compared with inbreeders in Brassicaceae. Alternatively, pollen tube attrition may simply depend on the maternal or paternal identities and the environments of the plants involved, and is even independent of the identity of the pollen (Plitmann 1993, Cruzan and Barrett 2015, Mazer et al. 2016, Swanson et al. 2016). For example, Swanson et al. (2016) indicated an association between the number of pollen and the degree of attrition. Cruzan and Barrett (2015) also demonstrated the associations between the amount of pollen loaded on the style and the attrition. These associations, as suggested by Swanson et al. (2016), could be due to antagonistic interactions or stylar constriction, a physical limitation on the width of the transmitting tissue in the style. Alternatively, there could be limitations on the nutrient supplies available to the pollen during heterotrophic growth, resulting in later pollen failing.

However, little study has focused on CSI/PSI in species with the SSI system, not to mention in hybridisation of two species with contrasting breeding systems (i.e., SI and SC species) (but see Plitmann 1993). Studies of Clarkia and other species in different families (Cruzan 1989, Yacine and Bouras 1997, Travers and Mazer 2000, Vassiliadis et al. 2000, Castro et al. 2009) are all conducted in self-compatible species although many species in this family are outcrossers. Such studies, as presented in Plitmann (1993), in which the incompatibility system either presents as GSI or SSI, did not exist at all.

Most importantly, CSI and PSI, by definition, operate within the style instead of at the stigma surface while in SSI system pollen rejection happens at the stigma surface, which makes it uncertain whether CSI and PSI could be detected in SSI species. Specifically, Partial SI or pollen tube attrition mainly happens in the style; therefore, partial SI may not exist in the SSI system as the recognition of SSI occurs at the stigma surface instead of in the style. Plitmann (1993) had done experiments on tens of species in Brassicaceae, in which 17 species are SSI, and detected stronger attrition in those SI species compared with inbreeders. However, since then, although many studies have been conducted on pollen tube growth rate and attrition but little studies have been done in this specific area.

Overall, as suggested by the results in Plitmann (1993), both CSI and PSI could be observed in SSI species when pollinated with compatible pollen types but relevant studies are rare. Therefore, this area is relatively unknown and further studies are needed. Cakile species could be a perfect target in which pollen tube attrition may be observed to some extent.

2.2.6 Inheritance of breeding system The plant breeding system is a well-studied aspect of plant evolution that has been reviewed in above sections, in which most post-pollen-arrival landmark events were introduced. If all barriers were overcome, it is interesting and of great value to explore that what the breeding system would be in the next few generations (i.e., inheritance of breeding system). In Section 2.2.2, I have mentioned the inheritance of SI within the same mating system. However, what

23 the scenario would be if the crossing happens between different mating systems (for example, SI × SC)?

In fact, the inheritance of plant breeding system, especially upon hybridisation, did not draw much attention until mid-20th century. As mentioned above (see details in Section 2.1), floral mechanisms produce a wide range of mating systems from obligate outcrossing to predominant selfing with flexibility between different systems (Goodwillie and Ness 2013). Most studies have focused on its evolutionary significances but failed to address the inheritance. It had not drawn any attention until Mather (1943) published his paper on the dominance of self-incompatible in hybrids of an SC and SI species, in which self-incompatibility was dominant but weaken in the backcrossed hybrids. Later on, as mentioned in Section 2.2.4, Lewis and Crowe (1958) had summarized the SI × SC rule but did not do any further investigation on the inheritance of breeding system in hybrids. After that, Martin (1961) did some further genetic investigation on the breeding system of the parental species, F1, F2, and even backcrossed generations.

Although the above-mentioned studies provided insights into plant breeding systems, follow-on studies mainly focused on the rationales behind hybridisation of species instead of the inheritance of breeding system. It is mentioned in previous sections that when closely related species co-occur there is a chance for them to hybridise (see details in Sections 1.2 and 1.3). it is found that breeding systems can influence the rate of hybridisation in a number of ways, such as promoting intraspecific cross-fertilisation and hybridisation (De Nettancourt 1997, 2001), favouring particular crossing directions as well as inhibiting certain hybridisations, e.g., SI × SC rule (reviwed in Goodwillie and Ness 2013).

In recent years, studies on the inheritance of breeding system in hybridisation have achieved some progress. Although the Dobzhansky-Muller model (see detailed reveiw in Lindtke and Buerkle 2015) predicted that hybrids are always sterile and even inviable, the origin of many plant species can be traced back to sexual hybridisation between more or less diverged species (Rieseberg et al. 1999, Baack and Rieseberg 2007), especially between closely related species that have similar genomes and chromosome complements (Rieseberg 2006, Baack and Rieseberg 2007). These F1 hybrids are not sterile at all; instead, they can reproduce normally and generate fertile recombinant progeny.

As a matter of fact, in some cases, when interspecific hybridisation involves self-incompatible species, the generation of self-fertile hybrids is normally dependent on the breakdown of SI (Lindtke and Buerkle 2015, Christe et al. 2016), which means that SI is lost in hybrids. Nasrallah et al. (2007) did a thorough study on the inheritance of SI in hybrids and the mechanisms of the breakdown, in which he hybridised an SI and SC species (Arabidopsis species), checked the loss of SI in F1 and F2 hybrids and found out that the breakdown of SI was due to the function loss in the stigma as well as in the pollen parts. The loss of SI allowed the establishment of stable self-fertile hybrid genotypes, promoting their reproduction and even speciation. Moreover, unlike mutation or deletion of SI causing irreversible switches from SI to SC, these losses are epigenetic and reversible. This reversibility may further promote the success of hybrids. Another study about the evolution of self-fertility in Spartina hybrids also revealed the loss of SI promoted hybrids establishment (Sloop et al. 2009). In contrast, some hybrids inherited SI from

24 their parental species. For example, a study of two closely related SSI species in Adrian et al. (2013) demonstrated that SSI is inherited and functions normally in hybrids.

Previous studies have also suggested that the hybridisation between species with different breeding systems and genomes, especially in the Brassica species, e.g., “U’s triangle”, could also contribute to the breakdown of SI in the hybrids. Three ancestral Brassica species hybridised with each other and created three of the common modern vegetables and oilseed crop species (U 1935, cited in Janick 2008). In this process, the three diploid parental species were mostly SI while the resulting three allotetraploid species were SC, which clearly indicates that interspecific hybridisation caused the breakdown of SI in the produced species (Chen et al. 2011, Cheng et al. 2014)

Overall, the inheritance of SI upon hybridisation is of great interest and value: it can promote our understanding of the interaction between hybridisation and plant breeding systems, which will, in turn, help us get deeper insights into plant species. However, relevant studies are relatively rare, especially compared with the research into the mechanisms of self- incompatibility, which has been largely revealed.

2.3 Progeny fitness In previous sections, I have reviewed processes of hybridisation, from floral development (how it happens) to the pollen tube growth (breeding system controlled); however, there is another important part involved: progeny fitness. After successful pollination and fertilisation, ovules start their further development in which seeds, if not abandoned, are formed, germinate and then grow into another life cycle. Processes involved after seed formation add up to the total fitness of an individual. Theoretically, fitness has both evolutionary significance, as an individual’s contribution to a population’s subsequent genetic composition, and ecological significance, as the numerical contribution to a population’s growth (Shaw et al. 2008).

Plant fitness is essentially the lifetime reproductive output of an individual, subsuming all the processes dealt with above. It is common, however, for researchers to estimate fitness in experiments by measuring the plant biomass at an arbitrary point in time, or seed production (Pedersen et al. 2007). These may be correlated with fitness but they are not valid measures. Therefore, in this section, fitness is introduced from both the population’s and individual’s point of view, including three main parts: seed traits, individual growth, and reproductive attractiveness and output.

2.3.1 Seed traits Seed development, after fertilisation, is a complex and highly coordinated process that involves the integration of many genetic, physiological, metabolic, and signalling pathways, affected by endogenous and environmental signals and stimuli. The development of the seed can be viewed as a discontinuous, stepwise process where several different phases occur in succession to ensure the formation of a functional reproductive unit. Generally, in flowering plants, seed development starts with a double fertilisation event, in which one of the two haploid pollen nuclei fertilises the haploid egg cell while the other one fertilises the diploid central cell within the female gametophyte inside the ovule. This syngamic process generates a diploid zygote and triploid primary endosperm nucleus, respectively (Agrawal and Rakwal 2012).

Normally, after fertilisation, seeds will be generated and develop. However, even if the pollen tube arrives at and penetrate the ovule, barriers may still exist, ceasing further development under the following typical situations: late-acting self-incompatibility and inbreeding depression.

• Late-acting SI (LSI), where SI occurs within the ovary, either prior to fertilisation or as the result of abortion of the ovule or fruit (see details about SI in Section 2.2.2). According to Seavey and Bawa (1986), there are 21 examples (from 16 families) known to be of this kind. Recent studies have not achieved much progress except for finding out many species in Bombacaceae are of this kind (Bittencourt and Semir 2006). Hiscock and Allen (2008) suggested that LSI is more widespread among the basal group of angiosperms. A more recent review by Gibbs (2014) indicated that there are at least 67 genera identified with the late-acting self-incompatibility syndrome. However, no study has reported whether LSI could occur in a hybridisation situation, especially when an SI species hybridises with an SC species. Therefore, it is a relatively unexplored area. • Inbreeding depression (ID). Seeds generated from inbreeding may suffer from inbreeding depression (see details about ID in Section 2.2.3). Recessive deleterious genes will form post-zygotic barriers, such as lethality and genetic death, and cause seed abortion.

It should be noted that distinguishing LSI from ID may be difficult under the circumstance that ID expresses in early seed development (i.e., early inbreeding depression, EID). Previous studies had noticed these difficulties and tried to suggest some criteria that might help distinguish LSI from EID (such as Seavey and Bawa 1986). However, these early criteria either were challenged by later studies or failed to be applied in other species. For example, embryonic lethal used to be explanation only for selfed pistil rejection while in some certain species it was caused by inbreeding depression (reviewed in Bittencourt et al. 2003). Moreover, most studies failed to pay attention to the histology of post-pollination events, which made it hard to distinguish LSD from EID.

Recent studies, from the histologic point of view, have provided some clear and new insights into the differences between LSI and EID: LSI, if any, shows at early embryo development while EID leads to abortion at any time of developing embryos. Bittencourt et al. (2003) investigated the reproductive biology and histology of post-pollination events in pistils of selfed and crossed flowers of Spathodea campanulata, a species known to have ovarian sterility or LSI, and found out that LSI (in selfed flowers) caused embryo abnormality in early embryo development (no later than the forming of three celled endosperm) while EID (in crossed flowers) led to the abortion of embryo at various stages of development. Hao et al. (2012) also suggested that EID generally involves the expression of recessive deleterious genes at multiple loci and hence often leads to a wide range of aborted seed sizes.

Besides LSI and EID, seed competition also occurs during seed development. This competition can be caused by intrinsic properties that determine whether one seed is inferior compared with another congenitally (e.g., deleterious genes passed by and caused inferiority or diseases), and parental causes, e.g. indirect impact of unsuccessful competition of parents on seeds (Scott et al. 1998). It also can be induced by environmental effects such as resource limitation and the

26 position of seeds (e.g., those seeds near the apex of a long pod or the end of a long raceme get fewer resources, thus limiting seed development).

Before seed maturation, some seeds may develop a mechanism of embryonic quiescence: dormancy. It is established during or after the accumulation of storage compounds that effectively prevents germination (Baskin and Baskin 2004, Finch-Savage and Leubner-Metzger 2006, cited in Agrawal and Rakwal 2012). In a hybridisation situation, many hybrids also show dormancy (Tanesaka et al. 2012) but no study has examined hybrid dormancy in Cakile species.

Meanwhile, during development or at maturation, seeds are at risk of predation, which is inevitable for wild plants. Therefore, many species have evolved a variety of ways to protect their seeds, including morphological defences (hiding seeds), chemical defences (employing chemicals in seeds), animal bodyguards (e.g. ants), temporal escape (escape from predators existing on a time scale), and spatial escape (see details in Willson 1983). Interestingly, in a hybridisation situation, due to differences in genotypes, hybrids can be more, less, or equally resistant to herbivores compared to their parental species (Cummings et al. 1999, Campbell et al. 2002).

Overall, seeds, whether hybrids or not, will experience a series of broad processes that are almost the same; however, variation in the vigour, dormancy and size of particular genotypes may affect their development and survival. Most comparisons of seed development between hybrids and their parents are, perhaps, confined to seed size and is limited to the F1 generation.

In the genus Cakile, the seeds are dimorphic in which the upper fruit segments bear seeds slightly larger and heavier on the average than the lower segments (see Figure 1.2). According to the measurements of Rodman (1974), seeds of C. maritima are slightly smaller than those of C. edentula although the fruit size is the opposite (the fruits of C. maritima are bigger), which is consistent with my observation. I also observed that some segments contain more than one seed, which has not been documented before. As for the development of hybrid seeds, no data is available. It is also intriguing that, in both species, seeds of many lower segments are not fully developed and even aborted, with an obviously higher abortion rate in C. maritima (unpublished data), which needs further investigation. Moreover, whether such differences will be observed in hybrids and backcrossed progeny is also an unexplored area.

2.3.2 Progeny growth and survival With proper conditions, seeds will germinate, followed by seedling emergence. However, seedling growth is subject to many factors in which the most important ones are its intrinsic properties and environmental constraints. For example, inbred progeny often show inferiority in growth compared to outbred progeny, which is primarily caused by the expression of deleterious recessive mutations in homozygous inbred progeny (Eckert et al. 2010). With inbreeding depression, plants may suffer a decrease in seedling survivorship, seedling growth rate, total flower number, and above-ground biomass etc. (Eckert et al. 2010). In some cases, inbreeding depression can also reduce seed production and the proportions of proximal and distal segments in species with heteroarthrocarpic fruits, such as C. edentula (Donohue 1998). Studies have revealed that the magnitude of inbreeding depression on progeny may vary due to various factors, such as the extent of self-fertilisation (Dart and Eckert 2013), expression of self-

27 incompatibility in a population (Vogler et al. 1999), and environmental variation (Hayes et al. 2005). However, if the progeny are hybrids, then their growth may vary across species. Plant breeders often observe the existence of hybrid vigour (heterosis), i.e., hybrids between genotypes grow better than their parents (Stuber et al. 1992), which has been reported frequently.(Burke et al. 1998, Hatfield and Schluter 1999, Arnold and Martin 2010). In other cases, however, hybrids may be inferior to their parents and thus are selected against (see details in Burke et al. 1998).

As for environmental constraints, due to the limitation of environmental resources, it is common to observe competition between individuals Competition is often classified as interspecific and intraspecific, although the mechanisms are often the same in plants (especially if species are closely related). Competition varies with the dynamics of populations, increasing and decreasing along with population density. When a few seeds arrive at a location where there are no other plants around (e.g. recolonization of a beach denuded by a storm), these single founders will experience no competition and grow to their maximum potential in that habitat (Mayr 1942, Provine 2004). However, their reproductive output may still be reduced due to pollen limitation (the Allee effect – see Section 2.2.3). If other species are present and close to the place of seedling emergence, there will be inter-specific competition and their growth will be reduced. As the population size increases over generations, both intra- and inter-specific competition will increase. The intensity of competition may also be cyclic: for example, in the case of Cakile, the siblings arising from proximal seeds of a plant and buried by sand will undergo intense competition in the next generation (inter-plant distances will be very short and density is extremely high (Donohue 1998, 2003)). When the sand is disturbed later, and the seeds redistributed, density (and competition) will be reduced again.

Recent studies have revealed another featured recognition, kin recognition, between closely related individuals. Kin competition has been an important theme in animal ecology (West et al. 2001, West et al. 2002, Johnstone and Kuijper 2014) and issues such as altruism have featured strongly (Hamilton 1963, 1964). Kin competition, however, can be intense in plants, as outlined in Cakile. Several recent studies have also confirmed kin competition in plants (Donohue 2003, Bhatt et al. 2011, File et al. 2011, Karban et al. 2013). It should be noted that Dudley and co- workers also have some contrasting findings (Dudley and File 2007, Murphy and Dudley 2009, Bhatt et al. 2011). They suggested that in C. edentula and several other species, although kin competition may be severe, kin recognition will overwhelm kin competition in the presence of another plant (Dudley and File 2007, Murphy and Dudley 2009, Bhatt et al. 2011, Karban et al. 2013, Johnstone and Kuijper 2014, Crepy and Casal 2015, Harsh 2015).

Data on growth and reproduction have been collected for C. edentula and C. maritima, but not for hybrids. Published estimates of seed production vary enormously within species, reflecting differences in methodology (many studies did not account for the seeds that have dispersed during growth, while others only recorded the fruit number rather than seed number) and growing conditions. Meanwhile, different genotypes of plants may differ in survival, not just in growth. For example, in an invasive grass, Phalaris arundinacea, the genotype strongly influenced its survivorship (Morrison and Molofsky 2000). Many other studies have also reported that difference in genotype will affect offspring mortality (Oostermeijer et al. 1994,

Kery et al. 2000). In Cakile, Barbour (1970) compared survival of C. maritima and C. edentula in California, finding that the former lived longer and had higher survivorship.

2.3.3 Reproductive attractiveness and output When seedlings manage to grow into the reproductive stage, they must ensure their reproductive output to pass genes onto next generations. As a result, the priority of plants in reproductive stage is to ensure the transportation of pollen and following pollination. There are many ways of flower pollination, which can be divided into three categories: self-pollination, pollination via abiotic (i.e., wind and water pollination) and biotic pollen vector (pollinator).

Flowering plants, regardless of SI or SC, are usually dependent on pollinators for reproductive output (Ward and Johnson 2013) and thus the attractiveness of a flower plays a crucial role in pollination. As a matter of fact, pollinators visit flowers for various reasons. A large proportion of pollinators just go to flowers for food, feeding on nectar and sometimes also on the pollen itself. Occasionally, flowers offer neither pollen nor nectar as a foodstuff to their visitors but yield other rewards; or they may offer these as “extras” in addition to some pollen (see details in Chapter 9, Willmer 2011). Therefore, a flower could enhance its attractiveness by contributing more energy to attributes of two categories: either attractants (advertising signals) or rewards (usually foodstuffs). These attributes can also be divided into a) features of individual flowers (e.g., colour, shape, scent, reward availability, or time of flowering) and b) features of a whole plant or groups of plants (e.g., flower density, flower number, flower height, or spatial pattern) (Willmer 2011).

Individual flowers can enhance their floral display and morphology (size, shape, structure, and colour) to increase the attractiveness to pollinator animals, thus maximizing pollinator attention and consequently seed set and fitness (Karron and Mitchell 2012, Glover 2014). Specifically, flower size is enormously variable across the angiosperms and changes in it are often correlated with changes in plant size, habitat, and even flower number. For example, there may have a significant negative correlation and trade-off between the size and the number of flower on a plant, which has been found in 251 species (Sargent et al. 2007). Moreover, recent studies, measuring the relationship between outcrossing rate, flower size, flower number and floral display, have shown that outcrossing rate is positively correlated with the product of flower size and flower number (Armbruster et al. 2002, Goodwillie et al. 2010, Karron and Mitchell 2012). Correspondingly, selfing species, which are less reliable on pollinators, tends to reduce energy allocation to attraction, and thus have smaller and less showy flowers. This phenomenon is termed as the “selfing syndrome”, including a reduced flower size, smaller pollen number and pollen-ovule ratio (Sicard and Lenhard 2011, Luo and Widmer 2013, Tedder et al. 2015), which indicates a shift of the breeding system or pollinator types. Thus, in a situation of hybridisation, the attractiveness of produced progeny, although barely studied, could be a great indicator for the changes in breeding system as well as for the reproductive output.

As for the features of a whole plant or a group, when pollinators do visit a certain patch, they will be influenced by the species composition in that patch (Kunin 1993, Hegland and Boeke 2006, Seifan et al. 2014). Even generalist pollinators may focus on certain plant species in that patch as a food source; perhaps the species is most abundant or provides the greatest rewards. For example, if a conspicuous species, is introduced in a patch, pollinators of the less

29 conspicuous species may decide to visit the introduced species more frequently or exclusively. This decision, of course, is a complex outcome of the pollinator’s innate preferences and its current learning experiences (Hegland and Boeke 2006, Hegland and Totland 2012). The conspicuous species may act as a magnet to pollinators and thus increase pollination visitation and pollination success rate of the less conspicuous species in their vicinity or compete for pollinators and reduce their possible pollinator visitation (Seifan et al. 2014).

Meanwhile, different pollinators have different preferences for colour. In general, based on observations and experiments, the colour preference of pollinators is a combination intrinsic preference and acquisition learning (Willmer 2011). For example, many generalist flies tend to visit yellow or yellow and white flowers, while more specialist flies like pink, purple, or blue flowers; honeybees prefer flowers towards the blue end of the spectrum but this is dependent on the background contrast; birds have no innate preference for red colours but learn this as a choice.

In Cakile species, the attractiveness to pollinators is a combination of visual and olfactory signal. The flower colour of Cakile species varies from white to lavender, violet, or pink-purple (Rodman 1974). Commonly-seen pollinators on Cakile species are a variety of bees, flies, butterflies and beetles. But there is no observation for hybrids. As for odour, I observed fresh flowers of C. maritima have a fragrant odour, which is consistent with the observation of Rodman (1974), while C. edentula does not have such odour. It should be noted that the above statement about odour is only from the human olfactory point of view. Although there are many previous studies on insect sense of volatile compounds from flowers (e.g., Kobayashi et al. 2012) in Brassicaceae crops species, no previous study has been conducted in Cakile.

In summary, a flower’s visual attractiveness (interaction of shape, size and colour) and olfactory attractiveness can facilitate pollination via increasing pollinator visitation, and thus promote its reproductive output. In contrast, selfing species tend to have a reduce flower size and less showy flowers as pollinators are not necessary. However, there is still a lack of studies about the attractiveness in hybrids and comparisons between hybrids and their parental species.

2.4 Conclusions and research questions In the previous sections, I have reviewed studies relevant to floral development, breeding system and progeny fitness, identifying the research gaps. It gave an overall view of what we have already known and indicated questions that remains to be answered. To summarise, the main under-researched questions are listed below.

First of all, what is the pollination window for natural hybridisation between two closely related species with contrasting breeding systems? Floral development has been studying for decades as well as its genetic pathways. The ABC model and its descendants, such as ABCD and ABCDE, that describe the genes controlling floral development have been demonstrated in many different species in different families. Meanwhile, morphology of flowers at or after flowering also been extensively studied, such as herkogamy and dichogamy. However, since most attention has been drawn to these areas, there are barely any studies about the basic floral development and its morphology and visualization in the early stages. Such studies can provide simple but useful information about the status and images of floral development before and

30 after flowering, which can be used as a guide of how flowers develop and what the landmark events are, especially in the early stages. Combining these areas can provide new insights into floral development, which is a useful in many aspects, especially for hybridisation. For example, in the early floral development of a selfer, it is not only crucial to estimate whether the stamens have any chance to contact the stigma surface before flowering but also important to know the timing of stigma receptivity and pollen ripeness. Even though the stamens could reach the stigma surface, neither non-ripe pollen on receptive stigma nor ripe pollen on non-receptive stigma can cause any contamination, thus providing a pollination window for outcrossing and hybridisation.

Secondly, what are the traits and inheritance of the breeding systems of hybrids and parental species in hybridisation? As soon as pollen is deposited on the stigma surface, the breeding system starts showing its direct effects. The breeding system in many species, e.g., Brassicaceae, has been extensively studied, in which self-incompatibility, a mechanism in plants to help recognise self-pollen as illegitimate and thus prevent self-fertilisation, plays a key role. The genetic basis of SI has been studied for decades and is well known, especially in Brassicaceae while partial SI and cryptic SI have also drawn some attention. There are also some studies about the unilateral incompatibility when hybridising species are SI x SC. However, this incongruity (the SI x SC rule) does not appear to be consistent in many species and needs further investigation. Meanwhile, an important area that seems never has been addressed before is the inheritance of the breeding systems when SI and SC species hybridise. To understand the traits of the breeding systems in hybridisation, it is necessary to know that a) whether unilateral incompatibility exists and possibly how it works, b) whether the inheritance of the breeding systems varies between different generations as well as different crossing directions, and c) whether partial SI will be observed if no obvious SI is detected.

Last but not least, what are the fitness costs of hybridisation between two closely related species with contrasting breeding systems? Fitness of plants, as reviewed, should include the lifetime reproductive output of an individual, subsuming all the processes instead of estimating fitness in experiments by measuring plant biomass at an arbitrary point in time, or seed production, which were usually adopted in previous studies. Moreover, such fitness studies are restricted with parental species while relevant studies on their progeny are largely neglected. Hybridisation between species and their hybrids could affect seed attributes such as viability, size, and morphology, and hence the growth and dispersal of the progeny. However, there are barely any reports about the fitness comparisons between hybrids and their parental species (except in crops breeding where the crop yield is optimized). As a matter of fact, the fitness costs of hybrids are of great value for understanding hybridisation, its interaction with breeding systems, and the role it plays in invasion. Considering the relative fitness between hybrids and parental species, there are several possibilities that could occur: a) hybrids show heterosis and thus are superior, b) hybrids show no superiority compared with their parents, and c) hybrids are inferior to their parental species. In all three scenarios, the interaction between fitness and the breeding system could possibly lead to and account for above results, which needs further investigation.

These main questions deserve attention on their own right, to advance our understanding of the processes involved in the dynamics of hybridising populations; however, they are particularly

31 pertinent to the situation in Australia, where two invasive Cakile species, differing in breeding systems, coexist for a while in which hybrids may form and then the self-compatible species disappears. This replacement has also happened in New Zealand and the west coast of North America but has not been fully explained. A hypothesis has recently been proposed in our lab (Mesgaran et al. 2016) whereby a previously arriving self-compatible species may in fact increase the spread of a subsequent self-incompatible invader, by overcoming the single-founder problem. Central to this hypothesis are a) the fitness costs of hybridisation, e.g., attractiveness of the species and their hybrids towards pollinators, and (b) and the interaction between different breeding systems in hybridisation.

Therefore, in the following several chapters, I will use Cakile as the model plants to address above questions. Chapter 3 presents the methodology of producing all the hybrids used in this thesis, which involves the protocols and designs of hybridisation. From Chapter 4 to Chapter 6, each chapter deals with a single main question raised above: Chapter 4 focuses on floral development, Chapter 5 addresses questions about the breeding system and its relevant discussions, and Chapter 6 mainly tackles problems with progeny fitness. Taking all things together, Chapter 7, the last chapter, provides the final discussions and conclusions for this thesis, including some suggestions for further investigation. .

Generation of F1, F2 and backcrossed hybrids

3.1 Introduction Interspecific hybridisation has been regarded as an important factor for speciation (James 2005, Elizabeth et al. 2015, Wu et al. 2015). Empirical evidence for the facilitation of speciation by hybridisation, however, is often mixed or incomplete, and some of the mechanisms proposed are poorly understood or controversial (see details in Servedio et al. 2013). It usually occurs between but not limited to closely related species, such as interspecific and intergeneric hybridisation (Zhou et al. 2014, Kyle et al. 2015) or hybridisation between weedy and cultivated species (Allison et al. 2001, Yukio and Sang 2014) etc.

Field observations and lab research (Ohadi et al. 2012, Cousens et al. 2013, Ohadi et al. 2015) has confirmed that hybrids of C. maritima and C. edentula exist in the wild. In the field, Ohadi et al. (2015) found out that both species can contain chloroplast genes from the other species, especially in C. maritima, using markers. Such results indicate the occurrence of hybridisation and introgression between the two species, especially from C. edentula into C. maritima, both in the two regions where they are currently sympatric and elsewhere. Thus, it is of great importance to produce artificial hybrids, which can help promote the understanding of hybridisation between the two species.

MM EE

M Hybridization E

Figure 3.1 A schematic diagram of producing all lines in this experiment, where “M” and “E” are two parental species while “H” stands for the F1 hybrids. “MM” and “EE” represent intra-specific outcrossing of M and E respectively; “BC” stands for backcrossing between the F1s and parental species. F2s are produced using F1 hybrids. In this chapter, I will present the plan for generating F1 and F2 hybrids as well as backcrossed progeny under artificial conditions; intra-specific outcrossing will also be included. F1s were produced and then used to generate further generations such as F2s and backcrossed progeny. It should be noted that F2s in this thesis were produced by crossing different F1 individuals instead of by classical Mendelian genetics (i.e., by selfing of F1 individuals), to avoid inbreeding depression, similar to Darwin’s intercrossing of F1 pigeons to produce F2’s (as reviewed in Secord 1981)(see details in Section 3.3). These intra-specific outcrossed progeny (“MM” and “EE”), were used in the fitness comparison experiments (see details in Chapter 6). All buds chosen for hand-pollination were emasculated two days before artificial crossing was conducted (see details in Section 3.2.4).

3.2 Producing F1 hybrids In this part, seeds from both species were sown and reciprocal crossings were conducted to produce F1 hybrids. To get fully outcrossed lines for further experiments, outcrosses within the parental species were also included in this experiment. Figure 3.1 briefly shows how these lines were produced. The main processes of producing all the hybrid lines and their corresponding outcrossed progeny will be presented below.

3.2.1 Seed collections The study area is in Tasmania, Australia. The invasion history of the two sea rocket species has been reviewed in Chapter 1 as well as in recent papers (Ohadi et al. 2012, Cousens et al. 2013, Ohadi et al. 2015). Currently, the invasion front of C. maritima is in south-eastern Tasmania and a sympatric zone of the two species exists there. Both species, although introduced at different times (based on herbarium records; seed details in Section 1.4), invaded Tasmania from north to south and the sympatric zone is currently between Raspins Beach and the outer suburbs of Hobart. From Raspins Beach northwards, C. maritima has completely replaced C. edentula (Cousens et al. 2013). Around Raspins Beach, both species co-exist and putative hybrids are found from morphological traits.

To include all morphologically available variation present in different populations as well as to avoid inbreeding depression (see details in Chapter 1; also see in 3.3.3), seeds were collected at four different locations (see Figure 3.2 for the sampling sites). For C. maritima, seeds were collected at 3 different locations. These locations were: a long distance from the invasion front; a short distance from the sympatric zone and right in the sympatric zone, aligned with its invasion history. Through multi-location collections, variation and diversity of C. maritima were included. For C. edentula, two locations were considered, i.e., sympatric zone and allopatric zone where only C. edentula exists.

Within a site, collections were made from different maternal plants and kept in separate paper bags. Fruits were dried and stored under cool & dry conditions. It should be noted that Raspins Beach is in the sympatric zone where hybridisation between the two species could occur and thus the seeds collected here, even from plants morphologically similar to each species, may be hybrids. Seeds that later turned out to be hybrids, they were excluded from any further experiments (see details in Section 3.2.2).

Ulverstone Marion Bay Raspins Beach Sloping Main (Sympatric Zone) C. maritima C. edentula

Replaced Not colonized

Figure 3.2 A schematic drawing of relative positions of sampling sites. Green and black arrows indicate the invasion fronts of C. maritima and C. edentula, respectively. Dashed line indicates a certain species is either replaced or not colonized. Narrowed solid lines indicate the sympatric zone of the two species.

3.2.2 Seed treatments Seeds from both species were treated with multiple procedures to improve germination. Firstly, the fruit coat was removed before sowing. Then seeds were put on moistened filter paper in Petri dishes, self-sterilized by hypochlorite sodium 1% for 30 seconds and rinsed with sterile distilled water followed by 70% ethanol (to decrease the possibility of disease infestation) (Oliver et al., 2001). Once imbibed, the seeds were carefully scratched using fine pins to break the seed coat on one side to ensure germination (Gormally and Donovan 2011). After that, seeds of both species were incubated in a cold room at 5 °C for 7-10 days. Five pre-germinated seeds (from the same individual parent) were randomly chosen, put into a biodegradable pot and thinned down to one in the seedling stage, labelled with population and species name. The biodegradable pots were then transferred into bigger plastic pots, which were placed in a Poly tunnel at the Burnley campus, the University of Melbourne.

In this experiment, to ensure the development of good root systems, pots with a diameter of 40 cm and a depth of approximate 30 cm were used. The potting mix was also used in all other experiments in this thesis and its components were:

• 20% Medium pine bark (practical size is between 4mm – 6mm). Supplier is Deco. • 40% Propagation/coarse washed Sand. Supplier is Daisy Garden Supplies. • 40% Filtered sand. Supplier is Rockla Quarry products. • Controlled Release Fertiliser (CRF), Debco Green Jacket No. 2. N: P: K is 16.5: 4.1: 9, at a rate of 4g /L potting mix.

Figure 3.3 Seedlings in biodegradable pots.

Figure 3.4 Seedlings transplanted into the plastic pots (left) and grew bigger (right). Insecticide (Bayer Confidor) was applied during plant growth to control pests. Pests commonly seen on Cakile species are aphids and thrips (Payne and Maun 1984, Maun et al. 1990, Davy et al. 2006), sucking and consuming the liquid from young buds and thus stopping plants flowering. A fungal pathogen, Alternaria brassicicola (Dermateaceae), can also cause severe infection in Cakile species (Oliver et al. 2001), which mainly occurs on leaves and fruits. The insecticide was sprayed once pests were found and identified. The minimum time interval between two sprays was two weeks to avoid any side effects that may be caused by the insecticide.

3.2.3 Crossing plan for F1s In this experiment, F1s were produced through reciprocal crossings between the two species and used to produce F2s and backcrossed progeny where three generations in total were generated (as described in 3.2.4), which involved huge amount of work. Therefore, when producing F1s, the number of parental plants used to produce F1s was minimized. The experiment consisted of only 10 experimental plants from each species, adding up to 20 individuals in total.

10 individuals in each species were from different populations. In C. maritima, as there were 3 sampling populations but only 10 plants were available, each population could not have the same number of individual plants. Therefore, Raspins Beach (M1-M3) and Ulverstone population (M4-M6) had three individuals, respectively (3 + 3 = 6); while Marion Bay had four (M7-M10), adding up to 10 plants in total. C. edentula comprised five plants from Sloping Main (E1-E5) and five from Raspins Beach (5 + 5 = 10).

Table 3.1 Population sources of individual used in the experiment.

Species Sloping Main Raspins Beach Ulverstone Marion Bay C. maritima - M1-M3 M4-M6 M7-M10 C. edentula E1-E5 E6-E10 - -

Individuals from the same population were arranged together (see below figure: M and E represents C. maritima and C. edentula, respectively). For a diallel crossing design, there would be as many as 200 crossing directions, not to mention the replicates within each crossing to ensure the availability of F1s. Consequently, although a randomised design would be preferable statistically, this layout made it easier to keep track of the large numbers of crossing. Pots were separated with a 100 cm distance to minimise competition. Plants were watered as needed.

Figure 3.5 Experimental layout of the two species. Left panel shows the arrangement of plants within the polyhouse and right panel shows an enlarged picture of a single plant in a pot with pollination cages. The white square represents the pollination cage covering the plant. The green region shows the area outside each pollination bag. The grey circle shows the position of the pot. Potential pollinators were excluded by putting pollination cages around the plants when buds emerged. The pollination cage had a dimension of 60 cm in height, 100 cm in width, and 100 cm in length and made of plastic pipes, covered with fine mesh and attached using hot glue gun (Bosch PKP18E Glue Gun). The cage had only one side uncovered and thus could be easily put on and removed from plants.

Figure 3.6 Plants covered by cages in the Poly tunnel.

Considering the sheer volume of this hand-crossing experiment, a partially reciprocal crossing plan was conducted, with incomplete combinations of the plants as shown in Table 3.2. To ensure availability of seeds, at least 100 buds in each crossing direction were artificially pollinated (see pollination procedures in 3.2.4). In such case, if a complete diallel crossing were performed, there would be more than 20,000 replicates in total, which is far beyond my ability to finish it. Consequently, a partially diallel crossing plan were applied.

It was mentioned in Section 3.2.1 that collected fruits from Raspins Beach could have been hybrids. If the plants grown from field-collected seeds turned out to be hybrids (based on plant morphology), they were not used in further experiments.

Outcrosses within species were also included in this design. To reduce work load, only one individual was randomly chosen from each population. To avoid possible inbreeding depression (see in 3.3.3 for principles of crossing), outcrosses were done between these chosen individuals, as plants derived from the same location would have been at least half siblings. Here, the chosen plants were M2, M5 and M7 for C. maritima, and E1 and E10 for C. edentula.

Table 3.2 Crossing plan to produce F1s and outcrosses within C. maritima and C. edentula. Left panel is the crossing plan for producing F1 hybrids where “×” indicates crossings without allocated labels; other crossings were labelled with corresponding labels from “F11-F115” (see detailed explanation in Section 3.3.1). “M1-M10” refer to the 10 C. maritima plants; “E1-E10” represents the 10 C. edentula plants. Each plant acted as both pollen donor and recipient. Right panel is the outcrossing plan for producing outcrossed progeny, which was numbered as shown in the right panel.

# M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 # E1 E10

E1 F14 F112 E1 EE2

E2 × F113 E10 EE1

E3 × F114

E4 F15 ×

E5 F16 F115

E6 F11 F17 # M2 M5 M7

E7 × F18 M2 MM3 MM5

E8 × F19 M5 MM1 MM6

E9 F12 F110 M7 MM2 MM4

E10 F13 F111

3.2.4 Methodology of pollination For each crossing between the two species, two to three racemes on each plant were chosen as the pollen recipients and labelled. Buds on the racemes were emasculated at stage 4 in edentula and stage 5 in maritima to prevent possible self-fertilisation (see details of floral development in Chapter 4). The main procedures for producing F1s and outcrossed progeny are described in the following diagram (Figure 3.7).

In this experiment, only stamens were removed when emasculating buds while petals and sepals were remained intact. This is because petals could act as an indicator for pollination, i.e., when petals start showing up and elongating, it is the right time for pollination (it can help avoid pollinating unripe stigma or breaking SI in early stigma)

After emasculation, buds were bagged with pollination bags to prevent accidental pollination. Although the pollination cages excluded most pollinators, there was still possibility for emasculated buds to be contaminated by either self- or alien pollen. Therefore, it was necessary to put small pollination bags to isolate the emasculated buds from others. Emasculated buds were pollinated after petals showed up (stage 9 in C. maritima and stage 10 in C. edentula; see details in Table 4.1 of Chapter 4). Freshly opened flowers from designated plants were chosen and stamens were picked up by tweezers to pollinate the target buds. Specifically, a fully opened flower on day 1 (just opened) from the designated pollen-donor plant was used as a source of pollen and an anther was picked up using fine forceps and carefully brushed across the targeted stigma, to spread enough pollen evenly. Pollination bags were always on the racemes and could also act as containers, preventing ripe fruits from falling off and missing. Fruits were harvested after ripening and stored in paper bags separately for later experiments.

Put pollination Pollinate Put pollination Choose Repeat on 100 cages on plants Emasculate flowers with bags on racemes and buds for each before buds corresponding emasculated tagged direction flowering pollen buds

Figure 3.7 Diagram of pollination processes. As only a few racemes were chosen during above experiments, many flowers were left intact and were possible to be self-fertilised in the cages. Therefore, these flowers could be used to observe whether C. maritima (although they are supposedly SI) (Boyd and Barbour 1993, Clausing et al. 2000, Davy et al. 2006) used can self-fertilise itself or not. It is quite simple and easy to observe whether there is seed initiation without hand pollination in the cages.

3.3 Producing F2 and backcrossed hybrids In this section, procedures about the production of further generations of hybrids from the F1s, including F2s and backcrosses, will be introduced. Basically, backcrosses were conducted in reciprocal directions between F1s, C. maritima, and C. edentula. F2s were produced by outcrossing within F1 individuals to avoid possible inbreeding depression (see details in 3.3.3). Therefore, F2s here are not classical F2 commonly seen in the literature.

3.3.1 Seed collections F1s with C. edentula as the maternal parent were used for producing F1s (i.e., ME; see details of direction code in Chapter 5) and backcrossed with the half siblings of the originally used parental plants (see details about those 20 plants in Section 3.2.1) to produce backcrossed progeny. The reasons why these lines were chosen are explained as below.

Firstly, the design could be much more complex than normal and beyond my ability if seeds of all the directions were used. As introduced before, although it was partially diallel reciprocal crossing, there were still 40 directions when producing F1s. Secondly, there could be a problem of seed availability. Due to unilateral incompatibility (as described in 2.2.4), if it exists in our target species, most of generated F1 hybrids should be the ones with C. edentula as the maternal parent and those from the other direction will only contribute to a small proportion of the seed collection. Finally, seeds from a certain pair of reciprocal crossings, as suggested, should have the same chromosomes regardless of their crossing directions. In a pair of reciprocal crossings, seeds have the same parental plants and thus no matter which direction the seeds are from, their nuclear genome should be the same. Meanwhile, the genes related to self-incompatibility mentioned in this thesis are all carried by nuclear genome (Hiscock and Dickinson 1993, Thrall et al. 2000, Igic et al. 2008, Kitashiba and Nasrallah 2014). Therefore, F1s with C. edentula as the maternal parent were used for the next stage experiments (see details of F1 crossing plan in Table 3.2).

Besides, Cakile species mentioned in this thesis are mostly annual plants, which barely can survive into a second year (Rodman 1986, Cousens et al. 2013). Consequently, siblings of the original parental plants, i.e., seeds from the same individual collections of the original parental plants were used. Individuals of the parentals species used in this experiment were labelled using subscript “sib”, such as “M1sib” and “E1sib”.

The details of seeds used in this experiment are listed in the following table (Table 3.3), in which their lineages are described. It should be noted that F1s derived from M2 and M3 were excluded in this table as M2 and M3 were identified as hybrids after the first experiment (as described in 3.2), since they had typical hybrid fruit shapes. F1 Hybrids between E4 and M9 was not available as M9 plant died before any pollination could be done. Thus, there were only 15 lines of F1 hybrids available. For convenience, all F1s used in this experiment were labelled as F11-F115.

Table 3.3 F1s used in producing F2s and their lineage. M2 and M3 related F1s were excluded as M2 and M3 were identified as hybrids. M9 died during the experiment and thus no seeds were producing between M9 and E4.

Hybrid F11 F12 F13 F14 F15 F16 F17 F18 F19 F110 F111 F112 F113 F114 F115 ♀ E6 E9 E10 E1 E4 E5 E6 E7 E8 E9 E10 E1 E2 E3 E4 ♂ M1 M4 M5 M1 M4 M5 M6 M7 M8 M9 M10 M6 M7 M8 M10

3.3.2 Seed treatments Seeds from above collections (see in 3.3.1) were used and the pre-germination treatments were the same; however, there were slightly differences compared with the procedures mentioned in Section 3.2.2.

Firstly, this experiment was performed in a temperature-controlled glasshouse instead of the Poly tunnel. Due to the time limitation of my PhD study, I had to conduct experiments

40 continuously to make sure that all the required hybrids were produced in time. Therefore, this experiment was performed during the winter of 2015 in a temperature-controlled glasshouse, in which limited space was available. Consequently, pots used here were 25 by 30 cm ones. Potting mix were the same as described in above sections. Glasshouse conditions were set as 25℃ at day and 15℃ at night; day length was set as 16 hours. All pre-germinated seeds of each line were transplanted into one pot and allowed to grow. Plants were thinned down to one in each pot a few weeks later.

3.3.3 Crossing plan for BCs When doing artificial backcrossing, there could be many factors affecting its result. To avoid any detrimental effect, several measures were taken into consideration, which will be explained below.

Firstly, inbreeding depression should be minimized. To minimize possible inbreeding depression when backcrossing, F1s should not be backcrossed with the parental individuals from the same population where their parents were collected to avoid potential inbreeding depression. For example, F12 should not be backcrossed with E6 sib - E10 sib and M1 sib.

Secondly, backcrossing scale was reduced to make sure that the research was achievable. There were two general rules to achieve this target. Firstly, when backcrossing, F1s potentially can be backcrossed with any available C. maritima and C. edentula sibling plants except those derived from the same populations as their parents. As a result, the experimental design could be messy and analysis of collected data would be a big challenge. To tackle this dilemma, all hybrids derived from the same C. edentula plants (e.g., F11 and F17) were only backcrossed with the siblings of a randomly chosen edentula plant (e.g., E6 sib) in another population (instead of different plants when doing backcrossing with C. edentula). This also applied to the backcrossing with C. maritima. For example, F11 and F14 have the same paternal plant M1 and thus they were backcrossed with M7. Secondly, when backcrossing, the individuals of the paternal species already used for backcrossing was not used again unless the F1s backcrossed with these individuals were derived from the same maternal or paternal plants. It is necessary that F1s derived from different maternal/paternal plants should be backcrossed with different individuals in the target population.

Meanwhile, for balancing the experimental design and reducing the complexity of crossing, M1 sib was treated as a part of the population of M4sib- M6sib. As M2 sib and M3 sib were excluded, there was only one individual from Raspins Beach (M1sib) while there were eight individuals

(M1sib, M4sib - M6sib and M7sib - M10sib, respectively) in total for C. maritima. Therefore, M1 sib,

M4 sib - M6 sib were in one group and M7 sib - M10 sib were in another one. As a compensation for the variation loss caused by M1 sib being integrated into the population of M4sib - M6sib, F11 and

F14 were backcrossed with M5sib as well. F1s derived from M4sib-M6sib were not backcrossed with

M1sib as too many crossings were involved. In accordance with above rules, the complete backcrossing plan is shown in the following table (Table 3.4).

Table 3.4 Crossing plan of producing backcrossed hybrids. Crossings listed in the table are all reciprocal.

Direction F11 F12 F13 F17 F18 F19 F110 F111

Backcrossed E1 sib E4 sib E5 sib E1 sib E2 sib E3 sib E4 sib E5 sib with E Backcrossed M7 sib M8 sib M9 sib M10 sib M1 sib M4 sib M5 sib M6 sib with M M5 sib

Direction F14 F15 F16 F112 F113 F114 F115

Backcrossed E6 sib E9 sib E10 sib E6 sib E7sib E8 sib E10 sib with E Backcrossed M7 sib M8 sib M9 sib M10 sib M1 sib M4 sib M6 sib with M M5 sib

3.3.4 Crossing plan for F2s As described in section 3.3.1, M2, M3 and their related hybrids were excluded from any further experiments and F11 - F115 were used to produce F2s. F1s, according to their lineage, can be divided into 6 different groups, in which the paternal plants were all derived from the same population (this also applied to their maternal plants). The following table shows the group details (Table 3.5).

Table 3.5 Groups divided by F1s lineage. In a single group, F1s’ parents share the same origin, i.e., derived from the same population.

Raspins Ulverstone Marion Bay Beach Parent M1 M4 M5 M6 M7 M8 M9 M10

Sloping Main E2

E3 G3 G4 G6 (F14) (F15 F16 F112) (F113 F114 F115)

Rasp E7

ins Beach E8 G1 G2 G5 (F11) (F12 F13 F17) (F18 F19 F110 F111)

E10

When producing F2s, inbreeding depression was also considered as a factor that should be minimized, which means that the F1s whose paternal or maternal plants derived from the same population were not crossed. In other words, if two groups shared the same maternal or paternal plants, they were also not used in crosses. In accordance with above crossing rules, the crossing plan for producing F2s between different groups was:

Table 3.6 Crossing plan between F1 groups to produced F2s. “×” indicates crossing to produce F2s.

# G1 G2 G3 G4 G5 G6 G1 × × G2 × × G3 × × G4 × × G5 × × G6 × ×

As there were many lines in a single group, it would be a complex crossing design for a single crossing direction if every line was used to be crossed with every single line in another group. Thus, to reduce the amount of work, only one F1 line within a group was randomly chosen when there were several F1 lines available in some groups. As indicated in Table 3.5, the F1s were divided into six groups:

• Group 1 (G1): F11

• Group 2 (G2): F12 F13 F17

• Group 3 (G3): F14

• Group 4 (G4): F15 F16 F112

• Group 5 (G5): F18 F19 F110 F111

• Group 6 (G6): F113 F114 F115

Here, F1s marked in red above were used to produce F2s and the crossing plan for producing F2s can be described as:

Table 3.7 Detailed crossing plan to produce F2s.

# G1 G2 G3 G4 G5 G6

G1 F11× F15 F11× F113

G2 F12× F14 F12× F113

G3 F12× F14 F14× F18

G4 F11× F15 F15× F18

G5 F14× F18 F15× F18

G6 F11× F113 F12× F113

Reciprocal crossings were conducted and crossing techniques such as emasculation, pollination etc. were the same as mentioned in previous crossings (as described in 3.2.4). Pollination bags were also used in these procedures and pollinated buds were left on their raceme. Successfully pollinated buds were allowed to mature naturally. Fruits were harvested when ripe and data were collected.

3.4 Conclusions In this chapter, I provided the rationale behind pollination and crossing designs that could be used in the experiments described in the following chapters.

The role of floral development in hybridisation between C. maritima and C. edentula

4.1 Introduction Hybrids may occur naturally where the habitats of closely related species are separated by short distances or where related species from geographically distinct regions are introduced into the same habitat. While morphologically recognisable hybrids may not be apparent at the current time, evidence for past hybridisation may be present within genomes as a result of past introgression (Ohadi et al. 2015). The significance of hybridisation for plant speciation, local adaptation and colonisation depends on many factors (Mesgaran et al. 2016), including the relative reproductive outputs of the two species, their attractiveness to pollinators, phenology, abundance, mating system (including the ability of the pollen from one species to fertilise ovules of the other), and the viability, vigour and fertility of hybrid offspring. At one extreme, hybrids may have no impact whatsoever on the long-term success of the species, while at the other extreme hybrid demography and adaptation could be essential for one or both (Mesgaran et al. 2016). Although the list of hybrid-derived invaders is growing (Schierenbeck and Ellstrand 2009), the relative importance of hybridisation, both inter- and intra-specific, as a driving mechanism has not been assessed rigorously and the advantages endowed remain hypothetical in most cases.

While mating systems have received much attention in discussions of individual species’ colonisation/invasion ability, exploration of their role in the dynamics of hybridising species has barely begun (Goodwillie and Ness 2013, Bouhours et al. 2017). If wild hybrids are detected, there must have been at least some level of compatibility between the species’ mating systems; the implications of these hybrid genotypes for transgressive segregation, introgression and speciation has been explored without breeding system information (Yakimowski and Rieseberg 2014, Feliner et al. 2017). It has long been appreciated in single species, however, that traits reducing self-fertilisation - temporal, physiological, mechanical or a combination of these (Barrett and Harder 2017) - are likely to be a demographic disadvantage if just one or a few offspring arrive at a distant location (Cheptou 2012, Grossenbacher et al. 2017). The reproductive assurance given by the ability to self-fertilise is thus argued to be the reason that greater relative frequencies of self-fertile species are found on islands compared to large land masses and in invasive species compared with native floras (Baker 1955, Petanidou et al. 2012). The genetics of physiological self-incompatibility (SI) at the S-locus, whereby a pistil rejects pollen from its own flowers, is particularly well-known in some families, while self-compatibility (SC) is a secondary adaptation resulting from a loss-of-function at the S-locus itself or occasionally of a different, modifier locus (Igic et al. 2008, Vekemans et al. 2014). Notwithstanding the almost universal acceptance of Baker’s rule, there are many successful, out-crossing invaders (Snow et al. 2001, Ridley and Ellstrand 2009). Most species have mixed mating systems, somewhere between complete out-crossing and complete selfing (Barrett and Harder 2017).

The role of plant mating systems in hybridisation can be complex. It is believed that the original flowering plants were hermaphrodite, i.e., they had stamens as well as pistils (Richards 1997,

Charlesworth 2006, Goodwillie et al. 2010) but were normally unable to be pollinated by their own pollen (Igic and Kohn 2001, Steinbachs and Holsinger 2002). Self-pollen is rejected biochemically as illegitimate, thus preventing self-fertilisation, i.e., “self-incompatibility” (SI) (Hiscock and McInnis 2003). SI in species could break down and thus become self-compatible (Vogler et al. 1999). The breakdown is commonly due to the loss-of-function mutations at the S- locus (Igic et al. 2008, Vekemans et al. 2014). Selfing or self-compatibility (SC), the counterpart of self-incompatibility, refers to the ability in some plants to be fertilised by their own pollen. Selfing is quite common and sometimes the norm in species (Igic and Kohn 2001, Igic et al. 2008). Studies on the phylogenetic analyses of S-Rnases suggest that SI was the ancestral state in the majority of dicots and thus SC plants derived secondarily from SI species (Igic and Kohn 2001, Steinbachs and Holsinger 2002). Mechanisms that promote selfing could be a potential barrier to hybridisation between species. When two species are sympatric, it is highly likely that a plant will receive pollen from both species. In the absence of any physiological incompatibility between them, hybrids could be formed, depending on which species’ pollen arriving on the stigma and fertilises the ovule first. However, cleistogamy, a type of autogamy caused by a complete failure of the flowers to open, could prevent hybridisation entirely. For example, Arabidopsis thaliana has small flowers and usually pollinates itself before its buds open (Richards 1997). Moreover, due to a reduction in their floral display, selfers are normally less attractive to pollinators (Willmer 2011) and the reduced visitation makes it less likely that pollen will be transported to another species (Seifan et al. 2014).

In the case of an SI species cohabiting with an SC species, although a generalist pollinator could visit both species, we might expect the hybridisation to occur more readily in one direction than the other. The lack of self-incompatibility, proximity of anthers to stigma and unattractiveness to pollinators would make it less likely for SI pollen to be the first to reach the SC’s stigma than the other way around. Conflicting with this, however, is the so-called unilateral incompatibility (UI), a well-known phenomenon whereby artificial hybrids between SI and SC species can only be obtained in one direction, typically with the SI species being the pollen donor and the SC species being the receiver (e.g., Hiscock and Dickinson 1993). Incompatibility in the other direction (i.e., pollen from SC to SI) is believed to be caused by a physiological barrier involving S alleles, with SI and UI perhaps sharing the same pathway (Kitashiba and Nasrallah 2014).

There are, however, many cases of successful artificial hybridisation between SI and SC species, in which selfers can act as either female or male parents. There are also many hybrids between SI and SC species of cultivated crops in the Brassicaceae (Klinger and Ellstrand 1994, Cummings et al. 1999, Ellstrand and Schierenbeck 2000, Snow et al. 2001, Ridley and Ellstrand 2009).

Two species of Cakile, C. edentula (SC) and C. maritima (SI), have colonized the same regions outside their native range and hybrids have been reported occasionally (Rodman 1974) and confirmed by using genetic markers recently (Ohadi et al. 2012, Ohadi et al. 2015). The earliest claim of hybrids in Australia dates back to Black and Robertson (1965). Rodman (1986) also suggested the occurrence of hybrids in Australia. This was further supported by artificial crosses and evidence of introgression based on nuclear markers (Ohadi et al. 2012, Ohadi et al. 2015, Mesgaran et al. 2016). Chloroplast gene sequence data also indicated that hybrids have occurred in both directions, with the selfer being either pollen donor or recipient (Ohadi et al. 2012, Ohadi et al. 2015).

Clearly, there seem to be no physiological barriers to cross-fertilisation and thus the following temporal and spatial possibilities may help account for the natural occurrence of bi-directional hybridisation. Firstly, it is possible that in neither species pollen is released before flowering; when the flowers first open, stigmas in both species are therefore clean and ready to be pollinated by either self- or foreign pollen with the help of pollinators. Secondly, even though the pollen could be released before flowering, the stigma does not become receptive before the flower opens. Self-pollen may shed onto its stigma but cannot germinate due to lack of stigmatic receptivity; after flowering, foreign pollen can be deposited on the stigma, compete with the pre-loaded self-pollen, and possibly create hybrids. Lastly, although pollen is released and the stigma is receptive before flowering, the stigma surface in a flower may always be above its anthers during floral development. Consequently, there is no opportunity for pollen contamination before flowering, even though pollen is ready to shed and the stigma is receptive.

The aim of this study was to determine the timing of pollen ripening, pollen dehiscence, stigma receptivity and structural positions of anthers and stigma in Australian populations of the Cakile species, to determine the nature of any window of opportunity for insects to facilitate hybridisation.

4.2 Materials and methods To achieve the aim of this chapter, it was necessary to determine the ripeness of the anthers, the receptivity of the stigmas and heights of floral structures at different stages. Cakile belongs to the Brassicaceae family: their flowers are regular, actinomorphic, tetradynamous and borne on elongating terminal racemes. The flowers of Cakile possess four small green nectar glands, two subtended by the unpaired stamens and another two between and subtending the paired stamens.

To incorporate inter-population variation, the seeds of both species used in this experiment were from two different locations/populations in Tasmania, i.e., Raspins Beach and Sloping Main. They were collected from plants that morphologically appeared typical of the species. Seeds were germinated on a moistened filter paper in Petri dishes and then incubated in a cold room at 5 °C for 7-10 days. Pre-germinated seeds were transferred into pots in a Poly tunnel for further growth. Each population had three pots and plants were thinned down to one in each pot after 4 weeks. For details of potting mix, see Chapter 3. Collected data were analysed using SAS Software 9.4 (TS1M4) (SAS Institute Inc., Cary, NC, USA).

4.2.1 Bud development in Cakile The first step was to produce a simple descriptive scale for floral development. Smyth et al. (1990) developed such a system for Arabidopsis (Brassicaceae), dividing its floral development into 13 different stages. The floral development was observed under SEM from the very early development of the primordia of the flower bud to the fully opened flowers. This was modified for use in Cakile. Through observation and dissection of the flowers in both Cakile species, the bud development was divided into corresponding stages in accordance with the stages mentioned in Smyth et al. (1990).

Buds along a raceme theoretically are in a temporal sequence of different stages, which makes sampling straightforward. Buds were removed from the very top to the bottom of a randomly

46 chosen raceme. These were dissected and the lengths of the pistil and stamens were measured. Pictures of the dissected buds were taken under a dissecting microscope (Leica M205A, Leica Microsystems Pty Ltd). This was repeated on three racemes for each plant.

The distance between the stamens and pistil was then calculated. Length data were analysed with SAS. The General Linear Model (PROC GLM in SAS) was used if there was no random effect or nested effect; otherwise a Mixed Linear Model (PROC MIXED in SAS) was used.

4.2.2 Anther dehiscence Anthers from a dissected flower in section 4.2.1 were categorised according to the bud’s stage number and then the dehiscence was assessed as an indicator of the pollen ripeness. Generally, most methods assess pollen ripeness by brushing anthers and germinating pollen on to a pollen growth medium; if germination is observed, it means the pollen is ripe(Rodriguez-Riano and Dafni 2000, Marshall and Diggle 2001, Boavida and McCormick 2007, Bou Daher et al. 2009). Currently, the most commonly used medium in the Brassicaceae is the solid pollen growth medium (SPGM) for Arabidopsis. Several commonly used SPGMs for Arabidopsis pollen were tested for their performance on Cakile species (for details of procedures, refer to Boavida and McCormick 2007). However, none of them induced more than 2-3 pollen tubes after 24 h. Therefore, anther dehiscence was used instead. Pollen grains are fully enclosed by the anther’s endothecium, which must rupture before any self-pollen can land on its stigma surface regardless of pollen ripeness. Anthers from each dissected bud were observed under a dissecting microscope to determine the dehiscence of the anther.

4.2.3 Stigma receptivity If a stigma is receptive, compatible pollen grains deposited on the stigma surface can germinate. Upon landing, compatible pollen grains start absorbing water from the stigma surface, inducing pollen tube growth into the style and eventually reaching the ovule (Nasrallah 2011, Kitashiba and Nasrallah 2014). The occurrence of pollen tube growth following an artificial application of pollen is therefore a simple but robust way to determine the receptivity of a stigma.

The dissected buds used in previous sections were used again to test their stigma receptivity. After dissecting, the emasculated bud was immediately planted into an agar pad, standing upright with the pedicel in the agar. This enabled buds to draw up water through the pedicel and the pistil thus could stay fresh for a few hours (see details in Edlund et al. 2016, personal communication as well). The agar pad was made from 1.5 % agarose in distilled water.

A fully opened flower on day 1, from another plant of the same species in the other population was used as the source of pollen, minimising the possible effects of self-incompatibility. An anther was removed using a pair of fine forceps and carefully brushed across the targeted stigma, to spread enough pollen evenly. Throughout, the recipient pistil was always plugged into the agar pad to keep it fresh.

Stigmatic receptivity was assessed using ABF (Aniline Blue Fluorescence) to stain the callose secreted by pollen tube when elongating, emitting blue fluorescence under UV light and thus showing the path of pollen tube growth. The procedures for detecting pollen tube growth in the style were revised from Li (2011a):

1. Two hours after pollination, the pistil was removed, placed in a 1.5 mL tube containing 150 μl of a fixing solution, and stored at 4 ℃. The fixing solution was made by adding distilled water into glacial acetic acid with the ratio of 9: 1. 2. The fixing solution was replaced with 1 M NaOH (Sigma-Aldrich) and heated at 50 ℃ for 1 h to soften the tissue and make the pollen tubes more accessible to the dye. The pistil

was washed gently three times with 50 mM KPO4 buffer. The pistil was then stained in 0.01% aniline ABF (Biosupplies, http://www.biosupplies.com.au) for 5-10 min. 3. The pistil was transferred to a microscope slide on which a drop of 50% glycerol was added and covered with a cover slip. The pistil was then squashed slightly so that it became flat. The pollen tube could be viewed under a microscope with UV light immediately or stored cold in the dark.

Pictures were taken using a camera attached to the fluorescence microscope (Leica DM 2500) and pollen tube growth was then assessed by the photographs. Massive pollen tube growth indicated the receptivity of the sample stigma; if no significant number of pollen tubes were observed, the stigma was then treated as not yet receptive.

The growth stages of pollen tubes were assessed and recorded using the following 4 levels, zero, low, medium and high. After that, the four levels were coded as 0, 1, 2, and 3, respectively when analysing data in SAS. A Generalized Linear Mixed Model (GLMM) (PROC GLIMMIX in SAS) was used to analyse the data with multinomial distribution and cumulative logit as the distribution and link function. Odds ratio and its differences between stages and species were estimated.

4.3 Results

4.3.1 Bud development Based on the observations and measurements from the collected inflorescences, I have divided the continuous process of flower development from the visible buds until the flowers start wilting into nine stages, as shown in Table 4.1. Although C. maritima had bigger floral structures compared with C. edentula (Rodman 1974), their floral development fitted these stages well.

Table 4.1 Summary of the stages of floral development in Cakile and the landmark events used to define each stage.

Stage Landmark Events 1 Tiny bud, hard to dissect 2 The stigma approximately levels with the upper stamens 3 The stigma extends above the upper stamens 4 The upper paired stamens almost reach the height of the stigma. The lower pair ones are shorter and below the stigma 5 The upper stamens are above the stigma with lower ones still being below the stigma 6 Petals are longer than sepals and flowers are about to open 7 Flowers open. The upper stamens keep growing rapidly 8 Flowers are pollinated and pistils expand quickly 9 Anthers wilted and petals have fallen off

From stage 2 onwards, buds were easy to be dissected in both species to do observations and measurements. Flower buds became visible from stage 1 but were too small to be dissected.

The stigma levelled with the upper stamens at stage 2 and then grew faster, extending above all stamens at stage 3. T upper stamens, however, grew rapidly after stage 3, levelling with the stigma at stage 4 and then staying above it until the end of stage 7. At stage 8, flowers were pollinated and thus the pistil expanded itself quickly, inducing the withering of flowers at stage 9. In all stages, the lower stamens didn’t exceed the height of the stigma.

Figure 4.1 Pictures of stages 1-4 taken under Leica M205A dissecting microscope using its built-in Z-stack function, showing landmark events in C. edentula (upper panel) and C. maritima (lower panel). In C. maritima, scale bar = 1 mm at stages 1-4; in C. edentula, scale bar = 0.5 mm at stage 1, scale bar = 1 mm at stages 2-4.

Figure 4.2 Pictures of stages 5-9 taken under dissecting microscope, showing landmark events in C. edentula (upper panel) and C. maritima (lower panel). In C. maritima, scale bar = 2 mm at stages 5-9; in C. edentula, scale bar = 1 mm at stages 5-6, scale bar = 2 mm at stages 7-9. Stages 1, 8 and 9 were excluded from any further data analysis. When analysing data, extremely small buds (stage 1) were discarded due to the failure of dissecting them under microscope. Flowers after fully opening (stages 8 and 9) were also not considered because during this period it was obvious that the stigma surface in both species was heavily covered with pollen.

During stages 1-7, both species had a similar developmental pattern: the stigma grew faster during earlier stages (stages 2-3), thus extending above the upper stamens. However, the upper stamens in both species grew rapidly after stage 4, elongating above the stigma by stage 5 (Figure 4.1and Figure 4.2).

The distance between the stamens and stigma varied from stage to stage. In both species, the lengths of the pistil and upper stamens were similar at stage 1; then the pistil became slightly longer than the upper stamens at stage 3; but after stage 4 at least part of the stamens over- topped the stigma surface (Figure 4.3). It should be noted that from stage 4 onwards, the upper stamens of C. edentula grew rapidly, resulting in a dramatically increased distance even after flowering (stage 7). In contrast, the distance between the upper stamens and the pistil in C. maritima seemed to be reversed back to a positive value instead of going further down. There were more than a quarter of the sampled buds in C. maritima showing positive values (Figure 4.3b), indicating that the stigmas in these buds were above the tallest stamen at stage 7.

Figure 4.3 Difference in length between the pistil and upper stamens in C. edentula (a) and C. maritima (b). Positive values indicate that the stigma is above the tallest anther.

Figure 4.4 Difference in length between the pistil and lower stamens in C. edentula (a) and C. maritima (b). Positive values indicate that the stigma is above the tallest anther. It is obvious that both lower and upper stamens shared the same growth pattern and extended above the stigma during later stages (stage 5 onwards) in C. edentula (Figure 4.3a and Figure 4.4a), which was not observed in C. maritima. Most lower stamens in C. maritima stayed below the stigma surface during the whole developmental period although some of them were recorded as over-topping the stigma at stages 5-7, as shown in Figure 4.4.

4.3.2 Anther dehiscence Under a dissecting microscope, it was easy to distinguish a dehisced anther from a non-dehisced one. For both species, in the early stages anthers were generally greenish and the two sacs were tightly closed. As time went by, anthers became larger and the pollen grains inside the sacs were visible due to the change in the transparency of the pollen sac wall in which pollen grains developed rapidly (Ma 2005). During this period, the anthers changed in colour to yellow, indicating that the pollen grains were nearly ripe, though the pollen sacs were still closed. In some cases, small gaps were formed but not large enough to allow pollen to be released (e.g., stage 5 of C. maritima in Figure 4.5). Finally, the pollen sac ruptured completely and copious amounts of pollen were released.

Figure 4.5 Anther dehiscence along with the development of buds in C. edentula (a) and C. maritima (b). Scale bars in C. maritima, stages 2-3 = 1 mm, stages 4-6 and stage 7 = 2 mm. Scale bars in C. edentula, stages 2-3 = 1 mm, stages 4 and 6-7 = 0.5 mm, stage 5 = 1 mm. Both species started the dehiscence process at stage 5 but the timing and the extent to which the pollen was released were completely different. In C. maritima, anthers usually did not release their pollen after stage 5 (Figure 4.5) while some of them remained closed until the end of stage 6. However, in C. edentula pollen grains were usually released early at stage 5 (see details in Figure 4.5). Moreover, when dissecting flowers, no pollen grains were found on the stigma surface in C. maritima until stage 7; in contrast, in C. edentula pollen grains could be found on the stigma surface at early stage 5.

To determine whether there was a single parameter that could predict anther dehiscence, the stage, the length of pistil, and the length of lower and upper stamens, were used as factors in a logistic regression using PROC LOGISTIC in SAS. Only the length of upper stamens explained a significant amount of variance. In C. maritima, the anthers were predicted to start releasing pollen grains after reaching a height of 5.25 mm (their final mean length was 6.15 mm). In contrast, a height of 3.85 mm in C. edentula was required for dehiscence to begin, compared with a final length of 4.90 mm.

Figure 4.6 Analysis of anther dehiscence by fitting logistic regression. Figures show predicted probabilities for anther dehiscence in C. maritima (left panel) and C. edentula (right panel). Upper stamens in C. edentula fitted better (P < 0.001) compared with that in C. maritima (P < 0.05).

4.3.3 Stigmatic receptivity Test results indicated that the stigmas in both species showed stigmatic receptivity in all stages. In C. edentula, stigmas at stage 2 induced pollen germination and pollen tube growth on their surface, but the pollen tubes mostly failed to elongate into the style (Figure 4.8). From stage 3 onwards, the pollen tubes grew into the style (Figure 4.7a), even reaching the ovules (Figure 4.8). In contrast, although the pollen in C. maritima germinated on the stigma surface at all stages (Figure 4.7b), it was obvious that the pollen tubes had not penetrated the style tract by the time when they were fixed (Figure 4.8). These differences in behaviour were clear when comparing pictures taken under light microscopy and UV light.

Further data analysis using odds ratio estimation, however, showed that above-mentioned differences were not even significant in most stages. There was no significant effect detected on the stigma receptivity between species (P > 0.05). Similarly, stigma receptivity did not vary much between different stages although it was significant that stigmas at stage 2 were less receptive compared with those at stages 3 and 7 (P < 0.05).

Figure 4.7 A close-up look of pollen grains (yellow arrows) that attached to the stigmatic papillae (black arrows) with pollen tubes (green arrows) elongating in C. edentula (a, stage 5) and C. maritima (b, stage 7) 2 h after pollination. Scale bar = 100 µm.

It should be noted that the vascular bundles also showed fluorescence, although not as obvious as the pollen callose. In many cases the pollen tubes mixed with or behind the bundles, which made it even harder to distinguish the pollen tubes from the vascular bundles (Figure 4.8).

Figure 4.8 Pollen germination and pollen tube elongation in C. edentula from stage 2 to 7 under a light microscope (before staining) and UV light (after staining). Yellow arrows indicate the pollen grains attached to the stigma surface; green arrows mark the pollen tubes growing in the style. Pictures from light microscopy were presented for a clear view of the embryos (green arrowhead) and vascular bundles (red arrows). Scale bar = 250 µm.

Figure 4.9 Pollen germination and tube elongation in C. maritima from stage 2 to 7 under a light microscope (before staining) and UV light (after staining). Pollen grains (yellow arrows) germinated, inducing pollen tube (green arrows) growth and elongation into the stigma. Pictures from light microscopy were presented for a clear view of the embryos (green arrowhead) and vascular bundles (red arrows). Stigmas were harvested 2 h after pollination. Scale bar, 250 µm.

4.4 Discussion At the beginning of this chapter, it was hypothesised that there could be a window of opportunity, when or after flowers first open, for pollinators to deposit foreign pollen before the flower release its own pollen. Such a window opportunity might be due to the timing of anther dehiscence (Ishii and Harder 2012), placement of anthers beneath the stigma (Takebayashi et al. 2006, Tedder et al. 2015) or delayed stigma receptivity (Lankinen et al. 2007). One or more of these would readily allow hybridisation between these two species with C. edentula as the pollen receiver. The results of this study, however, showed that such a window may not exist.

4.4.1 Anther dehiscence In both species anther dehiscence began before the flowers opened (at stage 5). In C. edentula the extent of dehiscence was sufficient to allow pollen shedding on to the stigma surface by the end of stage 5 whereas in C. maritima no pollen was observed on the stigma surface until its flowers opened.

This pattern of anther dehiscence in both species differs slightly from previous reports. Rodman (1974) stated that anthers dehisce introrsely (inwards) in C. edentula but somewhat latrorsely in the C. maritima. In my observations, all the anthers in both species dehisced introrsely. Meehan (1892) also suggested that the anthers often dehisced in bud just prior to flowering (i.e., stage 6 in Table 4.), which was not accurate through my observations: the anther in C. edentula dehisced much earlier, instead of just prior to flowering. However, instead of pressing stamens against the stigma and ensuring coverage with their own pollen before the bud opened (Meehan 1892), anthers of C. edentula were very close to the stigma but did not have any direct contact with it (Figure 4.5, Figure 4.10 & Figure 4.11).

4.4.2 Distance between stamen and stigma By the time anthers dehisced (stage 5), all stamens in C. edentula had elongated such that at least part of the anthers extended above the level of the stigma surface whereas most lower stamens in C. maritima stayed below the stigma surface (Figure 4.3 and Figure 4.4).

This placement of anthers above the stigma has been referred to as reverse herkogamy (Luo and Widmer 2013, Summers et al. 2015). Reverse herkogamy is an effective way to assist self- pollination in self-compatible species. With the increasing of reverse herkogamy, for example in Arabidopsis thaliana, the outcrossing rate may decrease from 10% to 0.3% (Luo and Widmer 2013). In C. edentula, it has been reported that the estimated selfing rate is over 50% (Donohue 1998). No paternity study, however, was carried out to get the precise outcrossing rate in a species in which reserve herkogamy occurred and hybrids were found. As it has been suggested that the pollination window may not exist, it is intriguing to perform paternity studies to determine the extent at which outcrossing within the species occurred in wild populations.

For a self-incompatible species, such as C. maritima, this will only be a problem if the coverage by self-pollen considerably reduces access for pollen from other plants. Indeed, reverse herkogamy has been argued to be a strategy that maximises pollen uptake and transportation, especially for Lepidopteran pollinators (Arroyo et al. 2002, Kulbaba and Worley 2012, Kissling and Barrett 2013). Even in predominant selfers such as Arabidopsis thaliana, however,

55 herkogamy does occur in natural populations and will presumably result in enhanced outcrossing (Luo and Widmer 2013, Glover 2014).

It has been reported that outcrossing rate is positively correlated with the degree of floral display (Armbruster et al. 2002, Goodwillie et al. 2010, Sicard and Lenhard 2011, Karron and Mitchell 2012, Tedder et al. 2015). Selfing species, which are less reliant on pollinators, tend to reduce their resource allocation to attraction and reward structures, and thus have smaller and less showy flowers. This “selfing syndrome” (Sicard and Lenhard 2011, Vekemans et al. 2014, Tedder et al. 2015) includes smaller and less showy flowers, smaller amount of pollen and a reduced pollen-ovule ratio. In C. edentula, the syndrome also shows as the reduction in the number of petals (Rodman 1974). For example, it is common to observe only 2-3 petals in a flower of my experimental populations (see details in Chapter 6).

4.4.3 Stigma receptivity Stigma receptivity changed gradually over the course of floral development in both species, though in both cases the stigmas were completely receptive before their flowers had opened. Even at very early stages pollen grains (ripe pollen from other plants) could germinate on the stigma, but the ability for the pollen tubes to grow towards the ovules and to penetrate them was later to develop.

Usually the stigma of a flowering plant becomes receptive after flowering and thus legitimate pollen can germinate upon deposition. For example, the stigmas in Streptanthus tortuosus (Preston 1991) and Arabidopsis thaliana (Luo and Widmer 2013) become receptive only at the time their flowers open. In some cases, species even show delayed stigma receptivity, especially in self-compatible species, which reduces the possibility of selfing and promotes outcrossing (Lankinen et al. 2007, Lankinen et al. 2015). The delay period varies between species, from a few hours to 2-3 days (Lankinen and Madjidian 2011).

However, there are few studies reporting that species can gain the stigmatic receptivity way before flowering. Normally, when pollen is deposited on the stigma surface at different stages during bud development, it will not germinate due to the lack of stigma receptivity until the bud opens (Lankinen et al. 2007). In contrast, as suggested in Lankinen and Kiboi (2007), some flowering species, especially in self-compatible ones, can regulate timing of stigma receptivity and induce early receptivity than usual. However, the timing of receptivity, according to previous studies (Lankinen et al. 2007, Lankinen et al. 2015), is commonly reported to be regulated no earlier than flowering, which means that the timing of receptivity is adjustable from flowering onwards. Thus, the whole-range stigma receptivity detected in Cakile species cannot simply be explained by the regulation of receptivity, which needs further studies.

4.4.4 Explanations for the occurrence of hybridisation The hybridisation in the field has already demonstrated that an SC species can fertilise the ovules of an SI species and produce viable seeds and vice versa (Ohadi et al. 2012, Ohadi et al. 2015, Mesgaran et al. 2016), which may be due to the pollinator behaviour. C. maritima, as an outbreeder, has larger floral display and thus is more attractive. Therefore, a pollinator visiting a plant of C. edentula is more likely to be carrying the pollen from C. maritima. For a pollinator carrying the pollen from C. edentula, it would be most likely to have come directly from a plant

56 of C. edentula. It is less like to happen when the site has many C. maritima plants except at the invasion front where there are very few C. maritima plants and lots of C. edentula. This would have to be the case even if a pollinator burrows into an unopened flower or “robs” the flower by going in through the calyx.

However, if there is no pollination window between the two species, hybridisation cannot occur even with the help of pollinators. But why are there still observed hybrids in both directions even though the pollination window between these species seemed not to be present as per the results?

Firstly, although the results suggested that it is most likely the C. edentula has been self- pollinated before flowering, pollen from C. maritima still has chances to fertilise the ovule in C. edentula. Many self-compatible plants possess physiological mechanisms in the pistil that reduce the performance of self-pollen compared to that of outcross-pollen (cryptic self- incompatibility, Bateman 1956, Carol et al. 2005, Cruzan and Barrett 2015, Lankinen et al. 2015, see details in Chapter 2). Thus, it allows plants to preferentially outcross when mixed pollen loads are applied and avoid inbreeding depression in the progeny generation. Although previous studies did not mention the outperformance of pollen from different but closely related species, it can be inferred that cryptic self-incompatibility may also present in the following scenario: mixed pollen loads on the stigma of two phylogenetically close-related species, C. maritima and C. edentula (Rodman 1976, 1980), and pollen discrimination occurs especially in self-compatible C. edentula. Even though the stigma in C. edentula, as suggested by experiment results, may be contaminated by self-pollen before flowering, the discrimination would be strong enough to ensure the occurrence of hybridisation. This area of study, including pollen competition and timing of pollination via different pollen donors, is complicated but intriguing, which is worth further investigation.

There is another possibility that pollen shedding in C. edentula happened at or after flowering although anther dehiscence occurs way before the flowers open. As mentioned in Section 4.3.2, pollen had not been released until flower fully opened although anther dehiscence was detected before flowering, which is also possible in C. edentula. When dissecting, most of the samples in C. edentula had been contaminated by self-pollen (Figure 4.10 A). It was also commonly to observe that the stigma surface was slightly covered by some pollen grains (Figure 4.10 B); However, some exceptions were also found, in which anthers were fully dehisced but the stigma surface was clean without any pollen shedding (Figure 4.10 C and D). For flowers shown in Figure 4.10 C & D, pollinators could readily transport pollen grains from C. maritima nearby and pollinate them. Therefore, it is highly possible for flowers from C. edentula, although having all anthers dehisced before flowering, to be pollinated by pollinators with deposited pollen grains from C. maritima.

Furthermore, dissecting procedures may also have directly side effects on the pollen dehiscence observation, especially in C. edentula, resulting in incorrect interpreting of the true story. Two panels in Figure 4.11 show a clue that there were influences of dissecting procedures on the process of pollen release.

Figure 4.10 Examples of pollen shedding onto the stigma at stage 7 in C. edentula. Pollen could completely cover the stigma surface (A) or a single area with complete clean stigma elsewhere (B), and even did not shed (C and D) despite fully flowering and anther dehiscence. Scale bar = 2 mm.

Figure 4.11 Pollen adhesion positions of C. edentula in different samples. All samples in the upper panel had pollen grains adhered mostly on both sides of the stigma whereas pollen was all over the stigma surface of the samples in the lower panel. Naturally, after anthers dehisced, the pollen should shed atop the stigma evenly and cover all the stigma surface (as shown in lower panel) instead of over a single and certain area, especially on both sides of the stigma. Scale bars: A and F, 1 mm; B, C and E, 2 mm; C, D, E and H, 500 µm. However, the upper panel in Figure 4.11 tells a completely different story, indicating the dissecting procedures forced the release of pollen grains. In the lower panel of Figure 4.11, it is obvious that contamination sites in all stigmas were located on both sides of the stigma surface, instead of randomly located on the stigma surface. Noticeably, there were no pollen grains on the stigma surface except those adhered at the contamination sites. Moreover, when dissecting, buds were held tightly and the holding location were the same as the sites where the pollen contamination happened. Thus, it is reasonable to suspect that the non-randomized pollen contamination of the upper panel in Figure 4.11 was caused by the holding and dissecting force

58 during dissecting processes. Thus, further in situ dissecting work may be of great importance to help understand the precise status of pollen shedding in C. edentula.

The inheritance of breeding system in hybrids between SI and SC species

5.1 Introduction The plant “breeding system”, or “mating system”, is a set of traits that a plant has evolved, spanning the reproduction processes from pollen arriving on the stigma to seed set, seed dispersal and seed germination. In its broad sense, all aspects of sex expression in plants that determine the relative genetic contributions to the next generation of individuals within species are included (Dafni 1992). In plants, there are two main breeding systems: sexual reproduction and asexual reproduction. Asexual reproduction produces new individuals without producing gametes by meiosis and subsequent gamete fusion; these are genetically identical to the parental plants unless mutations occur. Sexual reproduction produces offspring by the fusion of gametes, resulting in plants genetically different from their parents; this is the predominant mating system in flowering plants (Charlesworth 2006, Eckert et al. 2010).

The breeding system plays a key role in natural populations of flowering plants and has significant ecological and evolutionary consequences (Eckert et al. 2009). It is believed that the original flowering plants were all hermaphrodite, i.e., they had stamens as well as pistils (Richards 1997, Charlesworth 2006, Goodwillie et al. 2010). It is possible for self-pollen to act as “legitimate” pollen and fertilise the ovule, i.e., self-compatibility (SC) (see Chapters 2 and 4). SC, or selfing, acts as a reproductive assurance mechanism but results in increased homozygosity and thus restricts the opportunities for adaptation. Many plants, however, evolved mechanisms that recognise self-pollen as illegitimate, thus preventing self-fertilisation, i.e., they are “self- incompatible” (SI). This can eliminate the risk of inbreeding (Hiscock and McInnis 2003), increase genetic diversity and thus readily allow adaptation for changed conditions (Willmer 2011; Chapter 3).

There is a close relationship between SI and SC. Many modern self-incompatible species have self-fertile relatives. For example, the closest relatives of the inbreeder Arabidopsis thaliana are self-incompatible; this is common in the Brassicaceae as well as in other plant families (Charlesworth 2006). Phylogenetic studies have suggested that SI was the ancestral state in most dicots and SC species were derived secondarily from SI species (Igic and Kohn 2001, Steinbachs and Holsinger 2002). SI shifting to SC is generally ascribed to the action of single genes linked to the S-locus, i.e., the locus of self-incompatibility (SI) genes (Nasrallah et al. 2007); however, there are also cases in which the SC mutation is unlinked to the S-locus (Gaude et al. 1993). The shift from SI to SC (Vogler et al. 1999) is accompanied by changes in traits such as pollen size, pollen- ovule ratio, as well as floral display (Vekemans et al. 2014, Tedder et al. 2015), making up the selfing syndrome (see detailed review in Chapter 2).

Normally, when SI species hybridise with SC species, there are several possibilities for the inheritance of breeding systems. Firstly, it is common to observe that the SI is dominant over SC and thus inherited in all produced F1s (Zeng and Cheng 2014). Secondly, the SC overrides SI and thus F1 hybrids are self-compatible (Kakeda et al. 2000). However, in both situations, if the SC is a dominant/recessive trait ascribed to the action of single genes unlinked to the S-locus due to the separation of alleles, there could be coexistence of SI and SC plants in F2s (Nasrallah et al.

60 2007). Therefore, it is possible to observe the dominance of one breeding system in F1s while the segregation of SI and SC with an expected Mendelian ratio could be detected in produced F2s (Nasrallah et al. 2007). Moreover, in recent studies, there is evidence that small RNAs resulted in a complex system of regulation among alleles and determined the dominance level of S-alleles, which could also play a role in the inheritance of the breeding system in hybrids (Durand et al. 2014, Yasuda et al. 2017).

The breeding systems and their inheritance in hybrids determine not only the success of hybrids but also the interaction between hybrids and their parental species. Firstly, hybrids may not be generated due to the presence of SI in the parental species. It is possible that hybrids can only be obtained in one direction, typically with the SI species being the pollen donor and the SC species being the receiver (e.g., Hiscock and Dickinson 1993), termed as unilateral incompatibility (UI). Secondly, hybrids, if successfully produced, can be a major contributor to speciation and may result in the introgression of beneficial genetic variations into a species (Runquist et al. 2014, Baack et al. 2015). For example, hybridisation can help establish a stable SC species by inducing SC and promoting reproductive barriers between hybrids and their parental species. Besides, it can also allow transgenes and other crop alleles to spread to wild or weedy populations of related taxa although the success rate is low (Ellstrand and Schierenbeck 2000, Snow et al. 2010). Last but not least, interspecific hybrids with flexibility in their selection of the mating system may also contribute to the reproductive success of the parental species when facing uncertain ecological conditions (see details in Nasrallah et al. 2007), e.g., the establishment of invasive species, in which SI species may suffer from an Allee effect (see detailed explanation in Chapter 2) (Mesgaran et al. 2016).

In Australia, Cakile maritima (SI) and Cakile edentula (SC) have successively invaded along coastlines and hybridisation has been shown to have occurred (Ohadi et al. 2012, Ohadi et al. 2015). The North American C. edentula had already become widespread in Australia by the time C. maritima arrived from the Mediterranean or Western Europe (Rodman 1974, 1980). Despite the presence of self-incompatibility and facing an Allee effect upon first arrival (Mesgaran et al. 2016), Cakile maritima has completely replaced C. edentula in much of southern and eastern Australia (Rodman 1974, 1986). Field surveys have shown that the invasion fronts of C. maritima are currently in southern Tasmania, and in northern New South Wales and southern Queensland (Cousens et al. 2013). Hybrids between Cakile maritima and Cakile edentula have been found in the areas of overlap, along with evidence suggestive of introgression (Ohadi et al. 2012, Ohadi et al. 2015). Chloroplast marker data indicated that hybridisation has occurred in both crossing directions (Ohadi et al. 2015).

Rather than competition between the two species being a problem for C. maritima invasion, Mesgaran et al. (2016) have explored the possibility that hybridisation with the already-present SC species may have enabled it to overcome debilitating Allee effects resulting from it being unable to reproduce unless a suitable mate of its own species is present. The breeding system of the progeny, therefore, plays an important role in subsequent introgression. However, information on the inheritance of breeding systems in hybrid progeny is needed to test these ideas, as well as to better understand the evolutionary implications of SI x SC crossing in species in general.

61 Therefore, in this study, I used two Cakile species as a model system and performed artificial reciprocal crosses between C. edentula and C. maritima to produce F1, F2 and backcrossed hybrids, addressing the following questions: (1) Can offspring of the SI x SC hybridisation really occur in both directions? (2) Which breeding system is dominant in the resulting hybrids and do Mendelian principles apply in each generation? (3) Does the variation of breeding system in hybrids, if it exits, reflect specific traits of two sex organs (pollen and ovule), such as the pollen size and pollen-ovule ratio? (4) What are the implications of the variation of breeding systems on the likely importance of hybrids in an invasion.

5.2 Materials and methods Reciprocal crossings were performed between C. maritima (SI) and C. edentula (SC), as well as their progeny to produce F1, F2 and backcrossed progeny (BC). To incorporate as much genetic variation as possible, seeds of the two species from four populations were used (see Chapter 3 for population details).

All artificial crosses were performed at the Burnley Campus, the University of Melbourne in a glasshouse, except for producing F1s which was conducted in a poly tunnel in 2014. The glasshouse was temperature-controlled with 25℃ during the day and 15℃ at night; the day length was set as 16 hours.

For simplicity, when denoting the crossing directions and produced progeny, initials of species were used (E, C. edentula; M, C. maritima; H, F1 hybrids) with the male placed ahead of the female. [Although this is in opposition to the standard convention in plant breeding of placing the female first in the pedigree, it has no implications beyond this thesis.] For example, a cross introducing the pollen from C. maritima to the stigma of C. edentula was denoted as “ME”. Hereafter, I will use the initials to denote each crossing and its direction. These are listed in the following table. Collected data were analysed using SAS 9.4 (TS1M4) (SAS Institute Inc., Cary, NC, USA)

Table 5.1 Crossing directions and corresponding codes used in experiments. For details of seed collection and experiment design, please refer to Chapter 3.

Generation ♂ ♀ Code - - C. maritima Field collected P1 - - C. edentula C. maritima (M) C. maritima (M) MM Outcrossed P1 C. edentula (E) C. edentula (E) EE C. maritima (M) C. edentula (E) ME F1 C. edentula (E) C. maritima (M) EM F2 F1 (H) F1 (H) HH C. maritima (M) F1 (H) MH F1 (H) C. maritima (M) HM BC C. edentula (E) F1 (H) EH F1 (H) C. edentula (E) HE

After the first crosses of the two species, a random selection of F1s from the ME plant type were used to produce F2s and BCs in 2015. The EM plant type was not used due to the scarcity of

62 available seeds. All details, including methods and designs used to produce the progeny, are described in Chapter 3.

In this chapter, all produced hybrid lines, including the outcrossed P1 (MM and EE) as well as the original (C. maritima and C. edentula collected in the field) parental species, were used to assess the dominance of the breeding system and its related traits, i.e., pollen germination, pollen viability and pollen-ovule ratio. Each produced line as well as the parental individual had three replicates, which summed up to a total number of 381 plants.

5.2.1 Pollen germination and pollen tube growth The inhibition of self-pollen germination indicates what the breeding system of the recipient individual would be (i.e., SI) while different degrees of pollen germination and pollen tube growth can be used to demonstrate and describe the differences of pollen performance in the recipient’s style. Therefore, although fruit set was used for assessing the dominance of the breeding system, pollen performance was also measured. Pollen tube growth was measured when conducting artificial crossings using aniline blue fluorescence (ABF). Buds were emasculated before they opened (at stage 4 for C. edentula and stage 5 for C. maritima: see Chapter 4 for details). At this stage, only stamens were removed while petals remained intact. The emasculated buds were then covered with pollination bags and allowed to grow until open, which could be identified by the development of the remaining petals. Finally, pistils were pollinated with pollen from the appropriate pollen donors.

After pollination, samples were harvested at different intervals: 2 h after pollination (2 hap), 4 h after pollination (4 hap), and 6 h after pollination (6 hap). When sampling, three emasculated buds were collected at each interval (therefore nine samples on a single plant).

Samples were then stored in a 1.5 ml tube, containing 150 μl of the fixing solution (a mixture of ethanol and glacial acetic acid at a ratio of 9: 1). Tubes were stored in a fridge at 4 °C. When time was available, the preserved samples were taken out and stained with ABF. ABF can stain the callose secreted by the pollen tube when elongating, emitting blue fluorescence under UV light. The detailed procedures can be found in Chapter 4.

The intention was to collect samples from outcrossed progeny and hybrids, including F1s, F2s and BCs; however, C. edentula and some F1s ceased flowering too soon after the hybridisation treatments so that there were no buds available. The experimental design for pollen tube measurement was therefore an incomplete design.

Pollen germination and pollen tube growth were observed under fluorescence microscopy (Leica M250A). Pollen tube length was then assessed using ImageJ (https://imagej.nih.gov/ij/). Pollen tube germination was expressed as a categorical variable with four levels: zero, low, medium and high. A Generalized Linear Mixed Model (GLMM, PROC GLIMMIX in SAS) was used to analyse the data with a multinomial distribution and cumulative logit as the default distribution and link function, respectively. Odds ratios were compared between pollen tube growth levels for each cross. The number of germinated and the total adhered pollen grains were also recorded when producing F1s to assess any difference between the two crossing directions; the PROC GLIMMIX and “event/trial” syntax was used for analysis.

63 5.2.2 Fruit and seed set Fruit and seed set were recorded after artificial pollination in each generation; pollinated buds were covered with mesh bags and the total number of pollinated buds in each crossing was recorded.

Fully developed fruits were counted about two weeks after pollination and harvested after the plants matured. The fruits were then stored in paper bags for further use. Fruit set percentage

(Rf) for a given cross was calculated as:

푡표푡푎푙 푛푢푚푏푒푟 표푓 푓푟푢푖푡 푠푒푡 푅 = 푓 푡표푡푎푙 푛푢푚푏푒푟 표푓 푝표푙푙푖푛푎푡푒푑 푏푢푑푠

As Cakile fruits are dimorphic, the number of fully developed proximal (Np) and distal segments

(Nd) were also counted. The fruit wall in both segments was removed and fully developed seeds were counted; aborted or abnormal seeds (obvious to the naked eye) were omitted. The total number of fully developed seeds (Ns) was used to calculate the seed set rate as:

푁푠 푅푠 = 푁푝 + 푁푑

Data were analysed using the Generalized Linear Model (PROC GLM in SAS). Crossing direction in each generation was used as a fixed effect to model its impact on fruit and seed set rate. Differences between directions were estimated using the “MENAS” syntax and pairwise comparisons were based on Tukey’s post hoc test.

5.2.3 Pollen viability The viability of pollen in each produced plant type as well as the parental species cultivated in the field station was assessed using fluorescein diacetate (FDA) (Heslop-Harrison and Heslop- Harrison 1970). This method stains pollen grains for the enzyme as the vital indicator of membrane integrity and only fertile grains fluoresce under microscopic examination.

Fresh anthers were collected from open flowers on day 1 of flowering and carefully brushed onto the centre of a clean slide. The released pollen grains were then stained with a drop of stock solution of FDA (2 mg FDA/mL of acetone) and observed under an optical microscope (Leica M250A) in blue light (wavelength = 495nm). The viable pollen grains showed fluorescence (FCR+), which was normally light white to bright blue. Details of procedures can be found in Li (2011b).

The number of viable as well as total pollen grains were counted, and the percentage of viable pollen was calculated. All lines were used, and each line had three replicates. In each replicate, three samples were collected for pollen viability and their average percentage of the viable pollen was used. In some extreme cases, one replicate did not have any viable pollen; to deal with this, the averaged percentages in three replicates of all the plants were then averaged again to get the final averaged value, which was used to perform data analysis.

Data were analysed using the General Linear Model (PROC GLM in SAS). The differences of means between each cross were identified by Tukey’s post hoc tests, using the “means/tukey” syntax in SAS.

64 5.2.4 Determination of self-incompatibility SI prevents incompatible pollen, including self-pollen, from germinating and stops pollen tube growth. Therefore, the germination and tube growth of self-pollen on the stigma surface in a plant can be used to determine whether the SI is inherited in produced progeny (Nasrallah et al. 2007, Goodwillie and Ness 2013).

However, due to the high number of crosses, fruit set following artificial selfing, was used to determine the breeding system in this study (it has been used previously in Cakile maritima: A. Young, pers. comm. to R. Cousens). Determination of the breeding system was made on the same plants that were being used to generate the next generations of crosses. A raceme was randomly chosen for artificial selfing on each plant and covered with a pollination bag to exclude pollinators. Instead of emasculating and pollinating buds, newly-open flowers (day 1) of the bagged racemes were simply squeezed to make sure the stigma surface had been fully covered by self-pollen. This procedure was repeated on all flowers of the bagged racemes until no further flowers occurred. On each plant, at least 50 flowers were treated in this way. After plants senesced, the number of successfully set fruit in each plant was counted. Since even in SI species occasional seed production can occur, e.g., matromorphy, plants with less than two fruit set were treated as self-incompatible.

The proportion of self-compatible vs. self-incompatible plants was calculated in each generation to determine the dominance of the breeding system. As described above, it is possible that the SC gene (designated as a) is a single mutation unlinked to the S-locus but suppressed by the usual form of the gene (A) when they both occur in the genome. Therefore, it seems reasonable to assume that C. maritima has a genotype of AA while C. edentula is aa. If the two species are crossed, the genotype of all generated F1s should be Aa, indicating the presence of the SI phenotype in all produced F1s. Furthermore, if we take random crosses of these F1s, the genotypes of F2s are expected to segregate out as AA, Aa and aa at the ratio of 1:2:1. So there will be 3 A phenotypes for every 1 a (i.e. 3:1 SI:SC).

Data were analysed using a Generalized Linear Mixed Model (GLMM) (PROC GLIMMIX in SAS). The occurrence of an SC plant in each plant type was recorded as “1” and PROC GLIMMIX was used to model the probability of the event = “1” in each species. Binary distribution and logit function were invoked as the default distribution and link function, respectively. The “random _residual_” syntax was used to achieve the convergence and correct the over-dispersion of the model.

5.2.5 Pollen-ovule ratio In general, SC species tend to have lower pollen-ovule ratios than SI species (Tedder et al. 2015). In this section, pollen-ovule ratios of the parental species and their progeny were measured to help further evaluate the effects of hybridisation on the breeding system.

A freshly opened flower (day 1) was collected in the morning. As mentioned in Chapter 3, although anthers dehisced before flowering, in many cases, there were only a few pollen grains shedding on the stigma of C. edentula. Furthermore, it was also common to observe clean stigmas in C. edentula. Therefore, when collecting flowers in all plant types, I assumed that there

65 was no pollen loss despite the early anther dehiscence before flowering in C. edentula and its related hybrids.

All anthers in the flower were carefully removed, squeezed gently along the line of dehiscence to promote pollen release. Then anthers were put into a 1.5 ml Eppendorf tube with several drops of 90% ethanol and stored at room temperature for a few days to allow the pollen grains to be released. At this stage, samples could be stored up to several months. All plants were used and three flowers on each plant were sampled.

To prepare the pollen for counting, additional ethanol was added to each Eppendorf tube, making a final volume of 1 mL. Tubes were then agitated and vortexed to fully release and to suspend the pollen. A sample of 20 µl suspension was taken for pollen counting using a haemocytometer with a cover slip under a microscope at 50x magnification. For each sample, five counts were performed, among which the highest and lowest readings were discarded. The mean value of the remaining counts was used to estimate the total pollen production per flower based on the volume of the solution held in each tube.

To measure the ovule number per flower, pistils from the same samples were kept in a 1.5 mL Eppendorf tube. The pistils were immersed in 1 ml 90% ethanol for 4-7 days to soften the tissues. After softening, the pistils were placed onto a microscope slide and then dissected using a pair of fine forceps under a dissecting microscope to expose the ovules. The number of ovules in each sample was counted.

When analysing the ovule count data, PROC LOGISTIC in SAS was used to fit a cumulative logit model in which the probabilities were modelled and summed over the responses (ovule number) having the lower ordered values. The LSMEANS syntax was used to distinguish any differences of ovule number between crossing directions and generations.

Data for pollen number and pollen-ovule ratio were analysed using a Mixed Linear Model (PROC MIXED in SAS). In the analysis of pollen-ovule ratio, data were square root transformed after checking the residual plot of the original data. Cross (including parental species) was treated as a fixed effect while the “plant (cross)” syntax was used as a random effect since plants were nested within crosses.

5.3 Results

5.3.1 Pollen germination and pollen tube growth When producing F1s, the number of adhered pollen grains differed significantly (P < 0.001) with 30 ± 3.28 and 11 ± 2.39 pollen grains attached on the stigma in the crossing direction of ME and EM, respectively; pollen germinated rapidly after pollination in both reciprocal crossings and there was no significant difference in the number of germinated pollen grains between crossing directions (F = 4.38, P > 0.05). Three out of 10 plants in C. maritima, however, had no pollen of C. edentula germinated on their stigma surface, although some pollen grains successfully attached (Figure 5.1a). Such unsuccessful germination was not observed in producing ME (i.e. pollen from C. maritima on the stigma of C. edentula).

66 In all the samples that had pollen successfully germinated, the pollen tubes grew rapidly into the style within the first collection time interval (2 h) (Figure 5.1b & c). In some cases, the fertilisation process was captured under the microscope in which the pollen tubes had reached the ovule (Figure 5.1d).

Figure 5.1 Pollen germination and pollen tube growth under UV light 2 h after pollination when the pollen of C. edentula landed on the stigma of C. maritima (a and b) and vice versa (c and d). Figure a, b and c were under 5x magnification while d was under 10x magnification. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule.

Figure 5.2 Examples of pollen tube growth in the intraspecific outcrossing of C. edentula (a and b) and C. maritima (c and d). Figure a and b were under 10x and 20x magnification, respectively; while c and d were under 5x magnification, respectively. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule; SP, stigmatic papillae.

Figure 5.3 Pollen germination and pollen tube growth under UV light 2 h after pollination in the backcrossing of F1s with C. maritima (F1 was the pollen donor in a and b) and interspecific crossing of F1s (F1s were the pollen donor as well as recipient in c and d). All figures were taken under 5x magnification. Arrows in different colours indicate different tissues: PG, pollen grain; PT, pollen tube; OV, ovule. As shown in Table 5.2, the pollen tube growth was significantly worse when producing hybrids compared with that within the original parental species (i.e., producing outcrossed P1, EE and MM) (P < 0.05; see Table 5.1 for details about the pollen donor and recipient of each plant type). However, no significant difference was detected in pollen tube growth when producing F1s, F2s as well as BCs.

Table 5.2 Estimated odds ratios of (i.e., differences between) the degree of pollen tube growth when producing EE, MM, F1s, F2s and BCs. Estimated odd ratios were based on the higher value of pollen tube growth*. Significant comparisons are in bold (P < 0.05). Only parts of pollen tube growth when producing hybrids were sampled due to the limited availability of buds in each plant type. Therefore, pollen tube growth condition was not sampled when producing MH.

Pollen tube growth in producing Pollen tube growth in producing Odds ratio 95% Confidence Limits

HM EE <0.001 <0.001 0.104 HM EM 0.069 <0.001 12.229 HM ME <0.001 <0.001 >999.999 HM HH 0.082 <0.001 8.796 HM MM <0.001 <0.001 0.009 EE EM 288.756 0.415 >999.999 EE ME <0.001 <0.001 >999.999 EE vs. HH 341.185 0.417 >999.999 EE MM 0.164 <0.001 63.282 EM ME <0.001 <0.001 >999.999 EM HH 1.182 0.004 382.825 EM MM <0.001 <0.001 0.258 EM HH >999.999 <0.001 >999.999 EM MM >999.999 <0.001 >999.999 HH MM <0.001 <0.001 0.225 * PROC GLIMMIX was invoked and plants were treated as a random effect nested within plant types. To allow the procedure model perform estimation based on the higher values of the pollen tube growth, the “descending” syntax was used to reverse the order of the response categories (i.e., pollen tube growth). For example, the pollen tube growth when producing BCs (HM) compared with it was when producing EE had an estimated odds ratio of < 0.001, which means it was more than 1000 times (= 1 / 0.001) likely to have worse pollen tube growth when producing HM. Confidence limits where the value “1” are included indicate the corresponding comparisons are not significant.

5.3.2 Fruit and seed set Reciprocal crossings were conducted when producing F1s, F2s and BCs, in which fruit set was observed in all directions but the proportion varied in all plant types as well as within generations. Specifically, both directions of the cross between C. maritima and C. edentula (i.e., EM and ME) had seed set; however, the proportion of fruit set was significantly higher when C. edentula was the maternal parent (P < 0.001). Similarly, when producing BCs, the fruit set percentages of the four directions involved were different from each other, with the lowest proportion in the cross of EH (0.14 ± 0.05 (SE), P < 0.01). In contrast, the fruit set proportion of the two different crossing directions when producing F2s did not differ significantly, with a mean rate of 0.3 ± 0.05 (SE).

70 Fruit set Seed set proportion proportion

0.30 ± 0.05 0.28 ± 0.02

H H

M 0.25 ± 0.05 E M 0.68 ± 0.06 E 0.79 ± 0.02 0.85 ± 0.03

0.51 ±0.06 0.66 ± 0.02 0.75 ±0.03 0.67 ± 0.06

Figure 5.4 Proportion of fruit set and seed set (± SE) of all reciprocal crossings conducted in this study. M, C. maritima; E, C. edentula; H, F1 hybrids. There were two and four different crossing directions when producing F1s and BCs, respectively, i.e., EM and ME in producing F1s, and F1 hybrids backcrossing C. edentula (EH and HE) and C. maritima (MH and HM) in producing BCs. Arrows in different colours indicate different crosses: red, reciprocal crossing producing F1s; blue, backcrossing between F1s and parental species; black, outcrossing within parental species and F1s. The seed set in different crosses also varied and had a similar pattern compared with their fruit set. When producing F1s, the direction of “ME” again had a higher proportion of seed set than “EM” (P < 0.05). However, when producing BCs, the significance levels between crosses were different (the differences of HE vs. HM and HM vs. MH were not significant) although the pattern shown in fruit set also presented in the seed set.

5.3.3 Pollen viability The parental species and its outcrossed progeny did not show much superiority in pollen viability compared with their hybrids. Differences in the proportion of viable pollen between plant types were not significant except the following three pairwise comparisons: MM vs. EE, EM and HH (P < 0.05). Outcrossed C. maritima (MM) plants tended to have higher percentages of viable pollen although mostly it was not significant. In EM and F2s (HH), two lines (3 plants in each line) were found to be male sterile, i.e., no pollen grains showed fluorescence after FDA staining (Table 5.3)

Figure 5.5 Examples of pollen viability of C. martima (a), C. edentula (b), F1s (c, HM), F2s (d) and BCs (e, EH; f, MH). Pollen grains were stained with FDA and observed under UV light (wavelength = 495 nm). Bright pollen grains were judged as viable while those greyed out were not. Scale bar = 100 µm.

Table 5.3 The proportion of viable pollen in parental species and all produced hybrids. Pollen viability was assessed using FDA. Significance groupings (lower case letters) were based on Tukey’s post hoc tests (pair-wise comparisons with a significance level of P < 0.05). Means with the same letter are not significantly different. Minimum pollen viability with a value of zero is in bold.

Generation Plant type Pollen viability Max Min SE C. maritima 0.52 ab 0.82 0.19 0.07 MM 0.62 a 0.78 0.42 0.07 P1 C. edentula 0.53 ab 0.80 0.28 0.05 EE 0.25 b 0.32 0.18 0.07 EM 0.30 b 0.51 0.00 0.05 F1 ME 0.37 ab 0.53 0.14 0.03 F2 HH 0.30 b 0.59 0.00 0.06 EH 0.39 ab 0.83 0.07 0.08 HE 0.35 ab 0.61 0.06 0.04 BC HM 0.45 ab 0.83 0.15 0.05 MH 0.52 ab 0.82 0.35 0.04

5.3.4 Determination of self-incompatibility In the parental species, C. edentula was completely self-compatible while C. maritima was mostly self-incompatible (see self-pollen tube growth of C. maritima in Appendix 2- 1 & Appendix 2- 3) with just a few spontaneously set fruits on a single plant. Meanwhile, C. edentula had a high selfing rate. The selfing rate of the two treatments in C. edentula, natural selfing and artificial selfing in the glasshouse, did not vary significantly (P > 0.05) but was much higher than the natural selfing rate outside in the field station (P < 0.05).

72 Table 5.4 Proportion of fruit set in the artificial or natural selfing in C. edentula. All selected racemes were bagged to prevent potential pollinator’s visitation. Significance groupings (lower case letters) are based on Tukey’s post hoc tests (pair-wise comparisons with a significance level of P < 0.05). Means with the same letter are not significantly different.

Treatment Number of plant Mean SE Minimum Maximum Artificial selfing in the 10 0.82a 0.01 0.76 0.92 glasshouse Natural selfing in the 10 0.81 a 0.02 0.71 0.87 glasshouse Natural selfing in the field 12 0.48 b 0.04 0.28 0.70

Only a small fraction of the tested plants in hybrids showed self-compatibility and the SI: SC ratios observed did not obey the classic Mendelian segregation rules (Table 5.5). In F1s, most plants were SI except two plants of ME. Similarly, three out of 36 F2s were self-compatible. In contrast, the number of SC plants in the BCs tended to increase compared with that in F1s, especially in HE with six SC plants.

The selfing rate of the SC plants between plant types did not differ significantly (P > 0.05). However, there was a tendency that the selfing rate, although low in SC plant of F1s, increased in F2s as well as in the backcrosses with C. edentula, i.e., HE and EH. In contrast, the selfing rate in the backcrosses with C. maritima (HM and MH) seemed to be smaller than it was in F1s.

There was also a similar tendency in the proportion of SC plant in each plant type. Data analysis using GLMM models indicated that the occurrence of SC plants, although variable, did not differ significantly between plant types (P > 0.05). However, it seemed like that the proportion of SC plant increased in F2s and BCs, especially the backcrosses with C. edentula (HE and EH).

Table 5.5 Breeding system in the parental species and their progeny. The “selfing rate” column in the table indicates the selfing rate in those SC plants of each plant type. Selfing rate was estimated in the field station by artificial selfing on chosen racemes covered with pollination bags. “1: 0” and “0: 1” ratios (SI: SC) indicate that all plants in that row were self-incompatible (1: 0) or self-compatible (0: 1).

Generation Plant type Total SI plant SC plant SI: SC ratio Selfing rate (SE) C. maritima 27 27 0 1: 0 - MM 12 12 0 1: 0 - P1 C. edentula 33 0 33 0: 1 0.81 (0.05) EE 6 0 6 0: 1 0.50 (0.18) EM 42 42 0 1: 0 - F1 ME 45 43 2 43: 2 0.13 (0.03) F2 HH 36 33 3 11: 1 0.45 (0.15) EH 33 30 3 10: 1 0.22 (0.04) HE 45 36 9 4: 1 0.27 (0.15) BC HM 51 46 5 46: 5 0.10 (0.04) MH 51 48 3 16: 1 0.10 (0.03)

73 5.3.5 Pollen-ovule ratio

Pollen grain number in a flower Natural C. maritima produced many more pollen grains than any other plant type (P < 0.001) while the second highest pollen number was found in artificially outcrossed progeny of C. maritima (MM) (P < 0.001). F1s (EM and ME) and the backcrosses with maternal C. maritima (HM) produced more pollen grains per flower (P < 0.001) than other hybrids in which they had a similar number of pollen grains.

Table 5.6 Pollen grain number (back transformed) and pairwise comparisons between plant types. Significance groupings (lower case letters) are based on LSMEANS (pair-wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different.

Generation Plant type Pollen grain number C. maritima 12960 a MM 7854 b P1 C. edentula 2608 e EIC 2617 e EM 5320 c F1 ME 5426 c F2 HH 2557 e EH 3130 e HE 2944 e BC HM 4336 d MH 2901 e

Ovule number The variation of ovule number in plant types presented a different pattern compared with that of pollen grain number. In the parental species, C. edentula produced more ovules than C. maritima (including MM) (P < 0.01) and other plant types (though not all comparisons were significant). In hybrids, there was little difference between F1s, F2s or the backcrosses with C. edentula (EH and HE). However, significant differences were observed within BCs: EH and HE produced more ovules than HM and MH did, with no significant difference within these two groups (P > 0.05). Furthermore, the backcrosses with C. maritima (HM and MH) also produced fewer ovules compared with any other plant types, as shown in Table 5.7.

74 Table 5.7 Ovule number produced by each plant type. Significance groupings (lower case letters) are based on the differences of least squares means of the odds ratios in ovule number per fruit, estimated using PROC LOGISTIC. Groups with the same letters are not significantly different (P < 0.05).

Generation Plant type Mean SE Minimum Maximum C. maritima 2.70 bc 0.15 2 4 MM 2.33 cd 0.14 2 3 P1 C. edentula 3.17 a 0.13 2 4 EE 3.00 ab 0.37 2 4 EM 2.97 ab 0.09 2 4 F1 ME 2.73 b 0.10 2 4 F2 HH 2.94 ab 0.14 2 5 EH 3.14 a 0.14 2 4 HE 2.89 ab 0.10 2 4 BC HM 2.22 d 0.09 1 4 MH 2.41 cd 0.10 1 4

Pollen-ovule ratio Pollen-ovule ratio had a similar pattern of variation compared with pollen grain number. C. maritima, both natural and artificial (MM), had the largest ratio in all plant types while F1s and one type of BC (HM) had a significantly larger ratio than other hybrids.

In parental species, C. maritima had the largest pollen-ovule ratio (P < 0.001) (no significant difference between C. maritima and MM, P > 0.05) while C. edentula and EE had the smallest pollen-ovule ratio (Table 5.8).

F1s (EM and ME) and HM, although not significantly different from each other, had a significant larger ratio than any other plant types in hybrids, (P < 0.001). The pollen-ovule ratios in F1s and HM were also much larger than those in C. edentula and EE. F2s had the smallest pollen ovule: ratio in all produced hybrids although not significant when compared with the backcrosses with C. edentula (EH and HE).

Table 5.9 Pollen-ovule ratio (back transformed means). Means with the same letter are not significantly different.

Generation Plant type Pollen-ovule ratio C. maritima 5,017 a MM 3,495 a P1 C. edentula 845 d EE 918 cd EM 1,839 b F1 ME 2,057 b F2 HH 979 d EH 1,018 cd HE 1,066 cd BC HM 2,026 b MH 1,292 c

75 5.4 Discussion Hybridisation between Cakile species with contrasting SI and SC breeding systems could indeed be achieved in both directions and SI was dominant in the F1s. The produced F1s hybrids could be easily backcrossed with their parental species, generating backcrosses within the two crossing directions. Such results indicate that allopolyploidisation, a phenomenon frequently reported in Brassica species that may result in the breakdown of SI in hybrids (Chen et al. 2011, Cheng et al. 2014), did not exist in the hybrids between species. However, differences between crossing directions were detected within reproductive processes, e.g., pollen tube germination and growth, and post-fertilisation, such as fruit set and pollen viability in progeny. Furthermore, although segregation of breeding systems occurred in F2s and BCs, their segregation ratios did not obey the Mendelian rules, which suggested that the dominance of SI over SC is much more complicated than the proposed invariant dominance relationship. Besides, when testing pollen size and pollen-ovule ratio in hybrids, little difference was observed in the former but there were significant differences between generations in the latter. The implications and potential interactions of these results will be discussed in this section.

5.4.1 Breeding system barriers upon hybridisation As previously described in Chapter 2, Cakile species have an SSI (sporophytic SI) system, in which compatible pollen on its dry stigma adheres to the surface and then absorbs water from the stigmatic cells (Richards 1997, Willmer 2011). After hydration, the penetration of the pollen tube happens in the cuticle of the stigma and this grows in through the cells (Richards 1997). However, for incompatible pollen landing on the stigma surface, the secretion of factors for proper rehydration and germination is prevented. In some cases, even the adhesion of pollen can be inhibited, leading to the rejection of incompatible pollen before rehydration (De Nettancourt 1997, Hiscock and McInnis 2003).

In my experiment, I did not observe any complete SI rejection when producing all hybrids, with proper pollen adhesion and germination occurring in all reciprocal crossings. However, the number of pollen grains attached (although data were only collected for F1s) varied and the difference of pollen tube growth between the parental species as well as artificial crossing directions was detected. Fruits developed in all pollination directions but it varied significantly between directions. Furthermore, pollen viability appeared to vary with a tendency of decreasing in produced hybrids (Section 5.3.3) but this would require further investigation.

Such differences indicate the presence of hybridisation barriers related to the breeding system (see details in Chapter 2) in many aspects. Firstly, the difference in pollen germination and pollen tube growth when producing F1s suggested the presence of partial self-incompatibility, although fully functional SI was not detected. Partial SI was firstly introduced as a term by Cruzan (1989) and is indicated by pollen tube attrition in the style, in which pollen tube growth ceases. This stylar discrimination based on pollen tube attrition may share the same genetic basis as self-incompatibility in some species (Plitmann 1993, Cruzan and Barrett 2015, Mazer et al. 2016, Swanson et al. 2016). When producing F1s, the pollen tubes of C. edentula grew worse in the style of C. maritima than the pollen tubes of C. maritima in the reciprocal cross although SI did not prevent pollen germination and pollen tube growth. Similarly, it is not surprising to observe that the pollen tube growth in all artificial crossings (i.e., producing F1s, F2s and BCs) was worse

76 than in the outcrossing of the original parents (i.e., producing EE and MM) (P < 0.05) since SI was still dominant in most produced progeny (see Section 5.3.4).

Secondly, the difference of fruit and seed set between reciprocal crossings indicates the existence of post-zygotic barriers, though these barriers may only be partial. Normally, when fertilisation occurs, seeds will be generated and develop successfully. However, barriers may still exist after the pollen tube arrives at and penetrates the ovule, ceasing further development due to late-acting self-incompatibility (LSI) and inbreeding depression (ID) Gibbs (2014) (reviewed in Bittencourt et al. 2003). It should be noted that ID had been considered and avoided to a large extent in the experiment design (see details in Chapter 3). Therefore, the most likely reason for such differences in fruit and seed set rate was LSI, with SI occurring within the ovary, either prior to fertilisation or as the result of the abortion of ovules or fruits (see details about SI in Section 2.2.2). Hiscock and Allen (2008) suggested that LSI is more widespread among the basal group of angiosperms and a more recent review by Gibbs (2014) indicated that there are at least 67 genera with the late-acting self-incompatibility syndrome.

Thirdly, the variation in pollen viability also indicates that there are genetic barriers for the hybridisation between these two species. Reduction in pollen viability is commonly used as an indicator of potential hybridisation barriers (Burton and Husband 2000, Goodwillie and Ness 2013). In this study, although hybridisation was achieved, there was a reduction in the male fertility (i.e., pollen viability) to some extent, especially in the produced progeny. Moreover, there were several plants that had completely lost their pollen viability, i.e., they were male sterile (Table 5.3). Since pollen viability can affect the result of pollen germination and pollen tube growth to a large extent (Xiangwen et al. 2010, Swanson et al. 2016), such variation of pollen viability could in turn further help explain the relatively poor pollen performance observed when producing BCs and F2s (Xiangwen et al. 2010) in which pollen of the parental species and F1s was involved. In other words, the inferiority of the F1 pollen compared with the parental species in producing BCs and F1s was at least partially caused by its relatively low viability.

5.4.2 Inheritance of the breeding system Self-compatibility (SC) can be ascribed to a single suppressor or modifier gene unlinked to the self-incompatibility (SI) locus whose action seems to be strictly limited to the stigmatic tissue (Nasrallah et al. 1991, Nasrallah et al. 2004). Such unlinked SC genes and the consequent segregation of SI and SC have been reported previously (Gaude et al. 1993). In such cases, the SI is normally dominant over SC upon interspecific hybridisation (Nasrallah et al. 1992, Nasrallah et al. 2004). Therefore, all F1s should be SI while the segregation ratio of SI to SC in F2s should be 3:1 (Nasrallah et al. 2004).

However, things are rarely so simple in the real world. When SC genes are linked to the S-locus, the dominance and segregation may be much more complicated (Gaude et al. 1993). The action of linked SC genes (designated as S0) seem to be not limited to the stigmatic tissue; instead, they could also express in pollen and interact with other functional S-alleles (Sx) (Nasrallah et al. 2007). However, even though SC genes are linked to S-locus, theoretically the segregation ratio in F2s should still obey the Mendelian rules as single locus genes, i.e., if Sx is dominant over S0, the expected ratio should still be 3 Sx: 1 S0 in terms of phenotypes.

77 In this study, although the SI was dominant over SC in F1s, the segregation ratio in F2s was very different than expected (expected 3:1 vs. recorded 11:1), indicating the SC in these Cakile species may be not a simple recessive gene unlinked to the S-locus. Furthermore, segregation ratios in F2s as well as backcrosses also violated the expectation. This implies that the system involved in Cakile may not be simply explained by such mechanisms. So, why did the breeding system segregate but its ratios were not as expected? There are two possibilities that could account for the occurrence of abnormal ratios in this study.

Firstly, it could be due to the interspecific crossing per se. Interspecific crossing, compared with intraspecific crossing, is much more complicated as it involves both pre-zygotic and post-zygotic barriers for hybridisation (Roccaforte et al. 2015, Christe et al. 2016, Vallejo‐Marín and Hiscock 2016). Even though artificial hybridisation could be achieved in both directions, it is obvious (as discussed in Section 5.4.1) that both F1s and F2s suffered from poor pollen performance, indicating that there should be barriers upon hybridisation between Cakile species. Such barriers then could make the segregation of genes abnormal and not as expected. Mendelian rules are commonly applied in the case of alleles within the same species, where there are no barriers. It is not necessarily to be expected that the Mendelian rules will apply to a scenario that involves interspecific hybridisation in which crossing barriers could interrupt the segregation, leading to abnormal ratios (but see Nasrallah et al. 1992, Nasrallah et al. 2004).

Secondly, the segregation ratio in the F2s and BCs produced in this experiment could also be affected by the experimental design. By definition, the F2 generation is the result of the selfing or a cross between F1 individuals in which the chosen F1 individuals should be within the same line. In such a case, the classic Mendelian segregation ratio of a typical trait, e.g., SI or not, should be 3:1 provided SI is dominant. The same rationale is applied when producing BCs. In contrast, the “F2s” (HH) in this experiment were produced by crossing different lines of F1s to avoid possible effects of inbreeding depression on further fitness estimation (see experimental design in Chapter 3). Mendelian rules about the segregation ratios of a single allele, therefore, at least partially, could not be applied in this study since different lines were involved. Combining the experimental results with the discussion presented above, it is reasonable to speculate that SI in these Cakile species is dominant over SC and should be a trait controlled by multiple alleles instead of a single pair of alleles. The speculation is based on and can be explained by three different aspects: dominance of SI plants, variation in the proportion of SC plants and the selfing rate of SC plants within each plant type. Firstly, most plants in the hybrids of different generations were SI although several plants were SC, implying that SI was overall dominant to SC. Secondly, although the majority was SI, the percentage of SC plants in F1s (2%) were slightly lower compared with F2s (9%). Such a trend also occurred within backcrosses: the backcrosses with C. edentula (EH and HE) had a higher percentage (15%) of SC plants than those with C. maritima (MH and HM) (8%) although no statistical significance test was conducted. Such a difference suggests that the SI in Cakile may have been degraded from F1 to F2 due to the recombination of recessive alleles; furthermore, the backcrosses with C. edentula (SC species) could even accelerate the degradation process. However, the SI could also be reversed after generations of backcrossing with C. maritima, i.e., an SI species. Thirdly, the selfing rate in the SC plants of each generation may help further explain the enhancement. In F1s, there were only two SC plants with an extremely low selfing rate (0.13 ± 0.03); however, with an increasing

78 proportion of SC plants, the selfing rate tended to increase in those self-compatible F2s (0.45 ± 0.15) as well as the backcrosses with C. edentula (i.e., EH and HE; 0.22 ± 0.04 and 0.27 ± 0.15, respectively). These increases in the proportion of SC plants and the selfing rate indicate that the SC in the population had been enhanced. Therefore, the co-variation of the proportion of SC plants and their selfing rates in each generation could serve as a proxy to estimate the strength of SI and SC in each generation after hybridisation: a degradation of SI may be commonly accompanied by an increasing proportion of SC plants in which the selfing rate increases as well.

5.4.3 Impact of the variation of breeding system on sex expression Plant breeding system, in its broad sense, has vital impacts on all aspects of sex expression in plants and determines their relative genetic contributions to the next generation (Dafni 1992). Therefore, in accordance with the variation of the breeding systems in produced progeny, there should be a variation in the related sex expression, e.g., pollen-ovule ratio and pollen size. Generally, SI species rely on animals for pollination and thus have larger flowers, along with a greater pollen-ovule ratio and a larger pollen grain number to counteract the inefficiency of the process (Armbruster et al. 2002, Goodwillie et al. 2010, Karron and Mitchell 2012). Correspondingly, selfing species, which are less reliant on pollinators, tend to reduce their allocation to attraction and thus have smaller and less showy flowers. Meanwhile, the greater reproductive assurance leads to a smaller pollen number and pollen-ovule ratio (Sicard and Lenhard 2011, Luo and Widmer 2013, Tedder et al. 2015). Such variations, in turn, could be an indicator of the potential variation between different breeding systems.

The variation of pollen grain number and pollen-ovule ratio in this study was largely consistent with the variation of the breeding systems between the parental species. In Section 5.4.2, I presented and discussed the inheritance of SI in all hybrids. Similarly, the pollen-ovule ratios in hybrids showed corresponding results (Table 5.9): although only two generations were tested, there could be a general tendency that both pollen-ovule ratio and the number of pollen grains reduced generation by generation (i.e., F1s to F2s), which could be accelerated or reversed via backcrossing with the SI parent. Therefore, the consistency in this study of the sex expression traits (i.e., pollen-ovule ratio and pollen grain number) and the breeding system confirmed that these two parameters can be used in predicting and reflecting the variation of the breeding system to some extent.

It should be noted, however, that there were several inconsistent results in the results, mostly in the BCs, which could be due to the short time scale involved in this study. In most previous studies, these parameters have been measured in species that have been in self-compatible or self-incompatible state for generations (Tian et al. 2014, Tedder et al. 2015). The impact of the variation in the breeding system on sex expression may still be at its early stages. It is possible that the breeding system in hybrids is not fixed and thus needs time before showing significant and steady impacts on related sex expression traits.

5.4.4 Ecological implications In this study, I determined that the SI in Cakile species could be a trait controlled by multiple alleles and inherited in hybrids. It could be enhanced or weakened by different crossing directions due to recombination of alleles. Such results may have important ecological implications.

79 Firstly, the speculation of the inheritance of SI as a trait controlled by multiple alleles could help explain how C. maritima, as an SI species, successfully established its populations and invaded along fragmented coastlines despite a possible lack of mates. The occurrence of hybridisation with an SC species already present may provide isolated plants of the SI species (i.e., C. maritima in this case) with mates, at least temporarily, and thus help its establishment within the new environment (Mesgaran et al. 2016). Mate availability can be a lethal problem for a new arrival, especially in an SI species (Mayr 1942, Provine 2004). The compatibility of the SI and SC species could potentially provide a flexible breeding system, acting as a bridge to overcome demographic constraints such as Allee effects (Allee 1931, cited by Stephens et al. 1999), e.g., pollinator limitation (see details in Stephens et al. 1999, Rodger et al. 2013). Subsequent backcrossing may have the added benefits of the introgression of greater variation into the incoming SI species, counteracting a genetic bottleneck and thus helping it to overcome environmental constraints.

The flexibility of the breeding system and its related sex expression traits in Cakile hybrids may indicate that hybridisation between closely related species with contrasting breeding systems has been overlooked in previous studies. Specifically, if hybridisation occurs, the hybrids tend to be SI and have intermediate sex expression traits, which could be enhanced towards their parental species. Due to the presence of SI, F1s mainly rely on pollinators for pollination and have higher chance of outcrossing as well as backcrossing with their parents, especially the SI one with a showy floral display (see in Chapter 6). Therefore, after repeated backcrossing, the breeding system and its expression traits in the progeny could either be completely SI or SC with corresponding traits. Such hybrid progeny would be barely distinguished from the parental species since they have the same phenotypes and breeding systems. Support for this comes from the finding that phenotypes identified as the two distinct Cakile species in the wild can contain genetic material from both species (Ohadi et al. 2015). In their native ranges, it is probably safe to assume that the species’ phenotypes reflect a pure parental genotype; however, if both species have invaded a region, the phenotype may hide all evidence of hybridisation that occurred early in the invasion. Cody and Cody (2004) and others have suggested that some leaf and fruit shapes are evidence of past hybridisation in Australian C. maritima, although Cousens et al. (2013) argued that it simply reflects the diversity found within C. maritima, perhaps associated with multiple introductions from different geographic regions. Therefore, it is worthwhile conducting more surveys using species-specific markers at different distances from current invasion front in Australia to deeper understand their invasion and hybridisation, which is also necessary in North American & New Zealand.

80 Fitness costs of hybridisation between two invasive plants with contrasting breeding systems

6.1 Introduction It is generally recognized that fitness is determined by the complete survival and reproductive schedules of individual organisms (McGraw and Caswell 1996). Theoretically, fitness has both evolutionary significance, as an individual’s contribution to a population’s subsequent genetic composition, and ecological significance, as the numerical contribution to a population’s growth (Shaw et al. 2008). It is common, however, for researchers to estimate fitness in experiments by measuring plant biomass at an arbitrary point in time, or seed production (Pedersen et al. 2007). These may be correlated with fitness, but they are not valid measures; instead they are just components of the total fitness.

Hybridisation and its fitness costs have been of interest for many decades (Lewis and Crowe 1958, Sork and Schemske 1992, Klinger and Ellstrand 1994, Arriola and Ellstrand 1997, Burke et al. 1998). When hybridisation occurs, the offspring often show transgressive segregation, i.e., a phenomenon specific to segregating hybrid generations and refers to the fraction of individuals that exceed parental phenotypic values in either a negative or positive direction (Stuber et al. 1992, Rieseberg et al. 1999, Allison et al. 2001, Vallejo‐Marín and Hiscock 2016). On one hand, hybrids may have increased mean fitness compared with their parents, termed as “heterosis” (Goodwillie and Ness 2013). This is most pronounced in F1s and may be transient and declines in later generations if there is no further outcrossing involved (Rieseberg et al. 1999). It can facilitate the establishment of hybrid lineages that are successful in either the same environment or a different one (Rieseberg et al. 1999). On the other hand, it is also common for hybrids to have reduced fitness than their parents (Christe et al. 2016), e.g., hybrid inviability and sterility in extreme cases (Burton and Husband 2000, Roccaforte et al. 2015). The reduction in fitness will see those lineages eventually purged from the population. However, some hybrids may survive despite their reduced fitness (Veen et al. 2001) and even become successful new species in some cases (Abbott et al. 2013, Christe et al. 2016) as long as they have differentiated niches compared with the parental species (Rieseberg et al. 1999).

The variation of fitness in hybrids may be complicated by the presence of different breeding systems in the parental species. The physiological aspects of their compatibility might clash in some way, i.e., self-incompatible and self-compatible; indeed, it has been suggested that the pollen from SC species will be rejected by the stigma of SI species but not vice versa (see detailed review in Chapter 2). Meanwhile, self-incompatibility may not reveal its impacts until seed development. For example, although ovules start to develop after successful pollination, it is still possible that the resulting seeds are abnormal or abandoned (Lewis and Crowe 1958, Martin 1961, Hiscock and Dickinson 1993) due to “late-acting SI (LSI)” (Hao et al. 2012, Gibbs 2014). Moreover, because self-compatibility negates the need for pollination, inbreeding species tend to have evolved reduced attractiveness and rewards, perhaps reducing energy allocation to floral display (Knight et al. 2005, Nasrallah 2011). Thus, hybrids are expected to show a range of intermediate attractiveness to insects which may affect seed production and therefore their fitness.

81 Theoretical models have been used to predict the outcomes of fitness costs on hybrids over many generations (Mesgaran et al. 2016) but most empirical studies only focus on the fitness costs in just the first generation (F1s) (Burton and Husband 2000, Lopez et al. 2000, Abbott et al. 2013). Recent empirical studies tended to use genetic evidence of past gene flow between parental species to infer the causes and consequences of hybridisation (Hamilton and Miller 2016), which may be less time-consuming and more effective in estimating the net outcomes of hybridisation. However, upon the presence of contrasting breeding systems in parental species, the inheritance of the breeding systems in hybrids may differ in different generations, depending on the number of genes involved, their locations and dominance (Nasrallah et al. 2007). The benefits of reproductive assurance gained by an SI species through hybridisation with an SC species may be short-lived or even detrimental in the long term. If we are to predict the consequences of hybridisation, empirical data are required for both model formulation and model testing.

Two invasive Cakile species are an excellent model for studying such interactions. Cakile maritima (SI) and Cakile edentula (SC) have successively invaded along Australian coastlines and have been shown to hybridise in the wild (Ohadi et al. 2015) as well as under glasshouse conditions (see Chapters 4 & 5; Ohadi et al. 2012). The North American C. edentula had already become widespread in Australia by the time C. maritima arrived from the Mediterranean or Western Europe (Rodman 1974, 1980). Rather than competition between the two species being a problem for C. maritima invasion, Mesgaran et al (2016) have explored the possibility that hybridisation with the already-present SC species may have enabled the newcomer to overcome debilitating Allee effects resulting from it being SI, in which the fitness and breeding system of the progeny plays an important role.

In this chapter, I compared traits contributing to the fitness of Cakile maritima, Cakile edentula and their progeny, including F1s, F2s and backcrossed hybrids. Due to the resource and time constraints of a single PhD project, not every component of plant lifetime fitness was measured; instead, measurements were confined to seeds (i.e., seed size, seed mass and seed germination), seedling development and survival, floral display (size of flower), biomass at maturity and total reproductive output. Together with the results from previous chapters, these allowed the exploration of the fitness costs upon hybridisation and the implications for co-invasion of related species – specifically by testing the assumptions made by Mesgaran et al (2016) in their model.

6.2 Material and methods F1s, F2s and backcrosses (BCs) with the parental species were produced, with both parents being used reciprocally as either the pollen donor or recipient. The experimental design for this experiment was given in detail in Chapter 3. Seeds were collected from each generation and stored in paper bags under laboratory conditions until they could be used in this experiment. It should be noted that, due to the lack of asexual propagation in the species, seeds used here for the parental species were those from the original individual field collection (see Appendix to Chapter 6 for details of seed collection), which were the siblings of the parental plants used in producing F1 hybrids.

82 Seed size, mass and germination were measured in the laboratory; the same seeds were then sown in pots and grown in the field station at the Burnley Campus, the University of Melbourne, for a common garden experiment. The experiment lasted from August 2015 to February 2016.

6.2.1 Seed mass and germination Ten proximal and ten distal segments (see fruit shapes in Figure 1.2) from each combination of parents were selected at random; if there were less than 20 segments available in a crossing line, then all seeds were used. The fruit coat was carefully removed and the seeds inside were weighed (see seed collection used in this experiment in Appendix to Chapter 6).

Seeds were put into Petri dishes containing one layer of filter paper (Whatman No.1) saturated with distilled water. They were stratified in a cold room at 5 ℃ for 10 days (Ignaciuk and Lee 1980, Rodman 1986, Maun et al. 1990). Once imbibed, the seeds were carefully scratched using fine pins to break the seed coat on one side to ensure germination (Gormally and Donovan 2011) and then the dishes were transferred into a glasshouse at the Burnley Campus. Germination was estimated over a period of 15 days in the glasshouse at a constant temperature of 25 ℃. A seed was considered as germinated when its radicle was first observed. The number of germinated seeds was counted every day. Seeds that had not germinated after 15 days were tested for viability using tetrazolium chloride (TTC, see Peters 2000).

A Generalized Linear Mixed Model (GLMM) and the General Linear Model (GLM) (PROC GLIMMIX and GLM in SAS, respectively) were used to analyse the data. Seed mass was analysed using PROC GLM. In contrast, data on the final number of germinated seeds were analysed using PROC GLIMMIX in which seed line (seeds with different paternities were treated as different seed lines) was used as a fixed effect and the “random _residual_” syntax was applied. The binomial distribution was assumed (dist = binomial) with its default link function, the logit (link = logit).

6.2.2 Seedling survival As they germinated, seedlings were distributed evenly into three small biodegradable pots, as shown in Figure 6.1a, such that every pot contained the same range of germination times. The biodegradable pots were transferred into plastic pots (20 cm in width and 30 cm in depth) after germination experiments were completed (Figure 6.1b).

All pots were randomly placed outside on a rectangular site covered with weed suppression mat. This experiment could not be done in the natural (beach) habitat because of lack of security and the unacceptable probability of being washed away. Distance between each pot was roughly 0.5 m to minimise above-ground competition between neighbours. An overhead irrigation system with high-pressure sprinklers was used to water plants automatically twice a day in the first two weeks and as required from then onwards. The number of surviving seedlings and any new emergence were recorded in every pot daily for the first 30 days. After that, the seedlings were thinned down to only one individual in each pot to avoid competition.

Figure 6.1 Common garden experiment. Germinated seeds were firstly transplanted into biodegradable pots (a) for a short period and then transferred to plastic pots (b). All pots then were randomly placed in the field station (c) at a spacing of around 0.5 m. Data were initially analysed using the General Linear Model (GLM) but this failed to converge. Therefore, a Generalized Linear Mixed Model (GLMM) (PROC GLIMMIX in SAS) was used in which each seedling was treated as a sample and assigned a score of “1” if it was still alive or “0” if it was dead. Plant type was used as a fixed effect while the sample was treated as a random effect. The distribution and link function syntax used was “dist = binary” and “link=logit”, respectively.

6.2.3 Plant development To estimate the rates of development, the time in days from germination to the onset of the following stages were recorded: (1) the first pair of true leaves, (2) the second pair of true leaves, (3) the third pair of true leaves, (4) the first visible bud, and (5) the first open flower. Since water was automatically supplied via an overhead irrigation system and slow-release nutrients were provided in the potting mix, plants could survive for a long time and measurements of time to maturity or death could be misleading (C. maritima is relatively indeterminate compared with C. edentula). In order to measure the “final” plant size and fruit production at an equivalent time (see details in Section 6.2.5 and 6.2.6), the water supply was shut down and plants were allowed to senesce.

Data were analysed using a Mixed Linear Model (PROC MIXED in SAS). “plant type” (i.e. each parental species; hybrids produced within the same crossing direction also belong to a single plant type, e.g., HM) and “stage” were treated as fixed variables; time for the thinned plants to reach each stage was the dependent variable. However, since the stages lacked independence, they could not be compared with each other.

6.2.4 Floral display Flowers attract pollinators through scent and visualization; however, in this study, only the traits affecting visualization, i.e., floral display, were measured due to time limitation. Three main

84 components were recorded: individual flower traits, flower number per plant open on a single day and the total flower number produced over the plant’s lifetime.

Petal number (n) and petal size (width, W; length of the flat and open part, L) were assessed. Three fully opened flowers were randomly collected on each plant and the number of petals was counted. The size of one petal on each flower, again at random, was measured using digital callipers. The approximate area of a single petal (A) was estimated using the equation 퐴 = 퐿푊휋/4, since petals have an ellipse-like shape (Emel et al. 2017). The total display area per flower can be estimated by multiplying the petal number in a flower by the area of a single petal.

To estimate the total floral display on a single plant on any one day, flower number was counted three times fortnightly after all plants had begun flowering. Then the total floral display was calculated as the average daily flower number multiplied by the total display area per flower.

Finally, total flower number during a plant’s lifetime was estimated from the number of floral nodes on each plant at the end of the experiment. A plant in its whole life cycle, especially in C. maritima, can produce thousands of flowers (Rodman 1974, 1986). Therefore, it was not practical to count every single node; instead, I counted the total node number on three randomly chosen branches of different orders. Then the total number of flowers was their average multiplied by the total number of branches of a plant.

Petal number was analysed using a Generalized Linear Mixed Model (GLMM) (PROC GLIMMIX in SAS), while A Mixed Linear Model (PROC MIXED in SAS) was used to analyse the petal size. When analysing the petal size, the individual plant itself was used as a random factor nested within “plant type”, while “plant type” was assumed to be a fixed factor. The LSMEANS syntax was used to estimate the marginal means, thus enabling pairwise comparisons between plant types.

A Mixed Linear Model (PROC MIXED in SAS) was used to analyse the data of flower number and floral display. “plant type” was treated as a fixed effect and “plant” (i.e., pots in the experiment) was treated as a nested random effect within “plant type”. Raw data were square root transformed after checking the residual plots.

6.2.5 Plant projected area To accurately estimate the lifetime biomass of a plant, aboveground and belowground biomass, including the leaves, stems and roots should be weighed. However, leaves and ripe fruit segments on each plant fell off almost every day and there were over 350 individuals to assess. Consequently, it was impractical to measure the biomass of each plant (although previous studies of this and other species have commonly ignored such losses).

Instead, in this experiment, plant projected area was considered as a surrogate for biomass, as plants tend to become more prostrate over time. A diameter tape (as used by foresters) was used to measure the circumference of the plant and convert this to a diameter under the assumption that the plant projected area is circular. Although plants had somewhat irregular shapes, for mature plants their outline could be roughly approximated by a circle. The tape was wrapped and tied on to a steel wire (with a diameter of ca. 4 mm) and bent to follow the outline of the plant. Plant projected area was then estimated using the equation of the area of a circle with the measured reading.

85 Plant projected area was analysed using a Mixed Linear Model (PROC MIXED in SAS). Data were used directly without any transformation as residual plots were considered acceptable. “plant” was treated as a random and nested effect within “plant type”.

6.2.6 Reproductive output All plants were allowed to set seed under natural conditions (i.e. they were open-pollinated). The number of fully developed fruits on each plant remaining at the end of the experiment was counted and seed number was estimated as a proxy of an individual’s reproductive output (since many fruits would already have been lost, this is clearly a gross underestimate).

As in 6.2.4, rather than counting all the fruits on the plant, three branches were sampled at different orders and the number of fully developed fruits on three sample branches was counted to estimate the total number of fruits. Hence, the number of fruits remaining on a single plant was calculated as: 퐹푇 = (퐹1 + 퐹2 + 퐹3)푁푏/3, where FT was the total number of fruits; F1, F2 and

F3 were numbers of fruits on the three chosen branches, respectively; Nb was the number of total branches.

Meanwhile, seed number per fruit was also calculated. Specifically, twenty fruits from each experimental plant were randomly selected in each plant type; the fruit coat was removed and seed number was counted. Then the number of the seeds set per fruit was calculated as the average seed number in each plant type divided by twenty. Therefore, seed number was estimated as the number of totally produced fruits in the season (as estimated above) multiplied by the number of seeds set per fruit of each plant type.

Seed number was analysed using a Generalized Linear Mixed Model (GLMM), i.e., “PROC GLIMMIX” in SAS. Plant type was as assumed to be a fixed effect while plants (i.e., pots) were treated as a random and nested effect within plant types. Differences between plant types were estimated using LSMEANS syntax in SAS.

6.2.7 Cumulative fitness To summarize the overall relative fitness of all plant types, all measured fitness components were combined using the method described by Silvertown and Lovett-Doust (1993). The method combines the results from multiple experiments and different life stages, allowing comparisons of the lifetime fitness. The arithmetic means were generated by PROC MIXED in SAS if any random effect involved. For each component (i.e., seed germination, seedling survival, seedling growth, plant projected area or plant size, floral display and fruit set), the mean for each plant type was divided by the largest mean for that trait. In this way, there was always at least one species with a relative fitness score of one for that component, and all other species had a fitness score that was some fraction of one (Johnston et al. 2003). The overall fitness of each plant type was calculated by multiplying the values obtained for every component (Burton and Husband (2000).

The cumulative (relative) fitness estimated here is not the true overall fitness, as the reproductive outputs were only measured from the seed produced at maturity and there are still parts of the lifecycle not included in the calculation. It may, however, indicate trade-offs and complementarity between the various components.

86 6.3 Results

6.3.1 Seed mass and germination Generally, individual seed mass varied little between the parental species and hybrids (Table 6.1). In the parental species, seed mass in C. edentula tended to be greater than in C. maritima (but not significantly, P > 0.05). Seeds produced by intraspecific outcrossing between populations of C. maritima (MM) had larger seeds than C. maritima itself (P < 0.05). Seed mass differed significantly within F1s, with a much smaller seed mass in EM than in ME; however, no significant difference was found between F1s and the parental species (except ME vs. MM). The seed mass in F2s did not show a significant difference compared with that in F1s sired by C. maritima (ME) but was significantly greater than it was in F1s sired by C. edentula (EM). Similarly, the seed mass of backcrosses, except HE, was not significantly different compared with that of their F1 parent (i.e., ME). HE had the largest seed mass in all backcrosses, which was also the largest in all hybrids (P < 0.05).

Table 6.1 Means of seed mass (mg) of all plant types. Lower case letters are based on Tukey’s post hoc test (pair- wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different. M, E and H indicate C. edentula, C. maritima and the F1 hybrids used in producing backcrosses (see details about crossing plan in Chapter 3), respectively. When the initials combine, the first letter indicates the pollen donor; for example, EM and ME indicate two reciprocal crossings used to produce F1s (the first with E as the pollen donor, and the second with M as the pollen donor). MM and EE are outcrossed seeds within natural C. maritima and C. edentula, respectively.

Generation Plant type Seed mass (mg) SE C. maritima 4.68 d 0.11 MM 7.99 bc 0.78 P1 C. edentula 7.20 bcd 0.40 EE 5.92 cd 0.39 EM 4.70 d 0.33 F1 ME 8.83 bc 0.94 F2 HH 8.37 bc 0.43 EH 7.53 bcd 0.70 HE 12.91 a 0.46 BC HM 7.01 bcd 0.83 MH 9.95 ab 0.38

Most seeds germinated (Table 6.2) but significant differences between plant types were detected (F = 5.24, P < 0.001; see in Appendix 3- 7). Specifically, no difference was detected between the parental species (P > 0.05). F1s had significantly lower germination percentages than the parental species, which was also significant when compared with backcrosses and F2s (P < 0.05). Moreover, F1s sired by C. maritima (i.e., ME) had the lowest number of germinated seeds in all plant types (P < 0.01). No significant difference was found within backcrosses; however, the germination in C. edentula was inferior to most BCs (P < 0.05), which was not the case for C. maritima (P > 0.05).

Although all seeds of EH germinated (Table 6.2), germination was not significantly greater than any other plant types, since there were only a few seeds available for this plant type (average n = 7 in all replicates), inducing large standard error terms and thus non-significant P values.

87 However, since it did not affect significance levels, it was not excluded from those pairwise comparisons.

It should be noted that most un-germinated seeds were inviable. Only six out of 117 seeds showed a positive TTC result (last column in Table 6.2).

Table 6.2 Average germination percentage in each plant type. Twenty seeds, if applicable, of each replicate within a plant type were used. The number of germinated, un-germinated and TTC viable seeds is the sum of all replicates in each plant type. Tetrazolium chloride (TTC) was used to test the viability of un-germinated seeds.

Generation Plant type Germination (%) Germinated Un-germinated TTC viable C. maritima 85.56 180 26 1 MM 97.50 80 2 0 P1 C. edentula 94.09 240 16 0 EE 87.50 40 5 0 EM 99.49 177 1 0 F1 ME 80.70 288 57 4 F2 HH 98.04 166 4 0 EH 100.00* 77 0 - HE 99.67 298 1 0 BC MH 99.41 257 2 1 HM 99.12 332 3 0 * EH had 100% germination rate but seed availability was extremely limited: only a few seeds were available for germination test in all replicates.

6.3.2 Seeding survival The survival of seedlings did not differ significantly between generations (F = 1.15, P > 0.05). Similarly, no significant difference was found within plant types of each generation (P > 0.05).

Table 6.3 Means of the proportion of seedling survival in different plant types. Multiple comparisons between different plant types indicated that none of them were significantly different from any others. Numbers are combined across all replicates; survival was calculated separately for each replicate.

Generation Plant type Survival SE Total seedling Survived seedling C. maritima 0.82 0.04 154 130 MM 0.85 0.07 59 50 P1 C edentula 0.77 0.04 207 158 EE 0.80 0.08 35 28 EM 0.77 0.07 176 147 F1 ME 0.72 0.06 231 180 F2 HH 0.86 0.05 162 131 HE 0.87 0.03 297 259 HM 0.88 0.03 329 288 BC MH 0.84 0.05 255 223 EH 0.81 0.08 77 63

88 6.3.3 Plant development There were no significant differences between plant types in the lengths of early phenological stages; however, the rates of reproductive development varied significantly between plant types at later stages (P < 0.001) (Table 6.4).

Table 6.4 Test of the differences between plant types within each stage in phenology. Significant P values are in bold. Least Square Means were used to estimate the differences within stages. Stages are all measured as days since germination.

Interaction Stage (#) F Value Pr > F Plant type*stage First pair leaves (stage 1) 0.94 0.5077 Plant type *stage Second pair leaves (stage 2) 0.90 0.5496 Plant type *stage Third pair leaves (stage 3) 0.91 0.5384 Plant type *stage Bud emergence (stage 4) 3.70 <.0001 Plant type *stage First open flower (stage 5) 3.70 <.0001

In the parental species, seedlings from field-collected C. edentula and its outcrossed progeny (EE) required much longer to reach bud emergence and flowering than any other plant types (P < 0.01) except that C. maritima > C. edentula and C. maritima > EE in bud emergence and C. maritima > EE in first flower opening. Outcrossed C. maritima (MM) had the highest mean rate of development in all the plant types; however, this was only significant when compared with field-collected C. maritima and C. edentula, EE and one of the backcrosses (MH) and F1s (EM) (P < 0.05).

All hybrids, including BCs, F1s, and F2s, had a significantly faster reproductive development (either to stage 4 or 5) than C. edentula (P < 0.05). Some hybrids (e.g., HE) also had faster development in reaching stage 4 and 5 compared with C. maritima (P < 0.05) (Table 6.5). However, there were no significant differences in reproductive development between produced hybrids.

Table 6.5 Time (days) required to reach the stages of bud emergence (stage 4) and first open flower (stage 5). Significance levels are indicated by lower case letters, based on Tukey’s post hoc test (pair-wise comparisons with a significance level of P < 0.05); means with the same letter are not significantly different.

Generation Plant type Stage Time (days) SE Stage Time (days) SE C. maritima 4 41.13 ab 1.60 5 50.75 ab 1.60 MM 4 33.92 c 2.26 5 44.08 c 2.26 P1 C. edentula 4 45.17 a 1.44 5 55.00 a 1.44 EE 4 46.50 a 3.20 5 56.17 a 3.20 EM 4 37.97 bc 1.31 5 47.94 bc 1.31 F1 ME 4 39.14 bc 1.21 5 48.21 bc 1.21 F2 HH 4 38.10 bc 1.32 5 47.89 bc 1.33 EH 4 38.73 bc 1.49 5 47.85 bc 1.56 HE 4 36.80 c 1.17 5 45.81 c 1.17 BC HM 4 37.76 bc 1.10 5 47.53 bc 1.10 MH 4 40.75 ab 1.10 5 50.42 ab 1.10

89 6.3.4 Floral display

6.3.4.1 Petal area in a single flower C. maritima had the largest petal area while C. edentula had the smallest. All hybrids, including F1s, F2s and BCs, had an intermediate mean petal area compared with the parental species (P < 0.05). F1s in both crossing directions had a significantly smaller petal area when compared with the backcrosses with C. maritima, i.e., HM and MH. These two plant types also had a significantly larger petal area than any other hybrids (P < 0.001).

Table 6.6 Means of petal area per flower (mm2) in each plant type. Means with the same letters (lower case) are not significantly different. Significance level was P < 0.001.

Generation Plant type Petal area per flower (mm2) SE C. maritima 64.21 a 3.50 MM 50.91 b 5.25 P1 C. edentula 10.08 d 3.20 EE 8.97 d 7.43 EM 36.21 c 3.05 F1 ME 33.29 c 2.81 F2 HH 35.69 c 3.18 EH 31.27 c 3.60 HE 35.84 c 2.72 BC HM 55.37 b 2.55 MH 51.72 b 2.63

Petal numbers in a single flower had a similar pattern compared with petal area. In parental species, all four petals were always present in C. maritima and its outcrossed progeny (MM) while it was common to observe the absence of one or two petals in a flower of C. edentula and outcrossed EE. In extreme cases, there was only one petal fully developed and visible in C. edentula. F1s sired by C. edentula (EM) had little petal absence (only four out of 105 collected flowers), while it was slightly more common in ME (15 out of 126). In contrast, all backcrosses with C. edentula (EH and HE) produced high proportions (29.5% and 40%, respectively) of flowers with missing petals compared with the backcrosses with C. maritima. There were two aberrant flowers with five and six petals in HM and three flowers without any petal in F2s (Table 6.7).

90 Table 6.7 Frequency distribution of petal number in all plant types. The column “total” presents the total number of samples collected in each plant type.

Petal number Generation Plant type Total 0 1 2 3 4 5 6 C. maritima 0 0 0 0 81 0 0 81 MM 0 0 0 0 36 0 0 36 P1 C. edentula 0 23 34 32 4 0 0 93 EE 0 3 12 0 3 0 0 18 EM 0 0 1 3 101 0 0 105 F1 ME 0 0 0 15 111 0 0 126 F2 HH 3 1 6 16 67 0 0 93 EH 0 0 6 12 45 0 0 63 HE 0 1 15 37 79 0 0 132 BC HM 0 0 1 4 146 1 1 153 MH 0 1 1 3 139 0 0 144

6.3.4.2 Flower production C. edentula and outcrossed EE produced the fewest flowers open on any one day compared with other plant types (P < 0.05). F1s tended to produce fewer daily flowers than C. maritima and MM, which was only significant in ME; however, there was no significance different within F1s. In F2s, the daily flower number was not significantly less than in F1s. In contrast, within backcrosses (BC), significantly different daily flowers were produced. The backcrosses with C. edentula (EH and HE) always produced fewer daily flowers than the backcrosses with C. maritima (MH and HM) (P < 0.05). Moreover, HM produced more daily flowers than all other hybrids as well as C. maritima (P < 0.05); however, this was not significant when compared with MM (P > 0.05).

Table 6.8 Means of daily flower number in each plant type. Means with the same letters are not significantly different (P > 0.05).

Generation Plant type Daily flower number SE C. maritima 75.01 bc 2.96 MM 77.83 ab 4.38 P1 C. edentula 42.42 f 2.73 EE 33.17 g 6.20 F1_EM 69.97 bcd 2.53 F1 F1_ME 67.69 de 2.34 F2 HH 63.62 de 2.61 EH 60.95 e 3.09 HE 57.62 e 2.26 BC HM 82.69 a 2.13 MH 72.26 bc 2.16

The variation in the total lifetime flower production had a similar pattern to the number of flowers open on any given day, i.e., most hybrids had an intermediate lifetime flower number while C. edentula and EE generated the fewest flowers in all plant types, which was only ca. 20%

91 of the value in C. maritima (Table 6.9). It should be noted that HM again produced the most lifetime flowers (P < 0.05). Meanwhile, the backcrosses with C. edentula (EH and HE) and F2s did not produce significantly fewer total flowers compared with C. maritima and MM (although more flowers were open on a given day in C. maritima and MM) (P > 0.05).

Table 6.9 Means of the total lifetime flower number in each plant type. Means with the same letters are not significantly different (P > 0.05).

Generation Plant type Total flower number (back transformed) Sqrt (total flower number) SE C maritima 5,623 74.99 cd 5.53 MM 4,767 69.04 cd 8.19 P1 C. edentula 1,286 35.86 e 5.04 EE 1,136 33.71 e 11.58 EM 7,314 85.52 bc 5.18 F1 ME 6,206 78.78 bcd 4.38 F2 HH 5,219 72.24 cd 4.89 HE 4,572 67.62 d 4.23 EH 4,276 65.39 d 5.66 BC HM 9,340 96.90 a 3.97 MH 8,046 89.70 ab 4.04

6.3.4.3 Daily floral display area Despite of the variation in the flower number of a single day, the daily floral display still presented a similar variation pattern to the petal area of a single flower (Table 6.10). In the parental species, C. maritima had the showiest floral display (although not all comparisons were significant) while C. edentula was the least attractive in all tested plant types (P < 0.05). In produced hybrids, no significant differences were observed between or within F1s, F2s and the backcrosses with C. edentula (i.e., EH and HE) in terms of their floral display; however, the backcrosses with C. maritima (MH and HM) were significantly showier than EH and HE (P < 0.05). It should be noted that MH and HM were nearly as attractive as C. maritima and MM were.

Table 6.10 Means of the daily floral display area (mm2) in each plant type. Means with the same letters are not significantly different (P > 0.05). When analysing, data were square root transformed.

Daily floral display area (mm2) Generation Plant type SE Back transformed Sqrt C. maritima 6,091 78.04 a 3.29 MM 4,995 70.67 ab 4.91 P1 C. edentula 596 24.42 d 3.02 EE 355 18.85 d 6.94 EM 2,526 50.26 c 2.85 F1 ME 2,343 48.41 c 2.66 F2 HH 2,266 47.60 c 3.06 EH 1,959 44.26 c 3.46 HE 2,450 49.50 c 2.54 BC HM 5,640 75.10 a 2.38 MH 4,559 67.52 b 2.45

92 6.3.5 Plant projected area In the parental species, C. maritima and its outcrossed progeny (MM) were much bigger at maturity than C. edentula and EE. Most offspring, including F1s, F2s and part of the BCs (EH and HE) had an intermediate plant size. No significant difference was detected between F1s and F2s or within F1s (P > 0.05). However, the backcrosses with C. maritima (HM and MH) not only outperformed their counterpart BCs (HE and EH) but also were bigger than F1s and F2s (Table 6.11). Moreover, HM even had a significantly bigger plant size than C. maritima (P < 0.05). Such differences had also been presented in the previous sections.

Table 6.11 Means of plant projected in all plant types at maturity. Means with the same letters are not significantly different (P > 0.05).

Generation Plant type Plant area (m2) SE C. maritima 0.45 bc 0.03 MM 0.49 ab 0.05 P1 C. edentula 0.15 g 0.03 EE 0.09 g 0.05 EM 0.40 de 0.03 F1 ME 0.37 de 0.03 F2 HH 0.33 ef 0.03 EH 0.29 f 0.03 HE 0.27 f 0.02 BC HM 0.55 a 0.02 MH 0.42 cd 0.02

6.3.6 Seed number Seed number had a much more complicated pattern compared with any other fitness components. No significant difference was detected between the parental species, including their outcrossed progeny, EE and MM (P > 0.05). F1s (EM and ME) showed no significant difference in seed production compared with F2s (P > 0.05). However, both generations produced much fewer seeds than the original parental species (P < 0.05). There was no difference within the F1s.

In backcrosses, crossing direction affected their reproductive output. The backcrosses with C. maritima (HM and MH) produced many more seeds than the backcrosses with C. edentula (HE and EH) did. Although the seed number in HM and MH was not significantly different from C. maritima, it was indeed larger than that of F1s and F2s (P < 0.05). In contrast, HE and EH did not have a significantly different seed number compared with F1s and F2s (P > 0.05).

93 Table 6.12 Means of seed number at maturity in different plant types. Seed numbers with the same grouping letters (lower case) are not significantly different (P > 0.05).

Seed number Generation Plant type Seed set per fruit SE Sqrt (seed number) SE (back transformed) C. maritima 1.39 0.04 958 30.95 a 3.14 MM 1.39 0.08 587 24.23 ab 4.66 P1 C. edentula 1.10 0.06 588 24.25 ab 2.86 EE 1.05 0.02 585 24.18 ab 5.60 EM 1.13 0.12 126 11.21 c 2.69 F1 ME 1.26 0.05 187 13.69 c 2.49 F2 HH 1.05 0.07 76 8.73 c 2.77 EH 0.87 0.17 73 8.57 c 2.99 HE 1.43 0.10 104 10.20 c 2.41 BC HM 1.48 0.07 867 29.44 a 2.26 MH 1.13 0.08 510 22.59 b 2.29

6.3.7 Cumulative fitness C. edentula and its outcrossed progeny (EE) had the lowest cumulative relative fitness in all plant types. In hybrids, most plant types had intermediate cumulative fitness compared with the original parental species. The relative fitness in both F1 plant types was relatively low, especially in ME, which was slightly higher than that in C. edentula (0.08 vs. 0.03). Such low fitness was also observed in F2s and the backcrosses with C. edentula (i.e., EH and HE). However, when BCs were produced by C. maritima, their fitness increased dramatically. In fact, the plant type that had the highest cumulative fitness was the backcrosses with C. maritima as the maternal plant, i.e., HM (Table 6.13).

Table 6.13 Relative fitness at different stages. Within each column, all values for that fitness component were divided by the largest value. Thus, a fitness score of 1.00 was assigned to the highest value in the column (in bold), and a proportion to all other plant types. No statistical significance is implied by these numbers (Johnston et al. 2003).

Daily floral Plant Seed Cumulative generation plant type germination survival Development display size number fitness C. maritima 0.86 0.83 0.87 1.00 0.82 1.00 0.51 MM 0.98 0.88 1.00 0.91 0.89 0.78 0.54 P1 C. edentula 0.94 0.61 0.80 0.31 0.27 0.78 0.03 EE 0.88 0.70 0.78 0.24 0.16 0.78 0.01 EM 1.00 0.80 0.92 0.64 0.72 0.36 0.12 F1 ME 0.81 0.61 0.91 0.62 0.67 0.44 0.08 F2 F2 0.98 0.81 0.92 0.61 0.60 0.28 0.08 EH 1.00* 0.75 0.92 0.57 0.53 0.28 0.06 HE 1.00 0.98 0.96 0.63 0.50 0.33 0.10 BC HM 1.00 1.00 0.93 0.96 1.00 0.95 0.85 MH 0.99 0.96 0.87 0.87 0.77 0.73 0.40 * Although all seeds in EH germinated, there were only a few seeds successfully produced for estimating the germination in each replicate within the plant type (see details in Chapter 5). Therefore, the germination result in this experiment was limited and may not fully represent the traits of seed germination in HE.

94 6.4 Discussion Hybridisation between the two Cakile species clearly affected plant fitness. Heterosis was obvious in some hybrids after plants germinated, grew and reached later stages of development, while most hybrids showed intermediate fitness. No significant fitness costs were found in fitness components in plant early lifecycle, i.e., seed germination, seedling survival and development. Meanwhile, the variation of fitness components indicates maternal effects within hybrids to some extent. In this section, I will discuss the fitness costs over generations during plant lifetime and its implications.

6.4.1 Fitness costs at early life stages It has been previously reported that the seedlings are more vulnerable than adult individuals to adverse environmental conditions (Burke et al. 1998). As suggested in Johnston et al. (2003), seed germination, seedling survival and seedling development are crucial parts for moving seedlings out from relatively dangerous early stages. Therefore, if hybrids did not show significantly inferiority in fitness compared with their parental species, their seeds and seedlings are equally likely to pass through environmental sieves as seeds and seedlings of the parental species (Johnston et al. 2003).

At early life stages, hybrids in this study did not show lower averaged fitness compared with the parental species; instead, both hybrids and the parental species had high average mean fitness in each component with even higher germination in some hybrids, indicating that hybrids seeds and seedlings have experienced the same or even lower level of environmental constraints. Most hybrids as well as the parental species in this study showed a relatively high proportion of germinated seed and seedling survival, which was much better than reported in previous studies (Barbour 1972, Payne 1980, Payne and Maun 1984) of a single Cakile species (but see Zhang 1993). For example, seedling survival in all plant types in this study was over 70%, whereas Cakile species have been reported with low seedling survival in the wild (Barbour 1970, Payne 1980, Zhang and Maun 1992). Such huge differences in germination and survival between current and previous studies were probably caused by the removal of some environmental constraints. All seeds used in my experiments were not only stratified but also treated with seed coat removal, specifically in order to improve germination. Meanwhile, all germination experiments were performed in a glasshouse instead of on a sand dune system where germination would be affected by salinity, extremes of temperature and moisture, and other environmental factors (Zhang and Maun 1992). Normally, for species in coastal and lacustrine sand dune system where nutrient deficiency, lack of moisture and sand accretion are the norm, it is common to observe relatively low seedling survival proportion (Maun 1994). It should be noted that, despite the removal of environmental constraints occurring in the wild, the results in this study were still valid as all seeds and seedlings experienced a uniform environment.

Many studies have reported that seed mass is strongly and positively correlated with the overall fitness of their offspring in many aspects, especially at early stages, such as faster germination and development (Giles 1990, Castro 1999). This covariation was not detected either in seed germination, seedling survival or in early development of all plant types, although it was reported in C. edentula (Zhang 1996). Specifically, hybrids did not show any inferiority compared with their parental species in seedling survival or early development. As a matter of fact, the only inferiority found in hybrids was the germination percentage of ME. Therefore, the impacts

95 of seed mass in Cakile species on plant early lifetime fitness did not seem to be so important or not significant enough to be detected. Furthermore, previous studies also mentioned that seed mass was largely affected by maternal effects, which has been found frequently in other species (Khan 2004, Li et al. 2015). In this study, maternal effects seemed to have strong influences on seed mass within generations. Seeds with C. maritima as the maternal parents tended to have larger seeds, e.g., within F1s (EM vs. ME) and BCs (HE vs. HM). Since all experiments were performed under the uniform conditions in a glasshouse, this tendency was more likely due to the different maternal genotypes observed in the parental species, i.e., maternal effects.

Overall, despite some variations, most hybrids in this study showed an equal and even superior capability (e.g., in seed germination) to escape from the environmental constraints at early life stages and thus contributed equally to their late life stage. However, since the same number of seeds from each plant type were used and all seeds were treated with stratification and seed coat removal, the effect of seed bank, which has been frequently discussed in Cakile species (Ignaciuk and Lee 1980, Payne 1980, Rodman 1986), was not taken into consideration when dealing with early lifetime fitness. In reality, the effect of seed bank on the overall early lifetime fitness could be vital as it directly determines how many seeds will be available for the next generation, which could be of great value and worth further investigations in future studies.

6.4.2 Fitness costs at late life stages Despite all plant types in this study being equally capable of passing through early environmental sieves, most hybrids showed intermediate fitness at late stages compared with the parental species. Specifically, the daily floral display did not vary between generations in most hybrids but showed significant inferiority compared with that in C. maritima, as shown in Table 6.10, although it was considerably larger than in C. edentula (normally, hybrids had a four-fold larger daily floral display than C. edentula). The differences in daily floral display were due to the smaller plant size (they cannot produce more flowers in a single day; Table 6.8) and the smaller petal size in a single flower (consequently less attractive as a single flower to pollinators; Table 6.6). As described in Chapter 5, most hybrids are self-incompatible and thus theoretically need pollinators for pollination; however, they are less attractive than the self-incompatible C. maritima when sympatric. Consequently, pollinators may have often visited the more attractive C. maritima instead of the less attractive hybrids in the field experiments where they were randomly arranged in the same patch, inducing larger number of pollinated flowers and thus possibly larger seed set number in C. maritima. These findings correspond with previous studies where the importance of floral display was discussed. For example, Rodman (1974) firstly measured and declared the size difference of flowers between the two parental species and then further concluded that difference of flower size between species was one of the possible explanations for the replacement of Cakile species (Rodman 1986). Many studies about the establishment and invasion of Cakile species also implied the difference of flower size between species and its importance (Boyd and Barbour 1993, Aigner 2004, Mhemmed et al. 2008, 2012, Mills 2013).

In contrast, heterosis was also observed, especially in the backcrosses with C. maritima. The backcrosses with C. maritima, especially HM, had equal and even higher fitness components than their parents at late life stages. Despite the significantly smaller size of a single flower, the superiority of HM in other fitness components contributed to a large daily floral display in HM

96 as in C. maritima (Table 6.10). Therefore, there was barely any difference in the final seed number between HM and C. maritima. Moreover, significant differences were also observed within the backcrosses with C. maritima (i.e., HM vs. MH) in every fitness component at late stages, including development, floral display, plant size, and seed set. These differences, therefore, indicate the existence of maternal effects in BCs. Consequently, compared with MH, the heterosis in HM could be partially due to and enhanced by the maternal effects from C. maritima, providing HM with advantages upon competition and thus outperforming other hybrids and even C. maritima. This means that crosses and backcrosses with C. maritima in a population would be more attractive as well as more fecund, as assumed in Mesgaran et al. (2016), making the model predictions about Allee effects and introgression plausible.

Overall, the results in this chapter imply that, despite having equal capability of passing through early environmental dilemmas compared with their parental species, most hybrids showed intermediate fitness at late lifetime stages. However, heterosis was found in certain hybrids, especially the backcrosses with C. maritima, with large daily floral display and plant size, indicating that the interaction between backcrosses and the SI species could be more common than between other plant types, and thus leading to possible introgression.

6.4.3 Implications for co-invasion of related species Mesgaran et al. (2016) suggested a new mechanism of plant invasion: unlike traditional invasion, early-generation interspecific crosses (i.e., hybrids) between related species could enable the colonizer genes to establish. Pure colonizer-type individuals can subsequently arise through crossing among hybrid lineages or repeated backcrossing with the colonizer parents. Through modelling with Cakile species, they also suggested that this hybridisation-rescue effect is more likely to eventuate if the new species and hybrids are more attractive to pollinators or when the hybrids are more compatible (i.e., more likely to produce viable offspring) with the newcomer than with the established species. That is, after sufficient generations of asymmetric breeding (backcrossing to the new colonizer), plants in the population will increasingly come to resemble the original newcomers. The arriving species will essentially have been reconstituted (Mesgaran et al. 2016). The main assumptions underlying their predictions (Mesgaran et al. 2016) have been further supported by the results of this study, which has valuable ecological implications for co-invasion of related species.

Table 6.14 The main assumptions and parameters used in the model of Mesgaran et al. (2016). “-” means the assumption has not been tested or confirmed in this study.

Parameter Assumption Confirmed Seed bank C. maritima has a larger seed bank while C. edentula has a smaller one. The seed - bank in hybrids is intermediate. Seedling Both the parental species and hybrids have the same mortality rate at early stages √ survival Individual Plants tend to produce a reduced number of flowers upon the increase of - competition population density Selfing rate Selfing rate in C. edentula is high while there is barely selfing in C. maritima. √ Hybrids has an intermediate selfing rate. Attractiveness Some hybrids are more attractive to pollinators with the newcomer (C. maritima) √ than with the established species.

97 Firstly, most hybrids in this study did not show lower fitness at early life stages than the original parental species. Thus, hybrids can easily pass through the environmental sieves with the coexistence of their parental species (Campbell 2003). The differences were then significant in the development and other fitness components at late stages. All produced hybrids showed a similar development pattern: most hybrids developed faster than C. edentula but slower than C. maritima; however, some hybrids developed significantly fast than both parental species, e.g., one of the backcrosses with C. edentula (HE). Previous studies have also confirmed that C. edentula grew significantly slower than C. maritima (Barbour 1970, Rodman 1986, Boyd and Barbour 1993). Therefore, all hybrids could grow relatively fast and reach their reproductive stages earlier and induce potential advantages against C. edentula.

Secondly, the maternal effects and heterosis observed in the backcrosses from reproductive stages onwards, as discussed in previous sections, also clearly suggest that the preference in crossing directions and the attractiveness in hybrids. The maternal effects can be summarized in the following aspects: the fitness components of late development (Table 6.5), daily floral display (Table 6.10), plant size (Table 6.11) and seed number (Table 6.12), as well as the relative fitness score shown in Table 6.13. All these identified maternal effects indicate that backcrossing with the SI species (C. maritima) would be the favourable crossing direction upon hybridisation. Meanwhile, the backcrosses related to C. maritima (MH and HM) had a much larger floral display compared with the backcrosses with C. edentula (EH and HE), indicating that such hybrids are more attractive to pollinators and thus inducing a preferability in crossing directions.

The cumulative fitness of each species, calculated by multiplying each component of the relative fitness, would further provide support to the modelling results in Mesgaran et al. (2016) as well as insights into the overall fitness, indicating the possible outcomes of the evolution of species, e.g., hybridisation and speciation. For example, HM hybrids produced by backcrossing with C. maritima (SI species, known as the “newcomer” in Mesgaran et al. 2016) had the highest cumulative fitness in all species, which implies that these backcrosses with C. maritima would be favoured by natural selection and thus in turn promote their interaction with C. maritima after hybridisation, helping the establishment of the SI species (C. maritima), which is one of the core predictions in Mesgaran et al. (2016). Furthermore, it also implies that these backcrosses with C. maritima would be favoured by natural selection and thus may persist after hybridisation. Such speculations are supported by the genetic findings that the chloroplast genes in C. edentula were found in a proportion of C. maritima in some regions, indicating that some current phenotypes visually resembling C. maritima and displaying C. maritima patterns of life-history may actually be hybrid-derived (Ohadi et al. 2015). In contrast, the C. edentula and its outcrossed progeny (EE) had extremely lower cumulative fitness than all other plant types, indicating that they probably will die out gradually if co-occurring with other congeneric species, which is also supported by the fact that C. edentula has almost been replaced by C. maritima in Australia in the past century (Rodman 1986, Cousens et al. 2013, Mills 2013).

However, there are potential drawbacks using the cumulated fitness without assigning a weight to each fitness component. Relying on the cumulative fitness using the method in Burton and Husband (2000) to perform speculations is arbitrary and sometimes unreliable due to overestimation or underestimation of particular components of fitness. For instance, it is normal that a selfer (e.g., C. edentula) has a small flower number and plant size but a high seed number

98 (Table 6.13). However, the cumulative fitness in C. edentula was extremely low due to the low score in the daily flower display (number of flower multiplied by petal area in a single flower), thus overestimating its impact on selfers. Although the cumulative fitness may be able to provide useful prediction of the outcome of species dynamics (i.e., replacement of C. edentula), it is still biased and even unreliable in such a case. Besides, the cumulative overall fitness was incomplete since seed production, as mentioned in 6.2.7, was only estimated at maturity instead of for its whole lifetime and thus not all life cycle stages were included in the estimation. Therefore, further studies are needed in this area to address the problems mentioned above.

Overall, most hybrids derived from the hybridisation between these two Cakile species showed intermediate fitness; in contrast, backcrosses with the SI species had equal or even higher fitness compared with their parents, suggesting that those better fitted hybrids and SI species would persist, between which the follow-on interactions would occur and thus help their establishment. Such results provide supports for the modelling assumptions in Mesgaran et al. (2016). However, neither this study nor the model in Mesgaran et al. (2016) addressed the following two questions: was the hybridisation just incidental or prevalent in the wild? Did the introgression improve the fitness of “C. maritima” in Australia? Therefore, further studies are needed to explore these questions, which can be extremely helpful in shaping our understanding of plant hybridisation and invasion.

99 General conclusions The aims of this thesis were to determine whether hybridisation is likely to happen with both species as the pollen donor and the roles that hybridisation plays in plant invasion where there is an SI and SC species. Accordingly, the following specific research questions were asked: what is the pollination window for natural hybridisation between two closely related species with contrasting breeding systems? What are the traits and inheritance of the breeding systems of hybrids and parental species in hybridisation? What are the fitness costs of hybridisation between two closely related species with contrasting breeding systems? Corresponding experiments were conducted and the major findings are listed as below:

• No direct pollination window existed between the two species. The floral development in both species shared the same pattern, which could be divided into nine continuous stages from visible buds to wilting flowers. Stigmas were receptive at all dissected stages (2-8) in both species, while anther dehiscence occurred much earlier in C. edentula (early stage 5) than in C. maritima (immediately before flower opening), indicating the potential for cleistogamy in C. edentula. However, cleistogamy does not seem to be prevalent in C. edentula: dissections showed that stigmas in many flowers (stage 5 and 6) were clean without self-pollen on their stigma surface, providing a possible time window after flowering for open pollination to occur. • Hybrids were produced in both directions and SI was dominant in most hybrids. F1 hybrids could be produced in both crossing directions; F2s and BCs were also readily produced with reciprocal crossing, indicating that the SI x SC rule, i.e., pollen from SC species is generally rejected by the stigma of SI species, did not apply. However, pollen from different plant types performed differently, suggesting the discrimination in the style. SI was dominant in hybrids, with the occurrence of atypical segregation ratios of the breeding systems in F2s and BCs. • Hybridisation between these Cakile species resulted in reduced plant fitness in most hybrids in which the maternal effects were obvious. However, heterosis was also found in some hybrids. Fitness components of hybrids at early plant life, i.e., seed germination, seedling survival and seedling development, differed little from the parental species; in contrast, at later stages, F1s, F2s and the backcrosses with C. edentula showed intermediate fitness compared with the parental species, while the backcrosses with C. maritima almost had the same fitness and even heterosis compared with C. maritima in daily floral display, plant size and seed set.

These findings provide new insights into the role that hybridisation may play in plant invasion and its interaction with breeding systems; meanwhile, they have clear ecological implications for hybridisation between invading species, especially when different breeding systems are involved. This chapter will discuss these potential implications, raise areas for further studies, as well as consider the limitations of this thesis.

7.1 Window of opportunity for hybridisation As shown in Chapter 4, anther dehiscence occurred much earlier in C. edentula than in C. maritima. By the time anthers dehisced, all stamens in C. edentula had elongated such that at least part of the anthers extended above the level of the stigma surface, whereas most stamens

100 in C. maritima stayed below the stigma surface. Such facts indicate the possibility of cleistogamy in C. edentula.

Consequently, C. edentula is expected to have high selfing rate, especially in the absence of pollinators. Many self-compatible species tend to have a low selfing rate often due to the selfing syndrome and thus not attracting pollinators (Sicard and Lenhard 2011, Luo and Widmer 2013, Tedder et al. 2015) as well as inbreeding depression under natural conditions (Briscoe Runquist et al. 2017), although they may achieve a relatively high selfing rate due to the lack of pollinator service or pollen limitation (Knight et al. 2005). In contrast, it is believed that C. edentula could achieve a high selfing rate (Rodman 1974, 1986), which seems to be supported by my results. Donohue (1998) used a model to estimate the selfing rate of C. edentula and suggested that the selfing rate could be significantly larger than 0.5. Similarly, I recorded a high fruit-set rate (mean = 0.48) in C. edentula in the field although it was much lower than in the glasshouse (mean = 0.81) without any pollinators.

However, previous studies have indicated that outcrossing between C. edentula and other species occurred under natural conditions. Black and Robertson (1965) firstly reported that putative hybrids were found occasionally between C. edentula and C. maritima. Rodman (1986) also suggested the occurrence of hybrids in Australia as did Cody and Cody (2004). Chloroplast gene sequence data also indicated that hybrids have occurred in both directions, with the SC species being either the pollen donor or recipient (Ohadi et al. 2012, Ohadi et al. 2015). All the evidence indicates that outcrossing could occur between C. edentula and C. maritima despite the high selfing rate of the SC species.

Therefore, there must be a pollination window between the two species, allowing pollen flow between C. edentula and other species, and thus outcrossing. Although it has been suggested in Chapter 4 that anther dehiscence occurred well before the flowers opened in C. edentula, it is still possible that not all stigmas in C. edentula has been contaminated by self-pollen before flowering. As a matter of fact, in many dissected buds, despite the early anther dehiscence, their stigmas were found out to be clean and free from any pollination contamination, indicating that pollen shedding in many buds of C. edentula could happen at or after flowers open and thus provide a pollination window for outcrossing (see details in Section 4.4.4). Another possibility might be insect pollination before the flowers open, especially ants. Ants were found to be extremely active on those experimental plants in the field station, trying to find foodstuffs (personal observation). Like the pollination via bees, massive pollen grains could be carried over by foraging insects to other flowers, either on the same or different plants, inducing pollination (Willmer 2011). It is also very likely that, during their foraging, these insects could probe into an unopened C. edentula flower, pollinating it with carry-over pollen grains. These speculations can be tested in future studies using time lapse footage video captured by cameras focusing on a developing bud.

However, the above discussion has overlooked an important fact that all selfing data of C. edentula in this thesis were collected within pollination bags, excluding the possibility of outcrossing. The small floral display vs. the large seed set number in the open pollinated C. edentula, as discussed in 6.4.2, implies that the selfing in C. edentula could be considerably high even though it was under open pollination conditions. So, what is the real proportion of

101 outcrossing in C. edentula in a natural environment without any intervention? Using genetic markers to identify the paternity could be the easiest way to tackle such problems. Genetic markers in chloroplasts were used in Cakile species to determine the purity of these model species, which was successful (Ohadi et al. 2012, Ohadi et al. 2015). Similarly, specific paternal markers can be also used in Cakile species to determine the paternity of produced seeds under natural conditions, thus providing robust estimates of the proportion of selfing vs. outcrossing, including the proportion of outcrossing between and within species (Wang et al. 2012). Such estimation is extremely useful for broadening our understanding of pollination processes under natural conditions, and potentially could be an important area for future studies.

7.2 Breeding system upon hybridisation Upon hybridisation between the two Cakile species, pollen germination and pollen tube growth were not rejected, leading to the formation of hybrids in both crossing directions, which indicates that the SI x SC rule was not applicable. The SI x SC rule was firstly introduced by Harrison and Darby (1955), defined as the phenomenon that pollen from SC species is normally rejected by the stigma of SI species but not vice versa, and is a part of unilateral incompatibility (UI) (Lewis and Crowe 1958). Recently, it has been suggested that the UI and SI may share a universal genetic pathway; therefore, not only self-pollen but also incompatible alien pollen will be rejected if SI is functionally operating in the pollen recipient species in some cases (see details in Kitashiba and Nasrallah 2014). Consequently, it is reasonable to speculate that the SI in Cakile species did not treat congeneric pollen as incompatible, leading to the occurrence of hybrids in both directions upon hybridisation. Such a speculation was also confirmed by the backcrossing results, in which pollen germination and pollen tube growth on the stigma of C. maritima were not rejected either.

Despite successful pollen germination in both crossing directions, differences in pollen tube growth were found. In all samples that had pollen successfully germinated, the pollen tubes grew rapidly into the style within the first collection time interval (2 h) but the pollen tube growth was significantly worse in the process of producing hybrids than that within the original parental species (i.e., producing outcrossed P1, EE and MM) (P < 0.05; see Table 5.1 for details about the pollen donor and recipient of each plant type). However, no significant difference was detected in the pollen tube growth between the processes of producing F1s, F2s and BCs. Such results indicate that a) interspecific pollen upon hybridisation performed worse than intraspecific pollen in intraspecific crossing; b) after hybridisation, pollen grains from hybrid tended to show their inferiority in pollen germination and pollen tube growth compared with the pollen grains of the original parents, either in crossing between or within plant types.

It should be noted that the observed difference in pollen germination and pollen tube growth was only examined under the single donor conditions, i.e., no pollen mixture was considered when conducting pollen germination experiments. In contrast, under natural conditions, mixed pollen load from different plant types of hybrids or parental species is the norm (Sork and Schemske 1992, Swanson et al. 2016). Although the seeds in Cakile tend to spread and germinate sparsely at different distances from their mother plants under natural conditions, most of the seedlings are restricted within a small radius (Cordazzo 2006). Consequently, if hybrids are formed, it is very likely that they will be in sympatry with their parents and mixed pollination can be achieved by foraging pollinators. Pollen competition can occur if a mixture of compatible

102 pollen is carried and deposited onto the stigma by pollinators. The order of the arrival of different pollen types determines which one gains preferential access, beginning pollen germination ahead of others. When pollen load is a mixture of different pollen types arriving at the same time, there will also be competition between pollen tube growth, in which pollen with faster pollen tube growth gets a higher chance to fertile the ovules (Rieseberg et al. 1995, Swanson et al. 2016). Many self-compatible plants possess physiological mechanisms in the pistil that reduce the performance of self-pollen relative to outcross-pollen (cryptic self- incompatibility; Bateman 1956, Carol et al. 2005, Cruzan and Barrett 2015, Lankinen et al. 2015; see details in Chapter 2). Thus, it allows plants to preferentially outcross when mixed pollen load is applied, avoiding inbreeding depression in the progeny generation.

After hybridisation, the breeding system of hybrids was checked in which SI showed the dominance over SC in all hybrid generations with the occurrence of several SC plants in each generation. It is proposed in this thesis that the SI and SC present in the model species cannot be explained by the presence of a single pair of alleles; instead SC could consist of multiple alleles and is recessive to SI. Thus, SI is dominant over SC in produced F1s. However, with recombination between alleles, the probability to get a hybrid with more recessive SC genes becomes higher, i.e., SI tends to be weakened generation by generation. Furthermore, introducing backcrossing with the SC species could accelerate the process and enhance the SC in the produced population. Such process could also be reversed if hybrids are backcrossed with SI species (e.g., HM in backcrosses in this thesis).

However, the genetic basis of the breakdown and the tendency of the reversibility of SI in hybrids mentioned above have not been explored in this study. Previous studies have proven that there is a close relationship between SI and SC, and many modern self-incompatible species have self-fertile relatives. Phylogenetic studies also have suggested that SI was the ancestral state in most dicots and that SC species were derived secondarily from SI species (Igic and Kohn 2001, Steinbachs and Holsinger 2002) due to the breakdown of SI, which is commonly caused by the loss-of-function mutations at the S-locus (Igic et al. 2008, Vekemans et al. 2014). Normally, the breakdown in these patterns is irreversible and thus the loss of SI appears far more common than its gain. In contrast, the mechanisms of the breakdown of SI may be different when the hybridisation between SI and SC species is involved. Nasrallah et al. (2007) showed that SI was lost in hybrids in which aberrant processing of S-locus receptor kinase gene transcripts and suppression of the S-locus cysteine-rich protein gene were observed. These are two reversible mechanisms by which SI might break down upon interspecific hybridisation to generate self- fertile hybrids in nature (Nasrallah et al. 2007). According to his results, the breakdown of SI could be reserved by backcrossing F1 hybrids with the SI parents, which corresponds to the results in this study to some extent.

Therefore, the following topics are of great value and need further attentions from researchers., Firstly, considering the results presented in Chapter 4 that some of the stigmas in C. edentula may be contaminated by self-pollen before flowering while others may be not, it would be much more complicated as well as interesting to determine the interactions between the timing of pollination and pollen competition, which is an area worth further investigation. Secondly, as described in Chapter 5, the inferiority was already observed even in the single pollen donor experiments upon hybridisation. Consequently, it can be inferred that cryptic self-

103 incompatibility may also present if mixed pollen loads on the stigma of two phylogenetically close-related species, e.g., C. maritima and C. edentula (Rodman 1976, 1980), and pollen discrimination could occur especially in self-compatible C. edentula. Such speculations need to be verified in further studies. Thirdly, it has been mentioned in Chapter 6 that there was a tendency for the SI to be enhanced via backcrossing with SI species, inducing reduced proportion of SC hybrids with lower selfing rates. However, whether such a breakdown observed in this study shares the same the genetic basis as speculated in Nasrallah et al. (2007) and thus is reversible is yet to be determined in future studies.

7.3 Fitness costs of hybridisation In this study, both the separate fitness components and cumulative fitness suggested that hybridisation between Cakile species with contrasting breeding systems did contribute to the differences of plant fitness; however, the difference was not significant until plants became reproductive.

Specifically, hybrids did not show any significant inferiority compared with their parental species at early stages, indicating that there was no significant cost of hybridisation between the two Cakile species before hybrids became reproductive. Since seedlings are more vulnerable than adult individuals to adverse environmental conditions (Burke et al. 1998), seed germination, seedling survival and seedling development are crucial parts for moving seedlings out from relatively dangerous stages. Therefore, such results suggest that the seeds and seedlings of hybrids between the two Cakile species are equally likely to pass through environmental sieves as seeds and seedlings of the parental species (Johnston et al. 2003).

In contrast, significant differences in fitness were observed at late stages of plant life cycle: most hybrids showed intermediate fitness while some of them, especially the backcrosses with the SI species (C. maritima), had a comparable and even larger floral display, plant size as well as number of seed set. Most hybrids, as mentioned in Chapter 5, were self-incompatible and theoretically cannot be pollinated without pollinators while they were less attractive than the self-incompatible C. maritima. Consequently, pollinators may have often visited C. maritima and the more attractive backcrosses (i.e., HM and MH) instead of those less attractive hybrids in the field experiments where they were randomly arranged in sympatry, inducing larger number of pollinated flowers and thus potentially larger seed set number in C. maritima and the backcrosses with C. maritima. In other words, the gene flow is most likely to occur between C. maritima and the backcrosses with C. maritima, leading to possible introgression between the SI species and its backcrosses, and potentially helping their establishment.

These findings confirm the assumptions made in Mesgaran et al. (2016) and thus provide support for their model results. Based on previous studies about the establishment and invasion of Cakile species (Boyd and Barbour 1993, Aigner 2004, Mhemmed et al. 2008, 2012, Mills 2013) as well as their own lab work (Ohadi et al. 2012, Cousens et al. 2013, Ohadi et al. 2015), Mesgaran et al. (2016) developed a model in which it was suggested that hybridisation between related species could enable the colonizer genes to establish. They also used the two Cakile species as the model species and suggested that this hybridisation-rescue effect is more likely to eventuate if the new species and hybrids are more attractive to pollinators than the established species. The results in this study confirm that some of the hybrids were equally or

104 even more attractive than their parental species, inducing the interaction between the backcrosses and the SI parent and thus enabling the hybridisation-rescue effect. Similarly, other main assumptions made in their model have also been confirmed by the results of this study (see details in Table 6.14). Therefore, the results in this study, together with the model of Mesgaran et al. (2016), suggest the existence of the hybridisation-rescue effect in which related species can co-invade and benefit from hybridisation.

However, how widespread such co-invasion and benefits from hybridisation are likely to be has not been addressed, either in this study or in the model of Mesgaran et al. (2016). Is this merely an exception and the only case in the world where it might occur, or could it be widespread? This question has been discussed in Mesgaran et al. (2016) and they believed that it may be a common phenomenon in nature. Since pollinators normally discriminate among plant types, they argued that, as indicated by their model results, the presence of another species can alleviate pollen limitation and reduce Allee effects through hybridisation with related species and the asymmetry in the direction of introgression (i.e., backcrossing) due to pollinator discrimination will enhance such benefits. Such speculations indeed have been supported by the results of this study but remain to be questioned unless more similar cases will have been reported in future studies.

7.4 Limitations During the past four years, although I have tried my best to design and conduct the experiments, analyse the data, and write the thesis, it is still only an in-depth study into a very tiny aspect of the field. Additionally, it is also limited by virtue of time, energy and personal capability. In this section, I will discuss the major limitations in this research.

Firstly, the experimental design, instead of a fully diallel one, was a partially diallelic and reduced crossing plan. Normally, to incorporate genetic variations as many as possible, a diallel plan should be taken into consideration when doing any crossing on plants (Hiscock and McInnis 2003, Goodwillie and Ness 2013, Gibbs 2014). However, to ensure availability of seeds in each direction, at least 100 buds in each crossing direction were artificially pollinated in this study. Therefore, if a complete diallel crossing were performed, there would be more than 20,000 replicates. This is only for a single generation whilst there would be three hybrid generations in total. Considering this, although it could be disadvantageous, a partially diallel crossing plan was applied in this experiment (Table 3.2).

Secondly, there were only four generations (P1 and BC inclusive) included in this study, which could be risky to draw conclusions from the evolutionary point of view. Evolutionary conclusions & implications involved in this study were only based on the data collected in four generations. Although this phenomenon is common in experimental studies (Briscoe Runquist et al. 2017), it could be much more robust to include as many generations as possible to draw evolutionary conclusions. However, this study dealt with two annual plant species whose life cycle from seed germination to seed set usually takes more than half a year. Consequently, three hybrid generations and the final common garden experiment almost took two years in total, which did not include the experiment preparation time. Moreover, one generation was destroyed by parasites in the glasshouse, which wasted another half a year since relevant experiments had to be prepared all over again. Therefore, no more generations were considered in this study.

105 Thirdly, genetic work, especially related to the breeding system, was absent in this study. Self- incompatibility is the key component of the plant breeding system and its genetic basis has been extensively studied in recent years (Nasrallah et al. 1992, Kitashiba and Nasrallah 2014) (see details in Section 2.2.2). Such genetic studies depict the mechanism thoroughly and help fully understand how it works. Although there are many studies about how SC was generated as a secondary state from SI (Sloop et al. 2009, Dwyer et al. 2013), there are barely genetic studies about the mechanisms of the inheritance of the breeding system when SI encounters SC upon hybridisation. Such genetic studies are of great value and would involve a large amount of work, which could be a new PhD project.

7.5 Conclusions At the start of this study, it was questioned whether and how the hybridisation between these two Cakile species with contrasting breeding systems would occur, whether the inheritance of SI would show up in hybrids, and what the fitness costs upon hybridisation would be. In order to answer these questions, during the past four years, enormous time and energy have been invested into this research.

Despite all the limitations mentioned above, this thesis has demonstrated that although anther dehiscence occurred before flower opening, hybridisation between Cakile species with contrasting breeding systems was achieved in both directions and the SI was inherited in most hybrids. Meanwhile, backcrosses with the SI parental species showed higher fitness compared with other hybrids, leading to the possibility that the interaction is more likely to occur between such hybrids and the SI parental species. Therefore, unlike traditional mechanisms of plant invasion, the hybridisation with closely related species and the following interaction between hybrids and their parents can provide the newly arriving species with increased mate availability when it spreads into a new environment, easing the Allee effect and thus helping its establishment.

106 Appendices

Appendix to Chapter 4

Appendix 1- 1 Differences of upper-stamen length between the parental species within each floral development stage.

Species Stage _species _stage Estimate SE Pr > |t| edentula 2 maritima 2 -0.59 0.21 0.0049 edentula 3 maritima 3 -0.63 0.21 0.0031 edentula 4 maritima 4 -0.91 0.22 <.0001 edentula 5 maritima 5 -1.34 0.23 <.0001 edentula 6 maritima 6 -1.88 0.21 <.0001 edentula 7 maritima 7 -1.22 0.23 <.0001

Appendix 1- 2 Differences of lower-stamen length between the parental species within each floral development stage.

Species Stage _species _stage Estimate SE Pr > |t| edentula 2 maritima 2 -0.39 0.19 0.0363 edentula 3 maritima 3 -0.57 0.19 0.0026 edentula 4 maritima 4 -0.75 0.19 0.0002 edentula 5 maritima 5 -0.53 0.21 0.0107 edentula 6 maritima 6 -1.09 0.19 <.0001 edentula 7 maritima 7 -1.14 0.21 <.0001

Appendix 1- 3 ANOVA table of the development of stamens in different species. “stage * population” means the interaction between the two effects. Interaction between stage and population in C. edentula was dropped as it showed non-significance for both lower and upper stamens. Non-significant comparisons are in bold.

Species Position Source DF Type III SS Mean Square F Value Pr > F Stage 5 174.31 34.86 419.12 <.0001 Upper Population 1 0.18 0.18 2.13 0.1473 edentula Stage 5 112.87 22.57 237.94 <.0001 Lower Population 1 0.05 0.05 0.53 0.4681 Stage 5 230.03 46.01 586.33 <.0001 Upper Population 1 3.44 3.44 43.78 <.0001 Stage*population 5 2.54 0.51 6.49 <.0001 maritima Stage 5 118.01 23.60 350.93 <.0001 Lower Population 1 2.69 2.69 40.06 <.0001 Stage*population 5 1.38 0.28 4.11 0.0022

107

Appendix 1- 4 ANOVA table of length difference between paired and unpaired stamens within the same stage of Cakile edentula.

Stage DF SS MS F Pr > F 2 1 0.23 0.23 7.22 0.0106 3 1 0.60 0.60 7.85 0.0081 4 1 0.56 0.56 19.55 <.0001 5 1 0.72 0.72 24.13 0.0002 6 1 4.42 4.42 43.93 <.0001 7 1 6.48 6.48 28.05 <.0001

Appendix 1- 5 ANOVA table of length difference between upper and lower stamens within the same stage of Cakile maritima.

Stage DF SS MS F Pr > F 2 1 1.20 1.20 14.15 0.0006 3 1 0.95 0.95 13.64 0.0007 4 1 0.95 0.95 7.08 0.0143 5 1 12.98 12.98 287.14 <.0001 6 1 20.59 20.59 90.19 <.0001 7 1 2.40 2.40 7.58 0.0204

Appendix 1- 6 Stamen length (mm) of each population in C. maritima. RM: C. maritima of Raspins Beach; UM: C. maritima of Ulverstone.

Stage Population Lower stamen SE Stage Population Upper stamen SE

2 RM 2.34 0.08 2 RM 2.72 0.09 2 UM 2.63 0.08 2 UM 2.94 0.09 3 RM 2.65 0.08 3 RM 2.93 0.09 3 UM 2.83 0.08 3 UM 3.17 0.09 4 RM 3.20 0.10 4 RM 3.61 0.11 4 UM 3.78 0.12 4 UM 4.16 0.13 5 RM 4.27 0.10 5 RM 5.55 0.11 5 UM 4.24 0.08 5 UM 5.45 0.09 6 RM 4.86 0.09 6 RM 6.21 0.09 6 UM 5.47 0.08 6 UM 7.06 0.09 7 RM 5.48 0.18 7 RM 6.26 0.20 7 UM 6.13 0.13 7 UM 7.08 0.14

108

Appendix 1- 7 Differences of stamen length between populations in C. maritima. Comparisons were only performed within stages and between populations. Cross-stage comparisons were omitted. Non-significant P values are in bold.

Population and stage interaction within stages of lower Population and stage interaction within stages of upper stamens stamen Stage SS F Value Pr > F Stage SS F Value Pr > F 2 0.42 6.27 0.0142 2 0.24 3.04 0.0849 3 0.15 2.29 0.1341 3 0.29 3.72 0.0571 4 1.00 14.92 0.0002 4 0.88 11.26 0.0012 5 0.00 0.07 0.7974 5 0.04 0.51 0.4791 6 1.78 26.46 <.0001 6 3.38 43.13 <.0001 7 0.58 8.57 0.0044 7 0.92 11.66 0.001

Appendix 1- 8 ANOVA table of differences in pistil length (mm) of C. martima between populations within each stage, Non-significant comparisons are in bold.

Effect UM RM stage Num DF Den DF F Value Pr > F

stage*population 2.96 2.71 2 1 82 3.63 0.0602 stage*population 3.43 3.14 3 1 82 5.30 0.0239 stage*population 4.22 3.64 4 1 82 11.98 0.0009 stage*population 4.65 4.47 5 1 82 1.63 0.2052 stage*population 6.01 5.05 6 1 82 53.81 <.0001 stage*population 6.52 6.29 7 1 82 0.88 0.3509

Appendix 1- 9 Differences of pistil length (mm) between species within stages.

Stage C. edentula C. maritima F Value Pr > F

2 2.27 2.84 6.02 0.0151 3 2.72 3.28 6.01 0.0151 4 3.01 3.91 14.42 0.0002 5 3.39 4.54 22.08 <.0001 6 3.59 5.54 70.92 <.0001 7 4.02 6.37 88.38 <.0001

109

Appendix 1- 10 Scatter plots of stamen and pistil length in C. edentula and C. maritima with 95% prediction ellipse. Correlation coefficient and P value are marked on the top corner of each graph. All scatter plots in C. maritima showed very strong (> 0.96) correlation while only one scatter plot in C. edentula had a correlation coefficient larger than 0.96, i.e., correlation between lower and upper stamens, which also was the strongest in C. maritima.

110

Appendix 1- 11 Linear regression between lower and upper stamens and pistil in C. edentula (upper panel) and C. maritima (lower panel). 95% confidence and prediction limits were indicated in each linear regression plot.

111

Appendix to Chapter 5

Appendix 2- 1 Bud pollination with self-pollen in C. maritima and pollen germination on the stigma of C. maritima Fluorescence of pollen tube under UV (left panel) and visible light + UV (right panel) 2 hours after pollination in a style. Black arrow indicates unsuccessfully rehydrated pollen grain. Red and green arrows represent pollen tube and pollen grains respectively. Scale bar = 500 µm. Pollen tube growth of the self-pollen on the stigma of C. maritima indicate that the SI upon bud pollination was not fully functional.

Appendix 2- 2 Pollen tube growth 2 hours after bud pollination. A, style under light microscope. B, pollen tube under UV light. C and D, enlarged pictures of rectangular area in A and B, respectively. Vb, vascular bundle. Pt, pollen tube. Scale bar = 500 µm. Pollen tube growth on the stigma indicate that the SI upon bud pollination was not fully functional.

112

Appendix 2- 3 Pollen germination on fully developed but pre-emasculated open buds. A and B are styles under light microscope. C and D are the corresponding views under UV light. Scale bar = 500 µm. The inhibition of pollen tube growth on the fully developed but pre-emasculated open buds indicate that the SI was fully functional.

113

Appendix to Chapter 6

Appendix 3- 1 Parental individuals used in fitness experiment. M2 and M3 have been excluded from last experiment because there phenotypically resembled C. edentula and were also been removed here

Species Cakile maritima Cakile edentula

Individuals M1sib M4sib M5sib M6sib M7sib M8sib M9sib E1sib E2sib E3sib E4sib E5sib E6sib E7sib E8sib E9sib E10sib M10sib

Appendix 3- 2 Parental individuals of intraspecific crossings used in common garden experiment. M2 individual had been excluded from the experiment and thus intraspecific MM1 and MM2 were not included. Thus 6 crossings of interspecific crossings in C. maritima were available. “→” indicates the pollen flow direction when producing seeds. For example, “M2→M5” means that pollen from M2 sire the ovule of M5.

Population Raspins Beach, TAS Ulverstone, TAS Marion Bay, TAS Raspins Beach, TAS C. maritima N/A MM3 (M2→M5) MM5 (M2→M7) N/A MM4 (M7→M5) MM6 (M5→M7)

C. edentula EE1 (E10→E1) N/A N/A EE2 (E1→E10)

Appendix 3- 3 F1s (with C. maritima as the maternal plant) used in common garden experiment.

Hybrid F117 F118 F120 F121 F122 F123 F124 F125 F126 F127 F128 F129 F131 ♂ E9 E10 E4 E5 E6 E7 E8 E9 E10 E1 E2 E3 E4 ♀ M4 M5 M4 M5 M6 M7 M8 M9 M10 M6 M7 M8 M10

Appendix 3- 4 F1s (with C. edentula as the maternal plant) used in common garden experiment.

Hybrid F11 F12 F13 F14 F15 F16 F17 F18 F19 F110 F111 F112 F113 F114 F115 ♀ E6 E9 E10 E1 E4 E5 E6 E7 E8 E9 E10 E1 E2 E3 E4 ♂ M1 M4 M5 M1 M4 M5 M6 M7 M8 M9 M10 M6 M7 M8 M10

Appendix 3- 5 Backcrosses used in common garden experiment and their lineages. For backcrossing, we used “H+M/E+#” and “M/E +H+#” to label the seeds from different crossing directions when hybrids (F1s) were the pollen donor and pollen recipient respectively. For example, HM1 indicates the backcrossing seeds sired by E1 with F11.

# F11 F12 F13 F14 F15 F16 F17 E HE1; HE2; HE3; HE4; HE5; HE6; HE7; EH1 EH2 EH3 EH4 EH5 EH6 EH7 M HM1; HM15; HM2; HM3; HM4; MH4 HM5; HM6; HM7; MH1; MH1-5 MH2 MH3 HM4-1; MH4-1 MH5 MH6 MH7

# F18 F19 F111 F110 F112 F113 F114 F115 E HE8; HE9; HE11; HE10; HE12; HE13; HE14; HE15; EH8 EH9 EH11 EH10 EH12 EH13 EH14 EH15 M HM8; HM9; HM11; HM10; HM12; HM13; HM14; HM15; MH8 MH9 MH11 MH10 MH12 MH13 MH14 MH15

114

Appendix 3- 6 F2s used in the common garden experiment and their lineages. Here, “HH1-HH12” were used to label the crossings. For example, the seeds from crossing F11× F15 were labelled as “HH1” and pollen donor was placed before the pollen recipient, i.e., F11 was the pollen donor F15 was the pollen recipient in this crossing.

F2 HH1 HH2 HH3 HH4 HH5 HH6

Lineage F15× F11 F113× F11 F14× F12 F113× F12 F12× F14 F18× F14

F2 HH7 HH8 HH9 HH10 HH11 HH12

Lineage F11× F15 F18× F15 F14× F18 F15× F18 F11× F113 F12× F113

Appendix 3- 7 Differences in germination between plant types, estimated using Least Squares Means. Only significant and marginally significant comparisons were retained in the table.

Plant type _plant type Estimated difference SE t Value Pr > |t| C. edentula HE -2.93 1.47 -1.99 0.049 C. edentula HM -1.93 0.91 -2.11 0.037 EE EM -3.22 1.57 -2.06 0.042 EE HE -3.75 1.57 -2.39 0.018 EE HM -2.75 1.06 -2.59 0.011 EE MH 2.90 1.21 -2.4 0.018 C. maritima HE -2.59 1.53 -1.69 0.094 C. maritima ME 1.71 0.63 2.73 0.007 MM ME 2.26 1.03 2.19 0.030 C. edentula ME 1.37 0.45 3.01 0.003 EM ME 3.77 1.43 2.63 0.010 HH ME 2.30 0.74 3.09 0.003 HE ME 4.29 1.43 3 0.003 HM ME 3.30 0.85 3.9 0.000 MH ME 3.45 1.02 3.37 0.001

Appendix 3- 8 Mean petal length and width and their correlation coefficient in each plant type. In most plant types, there was a strong correlation between the petal length and width. C. edentula and its outcrossed progeny (EE) had the worst correlation in all species (R2 < 0.51, n= 31 and 6). F1s (EM and ME) had intermediate correlation coefficients while C. maritima, BCs and F2 species had a much higher coefficient (R2 > 0.70). Petal lengths and widths overlapped extensively between most plant types except in C. edentula.

Generation Plant type Length (mm) Width (mm) Correlation coefficient C. maritima 4.71 4.29 0.87 MM 4.58 3.58 0.87 P1 C. edentula 2.97 1.88 0.34 EE 3.01 1.75 0.50 EM 3.71 3.13 0.56 F1 ME 3.54 3.06 0.64 F2 HH 3.57 3.14 0.91 EH 3.55 3.10 0.78 HE 4.01 3.18 0.74 BC HM 4.38 3.90 0.88 MH 4.36 3.82 0.84

115

Appendix 3- 9 Distribution of plants with fruit set and without fruit set in all plant types. Frequencies of the two types of plants were plotted against and stacked within plant types.

116

References

Abbott, R., D. Albach, S. Ansell, J. W. Arntzen, S. J. E. Baird, N. Bierne, J. Boughman, A. Brelsford, A. C. Buerkle, and R. Buggs. 2013. Hybridization and speciation. Journal of Evolutionary Biology 26:229-246. Adger, W. N., N. W. Arnell, and E. L.Tompkins. 2005. Successful adaptation to climate change across scales. Global Environmental Change 15:77-86. Adrian, C. B., A. H. Stephen, and J. H. Simon. 2013. The population genetics of sporophytic self- incompatibility in three hybridizing Snecio (Asteraceae) species with contrasting population histories. Evolution 67:1347-1367. Agrawal, G. K., and R. Rakwal. 2012. Seed Development: OMICS Technologies Toward Improvement of Seed Quality and Crop Yield. Springer. Aigner, P. A. 2004. Ecological and genetic effects on demographic processes: pollination, clonality and seed production in Dithyrea maritima. Biological Conservation 116:27-34. Allard, R., G. Babbel, M. Clegg, and A. L. Kahler. 1972. Evidence for coadaptation in Avena barbata. Proceedings of the National Academy of Sciences 69:3043-3048. Allen, A. M., C. J. Thorogood, M. J. Hegarty, C. Lexer, and S. J. Hiscock. 2011. Pollen–pistil interactions and self-incompatibility in the Asteraceae: new insights from studies of Senecio squalidus (Oxford ragwort). Annals of Botany 108:687-698. Allendorf, F. W. 1986. Genetic drift and the loss of alleles versus heterozygosity. Zoo Biology 5:181-190. Allison, A. S., L. U. Kristen, and M. C. Theresa. 2001. Fitness of hybrids between weedy and cultivated radish: implications for weed evolution. Ecological Applications 11:934-943. Armbruster, W. S., C. P. Mulder, B. G. Baldwin, S. Kalisz, B. Wessa, and H. Nute. 2002. Comparative analysis of late floral development and mating-system evolution in tribe Collinsieae (Scrophulariaceae s.l.). American Journal of Botany 89:37-49. Arnold, M. L., and N. H. Martin. 2010. Hybrid fitness across time and habitats. Trends in Ecology & Evolution 25:530-536. Arriola, P. E., and N. C. Ellstrand. 1997. Fitness of interspecific hybrids in the genus Sorghum: persistence of crop genes in wild populations. Ecological Applications 7:512-518. Arroyo, J., S. C. Barrett, R. Hidalgo, and W. W. Cole. 2002. Evolutionary maintenance of stigma- height dimorphism in Narcissus papyraceus (Amaryllidaceae). American Journal of Botany 89:1242-1249. Austen, E. J., and A. E. Weis. 2015. What drives selection on flowering time? An experimental manipulation of the inherent correlation between genotype and environment. Evolution 69:2018-2033. Baack, E., M. C. Melo, and L. H. Rieseberg. 2015. The origins of reproductive isolation in plants. New Phytologist 207:968–984 Baack, E. J., and L. H. Rieseberg. 2007. A genomic view of introgression and hybrid speciation. Current Opinion in Genetics & Development 17:513-518. Baker, H. G. 1955. Self-compatibility and establishment after'long-distance'dispersal. Evolution 9:347-349. Baker, H. G. 1967. Support for Baker's law-as a rule. Evolution 21:853-856. Barbour, M. G. 1970. Seedling ecology of Cakile maritima along the California coast. Bulletin of the Torrey Botanical Club 97:280-289. Barbour, M. G. 1972. Seedling establishment of Cakile maritima at Bodega Head, California. Bulletin of the Torrey Botanical Club 99:11-16. Barrett, S. C., and L. D. Harder. 1996. Ecology and evolution of plant mating. Trends in Ecology & Evolution 11:73-79.

117

Barrett, S. C., and L. D. Harder. 2017. The eology of mating and its evolutionary consequences in seed plants. Annual Review of Ecology, Evolution, and Systematics 48:135-157. Barrett, S. C. H. 2002. The evolution of plant sexual diversity. Nature Reviews Genetics 3:274-284. Bateman, A. J. 1956. Cryptic self-incompatibility in the wallflower: Cheiranthus cheiri L. Heredity 10:257-261. Baumel, A., M. Ainouche, M. Misset, J. Gourret, and R. Bayer. 2003. Genetic evidence for hybridisation between the native Spartina maritima and the introduced Spartina alterniflora (Poaceae) in South-West France: Spartina× neyrautii re-examined. Plant Systematics and Evolution 237:87-97. Bedsworth, L. W., and E. Hanak. 2010. Adaptation to climate change. Journal of the American Planning Association 76:477-495. Bernd, B., and N. Rolf. 1995. Evolution of Increased Competitive Ability in Invasive Nonindigenous Plants: A Hypothesis. Journal of Ecology 83:887-889. Bertin, R. I., and C. M. Newman. 1993. Dichogamy in angiosperms. The Botanical Review 59:112– 152. Bhatt, M. V., A. Khandelwal, and S. A. Dudley. 2011. Kin recognition, not competitive interactions, predicts root allocation in young Cakile edentula seedling pairs. New Phytologist 189:1135–1142. Billiard, S., V. Castric, and X. Vekemans. 2007. A general model to explore complex dominance patterns in plant sporophytic self-incompatibility systems. Genetics 175:1351-1369. Bittencourt, N. S., E. G. Peter, and S. JoÃO. 2003. Histological Study of Post‐pollination Events in Spathodea campanulata Beauv. (Bignoniaceae), a Species with late-acting self- incompatibility. Annals of Botany 91:827-834. Bittencourt, N. S., and J. Semir. 2006. Floral biology and late-acting self-incompatibility in Jacaranda racemosa (Bignoniaceae). Australian Journal of Botany 54:315-324. Black, J. M., and E. L. Robertson. 1965. Flora of South Australia. Government Printer. Boavida, L. C., and S. McCormick. 2007. Temperature as a determinant factor for increased and reproducible in vitro pollen germination in Arabidopsis thaliana. The Plant Journal 52:570-582. Bou Daher, F., Y. Chebli, and A. Geitmann. 2009. Optimization of conditions for germination of cold-stored Arabidopsis thaliana pollen. Plant Cell Reports 28:347-357. Bouhours, J., M. B. Mesgaran, R. D. Cousens, and M. A. Lewis. 2017. Neutral hybridization can overcome a strong Allee effect by improving pollination quality. Theoretical Ecology 10:319-339. Bowman, J. L., G. N. Drews, and E. M. Meyerowitz. 1991. Expression of the Arabidopsis floral homeotic gene AGAMOUS is restricted to specific cell types late in flower development. The Plant Cell 3:749-758. Bowman, J. L., D. R. Smyth, and E. M. Meyerowitz. 1989. Genes directing flower development in Arabidopsis. The Plant Cell 1:37-52. Boyd, R. S., and M. G. Barbour. 1993. Replacement of Cakile edentula by C. maritima in the strand habitat of California. American Midland Naturalist 130:209-228. Brandvain, Y., and D. Haig. 2005. Divergent mating systems and parental conflict as a barrier to hybridisation in flowering plants. The American Naturalist 166:330-338. Brennan, A. C., D. Barker, S. J. Hiscock, and R. J. Abbott. 2011a. Molecular genetic and quantitative trait divergence associated with recent homoploid hybrid speciation: a study of Senecio squalidus (Asteraceae). Heredity 108:87-95. Brennan, A. C., S. A. Harris, and S. J. Hiscock. 2006. The population genetics of sporophytic self- incompatibility in Senecio squalidus L. (Asteraceae): the number, frequency, and dominance interactions of S alleles across its British range. Evolution 60:213-224.

118

Brennan, A. C., S. A. Harris, and S. J. Hiscock. 2013. The population genetics of sporophytic self- incompatibility in three hybridizing Senecio (Asteraceae) species with contrasting population histories. Evolution 67:1347-1367. Brennan, A. C., D. A. Tabah, S. A. Harris, and S. J. Hiscock. 2011b. Sporophytic self-incompatibility in Senecio squalidus (Asteraceae): S allele dominance interactions and modifiers of cross- compatibility and selfing rates. Heredity 106:113-123. Briscoe Runquist, R. D., M. A. Geber, M. Pickett-Leonard, and D. A. Moeller. 2017. Mating system evolution under strong pollen limitation: evidence of disruptive selection through male and female fitness in Clarkia xantiana. The American Naturalist 189:549-563. Brys, R., J. Cauwenberghe, and H. Jacquemyn. 2016. The importance of autonomous selfing in preventing hybridisation in three closely related plant species. Journal of Ecology 104:601–610. Burke, J. M., S. E. Carney, and M. L. Arnold. 1998. Hybrid fitness in the Louisiana irises: analysis of parental and F1 performance. Evolution 52:37-43. Burton, T. L., and B. Husband. 2000. Fitness differences among diploids, tetraploids, and their triploid progeny in Chamerion angustifolium: mechanisms of inviability and implications for polyploid evolution. Evolution 54:1182-1191. Busch, J., and L. Delph. 2012. The relative importance of reproductive assurance and automatic selection as hypotheses for the evolution of self-fertilization. Annals of Botany 109:553- 562. Busch, J. W., T. Witthuhn, and M. Joseph. 2014. Fewer S ‐ alleles are maintained in plant populations with sporophytic as opposed to gametophytic self‐incompatibility. Plant Species Biology 29:34–46. Bushell, C., M. Spielman, and R. J. Scott. 2003. The basis of natural and artificial postzygotic hybridisation barriers in Arabidopsis species. The Plant Cell 15:1430-1442. Callaway, R. M., and W. M. Ridenour. 2004. Novel weapons: invasive success and the evolution of increased competitive ability. Frontiers in Ecology and the Environment 2:436-443. Campbell, D. R. 2003. Natural selection in Ipomopsis hybrid zones: implications for ecological speciation. New Phytologist 161:83-90. Campbell, D. R., M. Crawford, A. K. Brody, and T. A. Forbis. 2002. Resistance to pre-dispersal seed predators in a natural hybrid zone. Oecologia 131:436-443. Carol, G., K. Susan, and G. E. Christopher. 2005. The evolutionary enigma of mixed mating systems in plants: occurrence, theoretical explanations, and empirical evidence. Annual Review of Ecology, Evolution, and Systematics 36:47-79. Carrió, E., and J. Güemes. 2014. The effectiveness of pre‐ and post‐zygotic barriers in avoiding hybridisation between two snapdragons (Antirrhinum L.: Plantaginaceae). Botanical Journal of the Linnean Society 176:159-172. Cartwright, R. 2009. Antagonism between local dispersal and self-incompatibility systems in a continuous plant population. Molecular Ecology 18:2327-2336. Castro, J. 1999. Seed mass versus seedling performance in Scots pine: a maternally dependent trait. New Phytologist 144:153–161. Castro, S., S. Silva, I. Stanescu, P. Silveira, L. Navarro, and C. Santos. 2009. Pistil anatomy and pollen tube development in Polygala vayredae Costa (Polygalaceae). Plant Biology 11:405-416. Charlesworth, D. 2006. Evolution of plant breeding systems. Current Biology 16:R726-R735. Charlesworth, D., and J. H. Willis. 2009. The genetics of inbreeding depression. Nature Reviews Genetics 10:783-796. Chen, S., M. N. Nelson, A.-M. Chèvre, E. Jenczewski, Z. Li, A. S. Mason, J. Meng, J. A. Plummer, A. Pradhan, and K. H. Siddique. 2011. Trigenomic bridges for Brassica improvement. Critical reviews in plant sciences 30:524-547.

119

Cheng, F., J. Wu, and X. Wang. 2014. Genome triplication drove the diversification of Brassica plants. Horticulture research 1:14024. Cheptou, P. O. 2004. Allee effect and self‐fertilization in hermaphrodites: reproductive assurance in demographically stable populations. Evolution 58:2613-2621. Cheptou, P. O. 2012. Clarifying Baker's law. Annals of Botany 109:633-641. Christe, C., K. N. Stölting, L. Bresadola, B. Fussi, B. Heinze, D. Wegmann, and C. Lexer. 2016. Selection against recombinant hybrids maintains reproductive isolation in hybridizing Populus species despite F1 fertility and recurrent gene flow. Molecular Ecology 25:2482- 2498. Ciosi, M., N. Miller, K. Kim, R. Giordano, A. Estoup, and T. Guillemaud. 2008. Invasion of Europe by the western corn rootworm, Diabrotica virgifera virgifera: multiple transatlantic introductions with various reductions of genetic diversity. Molecular Ecology 17:3614- 3627. Clausing, G., K. Vickers, and J. W. Kadereit. 2000. Historical biogeography in a linear system: genetic variation of sea rocket (Cakile maritima) and sea holly (Eryngium maritimum) along European coasts. Molecular Ecology 9:1823-1833. Clegg, M. T., R. W. Allard, and A. Kahler. 1972. Is the gene the unit of selection? Evidence from two experimental plant populations. Proceedings of the National Academy of Sciences 69:2474-2478. Cody, M. L., and T. W. D. Cody. 2004. Morphology and spatial distribution of alien sea-rockets (Cakile spp.) on South Australian and Western Canadian beaches. Australian Journal of Botany 52:175-183. Coen, E. S., and E. M. Meyerowitz. 1991. The war of the whorls: genetic interactions controlling flower development. Nature 353:31-37. Colautti, R. I., J. L. Maron, and S. C. Barrett. 2009. Common garden comparisons of native and introduced plant populations: latitudinal clines can obscure evolutionary inferences. Evolutionary Applications 2:187-199. Colombo, L., J. Franken, A. R. Van der Krol, P. E. Wittich, H. Dons, and G. C. Angenent. 1997. Downregulation of ovule-specific MADS box genes from petunia results in maternally controlled defects in seed development. The Plant Cell 9:703-715. Cordazzo, C. 2006. Seed characteristics and dispersal of dimorphic fruit segments of Cakile maritima Scopoli (Brassicaceae) population of southern Brazilian coastal dunes. Brazilian Journal of Botany 29:259-265. Cousens, R. D., P. K. Ades, M. B. Mesgaran, and S. Ohadi. 2013. Reassessment of the invasion history of two species of Cakile (Brassicaceae) in Australia. Cunninghamia 13:275-290. Cousens, R. D., and J. M. Cousens. 2011. Invasion of the New Zealand Coastline by European Sea- Rocket (Cakile maritima) and American Sea-Rocket (Cakile edentula). Invasive Plant Science and Management 4:260-264. Crepy, M. A., and J. J. Casal. 2015. Photoreceptor‐mediated kin recognition in plants. New Phytologist 205:329-338. Cruzan, M. B. 1989. Pollen tube attrition in Erythronium grandiflorum. American Journal of Botany 76:562-570. Cruzan, M. B., and S. C. H. Barrett. 2015. Postpollination discrimination between self and outcross pollen covaries with the mating system of a self-compatible flowering plant. American Journal of Botany 103:568-576. Cummings, C. L., H. M. Alexander, and A. A. Snow. 1999. Increased pre-dispersal seed predation in sunflower crop-wild hybrids. Oecologia 121:330-338. Dafni, A. 1992. Pollination ecology: a practical approach. Nordic Journal of Botany 13:514. Dafni, A., and D. Firmage. 2000. Pollen viability and longevity: practical, ecological and evolutionary implications. Pages 113-132 Pollen and Pollination. Springer.

120

Dart, S., and C. G. Eckert. 2013. Experimental and genetic analyses reveal that inbreeding depression declines with increased self-fertilization among populations of a coastal dune plant. Journal of Evolutionary Biology 26:587-599. Darwin, C. 1877. The different forms of flowers on plants of the same species. John Murray. Davy, A. J., R. Scott, and C. V. Cordazzo. 2006. Biological flora of the British Isles: Cakile maritima Scop. Journal of Ecology 94:695-711. De Nettancourt, D. 1977. Incompatibility in angiosperms. Springer Berlin Heidelberg. De Nettancourt, D. 1997. Incompatibility in angiosperms. Sexual Plant Reproduction 10:185-199. De Nettancourt, D. 2001. Incompatibility and incongruity in wild and cultivated plants. Springer. Ditta, G., A. Pinyopich, P. Robles, S. Pelaz, and M. F. Yanofsky. 2004. The SEP4 gene of Arabidopsis thaliana functions in floral organ and meristem identity. Current Biology 14:1935-1940. Donohue, K. 1998. Effects of inbreeding on traits that influence dispersal and progeny density in Cakile edentula var. lacustris (Brassicaceae). American Journal of Botany 85:661-661. Donohue, K. 2003. The Influence of Neighbor Relatedness on Multilevel Selection in the Great Lakes Sea Rocket. The American Naturalist 162:77-92. Dudash, M. R., and D. E. Carr. 1998. Genetics underlying inbreeding depression in Mimulus with contrasting mating systems. Nature 393:682-684. Dudley, S. A., and A. L. File. 2007. Kin recognition in an annual plant. Biology Letters 3:435-438. Dufay, M., and E. Billard. 2012. How much better are females? The occurrence of female advantage, its proximal causes and its variation within and among gynodioecious species. Annals of Botany 109:505-519. Durand, E., R. Méheust, M. Soucaze, P. M. Goubet, S. Gallina, C. Poux, I. Fobis-Loisy, E. Guillon, T. Gaude, and A. Sarazin. 2014. Dominance hierarchy arising from the evolution of a complex small RNA regulatory network. Science 346:1200-1205. Dwyer, K. G., M. T. Berger, R. Ahmed, M. K. Hritzo, A. A. McCulloch, M. J. Price, N. J. Serniak, L. T. Walsh, J. B. Nasrallah, and M. E. Nasrallah. 2013. Molecular characterization and evolution of self-incompatibility genes in arabidopsis thaliana: The Case of the Sc Haplotype. Genetics 193:985-994. Eckert, C. G., and M. Allen. 1997. Cryptic self-incompatibility in tristylous Decodon verticillatus (Lythraceae). American Journal of Botany 84:1391-1397. Eckert, C. G., S. Kalisz, M. A. Geber, R. Sargent, E. Elle, P. O. Cheptou, C. Goodwillie, M. O. Johnston, J. K. Kelly, D. A. Moeller, E. Porcher, R. H. Ree, M. Vallejo-Marin, and A. A. Winn. 2010. Plant mating systems in a changing world. Trends in Ecology & Evolution 25:35-43. Eckert, C. G., B. Ozimec, C. Herlihy, C. Griffin, and M. Routley. 2009. Floral morphology mediates temporal variation in the mating system of a self-compatible plant. Ecology 90:1540-1548. Edlund, A. F., Q. Zheng, N. Lowe, S. Kuseryk, K. L. Ainsworth, R. H. Lyles, S. J. Sibener, and D. Preuss. 2016. Pollen from Arabidopsis thaliana and other Brassicaceae are functionally omniaperturate. American Journal of Botany 103:1-14. Elizabeth, W. M., E. A. Sarah, C. Lars, C. L. C. Steven, V. Robert, D. Steven, K. Sandra, J. K. Laura, W. C. Mark, T. B. Ian, K. Aleš, M. Corinne, T. Lin, and R. L. Andrew. 2015. The effect of polyploidy and hybridisation on the evolution of floral colour in Nicotiana (Solanaceae). Annals of Botany 115:1117–1131. Ellstrand, N. C., and K. A. Schierenbeck. 2000. Hybridization as a stimulus for the evolution of invasiveness in plants? Proceedings of the National Academy of Sciences 97:7043-7050. Emel, S. L., S. J. Franks, and R. B. Spigler. 2017. Phenotypic selection varies with pollination intensity across populations of Sabatia angularis. New Phytologist 215:813-824. Eyre-Walker, A., and P. D. Keightley. 2007. The distribution of fitness effects of new mutations. Nature Reviews Genetics 8:610-618. Favaro, R., A. Pinyopich, R. Battaglia, M. Kooiker, L. Borghi, G. Ditta, M. F. Yanofsky, M. M. Kater, and L. Colombo. 2003. MADS-box protein complexes control carpel and ovule development in Arabidopsis. The Plant Cell 15:2603-2611.

121

Feliner, G. N., I. Álvarez, J. Fuertes-Aguilar, M. Heuertz, I. Marques, F. Moharrek, R. Piñeiro, R. Riina, J. Rosselló, and P. Soltis. 2017. Is homoploid hybrid speciation that rare? An empiricist’s view. Heredity 118:513–516. Felker‐Quinn, E., and J. A. Schweitzer. 2013. Meta‐analysis reveals evolution in invasive plant species but little support for Evolution of Increased Competitive Ability (EICA). Ecology and Evolution 3:739-751. Figueroa-Castro, D. M., and T. P. Holtsford. 2009. Post-pollination mechanisms in Nicotiana longiflora and N. plumbaginifolia: pollen tube growth rate, offspring paternity and hybridisation. Sexual Plant Reproduction 22:187-196. File, A. L., G. P. Murphy, and S. A. Dudley. 2011. Fitness consequences of plants growing with siblings: reconciling kin selection, niche partitioning and competitive ability. Proceedings of the Royal Society B: Biological Sciences 279:209-218. Fisher, R. A. 1937. The wave of advance of advantageous genes. Annals of Human Genetics 7:355- 369. Fox, C. W. 2005. Problems in measuring among-family variation in inbreeding depression. American Journal of Botany 92:1929-1932. Galloway, L., J. Etterson, and J. Hamrick. 2003. Outcrossing rate and inbreeding depression in the herbaceous autotetraploid, Campanula americana. Heredity 90:308-315. Ganders, F. R. 1979. The biology of heterostyly. New Zealand Journal of Botany 17:607-635. Gaude, T., A. Friry, P. Heizmann, C. Mariac, M. Rougier, I. Fobis, and C. Dumas. 1993. Expression of a self-incompatibility gene in a self-compatible line of Brassica oleracea. The Plant Cell 5:75-86. Gibbs, P. E. 2014. Late‐acting self‐incompatibility – the pariah breeding system in flowering plants. New Phytologist 203:717-734. Giles, B. E. 1990. The effects of variation in seed size on growth and reproduction in the wild barley Hordeum vulgare ssp. spontaneum. Heredity 64:239–250. Glover, B. 2014. Understanding Flowers and Flowering Second Edition. OUP Oxford. Goodwillie, C., and J. M. Ness. 2013. Interactions of hybridisation and mating systems: A case study in Leptosiphon (Polemoniaceae). American Journal of Botany 100:1002-1013. Goodwillie, C., R. D. Sargent, C. G. Eckert, E. Elle, M. A. Geber, M. O. Johnston, S. Kalisz, D. A. Moeller, R. H. Ree, and M. Vallejo‐Marin. 2010. Correlated evolution of mating system and floral display traits in flowering plants and its implications for the distribution of mating system variation. New Phytologist 185:311-321. Gormally, C. L., and J. L. Donovan. 2011. Genetic structure of a widely dispersed beach annual, Cakile edentula (Brassicaceae). American Journal of Botany 98:1657-1662. Grant, V. 1994. Modes and origins of mechanical and ethological isolation in angiosperms. Proceedings of the National Academy of Sciences 91:3-10. Grossenbacher, D. L., Y. Brandvain, J. R. Auld, M. Burd, P. O. Cheptou, J. K. Conner, A. G. Grant, S. M. Hovick, J. R. Pannell, and A. Pauw. 2017. Self‐compatibility is over‐represented on islands. New Phytologist 215:469-478. Hamilton, J. A., and J. M. Miller. 2016. Adaptive introgression as a resource for management and genetic conservation in a changing climate. Conservation Biology 30:33-41. Hamilton, W. D. 1963. The evolution of altruistic behavior. The American Naturalist 97:354-356. Hamilton, W. D. 1964. The genetical evolution of social behaviour. I. Journal of Theoretical Biology 7:1-16. Hao, Y.-Q., X.-F. Zhao, D.-Y. She, B. Xu, D.-Y. Zhang, and W.-J. Liao. 2012. The role of late-acting self-incompatibility and early-acting inbreeding depression in governing female fertility in monkshood, Aconitum kusnezoffii. PLoS One 7:e47034. Harrison, B., and L. Darby. 1955. Unilateral hybridisation. Nature 176:982. Harsh, P. B. 2015. Shedding light on kin recognition response in plants. New Phytologist 205:4-6.

122

Hatfield, T., and D. Schluter. 1999. Ecological speciation in sticklebacks: environment-dependent hybrid fitness. Evolution 53:866-873. Hayes, C. N., J. A. Winsor, and A. G. Stephenson. 2005. Environmental variation influences the magnitude of inbreeding depression in Cucurbita pepo ssp. texana (Cucurbitaceae). Journal of Evolutionary Biology 18:147-155. Hegland, S. J., and L. Boeke. 2006. Relationships between the density and diversity of floral resources and flower visitor activity in a temperate grassland community. Ecological Entomology 31:532-538. Hegland, S. J., and Ø. Totland. 2012. Interactions for pollinator visitation and their consequences for reproduction in a plant community. Acta Oecologica 43:95-103. Heslop-Harrison, J., and Y. Heslop-Harrison. 1970. Evaluation of pollen viability by enzymatically induced fluorescence; intracellular hydrolysis of fluorescein diacetate. Stain Technology 45:115-120. Heyligers, P. 1984. Beach invaders: sea rockets and beach daisies thrive. Pages 212-214 Australian Natural History. Hiscock, S. J., and A. M. Allen. 2008. Diverse cell signalling pathways regulate pollen-stigma interactions: the search for consensus. New Phytologist 179:286-317. Hiscock, S. J., and H. G. Dickinson. 1993. Unilateral incompatibility within the Brassicaceae: further evidence for the involvement of the self-incompatibility (S)-locus. Theoretical and Applied Genetics 86:744-753. Hiscock, S. J., and S. M. McInnis. 2003. The diversity of self‐incompatibility systems in flowering plants. Plant Biology 5:23-32. Hobbs, R. J., and L. F. Huenneke. 1992. Disturbance, diversity, and invasion: implications for conservation. Conservation Biology 6:324-337. Hull-Sanders, H. M., R. Clare, R. H. Johnson, and G. A. Meyer. 2007. Evaluation of the Evolution of Increased Competitive Ability (EICA) Hypothesis: Loss of Defense Against Generalist but not Specialist Herbivores. Journal of Chemical Ecology 33:781-799. Hyten, D. L., Q. Song, Y. Zhu, I.-Y. Choi, R. L. Nelson, J. M. Costa, J. E. Specht, R. C. Shoemaker, and P. B. Cregan. 2006. Impacts of genetic bottlenecks on soybean genome diversity. Proceedings of the National Academy of Sciences 103:16666-16671. Igic, B., and J. R. Kohn. 2001. Evolutionary relationships among self-incompatibility RNases. Proceedings of the National Academy of Sciences 98:13167-13171. Igic, B., R. Lande, and J. R. Kohn. 2008. Loss of self incompatibility and its evolutionary consequences. International Journal of Plant Sciences 169:93-104. Ignaciuk, R., and J. Lee. 1980. The germination of four annual strand‐line species. New Phytologist 84:581-591. Ishii, H. S., and L. D. Harder. 2012. Phenological associations of within‐ and among‐plant variation in gender with floral morphology and integration in protandrous Delphinium glaucum. Journal of Ecology 100:1029-1038. Iwamoto, A., R. Izumidate, and L. P. De Craene. 2015. Floral anatomy and vegetative development in Ceratophyllum demersum: A morphological picture of an "unsolved" plant. American Journal of Botany 102:1578-1589. Izawa, T., T. Oikawa, N. Sugiyama, T. Tanisaka, M. Yano, and K. Shimamoto. 2002. Phytochrome mediates the external light signal to repress FT orthologs in photoperiodic flowering of rice. Genes & Development 16:2006-2020. Izawa, T., T. Oikawa, S. Tokutomi, K. Okuno, and K. Shimamoto. 2000. Phytochromes confer the photoperiodic control of flowering in rice (a short‐day plant). The Plant Journal 22:391- 399. James, M. 2005. Hybridization as an invasion of the genome. Trends in Ecology & Evolution 20:229-237. Janick, J. 2008. Plant breeding reviews. John Wiley & Sons.

123

Johnston, J. A., M. L. Arnold, and L. A. Donovan. 2003. High hybrid fitness at seed and seedling life history stages in Louisiana irises. Journal of Ecology 91:438–446. Johnstone, R. A., and B. Kuijper. 2014. Kin Competition and the Evolution of Sex Differences in Development Time and Body Size. The American Naturalist 183:537-546. Joshi, J., and K. Vrieling. 2005. The enemy release and EICA hypothesis revisited: incorporating the fundamental difference between specialist and generalist herbivores. Ecology Letters 8:704–714. Kakeda, K., H. Tsukada, and Y. Kowyama. 2000. A self-compatible mutant S allele conferring a dominant negative effect on the functional S allele in Ipomoea trifida. Sexual Plant Reproduction 13:119–125. Karban, R., K. Shiojiri, S. Ishizaki, W. C. Wetzel, and R. Y. Evans. 2013. Kin recognition affects plant communication and defence. Proceedings of the Royal Society B: Biological Sciences 280:2012-3062. Karron, J. D., and R. J. Mitchell. 2012. Effects of floral display size on male and female reproductive success in Mimulus ringens. Annals of Botany 109:563-570. Kery, M., D. Matthies, and H. H. Spillmann. 2000. Reduced fecundity and offspring performance in small populations of the declining grassland plants Primula veris and Gentiana lutea. Journal of Ecology 88:17-30. Khan, M. L. 2004. Effects of seed mass on seedling success in Artocarpus heterophyllus L., a tropical tree species of north-east India. Acta Oecologica 25:103-110. Kissling, J., and S. C. Barrett. 2013. Variation and evolution of herkogamy in Exochaenium (Gentianaceae): implications for the evolution of distyly. Annals of Botany 112:95-102. Kitashiba, H., and J. B. Nasrallah. 2014. Self-incompatibility in Brassicaceae crops: lessons for interspecific incompatibility. Breeding Science 64:23-27. Klinger, T., and N. C. Ellstrand. 1994. Engineered genes in wild populations: fitness of weed-crop hybrids of Raphanus sativus. Ecological Applications 4:117-120. Knight, T. M., J. A. Steets, J. C. Vamosi, S. J. Mazer, M. Burd, D. R. Campbell, M. R. Dudash, M. O. Johnston, R. J. Mitchell, and T.-L. Ashman. 2005. Pollen limitation of plant reproduction: pattern and process. Annual Review of Ecology, Evolution, and Systematics 36:467-497. Kobayashi, K., M. Arai, A. Tanaka, S. Matsuyama, H. Honda, and R. Ohsawa. 2012. Variation in floral scent compounds recognized by honeybees in Brassicaceae crop species. Breeding science 62:293-302. Kolbe, J. J., R. E. Glor, L. R. Schettino, A. C. Lara, A. Larson, and J. B. Losos. 2007. Multiple sources, admixture, and genetic variation in introduced Anolis lizard populations. Conservation Biology 21:1612-1625. Kondrashov, A. 1988. Deleterious mutations and the evolution of sexual reproduction. Nature 336:435-440. Korbecka, G. 2004. Genetic structure and post-pollination selection in biennial plants. Doctoral thesis. Leiden University. Kowyama, Y., H. Takahasi, K. Muraoka, T. Tani, K. Hara, and I. Shiotani. 1994. Number, frequency and dominance relationships of S-alleles in diploid Ipomoea trifida. Heredity 73:275-283. Kriticos, D., R. Sutherst, J. Brown, S. Adkins, and G. Maywald. 2003. Climate change and the potential distribution of an invasive alien plant: Acacia nilotica ssp. indica in Australia. Journal of Applied Ecology 40:111-124. Krizek, B. A., and J. C. Fletcher. 2005. Molecular mechanisms of flower development: an armchair guide. Nature Reviews Genetics 6:688-698. Kulbaba, M. W., and A. C. Worley. 2012. Selection on floral design in Polemonium brandegeei (Polemoniaceae): female and male fitness under hawkmoth pollination. Evolution 66:1344-1359. Kunin, W. E. 1993. Sex and the single mustard: population density and pollinator behavior effects on seed-set. Ecology 74:2145-2160.

124

Kyle, W. C., M. R. Fakhria, A. S. Connie, J. Tracey, and L. M. Sara. 2015. Bidirectional but asymmetrical sexual hybridisation between Brassica carinata and Sinapis arvensis (Brassicaceae). Journal of Plant Research 128:469-480. Lankinen, A., W. S. Armbruster, and L. Antonsen. 2007. Delayed stigma receptivity in Collinsia heterophylla (Plantaginaceae): genetic variation and adaptive significance in relation to pollen competition, delayed self-pollination, and mating-system evolution. American Journal of Botany 94:1183-1192. Lankinen, Å., and S. Kiboi. 2007. Pollen donor identity affects timing of stigma receptivity in Collinsia heterophylla (Plantaginaceae): a sexual conflict during pollen competition? The American Naturalist 170:854-863. Lankinen, Å., and J. A. Madjidian. 2011. Enhancing pollen competition by delaying stigma receptivity: pollen deposition schedules affect siring ability, paternal diversity, and seed production in Collinsia heterophylla (Plantaginaceae). American Journal of Botany 98:1191-1200. Lankinen, Å., H. G. Smith, S. Andersson, and J. A. Madjidian. 2015. Selection on pollen and pistil traits during pollen competition is affected by both sexual conflict and mixed mating in a self-compatible herb. American Journal of Botany 103:541-552. Lawrence, M. J. 2000. Population genetics of the homomorphic self-incompatibility polymorphisms in flowering plants. Annals of Botany 85:221-226. Levin, D. A., C. D. Kelley, and S. Sarkar. 2009. Enhancement of Allee effects in plants due to self‐ incompatibility alleles. Journal of Ecology 97:518-527. Levine, J. M., and C. M. D'Antonio. 1999. Elton revisited: a review of evidence linking diversity and invasibility. Oikos 87:15-26. Lewis, D., and L. K. Crowe. 1958. Unilateral interspecific incompatibility in flowering plants. Heredity 12:233-256. Li, N., W. Peng, J. Shi, X. Wang, G. Liu, and H. Wang. 2015. The natural variation of seed weight is mainly controlled by maternal genotype in rapeseed (Brassica napus L.). PLoS One 10:e0125360. Li, X. 2011a. Arabidopsis Pollen Tube Germination. http://www.bio-protocol.org/e73. Li, X. 2011b. Pollen fertility/viability assay using FDA staining. Bio-protocol 1:e75. Lindtke, D., and C. A. Buerkle. 2015. The genetic architecture of hybrid incompatibilities and their effect on barriers to introgression in secondary contact. Evolution 69:1987-2004. Lopez, G. A., B. M. Potts, and P. A. Tilyard. 2000. F1 hybrid inviability in Eucalyptus: the case of E. ovata× E. globulus. Heredity 85:242–250. Lowry, D. B., J. L. Modliszewski, K. M. Wright, C. A. Wu, and J. H. Willis. 2008. The strength and genetic basis of reproductive isolating barriers in flowering plants. Philosophical Transactions of the Royal Society B: Biological Sciences 363:3009-3021. Ludwig, S., A. Robertson, T. C. Rich, M. Djordjević, R. Cerović, L. Houston, S. A. Harris, and S. J. Hiscock. 2013. Breeding systems, hybridisation and continuing evolution in Avon Gorge Sorbus. Annals of Botany 111:563-575. Luo, Y., and A. Widmer. 2013. Herkogamy and its effects on mating patterns in Arabidopsis thaliana. PLoS One 8:e57902. Luu, D.-T., X. Qin, D. Morse, and M. Cappadocia. 2000. S-RNase uptake by compatible pollen tubes in gametophytic self-incompatibility. Nature 407:649-651. Lysak, M. A., M. A. Koch, A. Pecinka, and I. Schubert. 2005. Chromosome triplication found across the tribe Brassiceae. Genome Research 15:516-525. Ma, H. 2005. Molecular genetic analyses of microsporogenesis and microgametogenesis in flowering plants. Annual Review of Plant Biology 56:393-434. Mable, B., M. H. Schierup, and D. Charlesworth. 2003. Estimating the number, frequency, and dominance of S-alleles in a natural population of Arabidopsis lyrata (Brassicaceae) with sporophytic control of self-incompatibility. Heredity 90:422-431.

125

Mallet, J. 2005. Hybridization as an invasion of the genome. Trends in Ecology & Evolution 20:229- 237. Mandel, M., and M. F. Yanofsky. 1998. The Arabidopsis AGL9 MADS box gene is expressed in young flower primordia. Sexual Plant Reproduction 11:22-28. Marshall, D. L., and P. K. Diggle. 2001. Mechanisms of differential pollen donor performance in wild radish, Raphanus sativus (Brassicaceae). American Journal of Botany 88:242-257. Martin, F. W. 1961. The inheritance of self-incompatibility in hybrids of Lycopersicon esculentum Mill. x L. chilense Dun. Genetics 46:1443–1454. Masel, J. 2011. Genetic drift. Current Biology 21:R837-R838. Mather, K. 1943. Specific differences in Petunia. I. Incompatibility. Journal of Genetics 45:215- 235. Matioli, S. R., and A. R. Templeton. 1999. Coadapted gene complexes for morphological traits in Drosophila mercatorum. Two-loci interactions. Heredity 83:54-61. Maun, M. A. 1994. Adaptations enhancing survival and establishment of seedlings on coastal dune systems. Vegetatio 111:59-70. Maun, M. A., R. S. Boyd, and L. Olson. 1990. The Biological Flora of Coastal Dunes and Wetlands. 1." Cakile Edentula"(Bigel.) Hook. Journal of Coastal Research 6:137-156. Mayr, E. 1942. Systematics and the origin of species, from the viewpoint of a zoologist. Harvard University Press. Mayr, E. 1970. Populations, species, and evolution: an abridgment of animal species and evolution. Harvard University Press. Mazer, S. J., A. Moghaddasi, A. K. Bello, and A. A. Hove. 2016. Winning in style: Longer styles receive more pollen, but style length does not affect pollen attrition in wild Clarkia populations. American Journal of Botany 103:408-422. McCallum, B., and Chang. 2016. Pollen competition in style: Effects of pollen size on siring success in the hermaphroditic common morning glory, Ipomoea purpurea. American Journal of Botany 103:460-470. McGraw, J. B., and H. Caswell. 1996. Estimation of individual fitness from life-history data. The American Naturalist:47-64. Medrano, M., C. M. Herrera, and S. C. H. Barrett. 2005. Herkogamy and mating patterns in the self-compatible daffodil Narcissus longispathus. Annals of Botany 95:1105–1111. Meehan, T. 1892. Contributions to the life histories of plants, No. 8. Proceedings of the Academy of Natural Sciences of Philadelphia 44:366-386. Mena, M., B. A. Ambrose, R. B. Meeley, and S. P. Briggs. 1996. Diversification of C-function activity in maize flower development. Science 274:1537. Mesgaran, M. B., M. A. Lewis, P. K. Ades, K. Donohue, S. Ohadi, C. Li, and R. D. Cousens. 2016. Hybridization can facilitate species invasions, even without enhancing local adaptation. Proceedings of the National Academy of Sciences 113:10210-10214. Mhemmed, G., H. Kamel, and A. Chedly. 2008. Understanding the population genetic structure of coastal species (Cakile maritima): seed dispersal and the role of sea currents in determining population structure. Genetics Research 90:167-178. Mhemmed, G., H. Kamel, and A. Chedly. 2012. How selection fashions morphological variation in Cakile maritima: A comparative analysis of population structure using random amplified polymorphic DNA and quantitative traits. Journal of Systematics and Evolution 50:109– 118. Mills, K. 2013. Replacement of Cakile edentula with Cakile maritima in New South Wales and on Lord Howe Island. Cunninghamia 13:291-294. Mitchell, C. E., and A. G. Power. 2003. Release of invasive plants from fungal and viral pathogens. Nature 421:625-627.

126

Morrison, S. L., and J. Molofsky. 2000. Environmental and genetic effects on the early survival and growth of the invasive grass Phalaris arundinacea. Canadian Journal of Botany 77:1447- 1453. Murphy, G. P., and S. A. Dudley. 2009. Kin recognition: Competition and cooperation in Impatiens (Balsaminaceae). American Journal of Botany 96:1990-1996. Murren, C., and M. Dudash. 2012. Variation in inbreeding depression and plasticity across native and non-native field environments. Annals of Botany 109:621-632. Nasrallah, J. B. 2011. Self-incompatibility in the Brassicaceae. Pages 389-411 Genetics and Genomics of the Brassicaceae. Springer. Nasrallah, J. B., P. Liu, S. Sherman-Broyles, R. Schmidt, and M. E. Nasrallah. 2007. Epigenetic mechanisms for breakdown of self-incompatibility in interspecific hybrids. Genetics 175:1965-1973. Nasrallah, J. B., T. Nishio, and M. E. Nasrallah. 1991. The self-incompatibility genes of Brassica: expression and use in genetic ablation of floral tissues. Annual Review of Plant Biology 42:393-422. Nasrallah, M. E., M. K. Kandasamy, and J. B. Nasrallah. 1992. A genetically defined trans‐acting locus regulates S‐locus function in Brassica. The Plant Journal 2:497-506. Nasrallah, M. E., P. Liu, S. Sherman-Broyles, N. A. Boggs, and J. B. Nasrallah. 2004. Natural variation in expression of self-incompatibility in Arabidopsis thaliana: implications for the evolution of selfing. Proceedings of the National Academy of Sciences 101:16070-16074. Nores, M. J., H. A. López, and A. M. Anton. 2015. Contrasting models of the female reproductive tract in four o'clocks (Nyctaginaceae). American Journal of Botany 102:1026-3910. Ohadi, S., P. K. Ades, R. D. Cousens, and R. Ford. 2012. Genetic variation in invasive populations of sea rocket (Cakile maritima) in southern coastal habitats of Australia. Pages 373-374 in Proceedings of the 18th Australasian Weeds Conference. Ohadi, S., P. K. Ades, R. Ford, A. E. Strand, J. Tibbits, M. B. Mesgaran, and R. D. Cousens. 2015. Genetic structure along the strandline: Unravelling invasion history in a one dimensional system. Journal of Biogeography 43:451–460. Ohta, A. T. 1980. Coadaptive gene complexes in incipient species of Hawaiian Drosophila. The American Naturalist 115:121-132. Oliver, E. J., P. H. Thrall, J. Burdon, and J. Ash. 2001. Vertical disease transmission in the Cakile- Alternaria host-pathogen interaction. Australian Journal of Botany 49:561-569. Oostermeijer, J., M. Van Eijck, and J. Den Nijs. 1994. Offspring fitness in relation to population size and genetic variation in the rare perennial plant species Gentiana pneumonanthe (Gentianaceae). Oecologia 97:289-296. Paini, D. R., A. W. Sheppard, D. C. Cook, P. J. De Barro, S. P. Worner, and M. B. Thomas. 2016. Global threat to agriculture from invasive species. Proceedings of the National Academy of Sciences 113:7575-7579. Payne, A. 1980. The ecology and population dynamics of Cakile edentula var. lacustris on Lake Huron beaches. University of Western Ontario, London Ontario. Payne, A., and M. Maun. 1984. Reproduction and survivorship of Cakile edentula var. lacustris along the Lake Huron shoreline. American Midland Naturalist 111:86-95. Payseur, B. A. 2010. Using differential introgression in hybrid zones to identify genomic regions involved in speciation. Molecular Ecology Resources 10:806-820. Pedersen, B. P., P. Neve, C. Andreasen, and S. B. Powles. 2007. Ecological fitness of a glyphosate- resistant Lolium rigidum population: Growth and seed production along a competition gradient. Basic and Applied Ecology 8:258-268. Pejchar, L., and H. A. Mooney. 2009. Invasive species, ecosystem services and human well-being. Trends in Ecology & Evolution 24:497-504.

127

Pélabon, C., L. Hennet, G. H. Bolstad, E. Albertsen, Ø. H. Opedal, R. K. Ekrem, and W. S. Armbruster. 2016. Does stronger pollen competition improve offspring fitness when pollen load does not vary? American Journal of Botany 103:522-531. Pelaz, S., G. S. Ditta, E. Baumann, E. Wisman, and M. F. Yanofsky. 2000. B and C floral organ identity functions require SEPALLATA MADS-box genes. Nature 405:200-203. Petanidou, T., R. C. Godfree, D. S. Song, A. Kantsa, Y. L. Dupont, and N. M. Waser. 2012. Self- compatibility and plant invasiveness: comparing species in native and invasive ranges. Perspectives in Plant Ecology, Evolution and Systematics 14:3-12. Peters, J. 2000. Tetrazolium testing handbook. Association of Official Seed Analysts, http://www.aosaseed.com/tz_handbook_updates. Plitmann, U. 1993. Pollen-tube attrition as related to breeding systems in Brassicaceae. Plant Systematics and Evolution 188:65-72. Preston, R. E. 1991. The intrafloral phenology of Streptanthus tortuosus (Brassicaceae). American Journal of Botany 78:1044-1053. Provine, W. B. 2004. Ernst Mayr Genetics and Speciation. Genetics 167:1041-1046. Raunkiaer, C. 1934. The life forms of plants and statistical plant geography; being the collected papers of C. Raunkiaer. Richards, A. J. 1997. Plant breeding systems. Garland Science. Ridley, C. E., and N. C. Ellstrand. 2009. Evolution of enhanced reproduction in the hybrid-derived invasive, California wild radish (Raphanus sativus). Biological invasions 11:2251-2264. Rieseberg, L. H. 2006. Hybrid speciation in wild sunflowers. Annals of the Missouri Botanical Garden 93:34-48. Rieseberg, L. H., M. A. Archer, and R. K. Wayne. 1999. Transgressive segregation, adaptation and speciation. Heredity 83:363–372. Rieseberg, L. H., A. M. Desrochers, and S. J. Youn. 1995. Interspecific pollen competition as a reproductive barrier between sympatric species of Helianthus (Asteraceae). American Journal of Botany 82:515-519. Rieseberg, L. H., and J. F. Wendel. 1993. Introgression and its consequences in plants. Pages 70- 109 in R. G. Harrison, editor. Hybrid zones and the evolutionary process. Oxford University Press. Roccaforte, K., S. E. Russo, and D. Pilson. 2015. Hybridization and reproductive isolation between diploid Erythronium mesochoreum and its tetraploid congener E. albidum (Liliaceae). Evolution 69:1375–1389. Rodger, J. G., M. van Kleunen, and S. D. Johnson. 2013. Pollinators, mates and Allee effects: the importance of self-pollination for fecundity in an invasive lily. Functional Ecology 27:1023-1033. Rodman, J. E. 1974. Systematics and evolution of the genus Cakile (Cruciferae). Harvard University. Rodman, J. E. 1976. Differentiation and migration of Cakile (Cruciferae): seed glucosinolate evidence. Systematic Botany 1:137-148. Rodman, J. E. 1980. Population variation and hybridisation in sea-rockets (Cakile, Cruciferae): seed glucosinolate characters. American Journal of Botany 67:1145-1159. Rodman, J. E. 1986. Introduction, establishment and replacement of sea-rockets (Cakile, Cruciferae) in Australia. Journal of Biogeography 13:159-171. Rodriguez-Riano, T., and A. Dafni. 2000. A new procedure to assess pollen viability. Sexual Plant Reproduction 12:241–244. Runquist, B. R. D., E. Chu, J. L. Iverson, and J. C. Kopp. 2014. Rapid evolution of reproductive isolation between incipient outcrossing and selfing Clarkia species. Evolution 68:2885– 2900. Sargent, R. D., C. Goodwillie, S. Kalisz, and R. H. Ree. 2007. Phylogenetic evidence for a flower size and number trade-off. American Journal of Botany 94:2059-2062.

128

Schierenbeck, K. A., and N. C. Ellstrand. 2009. Hybridization and the evolution of invasiveness in plants and other organisms. Biological invasions 11:1093. Schierup, M. H., X. Vekemans, and F. B. Christiansen. 1997. Evolutionary dynamics of sporophytic self-incompatibility alleles in plants. Genetics 147:835-846. Scott, R. J., M. Spielman, J. Bailey, and H. G. Dickinson. 1998. Parent-of-origin effects on seed development in Arabidopsis thaliana. Development 125:3329-3341. Seavey, S. R., and K. S. Bawa. 1986. Late-acting self-incompatibility in angiosperms. The Botanical Review 52:195-219. Secord, J. A. 1981. Nature's fancy: Charles Darwin and the breeding of pigeons. Isis 72:163-186. Seifan, M., E. M. Hoch, S. Hanoteaux, and K. Tielbörger. 2014. The outcome of shared pollination services is affected by the density and spatial pattern of an attractive neighbour. Journal of Ecology 102:953–962. Servedio, M. R., J. Hermisson, and G. S. v. Doorn. 2013. Hybridization may rarely promote speciation. Journal of Evolutionary Biology 26:282-285. Shaw, R. G., C. J. Geyer, S. Wagenius, H. H. Hangelbroek, and J. R. Etterson. 2008. Unifying life- history analyses for inference of fitness and population growth. The American Naturalist 172:E35-E47. Sicard, A., and M. Lenhard. 2011. The selfing syndrome: a model for studying the genetic and evolutionary basis of morphological adaptation in plants. Annals of Botany 107:1433- 1443. Silvertown, J., and J. Lovett-Doust. 1993. Introduction to Plant Population Biology. Wiley. Sloop, C. M., D. R. Ayres, and D. R. Strong. 2009. The rapid evolution of self-fertility in Spartina hybrids (Spartina alterniflora× foliosa) invading San Francisco Bay, CA. Biological Invasions 11:1131–1144. Smyth, D. R., J. L. Bowman, and E. M. Meyerowitz. 1990. Early flower development in Arabidopsis. The Plant Cell 2:755-767. Snow, A. A., K. L. Uthus, and T. M. Culley. 2001. Fitness of hybrids between weedy and cultivated radish: implications for weed evolution. Ecological Applications 11:934-943. Sork, V. L., and D. W. Schemske. 1992. Fitness consequences of mixed-donor pollen loads in the annual legume Chamaecrista fasciculata. American Journal of Botany 79:508-515. Steinbachs, J., and K. Holsinger. 2002. S-RNase–mediated gametophytic self-incompatibility is ancestral in Eudicots. Molecular Biology and Evolution 19:825-829. Stephens, P. A., W. J. Sutherland, and R. P. Freckleton. 1999. What is the Allee effect? Oikos 87:185-190. Stevens, J., and Q. Kay. 1989. The number, dominance relationships and frequencies of self- incompatibility alleles in a natural population of Sinapis arvensis L in South-Wales. Heredity 62:199-205. Stuber, C. W., S. E. Lincoln, D. Wolff, T. Helentjaris, and E. Lander. 1992. Identification of genetic factors contributing to heterosis in a hybrid from two elite maize inbred lines using molecular markers. Genetics 132:823-839. Summers, H. E., S. M. Hartwick, and R. A. Raguso. 2015. Geographic variation in floral allometry suggests repeated transitions between selfing and outcrossing in a mixed mating plant. American Journal of Botany 102:745-757. Swanson, R. J., A. T. Hammond, A. L. Carlson, H. Gong, and T. K. Donovan. 2016. Pollen performance traits reveal prezygotic nonrandom mating and interference competition in Arabidopsis thaliana. American Journal of Botany 103:498-513. Takebayashi, N., D. E. Wolf, and L. F. Delph. 2006. Effect of variation in herkogamy on outcrossing within a population of Gilia achilleifolia. Heredity 96:159-165. Tanesaka, E., E. Umeda, M. Yamamoto, K. Masuda, K. Yamada, and M. Yoshida. 2012. Inheritance mode of seed dormancy in the hybrid progeny of sesame, Sesamum indicum, and its wild relative, Sesamum mulayanum Nair. Weed Biology and Management 12:91-97.

129

Tedder, A., S. Carleial, M. Gołębiewska, C. Kappel, K. K. Shimizu, and M. Stift. 2015. Evolution of the Selfing Syndrome in Arabis alpina (Brassicaceae). PLoS One 10:e0126618. Thrall, P. H., A. G. Young, and J. J. Burdon. 2000. An analysis of mating structure in populations of the annual sea rocket, Cakile maritima (Brassicaceae). Australian Journal of Botany 48:731-738 Tian, B., G. WenLing, S. Jie, and X. WeiJia. 2014. Flowering characteristics and breeding system of Rhododendron siderophyllum. Journal of West China Forestry Science 43:47-53. Tiffin, P., S. Olson, and L. C. Moyle. 2001. Asymmetrical crossing barriers in angiosperms. Proceedings of the Royal Society of London. Series B: Biological Sciences 268:861-867. Travers, S. E., and S. J. Mazer. 2000. The absence of cryptic self-incompatibility in Clarkia unguiculata (Onagraceae). American Journal of Botany 87:191-196. Ungerer, M. C., S. J. Baird, J. Pan, and L. H. Rieseberg. 1998. Rapid hybrid speciation in wild sunflowers. Proceedings of the National Academy of Sciences 95:11757-11762. Vallejo‐Marín, M., and S. J. Hiscock. 2016. Hybridization and hybrid speciation under global change. New Phytologist 211:1170–1187. Vassiliadis, C., J. Lepart, P. Saumitou-Laprade, and P. Vernet. 2000. Self-incompatibility and male fertilization success in Phillyrea angustifolia (Oleaceae). International Journal of Plant Sciences 161:393-402. Veen, T., T. Borge, S. C. Griffith, G. P. Saetre, S. Bures, L. Gustafsson, and B. C. Sheldon. 2001. Hybridization and adaptive mate choice in flycatchers. Nature 411:45-50. Vekemans, X., C. Poux, P. M. Goubet, and V. Castric. 2014. The evolution of selfing from outcrossing ancestors in Brassicaceae: what have we learned from variation at the S‐ locus? Journal of Evolutionary Biology 27:1372-1385. Vogler, Filmore, and Stephenson. 1999. Inbreeding depression in Campanula rapunculoides L. I. A comparison of inbreeding depression in plants derived from strong and weak self- incompatibility phenotypes. Journal of Evolutionary Biology 12:483-494. Wagenius, S., E. Lonsdorf, and C. Neuhauser. 2007. Patch aging and the S‐Allee effect: breeding system effects on the demographic response of plants to habitat fragmentation. The American Naturalist 169:383-397. Wang, J., Y. A. El-Kassaby, and K. Ritland. 2012. Estimating selfing rates from reconstructed pedigrees using multilocus genotype data. Molecular Ecology 21:100-116. Ward, M., and S. D. Johnson. 2013. Generalised pollination systems for three invasive milkweeds in Australia. Plant Biology 15:566-572. Ward, M., S. D. Johnson, and M. P. Zalucki. 2013. When bigger is not better: intraspecific competition for pollination increases with population size in invasive milkweeds. Oecologia 171:883-891.Weigel, D., J. Alvarez, D. R. Smyth, M. F. Yanofsky, and E. M. Meyerowitz. 1992. LEAFY controls floral meristem identity in Arabidopsis. Cell 69:843- 859. Weigel, D., J. Alvarez, D. R. Smyth, M. F. Yanofsky, and E. M. Meyerowitz. 1992. LEAFY controls floral meristem identity in Arabidopsis. Cell 69:843-859. West, S. A., M. G. Murray, C. A. Machado, A. S. Griffin, and E. A. Herre. 2001. Testing Hamilton's rule with competition between relatives. Nature 409:510-513. West, S. A., I. Pen, and A. S. Griffin. 2002. Cooperation and competition between relatives. Science 296:72-75. Willi, Y. 2009. Evolution towards self-compatibility when mates are limited. Journal of Evolutionary Biology 22:1967-1973. Willi, Y., and M. Fischer. 2005. Genetic rescue in interconnected populations of small and large size of the self-incompatible Ranunculus reptans. Heredity 95:437-443. Willmer, P. 2011. Pollination and floral ecology. Princeton University Press. Willson, M. F. 1983. Plant reproductive ecology. John Wiley & Sons. Wright, S. 1931. Evolution in Mendelian populations. Genetics 16:97-159.

130

Wu, Z., Z. Ding, D. Yu, and X. Xu. 2015. Influence of niche similarity on hybridisation between Myriophyllum sibiricum and M. spicatum. Journal of Evolutionary Biology 28:1465-1475. Xiangwen, F., C. T. Neil, Y. Guijun, L. Fengmin, and H. S. Kadambot. 2010. Flower numbers, pod production, pollen viability, and pistil function are reduced and flower and pod abortion increased in chickpea (Cicer arietinum L.) under terminal drought. Journal of Experimental Botany 61:335-345. Yacine, A., and F. Bouras. 1997. Self- and cross-pollination effects on pollen tube growth and seed set in holm oak Quercus ilex L (Fagaceae). Annals of Forest Science 54:447-462. Yakimowski, S. B., and L. H. Rieseberg. 2014. The role of homoploid hybridization in evolution: a century of studies synthesizing genetics and ecology. American Journal of Botany 101:1247-1258. Yanofsky, M. F. 1995. Floral meristems to floral organs: genes controlling early events in Arabidopsis flower development. Annual Review of Plant Biology 46:167-188. Yasuda, S., Y. Wada, T. Kakizaki, Y. Tarutani, E. Miura-Uno, K. Murase, S. Fujii, T. Hioki, T. Shimoda, and Y. Takada. 2017. A complex dominance hierarchy is controlled by polymorphism of small RNAs and their targets. Nature plants 3:16206. Yatabe, Y., N. C. Kane, C. Scotti-Saintagne, and L. H. Rieseberg. 2007. Rampant gene exchange across a strong reproductive barrier between the annual sunflowers, Helianthus annuus and H. petiolaris. Genetics 175:1883-1893. Yukio, K., and B. Sang. 2014. Interspecific and intergeneric hybridisation and chromosomal engineering of Brassicaceae crops. Breeding Science 64:14. Zeng, F., and B. Cheng. 2014. Self-(in) compatibility inheritance and allele-specific marker development in yellow mustard (Sinapis alba). Molecular Breeding 33:187-196. Zhang, J. 1993. Seed dimorphism in relation to germination and growth of Cakile edentula. Canadian Journal of Botany 71:1231-1235. Zhang, J. 1996. Seed mass effects across environments in an annual dune plant. Annals of Botany 77:555-563. Zhang, J., and M. A. Maun. 1992. Effects of burial in sand on the growth and reproduction of Cakile edentula. Ecography 15:296-302. Zhou, S., G. Yuan, P. Xu, and H. Gong. 2014. Study on lily introgression breeding using allotriploids as maternal parents in interploid hybridisations. Breeding Science 64:97–102.

131

Minerva Access is the Institutional Repository of The University of Melbourne

Author/s: LI, CHENGJUN

Title: Understanding the breeding systems of Cakile edentula, Cakile maritima (Brassicaceae) and their hybrids

Date: 2017

Persistent Link: http://hdl.handle.net/11343/207946

File Description: Understanding the breeding systems of Cakile edentula, Cakile maritima (Brassicaceae) and their hybrids

Terms and Conditions: Terms and Conditions: Copyright in works deposited in Minerva Access is retained by the copyright owner. The work may not be altered without permission from the copyright owner. Readers may only download, print and save electronic copies of whole works for their own personal non-commercial use. Any use that exceeds these limits requires permission from the copyright owner. Attribution is essential when quoting or paraphrasing from these works.