Interactions Between Floral Mutualists and Antagonists, And

University of Massachusetts Amherst ScholarWorks@UMass Amherst

Open Access Dissertations

2-2013 Interactions between floral mutualists and antagonists, and consequences for plant reproduction Nicole Leland Soper Gorden University of Massachusetts Amherst, [email protected]

Follow this and additional works at: https://scholarworks.umass.edu/open_access_dissertations Part of the Plant Biology Commons

Recommended Citation Soper Gorden, Nicole Leland, "Interactions between floral mutualists and antagonists, and consequences for plant reproduction" (2013). Open Access Dissertations. 708. https://doi.org/10.7275/dkm7-5k19 https://scholarworks.umass.edu/open_access_dissertations/708

This Open Access Dissertation is brought to you for free and open access by ScholarWorks@UMass Amherst. It has been accepted for inclusion in Open Access Dissertations by an authorized administrator of ScholarWorks@UMass Amherst. For more information, please contact [email protected].

Interactions between floral mutualists and antagonists, and consequences for plant reproduction

A Dissertation Presented

NICOLE L. SOPER GORDEN

Submitted to the Graduate School of the University of Massachusetts Amherst in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY

February 2013

Plant Biology

Interactions between floral mutualists and antagonists, and consequences for plant reproduction

A Dissertation Presented

NICOLE L. SOPER GORDEN

Approved as to style and content by:

______Lynn S. Adler, Chair

______Peter Alpert, Member

______Martha Hoopes, Member

______Anne Averill, Member

______Elsbeth Walker, Program Head Plant Biology

DEDICATION

In memory of my mother

ACKNOWLEDGEMENTS

I would like to thank my committee members and the members of the Adler lab who offered their time, ideas, and editing. Nancy Hanson and Hampshire Farm assisted with field space. Funding was provided by University of Massachusetts Natural History

Collections Group Jane H. Bemis Endowment Scholarship, the University of

Massachusetts Plant Biology Program, the University of Massachusetts Gilgut

Fellowship in Plant Biology, and National Science Foundation Doctoral Dissertation

Improvement Grant DEB-1011236.

Additionally, I have had numerous excellent undergraduate field and lab assistants, including Jonathan Alfredo López-Colón, Justin Van Goor, Lauren Bishop,

Karly Henry, Winston Anthony, Matthew Boyer, Isaac Han, Dennis Dietz, Kaycee

Farland, Tyler Davidson, Nadya Shlykova, Irina Showalter, Giavanni Garris, Kevin

Moran, Maria Servidone, Alexa Pfeiffer, Chelsea Merritt, Anthony Tse, and Thomas

Hernon.

Finally, I would like to offer a special thanks to my family and friends who consistently supported and believed in me throughout my dissertation.

ABSTRACT

INTERACTIONS BETWEEN FLORAL MUTUALISTS AND ANTAGONISTS,

AND CONSEQUENCES FOR PLANT REPRODUCTION

FEBRUARY 2013

NICOLE L. SOPER GORDEN, B.A, GUSTAVUS ADOLPHUS COLLEGE

Ph.D., UNIVERSITY OF MASSACHUSETTS AMHERST

Directed by: Professor Lynn S. Adler

While pollinators and leaf herbivores have been a focus of research for decades, floral antagonists have been studied significantly less. Since floral antagonists can be as common as leaf herbivores and have strong impacts on plant reproduction, it is important to understand the role of floral antagonists in the ecology and evolution of flowers. I conducted four experiments to better understand the relationship between plants, floral traits, floral antagonists, and other plant-insect interactions. First, I manipulated resources (light and soil nutrients) that are known to have impacts on plants and floral traits to test how they affect floral antagonists and other plant-insect interactions.

Plentiful resources increased the proportion of floral antagonists to visit flowers, but also increase tolerance of floral antagonists. Second, I manipulated flower bud gallers, a species-specific floral herbivore that destroys flowers, to test how it affected other plant- insect interactions, floral traits, and plant reproduction. Plants with flower bud gallers tended to have more pollinator visits, but this effect is due to a shared preference by gallers and pollinators for similar plants. Third, I manipulated florivory to examine how it affects subsequent plant-arthropod interactions, floral traits, and plant reproduction.

Florivory had systemic effects on other plant-insect interactions, including leaf

herbivores, and shifted the plant mating system towards more selfing. Additionally, I tested how several floral antagonists respond to floral attractive and defense traits to understand which floral traits are important in mediating antagonisms. Finally, I manipulated florivory, pollination, and nectar robbing to test for effects of multiple floral interactions on subsequent plant-insect interactions, floral traits, and plant reproduction.

There were significant many-way interactions between the three treatments on subsequent plant-insect interactions and reproduction, indicating that the effect of one interaction depends on what other interactions are present. Understanding the role that floral antagonists play in plant ecology can help scientists determine which interactions are most important, and may help determine why some floral traits exist in their current state. Together, this work represents some of the most comprehensive research on the community consequences of floral antagonists, as well as the interplay between floral traits and floral interactions.

vii

TABLE OF CONTENTS

Page

ACKNOWLEDGEMENTS ...... v

ABSTRACT ...... vi

LIST OF TABLES ...... x

LIST OF FIGURES ...... xi

INTRODUCTION ...... 1

CHAPTER

1. ABIOTIC CONDITIONS AFFECT ANTAGONISTS AND MUTUALISTS IN IMPATIENS CAPENSIS ...... 11

Abstract ...... 11 Introduction ...... 12 Methods...... 15 Results ...... 24 Discussion ...... 26 Tables ...... 35 Figures...... 38

2. INTERACTIONS BETWEEN FLOWER BUD GALLERS, POLLINATION, AND PLANT REPRODUCTION ...... 42

Abstract ...... 42 Introduction ...... 43 Methods...... 47 Results ...... 54 Discussion ...... 57 Tables ...... 64 Figures...... 67

3. FLORIVORY ALTERS PLANT-INSECT INTERACTIONS AND DECREASES PLANT REPRODUCTION ...... 71

Abstract ...... 71 Introduction ...... 72 Methods...... 76 Results ...... 84 Discussion ...... 86

viii

Tables ...... 96 Figures...... 98

4. CONSEQUENCES OF MULTIPLE FLORAL INTERACTIONS FOR SUBSEQUENT INTERACTIONS AND PLANT REPRODUCTION ...... 101

Abstract ...... 101 Introduction ...... 102 Methods...... 105 Results ...... 112 Discussion ...... 114 Tables ...... 124 Figures...... 126

BIBLIOGRAPHY ...... 132

LIST OF TABLES

Table Page

1.1. Loadings from significant principle components from principle component analysis on logical subsets of plant and flower data, including total variance explained by each PC...... 35

1.2. Results for MANCOVAs to test the effects of fertilizer addition, shade versus sun, and soil moisture on plant growth, pollination, nectar thieves, herbivory, and plant reproduction...... 36

1.3. Summary of significant results from individual ANOVAs following significant MANCOVAs (Table 2)...... 37

2.1. MANOVA (bumblebee and honey bee visits, herbivory) and ANOVA (bumblebee and honeybee time per probe) results from the manual manipulation of gall presence...... 64

2.2. MANOVA results testing the effects of gall manipulation on Impatiens capensis...... 65

2.3. Results from ANOVAs testing the gall manipulation treatments on plant traits, subsequent plant-insect interactions, and plant reproduction...... 66

3.1. GLIM regressions between early season Impatiens capensis floral traits and insect behavior...... 96

3.2. GLIM regression results between early season floral traits on late season floral traits in Impatiens capensis...... 97

4.1. Effects of supplemental florivory, nectar robbing, and pollination on Impatiens capensis growth, floral traits, plant reproduction, and subsequent plant- insect interactions...... 124

4.2. Standardized canonical coefficients of response variables from significant three-way MANOVAs testing the effects of manipulations of floral interactions (florivory, nectar robbing, and pollination) on plant reproduction in Impatiens capensis (see Table 4.1)...... 125

LIST OF FIGURES

Figure Page

1.1. Effect of fertilizer on plant growth and floral rewards in Impatiens capensis...... 38

1.2. Effects of fertilizer on Impatiens capensis floral visitors during 15 minute observations...... 39

1.3. Effects of fertilizer on herbivory in Impatiens capensis...... 40

1.4. Effects of fertilizer on Impatiens capensis reproduction...... 41

2.1. Pollination on Impatiens capensis plants comparing initial presence of galls and manual gall manipulations...... 67

2.2. Effects of gall manipulation treatments on the number of galls and size of galls on Impatiens capensis plants...... 68

2.3. Effects of Schizomyia impatientis galler manipulation on measures of Impatiens capensis flower attractiveness...... 69

2.4. Effect of Schizomyia impatientis galler manipulations on subsequent floral antagonists...... 70

3.1. Flower damage treatments, showing the amount of petal tissue removed from the lips and throat of flowers in each treatment...... 98

3.2. Effect of artificial florivory (no damage, 30% flower tissue removed, or 60% flower tissue removed) on plant antagonists in Impatiens capensis...... 99

3.3. Effect of artificial florivory (no damage, 30% flower tissue removed, or 60% flower tissue removed) on Impatiens capensis reproduction...... 100

4.1. Effect of supplemented florivory on the number of outcrossing (CH) flowers produced by Impatiens capensis plants...... 126

4.2. Three-way interaction between supplemented florivory, nectar robbing, and pollination on A) pollinator visits and B) nectar robber visits to Impatiens capensis plants...... 127

4.3. Effects of florivory, nectar robbing, and pollination supplementation treatments on nectar thief visits, tested using a generalized linear model (GLIM) with a Poisson distribution...... 128

4.4. Effects of florivory, nectar robbing, and pollination supplementation treatments on subsequent natural florivory, tested using a generalized linear model (GLIM) with a Poisson distribution...... 129

4.5. Effects of florivory, nectar robbing, and pollination on fruit production in Impatiens capensis...... 130

4.6. Effects of florivory, nectar robbing, and pollination on seed traits in Impatiens capensis...... 131

xii

INTRODUCTION

Species interactions play critical roles in determining individual fitness, species ranges, and the shape or magnitude of selection on traits. Some of these interactions are mutually beneficial. The role of mutualisms in structuring communities and facilitating evolution has been increasingly recognized in the last decade (e.g., Bronstein et al. 2006).

Many other interactions are antagonistic, including predation, parasitism, infection, and competition. Individuals must navigate in a realm of reducing the costs of antagonisms while still receiving the benefits of mutualisms. Since mutualists and antagonists can interact directly or indirectly and contribute to a diversity of selection pressures, we should not assume that increasing mutualisms and decreasing antagonisms will lead to a better outcome for organisms or communities. For example, a moderate mixture of both mutualist and antagonist interactions can increase community stability, while extremes of either may reduce community stability (Mougi and Kondoh 2012). Therefore, it is important that researchers consider both mutualists and antagonists to understand the ecology and evolution of organisms.

Because plants are largely sedentary, they can have unique challenges in terms of optimizing interactions with mutualists and antagonists. Most plants produce physical or chemical defenses to deter antagonists such as herbivores (Hopkins and Hüner 2004) and invest in advertisements and rewards including volatiles, colors, shapes, or nectar rewards to attract mutualists such a seed dispersers or pollinators (Hopkins and Hüner 2004).

Interactions between plants and both leaf herbivores and pollinators can be critical for determining plant fitness. For example, an estimated 87.5% of all flowering plants are animal-pollinated (Ollerton et al. 2011), including many of our most important crop

plants (Losey and Vaughan 2006, Klein et al. 2007). In fact, there has been a long- standing suggestion that pollinators are responsible for the evolution of flowers to their current diversity of forms due to selection on floral attractive and reward traits (Darwin

1862, Fenster et al. 2004). However, many antagonists can also be attracted to flowers.

Despite the fact that floral antagonists may have particularly strong effects on plant reproduction because flowers are so closely related to reproduction, floral antagonists are studied much less frequently than leaf antagonists (McCall and Irwin 2006, Irwin et al.

2010). Understanding the evolution of flowers and plant mating systems requires understanding selection pressure by both mutualists and antagonists acting concurrently.

In addition to the possibility for floral antagonists to have strong direct impacts on plant reproduction, floral antagonists may also have indirect effects through changing how other organisms interact with plants. Floral antagonists often remove tissue, nectar, or pollen, and may induce physiological changes in the damaged flower or in subsequently produced flowers on the same plant (e.g., McCall and Irwin 2006,

Hargreaves et al. 2009, Irwin et al. 2010), which can alter attractiveness to subsequent visitors. There have been many studies demonstrating the effects of leaf herbivory on subsequent antagonisms (e.g., Stinchcombe and Rausher 2002, Strauss and Irwin 2004,

Van Zandt and Agrawal 2004a), and the diversity of leaf damage on a plant can structure subsequent whole-plant interactions (e.g., Van Zandt and Agrawal 2004a, Johnson and

Agrawal 2007, Ohgushi 2008, Utsumi et al. 2009). However, we know little about the community consequences of floral antagonisms. Because floral antagonists interact with plant reproductive organs, any changes they cause to the community interacting with

plants could have potentially strong effects on plant reproduction, selection for floral traits, and evolution or coevolution between insects and plants.

Floral Antagonists. There are many arthropods that act as floral antagonists.

Cheaters of the pollination mutualism, including nectar robbers, may be some of the best studied floral antagonists, though they have still received little attention compared to pollinators or leaf herbivores. Nectar robbers take nectar through holes cut or pierced in the flower’s corolla, but rarely pollinate flowers (Inouye 1980, Irwin et al. 2010). In many systems, nectar robbing can occur at high rates, frequently with 75-100% of all flowers attacked (Irwin and Brody 1998, Navarro 2001, Irwin and Maloof 2002, Young

2008, Irwin et al. 2010). In many (though not all) cases, nectar robbing reduces plant reproduction, either directly by damaging reproductive parts or indirectly by altering interactions with pollinators (Irwin and Brody 1998, Navarro 2001, Burkle et al. 2007,

Irwin 2009, Irwin et al. 2010). In fact, many studies have demonstrated that nectar robbers make flowers or plants less attractive to pollinators (Irwin and Brody 1998,

Maloof and Inouye 2000, Temeles and Pan 2002, Richardson 2004). By damaging flowers, reducing nectar volume, and inducing changes in floral traits, nectar robbers have the potential to affect many other floral interactions beyond pollination, although these are less studied. Through their direct and indirect effects on plants, nectar robbers may play a strong role in selecting for floral traits and altering the community of interactions on flowers.

Florivores (herbivores that consume flowers) can be as common or more common than leaf herbivores (McCall and Irwin 2006) and also have the potential to strongly affect plant reproduction. Florivores can directly reduce plant reproduction by damaging

pollen or ovules (Krupnick and Weis 1999, Leege and Wolfe 2002, McCall and Irwin

2006). In severe cases, florivores can cause the near collapse of plant populations

(Washitani et al. 1996). Florivores can also have indirect effects on plant reproduction by changing how other insects, such as pollinators, respond to plants (McCall and Irwin

2006). For example, florivory can reduce nectar production (Krupnick et al. 1999), flower size (Mothershead 2000), or flower symmetry (McCall 2008). Subsequently, florivory can reduce attractiveness to pollinators (Krupnick and Weis 1999, Botto-Mahan and Ojeda-Camacho 2000, Leavitt and Robertson 2006, McCall 2008, Cardel and Koptur

2010, Sõber et al. 2010, Cares-Suarez et al. 2011). However, there is very little previous research on whether florivory affects other floral antagonisms or induces changes in floral traits. Additionally, there is almost no data on what traits confer resistance against florivores or attract florivores to plants, although it has been hypothesized that many traits that attract pollinators may also attract florivores (McCall and Irwin 2006).

Currently, our understanding of florivory is limited to studies of direct effects on plant reproduction and a small number of studies on effects on pollinators. Since florivores likely have much more far-reaching consequences via altering other plant-insect interactions or selection on plant traits, it is imperative that we study florivores in a more complete context to understand their full effects on plants.

Flower bud gallers are another guild of floral antagonists that may influence plant reproduction. Flower bud gallers are specialized herbivores that lay eggs in flower buds, co-opting the tissue to grow into a tumor-like larva nursery instead of a flower (Crespi et al. 1997). Gallers in general can reduce plant growth and health (Larson 1998, Yukawa

2000, de Souza et al. 2006, Tooker and De Moraes 2007), compete for resources with

plant sinks (Yukawa 2000, Leege 2006), and manipulate plant resources (Larson and

Whitham 1991, Bagatto and Shorthouse 1994, Hartley 1998) or chemical defenses

(Hartley 1998, Pascual-Alvarado et al. 2008), but their effects on plant fitness are still ambiguous (Fay et al. 1996, de Souza et al. 2006). Flower bud gallers may have a particularly strong effect on plant reproduction because they remove reproductive organs, but few studies have directly measured these effects. Additionally, there have been almost no studies testing how gallers affect subsequent plant-insect interactions. Since gallers have the ability to directly alter plant physiology and growth, it is potentially important to understand their impacts on other arthropods and indirect effects on plant fitness.

Floral Traits Can Mediate Interactions. Flowers have many traits that can mediate interactions with mutualists and antagonists, and different guilds have varying responses to floral traits. Therefore, understanding how floral traits mediate multiple interactions is necessary to predict the evolution of those traits in a given context. Pollinator preferences for floral traits have received the most attention; pollinators tend to prefer flowers that are larger (Conner and Rush 1996, Galen 1996, Martin 2004, Asikainen and

Mutikainen 2005), nectar- or pollen-rich (Cnaani et al. 2006, Brandenburg et al. 2009,

Aronne et al. 2012), or plants with a larger total floral display (Conner and Rush 1996,

Ishii et al. 2008, Karron and Mitchell 2012). Pollinator can also vary with floral color

(Waser and Price 1981, Frey 2004, Irwin and Strauss 2005, Campbell et al. 2012,

Malerba and Nattero 2012) or scent (Kawano et al. 1995, Andrews et al. 2007, Galen et al. 2011, Kessler et al. 2012). Information about pollinator preference has informed predictions about how plant-pollinator interactions will change across landscapes and

with changing climate. For example, climate change has caused snowmelt to occur earlier, leaving some alpine plants with less water; these plants produce smaller floral displays, which may cause pollinators to switch to other species (Boggs and Inouye

2012). In another example, higher temperatures can increase nectar production, leading to the testable prediction that pollinators may have more available food per plants

(Petanidou and Smets 1996, Hegland et al. 2009).

While a few studies have assessed which floral traits attract nectar robbers (e.g.,

Maloof and Inouye 2000, Galen and Cuba 2001, Galen et al. 2011), little is known about how floral traits mediate interactions with most other floral antagonists, including florivores and flower bud gallers. Given the strong negative effects of these antagonists on plant fitness (McCall and Irwin 2006, Agrawal et al. 2007, Irwin et al. 2010), our understanding of the evolution of floral traits may be significantly changed by including these interactions.

Resources, Floral Traits, and Interactions. Resource quality and quantity can have large effects on plant and floral traits (Mattson 1980, Verhoeven et al. 1996, Hopkins and

Hüner 2004, Gurevitch et al. 2006), as well as plant-insect interactions (Campbell and

Halama 1993, Adler et al. 2006, Muth et al. 2008, Onoda et al. 2008, Burkle and Irwin

2010). By altering interactions, resource availability may change selection for floral traits (e.g., Price 1991, Behmer and Joern 2008, Banta et al. 2010). However, abiotic effects have been considered mostly in the context of leaf herbivores or pollinators, with a few studies examining nectar larcenists. Other floral antagonists, such as florivores or flower bud gallers, have rarely been considered. Additionally, most studies manipulate only one resource at a time, even though there have been several examples of non-

additive effects of multiple resources on leaf herbivores (e.g., Kersch and Fonseca 2005).

Such non-additive effects may be common in plant-insect interactions, and can have large impacts on the selective pressures experienced by plants.

The Importance of Studying Multiple Interactions. While many floral antagonists individually reduce plant reproduction in some conditions, plants do not interact with only one type of antagonist at a time (McCall and Irwin 2006, Agrawal et al. 2007, Irwin et al. 2010). Understanding pairwise interactions may not be sufficient to predict fitness outcomes, community composition, coevolution, or shapes of selection in natural systems

(Thompson and Cunningham 2002, Agrawal and Van Zandt 2003, Agrawal et al. 2007).

While we have made great strides in understanding the community consequences of leaf herbivory for subsequent antagonisms (Stinchcombe and Rausher 2002, Strauss and

Irwin 2004, Van Zandt and Agrawal 2004a), we know little about the community consequences of floral antagonists. Because floral antagonists directly affect plant reproductive organs, their effects on subsequent interactions and plant traits may have particularly strong impacts on fitness.

Dissertation Outline. The goal of my dissertation is to better understand the relationship between plants, floral traits, floral antagonists, and other plant-insect interactions. The work is divided into four chapters that each approach the question from a slightly different perspective. In Chapter 1, I examined the effects of abiotic conditions

(light, soil nutrients, and soil moisture) on floral traits, floral interactions, and plant reproduction. I found that plants with supplemented soil nutrients grew larger, had more attractive flowers, and had more floral visitors. However, these high-nutrient plants also had an increased proportion of floral antagonist compared to floral mutualist interactions,

suggesting resource availability may increase the relative costs of floral antagonisms.

Despite increased antagonisms, high-nutrient plants still reproduced significantly more, suggesting that plentiful resources may increase tolerance to antagonists. Together, these results suggest that selection for tolerance instead of resistance traits may be higher in areas of high resources.

Chapter 2 includes two experiments that manipulate flower bud gallers to assess effects on subsequent plant-insect interactions. When galls were manually cut and moved between plants, I found that plants that originally had galls tended to receive more pollinator visits, even when the galls had been removed. To test whether this correlation between flower bud galls and pollinators was due to changes in plant traits caused by the galler or because both insects preferred the same plant traits, I manipulated flower bud gallers using gall supplementation and removal. Galls had no significant effect on pollinator choice or leaf herbivory. This result, coupled with the observation that both pollinators and flower bud gallers preferred fertilized plants (Chapter 1), suggests that both gallers and pollinators prefer similar plant traits, and that flower bud gallers do not alter plant attractiveness to pollinators. Additionally, there was no effect of the flower bud galler supplementation treatment on plant reproduction even though gallers remove whole flowers, suggesting that this galling species may not be antagonists or may only be antagonists under some conditions.

In Chapter 3, I manipulated artificial florivory and measured the effects on floral traits, plant-arthropod interactions, and plant reproduction. Plants with florivory had significantly less leaf herbivory and fewer flower spiders, and significantly more subsequent florivory, suggesting that floral damage can have systemic effects on plants.

Induced susceptibility to subsequent florivory in response to floral damage has not been previously observed. Florivory also reduced plant reproduction and significantly reduced the ratio of outcrossing to selfing reproduction, but only at intermediate levels of damage.

This suggests that there is a threshold where intermediate floral damage had significant effects on reproduction, but high levels of floral damage did not. I also calculated correlations between early season floral traits and arthropod choice to assess which floral traits are involved in attracting floral antagonists and mutualists. Florivores preferred to visit taller plants with more, larger flowers with less pollen, less red coloration, and lower levels of anthocyanins. Pollinators and nectar robbers preferred the opposite floral traits compared to florivores (with the exception of plant height), and flower bud gallers had no preference for any floral trait measured. This suggests that some floral traits may be effective means of resistance to florivores and nectar robbers but not flower bud gallers, and that there may be conflicting selection between pollinators and nectar robbers on floral traits.

I manipulated florivory, pollination, and nectar robbing to assess impacts on floral traits, subsequent plant-insect interactions, and plant reproduction for Chapter 4. I found many significant multi-way interactions of the three treatments on floral insect interactions, suggesting that interactions with flowers depend on what other species are present. In all cases, the florivory treatment seemed to have a dominant effect, reducing pollination, nectar robbing, and nectar thieves, and increasing subsequent florivory. Only when enhanced florivory was absent were the effects of the nectar robbing and pollination treatments on subsequent interactions discernible. This same pattern was apparent in the effects on plant reproduction, with florivory always having a dominant

effect reducing outcrossing and increasing selfing reproduction, which masked the positive effects of pollination and (in some cases) nectar robbing on outcrossing reproduction.

Understanding the role that floral antagonists play in plant ecology can help scientists determine which interactions are most important, and may help determine why some floral traits exist in their current state. For example, I found that high levels of floral anthocyanins both attracted pollinators and deterred florivores, which may reinforce selection for this trait. It is also important to understand how plants mitigate the damaging effects of floral antagonists. For example, while there was limited evidence for induced resistance to florivores, it seems more likely that plants induce tolerance to florivory by shifting their mating system towards a proportional increase in selfing. If this is common among plants with mixed mating systems, it may help explain why mixed mating systems are maintained, as well as why floral antagonisms often occur at such high rates. Finally, it is important to note that, with the exception of flower bud gallers, the presence of all manipulated floral interactions had significant effects on subsequent floral interactions. This suggests that selection for floral traits may be complex, and may shift depending on the context of which interactions are present in each location or year.

Together, this work represents some of the most comprehensive research on the community consequences of floral antagonists, as well as the interplay between floral traits and floral interactions.

CHAPTER 1

ABIOTIC CONDITIONS AFFECT ANTAGONISTS AND MUTUALISTS IN

IMPATIENS CAPENSIS

Abstract

While the effect of abiotic factors on leaf herbivory is well known, the relative importance of abiotic conditions influencing both mutualists and antagonists is less well understood. Species interactions could enhance or reduce the direct effects of abiotic factors, depending on how mutualists and antagonists respond to abiotic conditions. We manipulated soil nutrients and shade in a factorial design, and measured soil moisture in the annual Impatiens capensis. We then measured interactions with mutualists (two pollinating species) and antagonists (herbivores, florivores, nectar thieves, and flower bud gallers), as well as plant growth, floral rewards, and plant reproduction. Fertilizer increased plant growth, floral attractiveness, mutualist and antagonist interactions, and plant reproduction. Shade had no effects, and soil moisture was negatively associated with plant growth and reproduction. All effects were additive. Mutualist and antagonist floral interactions both increased on fertilized plants, but antagonists increased at a greater rate, leading to a larger ratio of antagonist to mutualist interactions on fertilized plants. Despite having more antagonists, fertilized plants still had significantly higher reproduction, suggesting higher tolerance to antagonists. The results from this study show that abiotic effects can have consistent effects on antagonists and mutualists, and on both floral and leaf antagonists. However, tolerance to antagonisms increased in favorable conditions. Thus, the direct positive effects of favorable abiotic conditions on

plants outweighed negative indirect effects via increased antagonisms, suggesting selection to grow in high-nutrient microsites in spite of increased herbivory.

Introduction

Plant ecologists have a long and rich history of studying how abiotic factors such as soil nutrients, light, and soil moisture affect plant growth and reproduction. Nitrogen and phosphorus are frequently the most limiting factors for plant growth and reproduction, and also increase allocation to attractive or defense traits (Mattson 1980,

Verhoeven et al. 1996, Hopkins and Hüner 2004, Gurevitch et al. 2006). Similarly, light quality and quantity can strongly affect plants. Species have different light requirements, and exhibit shade avoidance behaviors in too little light and wilting, sun scald, damage to photosynthetic machinery, or other heat stress in too much light (Ort 2001, Hopkins and

Hüner 2004, Sultan 2010). While water availability has varying effects, water often increases plant growth and stimulates plant physiological processes such as photosynthesis (Wu et al. 2011).

Abiotic factors can also have strong effects on interactions with mutualists and antagonists. Soil nutrients can change plant traits such as defenses, nutritional value, or attractiveness, frequently leading to decreased (but sometimes increased) herbivory

(Muth et al. 2008, Onoda et al. 2008) and increased pollinator visitation (Campbell and

Halama 1993, Adler et al. 2006, Burkle and Irwin 2010). Light can also affect insect behavior, both directly when insects are ectothermic or in danger of desiccation (Dudt and Shure 1994, Rossi and Stiling 1998, Gullen and Cranston 2005, Leege 2006,

Kilkenny and Galloway 2008, Muth et al. 2008) and indirectly by altering plant

attractiveness or defenses (Dudt and Shure 1994, Baraza et al. 2004, Kersch and Fonseca

2005, Muth et al. 2008, Ingersoll et al. 2010). Similarly, soil moisture can both directly and indirectly affect insects. Sap-sucking insects generally decrease on water-stressed plants, while foliar feeders are less predictable (Schowalter et al. 1999, Huberty and

Denno 2004). Water availability can affect flower number and rewards such as nectar traits (Villarreal and Freeman 1990, Carroll et al. 2001, Bissuel-Belaygue et al. 2002), and therefore may also alter interactions with pollinators.

Variation in the abiotic environment can have unintuitive effects on plant fitness by differentially affecting antagonists and mutualists. For example, fertilized Inga vera plants had more herbivory than non-fertilized plants in the sun but not the shade, due to differences in ant protection (Kersch and Fonseca 2005). In another example, Borrichia frutescens plants were protected from galling herbivores by large populations of parasitoids in nitrogen-enriched plots but not nitrogen-poor plots (Stiling and Rossi

1997). Examining differences in the effects of mutualists versus antagonists in varying conditions may be especially interesting because the abiotic environment can alter conditionality in interactions (Bronstein 1994, Gomulkiewicz et al. 2003, Kersch and

Fonseca 2005). For example, mycorrhizal associations with plant roots can be beneficial, neutral, or antagonistic depending on nitrogen availability and soil moisture (Heath and

Tiffin 2007, Kennedy and Peay 2007).

While there is a large literature addressing how abiotic factors can affect leaf herbivores (e.g., Coley et al. 1985, Price 1991, Hopkins and Hüner 2004), floral antagonists can be more detrimental to plant fitness because they attack reproductive organs (McCall and Irwin 2006, Irwin et al. 2010). Florivory can be as common as leaf

herbivory, and can remove an equivalent amount of tissue (McCall and Irwin 2006).

Despite the importance of floral antagonists to plant reproduction, we know of no experiments testing whether abiotic factors affect floral antagonists similarly to leaf herbivores.

While the effect of individual abiotic factors on plants and insects has been well studied, multiple abiotic factors also commonly have non-additive effects. For example, total phenols and hydrolysable tannins were unaffected by fertilizer addition in shade treatments, but decreased significantly with fertilizer in sunny open fields (Dudt and

Shure 1994). In non-nitrogen fixing plants, sufficient soil moisture is needed to make soil nitrogen available for root uptake, but excessive water can cause nitrogen to leach away (Mattson 1980, Cronk and Fennessy 2001, Hopkins and Hüner 2004). Thus, studies that examine multiple abiotic factors simultaneously, including effects on plant- insect interactions and plant reproduction, are needed to understand the role of habitat variation on community dynamics.

We examined the effects of three abiotic factors (soil nutrients, light, and soil moisture) on a suite of mutualist (two pollinator species) and antagonist (folivores, florivores, nectar thieves, and flower bud gallers) plant-insect interactions, as well as plant growth, floral rewards, and plant reproduction. We manipulated fertilizer addition and light in a factorial design and measured natural soil moisture across a gradient to determine: (1) how multiple abiotic factors affect plant-insect interactions; (2) if multiple abiotic factors differentially affect mutualist and antagonist insect interactions; and (3) how multiple abiotic factors affect plant growth and reproduction.

Methods

Study System and Location. Impatiens capensis Meerb. (Balsaminaceae) is an annual herb native to much of eastern North America. It grows in partial shade and moist, rich soil (Leck 1979). It has both selfing cleistogamous (CL) and open-pollinated chasmogamous (CH) flowers; the showy orange CH flowers lead to more seeds with better dispersal and survival (Mitchell-Olds and Waller 1985, Mitchellolds and Waller

1985, Eastman 1995). CH flowers are pollinated mostly by Bombus sp. and Apis mellifera (Apidae), both of which have been shown to be efficient pollinators (Rust

1977), and are incapable of selfing because the androecium covers the stigma until pollen shedding has finished (Rust 1977, Schemske 1978, Eastman 1995, Steets and Ashman

2004). Geitonogamy in I. capensis has been estimated at only 8.6% (Waller 1980). Both flower types produce capsule fruits with one to several seeds that dehisce explosively when mature. In Massachusetts, I. capensis generally germinates in late April or early

May, CL flowers begin to appear in May, and CH flowers last from mid July until mid

September. Seeds generally are not viable for more than one year (Simpson et al. 1985), resulting in little to no seed bank.

Impatiens capensis has many antagonists. There are a variety of insect herbivores, including true bugs (Hemiptera), grasshoppers (Orthoptera), katydids

(Tettigoniidae), aphids (Aphidoidea), and Japanese beetles (Scarabaeidae: Popillia japonica), and larger herbivores including deer (Eastman 1995, Steets and Ashman

2004). Agromyza borealis (Agromyzidae), a specialist leaf miner fly, attacks only I. capensis and I. pallida (Frost 1924, Eastman 1995). The flowers are robbed by several insect species, including Bombus pollinators and Vespula maculifrons (Vespidae), and

visits by small nectar thieves are common (Rust 1979, Eastman 1995, Young 2008). We follow previous work (Irwin et al. 2010) in defining nectar robbers as insects that pierce the flower or spur to take nectar, and nectar thieves as insects that are too small to contact reproductive parts. Popillia japonica beetles and other generalist herbivores also eat flowers (NLSG, pers. obs.). Finally, there are two species-specific Cecidomyiidae gallers: Cecidomyia fulva, a leaf midrib galler, and Schizomyia impatientis, a flower bud galler.

All field experiments were carried out at Hampshire Farm (N 42º 19’ W 72º 31’) at Hampshire College, Amherst, Massachusetts, USA during summer 2008. The site has a large natural population of Impatiens capensis. Study plots were located along the edge of a drainage field covered in a maple-dominated wood, facing northeast to northwest.

On May 7, we collected I. capensis seedlings less than 50 cm tall from the natural population adjacent to our study plots. Between May 7 and May 12, seedlings were transplanted into 10 cm diameter round pots (Fafard #2 potting soil, Conrad Fafard, Inc,

Agawam, Massachusetts, USA) and maintained in a greenhouse until they were planted in the field.

On June 6 and 7, I. capensis plants were planted under shade structures (see

Treatments below). Five plants were placed under each shade structure, with one focal plant in the middle and one plant in each corner of the structure. Measurements were taken only on the focal plant, but the border plants were included because I. capensis grows naturally in patches. Any conspecific plants growing naturally in the shade structures were removed to keep intraspecific competition constant. All plants were 45 -

60 cm tall at transplanting.

Treatments. Light and nutrients were manipulated in a 2x2 factorial design with

10 replicates in each light-nutrient combination, and natural soil moisture was measured.

Light was manipulated using 40 wooden structures (1.5m x 1.5m x 1.5m) with the top surface covered with 4 mil (0.1 mm thick) clear vinyl sheeting (Ace, Oak Brook, Illinois,

USA). Half of the plastic roofs remained clear to simulate sunlight and half were painted to simulate foliage shade using a pigment mixture adapted from Lee (1985) designed to mimic the R:FR typical of foliage shade. Briefly, one part Solavperm Yellow G and four parts Hostaperm Violet RL 02 (76 g and 303 g, respectively; Clariant Corporation,

Coventry, Rhode Island, USA) were mixed into 3.785 liters of clear satin polyurethane varnish (Ace, Oak Brook, Illinois, USA). This was then painted onto the vinyl sheeting using a paint roller (Lee 1985). Shade and light treatments were assigned to alternating structures to help account for natural gradients in soil and light along the forest edges, with twenty structures of each treatment. Four times during the summer on days with no clouds, photosynthetically active radiation (PAR) was measured as photosynthetic photon flux (ppf) once an hour under each of the structures using a basic quantum meter (model

BQM-S, Apogee Instruments, Logan, Utah, USA). Light levels spanned the range of

PAR that I. capensis naturally grows in, between 5 and 1500 ppf (Simpson et al. 1985,

Lechowicz et al. 1988), with shade treatments averaging 171.9 ppf and sun treatments averaging 906103 ppf (meanSE).

On June 11, Osmocote classic controlled release fertilizer (NPK 14-14-14; The

Scotts Company, Maryville, Ohio, USA) was added to each of the nutrient addition treatment plots, with 24 g applied evenly in a 20 cm diameter circle around each plant.

Plants were watered once two days after fertilizer addition due to excessively dry

weather, and thereafter received only natural precipitation. At the end of the season, soil samples were collected from each plot and tested in the University of Massachusetts Soil

Lab (Amherst, Massachusetts, USA) for nutrient content. Plots that were fertilized had significantly more nitrate, plant-available phosphorus, and potassium than unfertilized plots (ANOVA, P < 0.0001 for all), and all nutrient levels (fertilized and unfertilized) were within previously published natural levels for I. capensis plants (Simpson et al.

1985, Lechowicz et al. 1988).

Soil moisture for each plot was averaged from four measurements per plot throughout the summer using a Kelway soil acidity and moisture tester (model HB-2, Kel

Instruments, Wyckoff, New Jersey, USA), which measures moisture as the percent of total soil capacity. Soil moisture did not differ between shade treatments (mean ± SE: sun = 56.3 ± 1.85, shade = 56.1 ± 2.25; t-test, t = 0.06, df = 38, P = 0.95), or between fertilizer treatments (mean ± SE: fertilizer = 55.9 ± 2.01, no fertilizer = 56.5 ± 2.11; t- test, t = 0.22, df = 38, P = 0.82).

Plant Measures. Plant height, the number of nodes, internode length, percent of nodes that branched, and leaf area were measured four times during the summer to estimate plant growth. Leaf area was estimated by measuring the length and width of the four most fully expanded apical leaves; the product of leaf length and width is highly correlated with leaf area (n = 42, r2 = 0.998, P < 0.0001). Maximum measurements were used for plant height and number of nodes, and mean values were used for other measures. After the first heavy frost (October 17), all aboveground biomass was collected, dried, and weighed for each plant.

CH and CL flowers were counted two to three times per week to estimate the average number of each type of flower. We measured CH flower morphometrics on five dates on up to three flowers per focal plant, including total flower length, spur length, height and width of the corolla lip, and height and width of the corolla opening.

To measure nectar production, we bagged up to six CH buds per focal plant with polyester fiber bags (cut from Fibe-Air greenhouse sleeves, Kleen Test Products,

Milwaukee, Wisconsin, USA) and allowed the flowers to open. We measured nectar on half of the flowers in the male stage (approximately two to 24 hours after opening) and the other half in the female stage (approximately 36 to 48 hours after opening) because I. capensis has different volumes of nectar in male-phase and female-phase flowers (Rust

1979, Young 2008). Nectar volume was measured using microcapillary tubes, first through the corolla opening and then from the spur by snipping the end and squeezing nectar out. We measured sugar concentration using a handheld refractometer (Fisher

Scientific, Pittsburgh, Pennsylvania, USA). The refractometer only measured up to 62% sugar; nectar more concentrated was recorded as 63%.

Fruits were counted approximately every other day from June 12 until September

12, and mature fruits were counted at the same time beginning July 31. Total fruit production rather than average fruits per day can be counted from pedicel scars from dehisced fruits, but the process is extremely time-consuming. A mid-summer survey showed that the average number of fruits present per day was highly correlated with the total number of fruits produced up to that point (n = 40, r2 = 0.905, P < 0.0001), so average fruits per day was used as an efficient way to estimate total fruit production.

Additionally, six times during the summer all of the mature fruits were collected from

each plant to estimate seeds per fruit and seed mass. These were counted and identified as CH or CL, then stored in individual glassine envelopes at 5ºC until seeds were counted and weighed.

Insect Measures. Pollinators were surveyed approximately weekly between July

25 and August 5. We counted total open CH flowers (per plot and per focal plant), and only plots with at least one open CH flower on the focal plant were included. We recorded the identity of every floral visitor (to genus only for Bombus spp.) to the plot during 15 minute surveys. We also recorded the number of flowers probed and time per probe. When more than one pollinator visited at the same time, detailed observations were only recorded for the first visitor; the second was only counted and identified.

Weather and time of day were recorded for each visit, and had no significant effect on

2 pollinator behavior (ANOVA, F3, 1512 = 1.26, P = 0.28; and correlation, n = 1516, r =

0.01, P = 0.064, respectively), so neither was included in analyses.

Herbivores were surveyed approximately weekly between June 17 and August 13.

The number and identity of all herbivores was recorded. Herbivores were sorted into 11 functional groups (flower galls, small sap suckers, plant hoppers, true bugs, caterpillars, beetles, grasshoppers, moths, small flies, non-lepidopteran larvae, and internal feeders) for analysis of herbivore diversity. Since more species and more functional groups are likely to be found when more individuals are counted (Hurlbert 1971), we used rarefaction (rarefy() in R; R Development Core Team, 2.9.0, 2009) to calculate the expected diversity (the number of functional groups) using the smallest abundance measure seen (4 herbivores) as our expected population size. The goal of this analysis is to determine whether diversity would still be different across treatments if plants were

constrained to have only 4 herbivores per plant, subsampled from the actual herbivores recorded on each plant. Additionally, leaf area removed was estimated on the four topmost fully expanded leaves; since these were the newest leaves, any damage on these four leaves would have occurred since the last survey, ensuring we never double counted damage. Four additional surveys of flower galls were carried out between August 23 and

September 15, because galls are abundant late in the season when much other herbivory has declined. Florivory on CH flowers was recorded five times, measured as the proportion of flowers damaged and percent floral tissue missing.

Statistical Analyses. General Approach. We used several principle components analyses (PCA) to create composite variables reflecting plant growth, flower morphometrics, and plant flowering to reduce the number of variables and multicollinearity. All PCAs were performed using prcomp() in R (R Development Core

Team, 2.9.0, 2009) and all data used in PCAs were z-score standardized prior to PCA.

To test the effects of abiotic factors on logical subsets of data, we conducted separate

MANCOVAs on plant growth, floral rewards, floral visitors, herbivory, and plant reproduction; each of these is described in more detail below. All MANCOVAs were conducted using SAS (v 9.2, SAS Institute, 2008) with Type III sums of squares, including fertilizer, shade, and the fertilizer x shade interaction term as fixed factors, and soil moisture as a covariate. Individual ANOVAs were investigated when an overall

MANOVA was significant.

Plant Growth. The effects of shade, nutrients, and soil moisture on plant growth were tested with a MANCOVA. Several of the responses were principle components from analyses described immediately below. Response variables included PC1 and PC2

of the plant size PCA (representing overall plant size and amount of branching, respectively), PC1 of the flower morphometrics PCA (representing flower size), PC1 of the flower number PCA (representing number of flowers), and the ratio of CH to CL flowers.

A PCA representing plant size was calculated based on plant height, dry biomass, leaf area, percent of branching internodes, internode length, and plant lifespan (Table

1.1). Number of branches was squared and biomass and lifespan were square root transformed to improve normality. For PC1, a more positive value indicated an overall larger plant (taller, more biomass, larger leaves) while in PC2, a more positive value indicated greater branching (more internodes with more of them branching; Table 1.1).

Flower size was estimated using a PCA on the six measures of flower morphometrics (total flower length, spur length, height and width of corolla lip, and height and width of corolla opening), which yielded one PC that was strongly correlated with overall flower size (Table 1.1). Total flower length and spur length were squared to improve normality. In this PCA, plot 28 was a multivariate outlier, but removing it from the analysis did not qualitatively change results, so it was left in the PCA. A more positive PC value for flower size indicates a larger flower in all six measures (Table 1.1).

Flowering was estimated with a PCA on the number of CL and CH flowers and the Julian date of the first CH flower (Table 1.1). Number of flowers for both flower types was log transformed to improve normality. Plot 28 was removed because it never flowered during surveys. More positive values for this PC indicate more flowers and earlier flowering (Table 1.1).

Floral Rewards. The effect of abiotic factors on floral reward was tested using a

MANCOVA that included male and female phase nectar volume and concentration; these traits were not correlated (correlation, n = 37, r < |0.29|, P > 0.07 for all) and could not be reduced using PCA. Nectar volume of female phase flowers was log transformed to improve normality.

Floral Visitors. There were two pollinators seen during pollinator observations,

Bombus spp. and Apis mellifera, which we included separately in our model. Many other smaller species visited the flowers as thieves, including Augochlora spp. (Halictidae) and ants (most likely Formica spp, Formicidae); these visitors were lumped into one category of nectar thieves because visitation for individual species was relatively low. Only six nectar robbers were seen all summer, and thus were not prevalent enough to be analyzed.

Flower visitation was tested using a MANCOVA including the total number of

Bombus spp., Apis mellifera, and nectar thief probes per plot in 15 minutes as responses.

In addition to visitation, behavior during visits for Bombus spp., A. mellifera, and nectar thieves was tested using an individual MANCOVA for each, including the number of flowers probed and the proportion of flowers per plot visited. The number of probes provides an absolute value for total visitation to the plant, while proportion of flowers visited is scaled by how many flowers the plant produced; both measures are useful, since a plant with many flowers may have more absolute visitation but have a smaller proportion of its flowers pollinated. Time spent per flower was analyzed for Bombus spp. and A. mellifera but not nectar thieves as a group because this measure varies by species.

Herbivores. Herbivory was tested using a MANCOVA with herbivore abundance, diversity (number of functional groups), percent leaf damage, proportion of

flowers damaged, percent floral tissue missing, number of flower galls, and Julian date of the appearance of the first gall as responses. Flower galls were included as a separate response from the herbivore survey results because most gall surveys did not coincide with other herbivore surveys. Herbivore abundance, leaf damage, percent floral tissue missing, and number of galls were log-transformed to improve normality. Rarefied herbivore diversity was tested using a separate ANOVA with fertilizer treatment, light treatment, and soil moisture (as a covariate) as explanatory variables.

Plant Reproduction. Plant reproduction was measured as the average from weekly counts of the number of total fruits and mature fruits, seeds per CH and CL fruits, and seed mass for CH and CL fruits, and analyzed using a MANCOVA. Plot 28 was an outlier for seed mass; since plot 28 produced very few fruits, it was removed to improve normality.

Results

Plant Growth. Fertilizer increased plant growth and flowering (Tables 1.2, 1.3).

Individual ANOVAs showed that fertilizer increased plant size PCs 1 and 2 (representing plant size and branching, respectively), flower PC1 (representing more and earlier flowers), and induced proportionally more CH (outcrossing) flowers (Table 1.3, Figure

1.1). Plants growing in moister soil had fewer, later blooming flowers and proportionally more selfing cleistogamous flowers (Tables 1.2, 1.3). Flower size was not affected by any abiotic treatment. Shade had no effect on growth or flowering (Table 1.2).

Floral Rewards. Fertilizer marginally increased floral rewards (Table 1.2).

Fertilizer increased the volume of nectar in both male and female phase flowers (Table

1.3; Figure 1.1), but did not affect sugar concentration. Neither shade nor soil moisture affected floral rewards.

Floral Visitors. Fertilizer significantly increased floral visitation (Table 1.2).

Bombus spp. pollinators and nectar thieves both visited fertilized plants more than twice as often as unfertilized plants, while Apis mellifera pollinators showed no significant difference in visitation (Table 1.3; Figure 1.2). Overall, visitation by nectar thieves increased on fertilized plants more than visitation by both pollinator species combined

(125% vs. 70% increase), suggesting floral antagonists may respond more strongly to nutrient additions than pollinators.

Fertilizer also affected pollinator behavior during visits (Table 1.2). Bombus spp. and A. mellifera both probed a lower proportion of flowers per plant in fertilized plots compared to unfertilized plots (Table 1.3). However, Apis mellifera probed more flowers per visit on fertilized plants (Table 1.3), indicating that flower production increased more than the number of flowers probed per visit.

As with pollinators, nectar thief behavior during visits was affected by fertilizer but not shade or soil moisture (Table 1.2). Univariate analyses showed that nectar thieves probed a greater absolute number of flowers per visit, but a lower proportion of open flowers per visit, on fertilized plants than unfertilized plants (Table 1.3).

Herbivores. Fertilizer had a significant effect on overall herbivory (Table 1.2).

Fertilizer increased herbivore diversity, decreased the proportion of flowers with florivory, and resulted in more flower galls that appeared earlier (Table 1.3, Figure 1.3).

However, fertilizer had no effect on rarefied diversity (F4,35 = 1.01, P = 0.4137), suggesting that the number of herbivores may have driven the increased herbivore

diversity in fertilized plots. There was also no effect of fertilizer or any other abiotic factor on total number of herbivores, leaf damage, or amount of tissue removed from damaged flowers.

Plant Reproduction. Plant reproduction was significantly increased by fertilizer and negatively correlated with soil moisture (Table 1.2). Fertilizer increased total fruits, mature fruits, and seeds per fruit for CH but not CL fruits (Table 1.3, Figure 1.4). Total fruits, mature fruits, and seed mass for CH fruits were significantly negatively correlated with soil moisture (Table 1.3). Shade had no effect on plant reproduction.

Discussion

Increased antagonisms and tolerance. Fertilizer increased antagonisms but still led to greater plant growth and reproduction. Herbivores, flower galls, and nectar thieves all preferred fertilized plants (Figs. 2 and 3). This is consistent with both the Plant Vigor

Hypothesis, which states that plants with adequate water and nutrients should be more attractive and possibly more nutritious to insect herbivores than stressed plants (Price

1991), and the Resource Availability Hypothesis, which posits that plants that grow in low nutrient environments have higher levels of defensive compounds because their growth rate is limited and they therefore have a higher exposure time to herbivores

(Coley et al. 1985). This is the first evidence we are aware of that floral antagonists can be affected by abiotic factors in a manner similar to leaf herbivores. Previous work found increased florivores on Aspilia foliacea with extra nitrogen due to fire, but did not measure florivore damage (Prada et al. 1995).

Although both Bombus spp. pollinators and nectar thieves preferred fertilized I. capensis plants, A. mellifera had no significant preference for fertilized plants. This led to visitation by nectar thieves increasing at a greater rate than visitation by pollinators, so that the ratio of thieves to pollinators was greater on fertilized plants. Both flowering

(number and date) and nectar volume was significantly greater in fertilized plants (Table

1.3; Figure 1.1), suggesting that nectar thieves may respond more strongly than pollinators to flower number or rewards. This may mean that pollinators are relatively consistent in their visitation, whereas nectar thieves are opportunistic. Nectar thieves as a functional group may also be better at differentiating flowers with high and low nectar volumes than pollinators. Previous work suggests that Bombus pollinators of I. capensis cannot discern the amount of nectar rewards in flowers (Rust 1979), but the ability of nectar thieves to assess rewards is unknown (Irwin et al. 2010).

Fertilized plants had significantly more functional groups of herbivores per plant

(Table 1.3). Additionally, while fertilizer had no significant effect on the total number of herbivores (Table 1.3), there was high variation in herbivore numbers, ranging from 4 to

429 herbivores, with a trend for fertilized plants to have more total herbivores. When we calculated rarefied herbivore diversity with an expected population size of 4 herbivores, rarefied diversity of herbivores was no longer significantly different between fertilized and unfertilized treatments, suggesting that the differences seen in herbivore functional diversity are due to differences in herbivore number that are too variable to be significant in our MANOVA. Increasing herbivory is common in nutrient-rich plant communities; for example, fertilizing a grassland habitat led to greater herbivore abundance and richness (Siemann 1998). Similarly, bracken ferns with added nitrogen had increased

herbivore abundance and diversity (Jones et al. 2011). Both herbivore abundance and herbivore diversity can have strong negative effects on plants (Gullen and Cranston 2005,

Gurevitch et al. 2006, Behmer and Joern 2008), so access to nutrients may make plants more vulnerable to leaf damage.

Although fertilized plants had significantly more floral antagonisms and leaf herbivore diversity, I. capensis reproduced significantly more in fertilized plots. This suggests that I. capensis is more tolerant of antagonisms with adequate soil nutrients.

Soil nutrients, and nitrogen specifically, increase tolerance to leaf herbivory in many systems (Wise and Abrahamson 2008, Kohyani et al. 2009, Sun et al. 2010; but see

Katjiuna and Ward 2006, Marshall et al. 2008, Suwa and Maherali 2008), but only one study has measured tolerance to floral antagonists (nectar robbers; Irwin 2009) and there are no studies examining the effects of abiotic factors on tolerance to floral antagonists.

Increased plant growth and allocation to reproduction, both of which we see evidence for in this experiment, can serve as methods of tolerating herbivory in fertilized plots (e.g.,

Katjiua and Ward 2006, Suwa and Maherali 2008, Wise and Abrahamson 2008, Kohyani et al. 2009, Sun et al. 2010). Fertilized plants produced more nectar and more flowers, which have been suggested as possible methods of tolerating nectar larceny and florivory

(Irwin et al. 2008, Wise et al. 2008). Future research should address how well theories of resource allocation, resistance, and tolerance relevant to leaf herbivores (such as the

Optimal Defense Theory, the Plant Vigor Hypothesis, or the Resource Availability

Hypothesis) can be applied to floral antagonists.

Pollinators. Fertilizer increased pollinator visits by Bombus spp. but not A. mellifera (Table 1.3). This suggests that the native pollinators were more sensitive to

nutrient availability than the non-native A. mellifera. There were many plant-level traits affected by fertilizer (Tables 1.2, 1.3), suggesting that A. mellifera may forage at a larger scale than the plant (i.e. population or landscape). Others have found differences in pollinator response to scale. For example, Benjamin and Winfree (2011) found that smaller bodied bees tended to respond to landscape characteristics at a smaller scale (300 m radius) than larger bodied bees (1500 m radius). Factors other than body size, such as sensory ability or preferences for one species of plant, could also drive the scale at which pollinators respond. For example, A. mellifera may respond to all species of nectar- bearing flowers in the landscape, whereas Bombus spp. (individuals or colonies) may target I. capensis flowers. This could be tested by manipulating the presence and density of several flowering species in a landscape, and measuring preference by both pollinator species for I. capensis.

Effects of multiple abiotic factors. Fertilizer. Overall, fertilizer had the strongest effects on all of the insect interactions (florivores, folivores, pollinators, nectar thieves, bumblebees, and marginally honey bees) as well as on all plant measures, including growth, floral rewards, and reproduction (Table 1.2). This is not surprising, considering the wealth of past literature showing fertilizer effects on plant traits and insect interactions (e.g., Campbell and Halama 1993, Kersch and Fonseca 2005, Muñoz et al.

2005, Adler et al. 2006, Muth et al. 2008, Onoda et al. 2008, Burkle and Irwin 2010).

The effect of soil nutrients on plants or plant-insect interactions is not always consistent, sometimes decreasing defenses (Cornelissen and Stiling 2006) or shifting allocation from growth to reproduction, leading to smaller plants (Muñoz et al. 2005, Stiling and Moon

2005). Soil nutrients can also frequently have variable effects on herbivores, increasing

(Stiling and Rossi 1997, Stiling and Moon 2005, Cornelissen and Stiling 2006), directly decreasing (Cuevas-Reyes et al. 2004), or indirectly decreasing (Heil et al. 2001) herbivory. The effects of nutrients on pollinators has not been examined as frequently, but may range from positive (Muñoz et al. 2005, Burkle and Irwin 2009) to negative

(Muñoz et al. 2005) to neutral (Burkle and Irwin 2009). In this study, however, the effects of fertilizer were remarkably consistent.

Light. Although previous studies have suggested that light is the most important abiotic factor for the growth and reproduction of I. capensis (Simpson et al. 1985), with positive correlations between sunlight and plant size, branching, number of flowers, and

CH:CL flower ratio (Schemske 1978, Waller 1980, Schmitt 1993), there was no significant effect of shading on any measure of plant growth, floral rewards, plant reproduction, or insect interactions (Table 1.2). This is somewhat surprising, since I. capensis generally grows in shady areas (Lechowicz et al. 1988, Eastman 1995), probably because plants cannot close their stomata at midday and so may suffer high water loss in full sun (Schulz et al. 1993, Heschel et al. 2004). However, I. capensis also tends to be light-limited for photosynthesis and growth as well as reproduction, and can photosynthesize well despite wilting (Waller 1980, Schulz et al. 1993, Heschel et al.

2004). The combination of susceptibility to wilting and light-limited photosynthesis suggests that intermediate light levels are optimal for plant growth and fitness. By using only two light levels (high and low), this study may have missed an optimal intermediate light level that could be investigated in future work.

Previous work with I. capensis seedlings demonstrated that a decreased red to far red (R:FR) light ratio due to foliage shade increased height and length of internodes as an

etiolation response, while a high R:FR light ratio increased UV-blocking compounds

(Maliakal et al. 1999, Dixon et al. 2001, Heschel et al. 2004, Weinig et al. 2004, von

Wettberg and Schmitt 2005). However, in the current experiment, there was no effect of shade with a reduced R:FR ratio on plant height, growth, branching, or internode length.

Most previous work examining R:FR shade effects on I. capensis used seedlings, while in the current experiment seedlings grew under the same conditions until they were about 40 cm tall. The lack of shade effects on growth in our experiment suggest that the etiolation response may be strongest in germinating I. capensis. Since I. capensis germinates in dense, closely-spaced populations, there is high early-season intraspecific competition for light, and etiolation may be an important early-season response to outcompete neighbors

(Waller 1985, Schmitt 1993, Lively et al. 1995, Sultan 2010). By contrast, older plants are frequently shaded by overstory trees that have leafed out later in the season, leading to consistent shading for all plants and a negligible usefulness of etiolation. Previous research has shown that light has a stronger effect on shade avoidance early in a plant’s life (Augspurger et al. 2005, Mathews and Tremonte 2012) or early in the season

(Augspurger et al. 2005). Additionally, there has been some suggestion that an etiolation response early in a plant’s life primes the plant for etiolation later in life (von Wettberg et al. 2012), so it is possible that keeping our plants under fully lighted conditions early in growth may have prevented a strong subsequent etiolation response. The effects of abiotic factors may thus depend strongly on plant phenology, suggesting that it is important to examine abiotic effects at multiple plant stages.

Surprisingly, there was no effect of light on floral visitors or herbivores (Table

1.2). In other systems, pollinators frequently prefer sunny areas (Kilkenny and Galloway

2008). For example, Campanulastrum americanum plants received up to seven times as many pollinators in the sun than in the shade, leading to three times as much pollen removed and eliminating pollen limitation (Kilkenny and Galloway 2008). Pollinator preference for light has been shown under canopy gaps in I. capensis previously, with more floral visitors frequenting flowers in canopy gaps than under full canopy cover

(Walters and Stiles 1996). Similarly, herbivores are often more prevalent in sun than shade due to differences in temperature, plant defenses, and/or plant nutritional quality

(Karban et al. 1999). However, in this experiment we saw no effect of light on pollinator or herbivore preference. Summer 2008 was unusually warm, with June, July, August, and September each averaging 3.5°C above their respective monthly average temperatures (National Weather Service 2012). Overly warm temperatures, exacerbated by sun, can cause overheating or desiccation in insects, which could explain the lack of preference for sunny plots in this experiment.

Soil Moisture. Typically, I. capensis prefers wet areas (Leck 1979, Eastman

1995), often with soil moistures above 90% of soil mass (Simpson et al. 1985). Increased soil moisture can increase plant size, the CH:CL flower ratio, and flower nectar in I. capensis plants (Waller 1980, Marden 1984). However, plants in this experiment did not do as well in wetter soils. While plants in moister soils achieved the same size, they produced fewer flowers, a smaller proportion of outcrossing CH flowers, fewer fruits, fewer maturing fruits, and slightly smaller seeds (Table 1.2). There may be a trade-off where soils that are too dry allow plants to easily wilt (Schulz et al. 1993), but soils that are too moist keep roots from getting enough oxygen, increase soil acidity, decrease nitrogen availability, and can contribute to root rot (Abrahamson and Hershey 1977,

White 1984, Lechowicz et al. 1988, Hopkins and Hüner 2004). However, soil moisture was not directly manipulated in this experiment. Many other microhabitat features vary with natural levels of soil moisture, including soil acidity (at this site, correlation, n = 40,

P < 0.0001, r = -0.608; Soper Gorden, unpublished data), soil texture (at this site, sandy soils (low lying areas) = 63.09 ± 1.91% soil moisture, other soils (upland areas) = 51.13 ±

1.29%; mean±SE; Soper Gorden, unpublished data), oxygen available to roots, and nutrient availability, such as soluble nitrogen (Hopkins and Hüner 2004, Huberty and

Denno 2004). Any of these or other microhabitat features may have been the underlying cause of the negative correlation between soil moisture and I. capensis reproduction in this study.

Lack of non-additive effects. There was no interaction between light and fertilizer

(Table 1.2) for any response, even though there are many ways such an interaction could occur. Plants are frequently light- or nutrient-limited (Hopkins and Hüner 2004). If a plant is limited by one abiotic factor, then adding or removing a second non-limiting abiotic factor may have no effect on the plant until the critical level of the first abiotic factor has been reached, leading to non-additive effects. Other systems have shown interactions between shade and fertilizer. For example, fertilized Inga vera plants had more herbivory than non-fertilized plants in the sun, but not in the shade (Kersch and

Fonseca 2005). In both Liriodendron lulipifera and Cornus florida, chemical defenses

(total phenols, condensed tannins, and hydrolysable tannins) were unaffected by fertilizer addition in shade treatments, but total phenols significantly increased and both condensed tannins and hydrolysable tannins significantly decreased in sunny, fertilized fields (Dudt and Shure 1994). As I. capensis prefers shady habitats (Leck 1979, Eastman 1995), it

seems unlikely that plants were light limited; the general lack of shade treatment effects supports this observation. The lack of significant fertilizer by shade interactions may be due to the general lack of any shade effects, precluding interactions with fertilizer. It may be interesting to test how universal the lack of non-additive effects between light and soil nutrients is among shade-tolerant plants.

Conclusions. This study demonstrated the effects of multiple abiotic factors on a network of mutualist and antagonist plant-insect interactions. Specifically, soil nutrients strongly affected all interactions individually, and increased the ratio of antagonist to mutualist floral visits. Floral and leaf antagonists both responded positively to soil nutrients. Despite the indirect negative effects of increased antagonisms, fertilized plants still had higher growth and reproduction than unfertilized plants, suggesting higher tolerance in favorable conditions. In this system, positive direct effects of abiotic factors on plants may outweigh any negative indirect effects via insect interactions, suggesting selection for plants to grow in high-nutrient soils.

Tables

Table 1.1. Loadings from significant principle components from principle component analysis on logical subsets of plant and flower data, including total variance explained by each PC. PLANT SIZE FLOWER SIZE FLOWER NUMBER PC 1 PC 2 PC 1 PC 1 Height 0.381 0.386 Length 0.415 CH Flowers 0.557 Biomass 0.470 -0.156 Spur Length 0.379 CL Flowers 0.574 Leaf Area 0.390 0.264 Lip Height 0.398 CH Start Date -0.600 Branching 0.137 0.508 Lip Width 0.412 Nodes -0.146 0.670 Opening Height 0.422 Internode 0.469 -0.164 Opening Width 0.421 Length Lifespan 0.470 -0.152 Variance 0.611 0.209 0.730 0.806

Table 1.3. Summary of significant results from individual ANOVAs following significant MANCOVAs (Table 1.2). The full ANOVA model included main effects of fertilizer and shade treatments, their interactions, and soil moisture as a covariate. For clarity, only significant effects are showna. Individual ANOVA DF MANCOVA Factor F P response (N, D) PC 1 of plant size PCA 4, 33 15.06 0.0005 (overall size) PC 2 of plant size PCA 4, 33 10.64 0.0026 Fertilizer (branching) PC1 of flower number PCA Plant Growth 4, 33 16.85 0.0002 (more, earlier flowers) CH to CL Ratio 4, 33 10.59 0.0026 PC1 of flower number PCA Soil 4, 33 11.85 0.0016 (more, earlier flowers) Moisture CH to CL ratio 4, 33 7.37 0.0105 Floral Male nectar volume 4, 31 6.95 0.0130 Fertilizer Rewards Female nectar volume 4, 31 5.77 0.0225 Floral Number of Bombus spp. 4, 32 9.56 0.0041 Fertilizer Visitors Number of nectar thieves 4, 32 4.97 0.0329 Bombus spp. Fertilizer Proportion of flowers visited 4, 25 7.82 0.0098 Behavior A. mellifera Flowers per visitor 4, 19 6.26 0.0217 Fertilizer Behavior Proportion of flowers visited 4, 19 4.94 0.0385 Nectar Thief Fertilizer Proportion of flowers visited 4, 23 9.40 0.0055 Behavior Herbivore diversity 4, 31 4.41 0.0441 Proportion florivory 4, 31 13.32 0.0010 Herbivory Fertilizer Number of flower galls 4, 31 5.46 0.0260 Date of first flower gall 4, 31 6.32 0.0174 CH seeds per fruit 4, 32 19.34 0.0001 Fertilizer Number of fruits 4, 32 27.27 < 0.0001 Plant Mature fruits 4, 32 22.65 < 0.0001 Reproduction Number of fruits 4, 32 10.33 0.0030 Soil Mature fruits 4, 32 11.16 0.0022 Moisture CH seed mass 4, 32 5.11 0.0308 a Complete list of responses tested: Plant Growth. PCs 1 and 2 of plant size PCA (representing overall size and branching, respectively), PC1 of flower number PCA (representing more, earlier flowers), CH to CL flower ratio, PC1 of flower size PCA (representing overall size); Floral Rewards. male and female phase nectar volume, male and female phase nectar sugar concentration; Floral Visitors. number of visits Bombus spp., A. mellifera, and nectar thieves; Bombus spp. Behavior During Visits. proportion of flowers probed, probes per pollinator, time per probe; A. mellifera Behavior During Visits. proportion of flowers probed, probes per pollinator, time per probe; Nectar Thief Behavior. proportion of flowers probed, probes per thief; Herbivory. number of herbivores, herbivore diversity (number of functional groups present), leaf damage, proportion florivory, flower damage, number of flower galls, phenology of flower galls; and Plant Reproduction. number of fruits, number of mature fruits, seeds per CH and CL fruit, seed mass for CH and CL fruits.

Figures

2 A Fertilizer

*** No Fertilizer 1.5 *** 1 ** ** 0.5 0 -0.5

scorePCA or ratio -1 -1.5 Plant Branching Number CH:CL -2 Size of Flowers Ratio

Fertilizer 25 B No Fertilizer *

l) 20 μ

15 * 10

Nectar volume ( 5

MaleMale Phase Phase Flower FemaleFemale Phase Phase

Flower FlowerFlower

Figure 1.1. Effect of fertilizer on plant growth and floral rewards in Impatiens capensis. (A) Plant size (PC1 from the plant growth PCA), branching (PC2 from the plant growth PCA, representing number and length of internodes), number of flowers (PC1 of flowering PCA), and ratio of CH (outcrossing) to CL (selfing) flowers. (B) Nectar volume of male and female phase flowers. Error bars show standard error. * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

A *** 38 *** ** **

3 A ** 2.5

spp. Visits 1.5

1 Bombus 0.5

0 Fertilizer No Fertilizer

1.6 NS 1.4 B 1.2

Visits 1 0.8 0.6

0.4 Apis mellifera Apis 0.2 0 Fertilizer No Fertilizer

3.0 C *

2.0

1.0 NectarVisitsThief

0.0 Fertilizer No Fertilizer

Figure 1.2. Effects of fertilizer on Impatiens capensis floral visitors during 15 minute observations. Visits per plant by: (A) Bombus spp. pollinators, (B) Apis mellifera pollinators, and (C) nectar thieves. Error bars show standard error. * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

2.5 0.4 A * B *** 2 0.3 1.5 0.2 1

Proportion 0.1 0.5 DamagedFlowers

of Herbivores Diversity 0 0 Fertilizer No Fertilizer Fertilizer No Fertilizer

2.5 270 C D * 265 * 2 260 1.5 255 1 250 245

Number of Galls of Number 0.5 240

Gall First of Date Julian 0 235 Fertilizer No Fertilizer Fertilizer No Fertilizer Figure 1.3. Effects of fertilizer on herbivory in Impatiens capensis. (A) The diversity of herbivores, measured as the number of functional groups found on plants. (B) Proportion of flowers damaged by florivores. (C) Number of flower bud galls. (D) Date of first flower bud gall appearance. Error bars show standard error. * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

3 *** A 2.5

2 1.5

Seedsfruitper (CH) 0.5

0 Fertilizer No Fertilizer

25 B Fertilizer *** 20 No Fertilizer

15 ***

10 Fruits (number) Fruits 5

0 Average Mature Average Total

Figure 1.4. Effects of fertilizer on Impatiens capensis reproduction. (A) Number of seeds per fruit. (B) Average mature and total fruits. Error bars show standard error. * = P < 0.05, ** = P < 0.01, *** = P < 0.001.

CHAPTER 2

INTERACTIONS BETWEEN FLOWER BUD GALLERS, POLLINATION, AND

PLANT REPRODUCTION

Abstract

Herbivores can have strong direct effects on plants by changing patterns of growth, physiology, and reproduction, as well as indirect effects by changing interactions with other antagonists or mutualists. However, there are very few experiments testing how galling insects affect other plant-insect interactions. For example, flower bud gallers attack flower buds and prevent them from making flowers or setting fruit, but may also present a visual or chemical cue that could affect other plant-insect interactions. I conducted two experiments to test the effect of Schizomyia impatientis, a specialist flower bud galler of Impatiens capensis, on leaf herbivory, pollination, and plant reproduction. The first experiment manipulated the visual cue of galls by manually removing or adding bud galls to plants that initially did or did not have galls. Plants that initially had galls also had increased pollination, but there was no effect of gall presence on any other response. The second experiment manipulated the occurrence of galls by supplementing galling insects or removing galls. In this experiment, I found no effect of galling insects on pollination or leaf herbivory. Both experiments indicate that flower bud galls do not act as a visual cue or influence plant physiology in a manner that affects pollinator choice. Additionally, there was no significant effect of galling on plant traits or any measure of plant reproduction, although damage to plants via gall or bud removal tended to decrease nectar production, increase nectar concentration, and make flowers

bigger. These results indicate that, although S. impatientis co-opts floral structures and creates a large signal, this galler does not substantially influence other interactions or plant fitness. Altogether, my results suggest we should not assume that insects that damage flowers are antagonists, since plants may have strong tolerance mechanisms.

Introduction

Most herbivores have direct negative effects on the plants they feed on, including reduced photosynthesis (Delaney et al. 2008), changes in plant architecture (de Waal et al. 2012), allocation of resources to defense instead of reproduction (Coley 1986), water loss (Aldea et al. 2005, Nabity et al. 2009), and decreased biomass (Bigger and Marvier

1998). Herbivores can also have indirect effects on their host plants. Herbivore damage may affect damage by subsequent herbivores by changing attractiveness or inducing changes in allocation and defense (e.g., Zangerl and Rutledge 1996, Wackers and

Bezemer 2003, Poelman et al. 2008). Damage to plants may also make them less attractive to mutualists, such as pollinators or seed dispersers, by changing visual or volatile cues (e.g., Freeman et al. 2003, Kessler and Halitschke 2009, Lucas-Barbosa et al. 2011, Whitehead and Poveda 2011) or rewards (Adler et al. 2006, Brody and Irwin

2012). Herbivore damage may also attract predators, such as ants or parasitoid wasps, through changes in volatile emissions (Kessler and Baldwin 2001, Dicke and Baldwin

2010).

Galling insects are specialist herbivores that cause a persistent, swelling growth (a

“gall”) of plant tissue, inside of which galler larvae grow and feed (Crespi et al. 1997).

Galls are prevalent in most plant families and on many important crop species (Crespi et al. 1997). They can reduce plant growth and health, but their effects on plant fitness are still ambiguous (Fay et al. 1996, de Souza et al. 2006). Like many species, the effects of galling insects on plants may be conditional or context-dependent (e.g., Bronstein 1994).

Most gallers are species-specific and tissue-specific on their host plant (Crespi et al.

1997). In addition to changing plant morphology by inducing galls (Gagné 1989,

Shorthouse and Rohfritsch 1992), galling insects can also manipulate plant physiology.

Galls are strong sinks and may compete with plant sinks such as new leaves, flowers, or fruit for resources (Yukawa 2000, Leege 2006). Some galling insects manipulate nutrient transport in the vascular system (Larson and Whitham 1991, Bagatto and Shorthouse

1994, Hartley 1998), shunting nitrogen, carbohydrates, and other necessary resources from other plant parts to galls. Other galling insects directly manipulate plant defensive chemistry, either keeping the levels of defense in galls low to reduce exposure or storing them in high concentrations in the gall’s outer layer as a protection against herbivores and predators (Hartley 1998, Pascual-Alvarado et al. 2008). Gallers may also be able to alter or even suppress a plant’s volatile defenses (Tooker and De Moraes 2007).

Galling insects can have a wide range of direct effects on plants, from negative to neutral. Galls can reduce leaf area (de Souza et al. 2006), reduce photosynthetic rates

(Larson 1998, Yukawa 2000), stunt plant growth (Yukawa 2000, Tooker and De Moraes

2007), change plant architecture (Yukawa 2000), and reduce plant lifespan (Yukawa

2000). Galls can also lead to slower budburst and earlier leaf senescence (Larson 1998,

Foss and Rieske 2004). However, not all galling insects have a negative effect on plants,

and few studies have tested the effects of gallers on plant reproduction. Some galling insects can act as mutualists as well as herbivores; there are at least three examples of galling insects acting as pollinators, including the well-known fig wasps which gall fig ovaries while pollinating (Sakai 2002, Luo et al. 2010).

Although galling insects have ample opportunities to alter defensive or attractive traits, the effect of gallers on subsequent plant-insect interactions is almost unknown.

Galls may attract herbivores because of their high apparency and increased nutritional value (Abrahamson and Weis 1987, Schultz 1992). However, gall tissue also often has increased plant defensive compounds, such as phenolics, which may decrease herbivory

(Hartley 1998, Foss and Rieske 2004, Pascual-Alvarado et al. 2008). For example, ungalled branches and ungalled trees of Quercus palustris had significantly higher herbivory than galled branches or trees (Foss and Rieske 2004), which could be a function of galling inducing systemic changes in defenses, or of defenses accumulated in the galls themselves. There can also be competition between galling insects on the same plant (Kaplan et al. 2011).

Galling insects can change many traits that pollinators use as cues. Galls on branches or plants may lead to fewer flowers (Fay and Hartnett 1991, Fay et al. 1996,

Leege 2006), alter flowering phenology (Fay and Hartnett 1991), reduce pollen production (Rodriguez-Rajo et al. 2005), change volatile cues (Tooker and De Moraes

2007), and reduce plant height (Fay and Hartnett 1991, Fay et al. 1996), all of which may reduce pollinator preference. However, some galls produce nectar (Seibert 1993) or provide a showy visual cue (Stone and Schönrogge 2003) that could attract pollinators

instead. I am not aware of any published studies that test whether non-pollinating galling insects affect pollination, although one unpublished study found that plants with intermediate levels of galling had the highest pollination (John Tooker, personal communication).

Schizomyia impatientis (Cecidomyiidae) is a specialist flower bud galler that attacks flower buds on Impatiens capensis and I. pallida (Hummel 1956, Eastman 1995).

They are common in New England, and can occur in large numbers on Impatiens plants, yet little is known about how they affect subsequent plant-insect interactions or plant reproduction. These insects gall outcrossing chasmogamous (CH) flowers while selfing cleistogamous (CL) flowers are rarely galled; flower buds that are galled cannot produce fruits (Hummel 1956). The galls are large (averaging 7mm diameter, max over 20mm diameter; NLSG, unpublished data), roughly spherical, and often maintain the red or yellow colors of flowers (NLSG, personal observation). Since galls remain suspended from the pedicel that would have held the flower, they easily bob in the wind, adding to their conspicuousness. Galling insects may affect subsequent plant-interactions in one of two ways: 1) by changing visual cues on plants through the addition of large, colorful galls that may be attractive to pollinators or other insects, or 2) by preventing buds from becoming flowers and co-opting resources, leaving fewer resources for attractive traits, chemical defenses, and/or reproduction.

I conducted two experiments addressing the effects of S. impatientis galls on I. capensis plants. The first experiment compared the effect of initial gall presence or absence with manually manipulating galls as visual cues, to determine which factor

influenced pollination and leaf herbivory. Upon finding the initial presence of galls had the largest effect on pollinators, I conducted a second experiment manipulating galling on plants to determine whether galls were directly responsible for effects on pollinators, leaf herbivores, other floral antagonists, plant traits, and plant reproduction. These experiments addressed three questions: 1) do galling S. impatientis effect subsequent plant-insect interactions, including pollination and leaf herbivory? 2) do S. impatientis galls act as a visual cue to either pollinators or leaf herbivores? and 3) do S. impatientis gallers affect growth or reproduction of I. capensis?

Methods

Manual Manipulation of Galls. Field Site and Experimental Design. On 31

August 2007, 14 blocks were chosen within large natural populations of I. capensis in the

Prospect Hill area of Harvard Forest, Petersham, MA (N 42° 32’, W 72° 11’). Within each block, two plants were selected with at least two S. impatientis galls each, and two plants lacking galls. Within each block, one of the plants with galls was randomly chosen to have all galls removed. One of the plants without galls in each block was randomly chosen to have two galls sewn onto the stem using a #18 needle and transparent nylon thread, allowing galls to hang from thread similar to natural galls from pedicels; this resulted in a puncture in the gall, but only a knot of thread around the stem. Galls that remained naturally on one plant per block were punctured with the same needle to present equivalent damage. The other non-galled plant was not manipulated, as a control.

Plants that had galls before treatments were applied were considered plants with “initial

galls;” plants with galls that could be seen by insects, either naturally or tied on, were considered plants with “visible galls.” Altogether, this was a fully factorial experiment manipulating initial gall (2 levels) and visible gall (2 levels), resulting in four treatment combinations.

Responses. For pollinator observations, each plant was pruned before observations to have only two open flowers, chosen to be of similar age and height from the ground, to provide similar attractiveness to pollinators. On plants that had no flowers removed, two leaves were removed to simulate equivalent damage to the plant. Each block was observed for fifteen minutes each day for three days. The number of pollinator probes (defined as the pollinator entering the flower), pollination rejections (defined as a pollinator that spent less than one second at the flower), pollinator taxa, total time each pollinator spent at a flower during probes, and nectar robbers and/or nectar thieves was recorded. Nectar robbers take nectar by piercing the corolla, while nectar thieves visit flowers normally but are too small to regularly transfer pollen between flowers (Irwin et al. 2010). If one plant in a block did not have two healthy flowers on one day, no data was taken for that block on that day.

Herbivory was measured in three ways. Before treatments were applied, two undamaged leaves on each plant were randomly selected to measure herbivory. After treatments had been applied for three days, the damage to these leaves was tallied on a qualitative scale: 0 = no damage, 0.5 = unhealthy/chlorosis, but no leaf area missing, 1 =

<10% leaf area missing, 2 = >10% leaf area missing. Next, plants were censused three days after treatments were imposed, identifying and counting all insect herbivores.

Individual insects were counted to calculate density except in the case of aphids and scales, which were assessed as presence/absence, and the number of species present was counted for species richness.

Statistical Analysis. Three MANOVAs were performed to test the effects of gall treatments (PROC GLM; SAS Institute 2003): (1) Bombus spp. pollinator visitation, including number of probes, number of rejections, and nectar robbing by Bombus spp.,

(2) A. mellifera visitation, including number of honeybee probes and rejections, and (3) herbivory, including the herbivore damage score, herbivore density, and herbivore species richness. Additionally, I ran two ANOVAs to test time spent per probe by

Bombus spp. pollinators and by A. mellifera; because time per probe can only be recorded for plants that received pollinator visits, the sample sizes for these two responses were too small to include in the visitation MANOVAs. Each analysis included effects of initial galls, visible galls, their interaction, and block, all as fixed effects. Plot 12 was an outlier for pollination; removing it from the data set improved normality but did not qualitatively change the results, so was left in analyses. Plot 10 was an outlier for herbivore number, with a large aggregation of mating Agrosternum hilare, likely attracted by the sex pheromones of A. hilare which also act as aggregation pheromones

(Aldrich et al. 2007). Although A. hilare is a frequent herbivore on I. capensis, this unusually large number of individuals flew away after mating without feeding on the plant; therefore, plot 10 was excluded from statistical analyses of herbivory.

Galling Manipulation Experiment. Plant Propagation and Field Site. Seedlings were collected from a natural population of Impatiens capensis at Hampshire Farm on

Hampshire College, Amherst, MA (N 42º 19’ W 72º 31’) on 26 April 2009 and transplanted into 10 cm pots (Fafard #2 potting soil, Conrad Fafard, Inc, Agawam, MA) on 12 May 2009. Plants were maintained in a greenhouse and watered daily until they were about 0.5 m tall.

On 8 June 2009, 80 plants were planted at Hampshire Farm near a large natural population of I. capensis known to have S. impatientis. Plants were all at least 1.5 m away from each other, and were located in the understory of a wet, maple-dominated forest where S. impatientis galls are common.

Treatments. On 18 September 2008, several hundred naturally-formed galls were collected into cups of water. Schizomyia impatientis larvae emerge from galls that are suspended in water and are able to survive in water for several weeks without harm

(Hummel 1956). On 28 September 2008, the larvae were transferred to the moistened soil surface (Fafard #2 potting soil, Conrad Fafard, Inc, Agawam, MA) of 20 plastic

22.86 cm diameter (15.24 cm tall) pots where they were allowed to pupate naturally. In total, there were approximately 75 S. impatientis larvae in each of the 20 pots. To keep predators and competitors of S. impatientis out of the pots, each pot had the inside lined with two layers of nylon tulle before soil was added, and was topped with two layers of tulle tied in place. Sixty additional pots of the same size were set up in the same manner, with moistened soil but no galler larvae.

All 80 pots were buried to the lip of each pot on 11 November 2008 at Hampshire

Farm, along the northwest edge of a forest with a natural population if Impatiens capensis. All pots overwintered buried at the site.

On 16 June 2009, overwintered pots were dug up and re-buried with one next to each of the 80 experimental plants, which were randomly assigned to treatments. The pots with gall larvae were the “gall supplement treatment;” S. impatientis adults are weak fliers, and nearly always gall the plant nearest to their emergence (Hummel 1956).

Twenty plants served as a “control,” with no manipulation. Twenty plants were in a “gall removal treatment,” and any galls appearing on them were removed weekly by cutting their pedicel with dissecting scissors. The final twenty plants served as a “damage control treatment.” Each of these plants was paired with a plant of comparable height from the gall removal treatment, and each time galls were removed from the gall removal treatment, an equivalent number of flower buds were removed from the damage control plants. Flower buds were removed to replicate the effect of gallers, which co-opt flower buds, in an attempt to remove similarly strong resource sinks. Rarely (n=3), a plant in the damage control treatment would not have flower buds to remove; in this case, a lateral meristem was removed instead.

Responses. To test the effectiveness of treatments, the number of galls was measured weekly. I measured plant growth approximately every other week, including height, number of branches, and leaf area of four fully expanded leaves per plant

(estimated as leaf length times width, which is highly correlated with area: n=42, r2=0.998, P<0.0001). I also counted the number of CH and CL flowers per plant. I estimated flower size using total flower length, nectar spur length, lip size (height x width of the lip), and corolla opening size (height x width of the opening) on up to three flowers per plant approximately every other week. For up to three flowers per plant, I measured

pollen production and nectar traits in flowers that were bagged as buds with polyester mesh (cut from Fibe-Air greenhouse sleeves, Kleen Test Products, Milwaukee, WI).

Pollen production was determined by collecting whole androeciums into individual microcapillary tubes, dried at 45 °C for 24 hours, suspended in 70% ethanol, and counted on a hemocytometer six times per androecium for an average value. Nectar was sampled from up to three male-phase flowers per plant using microcapillary tubes, first collecting any nectar in the corolla tube, then nectar squeezed from the cut tip of the spur. Nectar volume was measured using microcapillary tubes and sugar concentration was measured with a handheld refractometer (Fisher Scientific, Pittsburgh, PA).

Herbivory was measured nine times during the summer as the percent leaf damage on the four most fully expanded leaves, the total number of herbivores present, and the richness of herbivore functional groups. Florivory was measured five times during the summer as the proportion of flowers with florivore damage and the percent floral tissue missing. Floral visitor surveys were conducted on eight days with 15 minutes per plant per survey, for a total of 90.5 hours of floral visitor observation. I recorded the number of floral probes, the identity of floral visitor (including whether they were pollinators, nectar robbers, or nectar thieves), proportion of flowers probed per visit, and time per flower probe.

Plant reproduction was estimated by counting the number of fruits approximately twice a week, which is highly correlated with total fruit production (Soper Gorden and

Adler, in review), beginning two weeks after treatments were applied (early enough to record all effects of treatments on fruits; Waller 1979) and lasting until the first hard frost

of the fall (plant death). I also collected mature fruits during surveys to measure seeds per fruit and seed weight. All reproduction methods were carried out for both CH and

CL fruits. Finally, I estimated the proportion of CH versus CL fruits to test whether flower bud gallers shifted plant mating system.

Statistics. Nine MANOVA tests (PROC GLM; SAS Institute 2003) were performed to test the effect of gall manipulation treatment on galling (to assess effectiveness of treatments), plant growth, flower size, floral rewards, floral visitors, pollinator behavior, nectar thief behavior, herbivory, and plant reproduction.

Galling included the date the first gall appeared and the number of galls. Number of galls was log transformed for normality.

Plant growth was measured as plant height, total aboveground dry biomass, number of branches, average leaf area (calculated as leaf length times width), and number of CH and CL flowers. Three plots were missing biomass values due to early die-off and were therefore dropped from the analysis, and biomass was log transformed for normality. CH flowers were square root transformed.

Flower size included total flower length, spur length, lip area, and corolla opening size. Spur length was x2 transformed for normality. The floral rewards analysis included pollen production, nectar volume, and nectar concentration. Nectar volume was square root transformed for normality.

Floral visitation was measured as the number of visits by pollinators and nectar thieves per 15 minutes. Pollinator visits were log transformed and nectar thief visits were square root transformed to improve normality. Pollinator behavior was measured as

flowers probed per pollinator, proportion of flowers probed per visitor, and average time per flower probe. Nectar thief behavior was measured as flowers per pollinator and proportion of flowers visited; time per probe could not be included because there were several species of thieves, each of which probes for a different amount of time. We measured both number of visitors and proportion of flowers probed per visitor as two different measures of visitation; the first gives information about the ability to attract pollinators to the plant, while the second provides information on how rewarding pollinators perceive flowers on that plant to be. Probes per pollinator and per nectar thief were both log transformed.

Herbivory was measured as average leaf damage, total number of herbivores, herbivore diversity, proportion of flowers with florivore damage, and percent flower tissue missing from damaged flowers. Leaf damage, total herbivores, and percent flower damage were log transformed, and proportion of flowers with florivore damage was square root transformed.

We measured plant reproduction as the number of CH and CL fruits, average number of seeds per CH and CL fruits, average weight per seed for CH and CL fruits, and the ratio of CH to CL fruits. Number of CL seeds per fruit and number of CH fruits were both log transformed to improve normality.

Results

Manual Manipulation of Galls. Only Bombus spp. and Apis mellifera visited I. capensis flowers, with over six times as many Bombus spp. visits as A. mellifera.

Bombus spp. were also the only species observed robbing I. capensis, and only robbed flower buds. No nectar thieves were observed. Initial gall plants had significantly longer bumblebee probe times (F1, 35=8.36, P=0.007; Figure 2.1a). Initial gall also had a marginally significant effect on honey bee pollination, while visible galls only had a marginally significant effect on honey bee pollination (Table 2.1). Plants with initial galls had significantly more probes by honey bees (ANOVA: F1, 36=4.45, P=0.04; Figure

2.1b). Plants with initial galls also had 2.5 times longer honey bee visits, 1.5 times more robbing, and 25% fewer bumblebee rejections. These effects were not significant, but are consistent with a pattern suggesting that overall floral visitors preferred plants with naturally occurring galls. Plants with visual galls tended to have more honey bee rejections (ANOVA: F1, 36=3.17, P=0.08; Figure 2.1c). There was no significant effect of either original or visual gall treatments on herbivory, and no significant effect of the initial gall by visible gall interaction on any response measured (Table 2.1). Insect behavior often varied with block (Table 2.1), including the number of honey bee probes

(ANOVA: F1, 36=3.80, P=0.0009), number of honey bee rejections (ANOVA: F1,

36=2.28, P=0.03), honey bee visit length (ANOVA: F1, 6=8.03, P=0.01), number of nectar robbers (ANOVA: F1, 36=2.39, P=0.02), herbivore density (ANOVA: F1, 38=2.42,

P=0.02), and herbivore richness (ANOVA: F1, 38=2.58, P=0.01).

Gall Manipulation Experiment. Treatment Effectiveness. Gall treatment significantly affected the number of galls per plant, but did not affect when flower galls first appeared (Tables 2.2, 2.3). The gall supplement treatment had significantly more

galls than the gall removal treatment, with the two controls showing intermediate numbers of galls (Figure 2.2).

Effects on Plant Traits. The gall manipulation treatments had no effect on any measure of plant growth or number of CH or CL flowers (Table 2.2). Out of the four measures of flower size, only corolla lip size had a marginally significant effect (Table

2.3), with the gall removal treatment having a larger lip area than the other three treatments (Figure 2.3a). There was no effect of treatment on pollen production, but flowers from the damage control treatment produced significantly less nectar that was marginally more sugar-rich than the unmanipulated plants, with the gall treatments intermediate (Table 2.3, Figure 2.3b, c).

Effects on Insects. Bombus spp. were the only pollinators observed; nectar thieves included halictids and other small bees. I did not witness any nectar robbing at this site during this year. Treatment did not affect the number of visits by either pollinators or nectar thieves or any measure of pollinator behavior during pollination (Table 2.3).

While there was no effect on the number of flowers nectar thieves visited, plants in the gall supplement treatment tended to have a smaller proportion of flowers visited by nectar thieves than the damage control treatment, with the unmanipulated plants and gall removal plants showing intermediate levels of nectar thieves (Table 2.3, Figure 2.4a).

There was no significant effect of treatment on leaf herbivory or percent flower damage, but the gall supplement treatment tended to have the lowest proportion of damaged flowers, while the damage control treatment had the highest proportion (Table

2.3, Figure 2.4b).

Effects on Plant Reproduction. There was no significant effect of gall treatment on any measure of CH or CL reproduction, or on the ratio of CH to CL reproduction

(Tables 2.2, 2.3).

Discussion

Plants that initially had flower bud galls also received more pollination, regardless of whether those galls were still on the plant when pollinators were observed.

Bumblebees spent significantly more time per flower and honey bees probed marginally more flowers on plants that initially had galls (Figure 2.1); every measure of pollinator preference increased on plants with initial galls, although most were not significant due to low sample size and high variance. This suggests that pollinators prefer naturally galled plants. One explanation for this result is that galling insects induce changes that make plants attractive. Alternately, because the presence of initial galls was not randomly assigned to plants, it is also possible that both gallers and pollinators prefer the same plants.

While there are instances of specialized relationships where galling insects act as pollinators, therefore affecting plant pollination directly (Sakai 2002, Luo et al. 2010), I know of no studies testing how galling insects affect other pollinators. Several studies have found that galling insects change traits that may affect pollinator choice, such as number of flowers, flower size, nectar production, and plant height (Leege 2006).

Additionally, galling insects have frequently had negative effects on plant reproduction

(Yukawa 2000). In fact, several studies have found that galls compete with other sinks

such as flowers and fruits for resources (Larson and Whitham 1991, 1997). Therefore, it seems likely that galling insects have the potential to change pollinator preference.

However, it is unintuitive and unexpected that a non-pollinating galling insect would increase pollinator visitation, since gallers generally reduce resource availability and change traits that would make plants less attractive.

One of the most obvious changes S. impatientis makes to I. capensis plants is the visual addition of the galls. These large, mobile galls often maintain some of the colors of the flowers themselves, with yellow, orange, and light red patches on the gall surface.

These traits make the galls conspicuous and may even help them mimic how flowers look. However, the visual gall treatment only had one marginal effect, leading to an increase in the number of rejections by honey bees (Figure 2.1c), suggesting visual cues are not the cause of the increase in pollinator visitation on galled plants.

The gall manipulation experiment allows me to distinguish correlations between pollinator and galling preference from causative effects of galls on plant reproduction by directly manipulating galling to randomly assigned plants. I found no effect of galling insects on plant pollination. This, combined with previous results showing that galling insects and pollinators both prefer plants with plentiful soil nutrients (Soper Gorden and

Adler, in review), suggests that galling insects and pollinators may both prefer healthy or large plants. A similar trend may happen in other systems as well. For example, in Salix spp., the same volatiles that attract pollinators also attract galling insects (Kehl et al.

2010). In species that gall flowers, it is common to find more galls on plants with more flowers (Hummel 1956, Sakai 2002, Kehl et al. 2010); similarly, pollinators often prefer

plants with more flowers (Schmid-Hempel and Speiser 1988, Brody and Mitchell 1997,

Ishii et al. 2008, Kilkenny and Galloway 2008). Overall, this study provides little evidence that these bud gallers make plants more attractive to pollinators in spite of the large, colorful galls produced. Rather, results suggest that galling insects and pollinators prefer similar plants.

Although galling insects are known to act as strong sinks, can directly change plant defense and nutrition traits, and could potentially compete with other insect herbivores (Larson and Whitham 1991, Bagatto and Shorthouse 1994, Crespi et al. 1997,

Yukawa 2000, Leege 2006), I saw no effect of gall rearrangement or gall manipulation on leaf herbivory (Tables 2.1, 2.2). Galling insects can interact with plants very differently than leaf herbivores. While chewing damage initiates induced responses via the jasmonic acid (JA) pathway, galling insects induce the salicylic acid (SA) pathway, like pathogens and sucking insects (Larson 1998, Walling 2000, Ollerstam and Larsson 2003). By inducing the SA pathway, galling insects may be more likely to affect sucking than chewing insects, and most of the herbivores found on I. capensis over the two years of this study were chewing leaf herbivores, which may not respond to any induced changes.

Alternately, since many galling insects act as strong sinks and have the ability to control the host plant’s physiology (Larson and Whitham 1991, Bagatto and Shorthouse 1994,

Crespi et al. 1997, Yukawa 2000, Leege 2006), it is possible that S. impatientis prevents induced responses that could affect leaf herbivores.

The gall manipulation treatments affected three flower traits, leading to decreased nectar volume and increased flower lip petal size and nectar sugar concentration in the

gall removal and damage control treatments (Figure 2.3). Effects were similar in the gall removal and damage control treatments, suggesting that removing flower tissue (either as galls or as flower buds) causes changes rather than gall-specific effects. Plants that had tissue removed had significantly less nectar that was marginally more sugar-rich.

Damage to plants, on leaves or flowers, can lead to significant water loss (Aldea et al.

2005, McCall and Irwin 2006, Nabity et al. 2009), which may have caused the reduction in nectar volume. Since nectar volume is often inversely related to sugar concentration

(e.g., Carroll et al. 2001, Petanidou et al. 2006), this could explain both nectar effects.

Alternately, damage may have caused a change in resource allocation. For example, damaged leaves often become strong sinks until they are repaired (Nabity et al. 2009), using resources that could otherwise be used for nectar. The marginal increase in lip petal size in the gall removal treatment, combined with reduced nectar, suggests that damaged plants may increase resources in attracting pollinators and decrease resources for rewarding pollinators. Since pollinators of I. capensis may not be able to discern how much nectar is in flowers without visiting (Rust 1979), this may be an effective way of keeping pollinator visitation high with fewer resources.

Gall supplementation significantly decreased the proportion of flowers affected by floral antagonists. The gall supplement treatment had a significant lower proportion of flowers visited by nectar thieves and florivores than the damage control treatment

(Figure 2.4). This change in proportion of flowers visited is not due to a reduction in flower number, since galling treatments did not affect the number of flowers per plant

(Table 2.3). Instead, it seems that plants with supplemented galls had a lower proportion

of flowers visited by floral antagonists, despite having the same number of nectar thief visitors as other treatments (Table 2.3), suggesting that florivores and nectar thieves left a plant with supplemental galls after probing fewer flowers than a plant without galls. This may mean that flower bud gallers make flowers less attractive to other floral antagonists, perhaps by changing allocation, nutrition, or defense properties of flowers. Alternately, since gall or flower bud removal damage decreased nectar volumes (Figure 2.3b), nectar thieves may visit more flowers on damaged plants to collect an equivalent amount of nectar. Previous studies have shown that leaf or flower damage can reduce florivory

(McCall 2006, McCall and Karban 2006); however, the current study’s results found the opposite effect. Previous research in I. capensis suggests that florivores prefer plants with low levels of flower damage (Soper Gorden and Adler, in prep), so this may explain the increase in florivory on the damage control plants, which had small numbers of flower buds removed.

Overall, there was no significant effect of the gall manipulation treatments on any measure of plant reproduction (Table 2.3), even though S. impatientis directly removes

CH flowers by galling them (Hummel 1956). This suggests that either plants can compensate for the lost flowers, or that flower bud galling occurs at a low enough rate to have no effect on plant reproduction. Impatiens capensis can tolerate other antagonisms, including leaf herbivory (Steets 2005, Steets et al. 2006a, Steets et al. 2006b), competition (Steets et al. 2006b), and florivory (Soper Gorden and Adler, in prep), by shifting towards a greater selfed reproduction. However, in plants with flower bud gallers, there was no increase in CL flowers or CL reproduction, suggesting plants are not

shifting mating system to tolerate galling. Schizomyia impatientis galling occurred at below-average levels in 2009 (average of less than 1 gall per plant vs. 6.5 galls per plant in 2008; Soper Gorden, unpublished data). This was likely due to extremely rainy weather in June and July, with over twice as much precipitation as normal (June + July =

44.75 cm in 2009 vs. 20.65 cm average; National Weather Service 2012), leading to late flowering and early arrival of galling insects (Soper Gorden, in prep). Nonetheless, removing an average of less than 1 gall per plant may have little effect on plant reproduction. Repeating this experiment in a year with higher galling could demonstrate greater fitness effects.

A limitation of this study is the number of galls on plants in each treatment

(Figure 2.2). While gall supplement plants did have significantly more galls than the gall removal treatment, they did not have significantly more galls than the unmanipulated control with natural levels of galling. The overall trend, however, is as expected, with the gall supplement treatment having the most galls, the gall removal treatment having the fewest, and the two controls (unmanipulated and damage control) having intermediate levels of galls. It is likely that any significant differences between treatments were washed out by the rainy weather, low total levels of galling, and high gall mortality.

There may also be differences in plant susceptibility to galling based on genetics or resource availability. For example, other plant species can prevent galling insects from ovipositing, or can kill the gallers before they can form a gall (McCrea and Abrahamson

1987, Anderson et al. 1989, Abrahamson et al. 1991, Ollerstam and Larsson 2003).

Therefore, it is possible that plants with more galling insects ovipositing on them do not always end up with more galls.

Conclusions. Plants that initially had S. impatientis galls were more likely to be pollinated, even when those galls had been removed. However, when galling was manipulated by supplementing galling insects, there was no effect of galls on pollinators.

This, coupled with previous work showing that both gallers and pollinators prefer plants with high resources, suggests that S. impatientis do not alter plant attractiveness to pollinators but instead that both gallers and pollinators prefer similar plants. This suggests that plants with plentiful resources may experience conflicting selection pressures from both antagonists, such as galling insects, and mutualists, such as pollinators. However, in this research I found no evidence that galling insects affect plant reproduction. Instead, the effect of galling insects on plants may range from neutral to negative, depending on the intensity of galling, the availability of resources, and levels of pollination. Thus, local geographic or temporal variations may play a large role in determining the strength of selection galling insects on plant traits.

Tables

Table 2.2. MANOVA results testing the effects of gall manipulation on Impatiens capensis. Significant values (P<0.05) are in bold; values where 0.05

Table 2.3. Results from ANOVAs testing the gall manipulation treatments on plant traits, subsequent plant-insect interactions, and plant reproduction. Bold values indicate significance; italic values indicate marginal significance. MANOVA Factor df (n, d) F P Date of first gall 3, 63 1.46 0.24 Flower Galls Number of Galls 3, 63 4.54 0.006 Plant height 3, 76 0.17 0.92 Branches 3, 76 0.57 0.64 Plant Leaf size 3, 76 0.73 0.54 Growth CL flowers 3, 76 1.36 0.26 CH flowers 3, 75 0.33 0.80 Nectar spur length 3, 49 0.43 0.73 Total flower length 3, 49 2.07 0.12 Flower Size Petal lip size 3, 49 2.48 0.07 Corolla opening 3, 49 1.84 0.15 Pollen production 3, 30 0.78 0.51 Floral Nectar production 3, 29 3.30 0.03 Rewards Nectar sugar concentration 3, 29 2.59 0.07 Floral Pollinator visitors 3, 56 1.79 0.16 Visitation Nectar thief visitors 3, 56 0.45 0.72 Pollinator time per pollinator 3, 35 0.38 0.77 Pollinator Probers per pollinator 3, 35 0.61 0.61 behavior Proportion of flowers probed per pollinator 3, 35 1.48 0.24 Nectar thief Probes per nectar thief 3, 25 1.86 0.16 behavior Proportion of flowers probed per nectar thief 3, 25 2.89 0.05 Percent leaf damage 3, 75 2.03 0.12 Total number of herbivores 3, 75 0.12 0.95 Herbivory Herbivore species richness 3, 75 0.95 0.42 Percent florivory damage 3, 23 0.60 0.62 Proportion flowers damaged 3, 47 2.58 0.06 CH fruits 3, 76 0.89 0.45 CH seeds per fruit 3, 58 1.13 0.34 CH average seed weight 3, 57 1.69 0.18 Plant CL fruits 3, 76 1.43 0.24 reproduction CL seeds per fruit 3, 74 0.11 0.95 CL average seed weight 3, 74 0.92 0.43 Ratio of CH to CL fruits 3, 76 0.76 0.52

Figures

Figure 2.1. Pollination on Impatiens capensis plants comparing initial presence of galls and manual gall manipulations. Error bars show standard error.

1.6 a 1.4 ab 1.2

1.0

0.8 b ab 0.6

0.4 Average Number of GallsofNumber Average 0.2

0.0 Gall Supplement Unmanipulated Gall Removal Damage Control

Figure 2.2. Effects of gall manipulation treatments on the number of galls and size of galls on Impatiens capensis plants. Different letters indicate treatments that were significantly different from one another. Error bars show standard error.

Figure 2.3. Effects of Schizomyia impatientis galler manipulation on measures of Impatiens capensis flower attractiveness. Letters indicate significant differences between treatments; panels without letters were marginally significant only. Error bars show standard error.

Figure 2.4. Effect of Schizomyia impatientis galler manipulations on subsequent floral antagonists. Letters indicate treatments that are significantly different. Error bars show standard error

CHAPTER 3

FLORIVORY ALTERS PLANT-INSECT INTERACTIONS AND DECREASES

PLANT REPRODUCTION

Abstract

While leaf herbivory and its effects on plants and plant-insect interactions have long been studied, floral herbivory is less well understood, especially in regards to its community consequences beyond pollinators and its relationship with floral attractive or defense traits. Since flowers are critical for angiosperm reproduction, the direct and indirect effects of florivory on plants and their subsequent interactions may have important effects on plant fitness and selection for floral traits. Additionally, floral attractive or defensive traits may mediate interactions with floral antagonists or mutualists, and may be altered by floral damage. We manipulated floral damage in

Impatiens capensis to mimic florivory and measured effects on floral attractive traits, floral chemical defenses, visitation by insect mutualists and antagonists, and plant reproduction. We also examined relationships between early-season floral traits and interactions with several floral antagonists and mutualists, to shed light on traits that may mediate a range of floral interactions that can impact plant reproduction. Finally, we examined whether early-season floral traits consistent or plastic over the flowering season. Florivory significantly decreased plant reproduction and increased the proportion of selfed reproduction, especially at intermediate damage levels. Additionally, florivory increased subsequent natural florivory and decreased leaf herbivory and flower spider abundance, suggesting that floral damage induces susceptibility to subsequent florivory but induces systemic resistance in vegetative tissues after florivory. Floral anthocyanins

and condensed tannins were associated with reduced levels of many floral antagonisms, including florivory, nectar larceny, and flower spider abundance, suggesting these traits play a role in floral resistance. While many floral traits were correlated across the season, floral defenses were not, suggesting they are more plastic in response to environmental cues. Overall, our results suggest that florivory may have community consequences for other plant-insect interactions, including leaf herbivory, and shape the evolution of mating systems.

Introduction

Flowers are essential organs for angiosperm reproduction, and many plant species require pollinators to set fruit. An estimated 87.5% of all flowering plants are animal- pollinated (Ollerton et al. 2011), including many of our most important crop plants

(Losey and Vaughan 2006, Klein et al. 2007). In fact, there has been a long-standing suggestion that pollinators are responsible for the evolution of flowers to their current diversity of forms due to selection on floral attractive and reward traits (Darwin 1862,

Fenster et al. 2004).

However, there are many antagonists that can also be attracted to flowers. For example, florivores (herbivores that consume flowers) can be as common or more common than leaf herbivores (McCall and Irwin 2006). Florivores can directly reduce plant reproduction by damaging pollen or ovules (Krupnick and Weis 1999, Leege and

Wolfe 2002, McCall and Irwin 2006). In severe cases, florivores can cause the near collapse of plant populations (Washitani et al. 1996). Florivores can also have indirect effects on plant reproduction by changing how other insects, such as pollinators, visit

plants (McCall and Irwin 2006). For example, florivory can reduce nectar production

(Krupnick et al. 1999), flower size (Mothershead 2000), or flower symmetry (McCall

2008), all of which can alter attractiveness to pollinators or other floral antagonists.

While a several studies have tested how floral damage affects pollinator visitation

(Krupnick and Weis 1999, Botto-Mahan and Ojeda-Camacho 2000, Leavitt and

Robertson 2006, McCall 2008, Cardel and Koptur 2010, Sõber et al. 2010, Cares-Suarez et al. 2011), there have been very few experiments testing how florivores affect other plant-insect interactions, including other floral antagonists. By contrast, community consequences of leaf herbivory for subsequent antagonisms are well known

(Stinchcombe and Rausher 2002, Strauss and Irwin 2004, Van Zandt and Agrawal

2004a), and the species of herbivore that first damages a plant can have cascading impacts on the entire community of subsequent consumers (Van Zandt and Agrawal

2004a, b).

Although florivory can have strong impacts on plant reproduction and population dynamics, little is known about how plant traits influence florivore choice. In dioecious or gynodioecious plants, florivores preferentially damage male (Cox 1982, Wolfe 1997,

Ashman 2002) or hermaphrodite (Ashman 2002, Ashman et al. 2004, Asikainen and

Mutikainen 2005, Lu et al. 2012) flowers over female flowers. Shorter plants with lower flowers (Held and Potter 2004) or plants with smaller or less conspicuous flowers

(Ashman et al. 2004) may have reduced florivore preference. In fact, it has been hypothesized that many of the traits that attract pollinators will also attract florivores; although this has been demonstrated for other floral antagonisms, such as nectar robbing

(Maloof and Inouye 2000, Galen and Cuba 2001, Galen et al. 2011), this hypothesis is largely untested for florivores (McCall and Irwin 2006).

Although chemical defenses are most commonly studied in leaves, such defenses are also often present in flowers, sometimes at higher concentrations than in leaves

(Zangerl and Rutledge 1996, Hause et al. 2000, Strauss et al. 2004, Damle et al. 2005,

Frölich et al. 2006, Frölich et al. 2007), and may influence florivore visitation. Petals often contain the same chemical defenses as leaves (Euler and Baldwin 1996, Adler et al.

2001, Strauss et al. 2004, Irwin and Adler 2006). Many nectars also contain chemical defenses, such as alkaloids and phenolics (Adler 2000, Irwin et al. 2004, Adler and Irwin

2005, Adler et al. 2006, Johnson et al. 2006). Because flowers are intimately related to plant fitness, optimal defense theory predicts that flowers should be well-defended (Zangerl and Rutledge 1996, Begon et al. 2006, McCall and Irwin 2006, McCall and Fordyce 2010).

A few previous studies have found that floral chemical defenses can be induced following vegetative damage (Euler and Baldwin 1996, Wackers and Bezemer 2003,

Adler et al. 2006, McCall and Karban 2006), and one study has shown that floral damage induces resistance to subsequent florivores (McCall 2006). However, no one has examined whether chemical defenses are induced in flowers following floral damage despite possible implications for resistance, allocation, and selection on plant traits, or the consequences of such damage and induction for subsequent interactions beyond pollination. Additionally, no one has considered whether damage to flowers can change defenses in leaves, despite the fact that florivores can often be generalists that feed on leaves as well (McCall and Irwin 2006). While flowers are often strong physiological sinks, which may make sending systemic vascular signals difficult (Hopkins and Hüner

2004), flowers are still capable of sending some vascular as well as volatile signals to the rest of the plant.

While florivores are generally considered antagonists, their effects on plant reproduction can vary from neutral to negative (McCall and Irwin 2006). However, most studies have focused on total plant reproduction without considering the quality of the resulting seeds. For example, plants with florivory may have equivalent total reproduction but increased selfing (Penet et al. 2009). Selfing, through geitogamy or self-pollination, can have negative impacts on population dynamics and gene flow, beyond reducing seed production. For example, selfed fruits can have more limited dispersal (Schmitt et al. 1985) or fewer pathogens (Koslow and Clay 2007). Selfing can also alter plant-insect interactions, with sometimes strong consequences for offspring fitness. Inbred plants often have fewer (Walisch et al. 2012), smaller flowers (Andersson

2012) and smaller leaves (Walisch et al. 2012), and may produce fewer or different volatiles (Ferrari et al. 2006). These changes may alter the plant’s attractiveness to pollinators or herbivores. For example, in a wild gourd, inbred plants produced lower levels of volatiles, attracted fewer herbivores, and therefore were less infected by a fatal wilt disease transferred by herbivores (Ferrari et al. 2006, Ferrari et al. 2007, Du et al.

2008).

We manipulated florivory to measure effects of floral damage on floral attractive and chemical defense traits, other plant-insect interactions, and plant reproduction.

Additionally, we examined associations between early-season floral traits and several mutualist and antagonist floral interactions to shed light on the role of these traits in attraction or resistance to a range of interactions. Finally, we tested whether early season

floral traits predict late season floral traits, or are more plastic in response to seasonality or environmental cues. Overall, this study provides a comprehensive investigation of the consequences of floral damage for floral traits, interactions, and plant reproduction.

Methods

Study System. Impatiens capensis (Balsaminaceae) is an annual native herb that grows in partial shade and moist soil (Leck 1979). It has a mixed-mating system with both selfing cleistogamous (CL) and open-pollinated chasmogamous (CH) flowers. CH flowers are protandrous, spending their first ~36 hours in a male phase and their final ~12 hours in a female phase; each phase produces different amounts of nectar (Soper Gorden and Adler, in review), so all nectar was measured in male phase flowers for consistency.

CH flowers are pollinated mostly by Bombus sp. and Apis mellifera, and are incapable of selfing due to flower anatomy and phenology (Rust 1977, Schemske 1978, Eastman

1995, Steets and Ashman 2004); geitonogamy has been estimated at only 8.6% (Waller

1980). Both flower types produce capsule fruits with one to several seeds that dehisce explosively when mature. In Massachusetts, I. capensis generally germinates in late April or early May, CL flowers appear in May, and CH flowers last from mid July until mid

September. Seeds generally are not viable for more than one year (Simpson et al. 1985), resulting in little to no seed bank.

While pollinators are the only known mutualists to interact with I. capensis flowers, there are many antagonists. The flowers are robbed by several insect species

(including Bombus spp. and Vespula maculifrons), and visits by nectar thieves such as ants and halictids are common (Rust 1979, Eastman 1995, Young 2008). Popillia

japonica (Scarabaeidae) beetles and other generalist herbivores act as florivores, eating petal and sepal tissue (NLSG, pers. obs.). There is a species-specific Cecidomyiidae flower bud galler, Schizomyia impatientis (Hummel 1956). Finally, there are a variety of insect leaf herbivores, including true bugs (Hemiptera), grasshoppers (Orthoptera), katydids (Tettigoniidae), aphids (Aphidoidea), and P. japonica (Eastman 1995, Steets and

Ashman 2004).

Several chasmogamous floral traits could be attractive or defensive. In I. capensis, there is significant variation (range, mean±SE for all) in the number of flowers

(0-634 flowers, 131.4±5.8 flowers, unpublished data), flower size (e.g. flower length 16-

31 mm, 26.0±0.1 mm, unpublished data; Schemske 1978, Steets and Ashman 2004), nectar production (0-68 μl, 13.5±0.45 μl, unpublished data; Marden 1984, Lanza et al.

1995), and pollen production (800-20,600 grains, 7440±540 grains, unpublished data).

Flowers can also be presented at different heights (29.5-140.4 cm, 67.3±1.9 cm, unpublished data). Flower color can vary from entirely yellow (no red spotting) to almost entirely red (extensive red spotting; NLSG, pers. obs.). While little has been published on the role of I. capensis defensive chemistry mediating species interactions,

Impatiens spp. contain anthocyanins and condensed tannins. Anthocyanins are the most common flavonoid pigments, and can attract pollinators (Delpech 2000, Koes et al. 2005) and reduce florivore preference (Johnson et al. 2008). Anthocyanins are present in

Impatiens spp. leaves, flowers, and stems, and cause the variable red spots on the lip petals of I. capensis (Aras et al. 2007). Condensed tannins are common in plant species that have anthocyanins, including Impatiens spp. (Waterman et al. 1983). Although condensed tannins are usually measured as vegetative defenses, they are also found in

floral tissue and have the potential to deter florivores (Gautierhion and Maisels 1994,

Burggraaf et al. 2008).

Study location. The experiment took place at Hampshire Farm on Hampshire

College, Amherst, MA (N 42º 19’ W 72º 31’). The site has a large population of wild I. capensis plants. Study plots were located along the northwest edge of a swampy stand of trees. On May 4, 2010, we collected naturally growing I. capensis seedlings from the site and transplanted them into 10 cm diameter pots (Fafard #2 potting soil, Conrad Fafard,

Inc, Agawam, MA). Seedlings were maintained in a greenhouse, with daily watering and weekly re-randomization of bench location.

On June 1, 2010, 200 plants were planted at Hampshire Farm, in four rows of 50 plants closely following the contours of the forest edge to maintain shady conditions.

Plants were 1 m apart. Wild growing I. capensis seedlings that were within a 25 cm diameter of experimental plants were removed to alleviate intraspecific competition, but all other wild plants were left in place. Transplant survival was high, and only four plants needed to be replaced in the first week due to mortality.

Treatments. We manipulated floral damage in three treatment groups: 0%

(control), 30%, or 60% flower tissue removed. All floral damage treatments were applied to every fourth CH flowers using dissecting scissors, removing lip and throat tissue without damaging the spur or reproductive parts (Figure 3.1). Treatments were based on previous data (Soper Gorden, unpublished data): on average, 25% of flowers are naturally damaged by florivores that remove approximately 30% of the floral tissue, with

60% floral tissue removal being well within the normal range. Natural florivory was allowed on all plants.

Floral Traits. All floral traits were measured on CH flowers, which are referred to simply as “flowers” hereafter. We counted total flower production per plant. Flower size was measured on up to three flowers per plant by taking six morphometric measurements

(lip height and width, spur length, total flower length, and opening height and width), which were highly correlated. A principal components analysis was used to reduce the six measures into one variable (prcomp() in R; R Development Core Team, 2.13.0, 2011), with the first PC, reflecting flower size, explaining 70% of the variance.

Nectar volume was measured on up to two male-phase flowers per plant. Buds were bagged with polyester mesh bags overnight to let flowers open. Nectar was collected into microcapillary tubes by inserting the tube into the flower’s throat to collect pooled nectar, then snipping the end of the spur and squeezing the remaining nectar into the tube.

Pollen production was estimated using anthers collected from flowers used for nectar measurements. Since the flowers were bagged as buds, no pollen could have been removed by pollinators. We collected the androecium upon initiation of dehiscence and excluded anthers that had shed pollen into the flower before collection. Pollen production was estimated by removing the entire androecium into a microcentrifuge tube, drying at 45°C for 48 hours, suspending in 1.0 ml 70% ethanol, and counting pollen samples six times per androecium on a hemacytometer.

We collected flowers from each plant twice to measure floral chemical defenses.

Two flowers (one for anthocyanins and one for condensed tannins) were collected from the first flowers produced by each plant before treatments began (July 16 – September

13); a second set of two flowers was collected after August 30 (131 plants) or when the

plant had had at least one treated flower (22 plants), whichever came later. All flowers were photographed with a digital camera for color analysis, then frozen at -80 C until defense extraction.

Anthocyanins were extracted from one early and one late flower per plant using modified methods from published sources (Mancinelli 1990, Neff and Chory 1998, Aras et al. 2007, Brussland 2007). Briefly, frozen flowers were ground, allowed to soak in 3 ml of acidified methanol for 48 hours, filtered, and quantified using a spectrophotometer

(Genesys 10S UV-Vis Spectrophotometer, Thermo-Scientific) at 530 nm and 657 nm.

Anthocyanins were calculated as abs(530nm)-0.25abs(657nm), as per Mancinelli (1990).

Absorbance was weighted by initial flower mass, providing relative anthocyanin amounts.

We extracted condensed tannins using a basic acid butanol extraction, modified from Hagerman (2002). Briefly, frozen flowers were ground with 70% acetone and sonicated, with a portion of the resulting supernatant added to acid butanol (5% HCl v/v) and 2% ferric ammonium sulfate in 2N HCl, boiled, and measured using a spectrophotometer (Genesys 10S UV-Vis Spectrophotometer, Thermo-Scientific) at 550 nm. Absorbance values were weighted by initial flower mass, providing relative condensed tannin amounts.

Flower color was quantified from photographs, measured as the percent area of the flower lip that was red versus yellow-orange using the threshold and measure features on ImageJ (v.1.43, National Institute of Health, 2010). We measured flower color on two flowers before and two flowers after treatments were applied, and calculated the change in flower color by subtracting the early average from the late average for each plant.

Insect Interactions. Leaf herbivory was assessed by estimating percent leaf damage on the four newest fully expanded leaves three times over the summer. Leaf herbivores, crab spiders, and flower bud galls were counted during herbivory surveys.

Florivory was measured five times during the summer as the proportion of flowers with florivore damage and the percent of floral tissue missing from damaged flowers, distinguishing between treatment and natural damage. This allowed us to test how our florivory manipulation affected subsequent natural insect florivory.

Pollinators, nectar robbers, and nectar thieves were measured during 15 minute surveys of each plant with open CH flowers on ten days throughout the flowering period.

All floral visitors were identified to interaction type (pollinator, robber, or thief) and species, and their probe lengths timed. Bumblebees and honey bees are both legitimate pollinators of I. capensis (Rust 1977, Eastman 1995, Steets and Ashman 2004). Smaller visitors (such as halictid bees and ants) were considered nectar thieves unless they were explicitly seen manipulating pollen. Nectar robbers (mostly Vespula maculifrons) could be seen chewing holes in nectar spurs and drinking.

Plant Growth and Reproduction. Plant growth was measured throughout the summer using plant height, the number of nodes, and average leaf size. Aboveground biomass was harvested as each plant died or on October 11 after the first frost, and dry biomass was measured.

Approximately every two weeks, the number of CH and CL fruits on each plant was counted. Total fruit production can be counted from pedicel scars from dehisced fruits, but the process is extremely time-consuming. Previous work showed that the average number of fruits per day was highly correlated with the total number of fruits

produced up to that point (n = 40, r2 = 0.91, P < 0.0001; Soper Gorden and Adler, in review), so average CH and CL fruits per day was used to estimate total fruit production.

Mature fruits were collected and stored at 4°C until seeds per fruit could be counted and average seed mass determined. Seed mass is highly correlated with germination in this species (Waller 1985). Because I. capensis plants can respond to decreased CH reproduction with increased CL reproduction without changing total female reproduction

(e.g., Steets and Ashman 2004, Koslow and Clay 2007), we also calculated the proportion of CH vs. CL fruits.

Statistical analyses. Effects of Experimental Florivory. The effect of florivory on logical sets of response variables (plant growth, floral defenses, floral attractiveness, and nectar and pollen production) were tested using 4 separate MANOVAs (v 9.2, SAS

Institute, 2008); individual ANOVAs were investigated when MANOVA results were significant. Effects of florivory on plant growth were analyzed using plant height, number of nodes, leaf size, and final dry biomass, with biomass log transformed to improve normality. Floral chemical defense was analyzed using floral anthocyanins and floral condensed tannins from late-season flowers, and both were square root transformed to improve normality. Floral attractiveness traits were analyzed using the total number of

CH flowers, flower size (using PC1 from the PCA on flower morphology), and the change in redness over the season; both the number of flowers and the change in flower redness were log transformed to improve normality. Because of a limited number of samples, nectar production and pollen production were tested in their own MANOVA; nectar volume was square root transformed to improve normality.

Many measurements of floral interactions and plant reproduction were highly non-normal, and were therefore tested using generalized linear models (GLIM): number of pollinator, nectar robber, and nectar thief visits; percent leaf herbivory and subsequent florivory; number of flower bud gallers and crab spiders; number of CH and CL fruits; number of CH and CL seeds per fruit; seed mass for CH and CL fruits; and proportion of

CH fruits. For traits measured more than once, the average value per plot was used

(rounded to the nearest integer for counts). All GLIMs were run in R using glm() with a

Poisson distribution and a log link function, comparing a priori contrasts between the two damage treatments and between the control and both (pooled) damage treatments (R

Development Core Team, 2.13.0, 2011). Because we conducted 14 separate tests, we used Bonferroni corrections to set our alpha at P=0.004.

Effects of Floral Traits. We used GLIM multiple regressions (glm(); R

Development Core Team, 2.13.0, 2011) to examine relationships between early-season floral traits and insect choice. All insect responses were highly non-normal and zero- inflated, and therefore we used a Poisson distribution with a log link function in our models. Our main GLIM regressions tested whether initial plant height, date of first flower, and early season measures of floral anthocyanins, floral condensed tannins, and flower redness affected each insect interaction in a separate analysis. Flower size, nectar production, and pollen production all had low sample sizes. Because of this, we conducted separate GLIMs testing relationships between flower size and insect interactions, and nectar production and pollen production (in one analysis, since they have the same sample size) and insect interactions. Altogether, we conducted 21 GLIM regressions, and used Bonferroni corrections to set our alpha at P=0.0024.

Finally, to examine floral trait constancy over the season, we used GLIM regressions to correlate early and late season floral traits. This was only possible for traits that were measured both early and late in the season (plant height, flowering, floral anthocyanins, floral condensed tannins, and flower redness). For each trait, we regressed the late season trait value on the early season trait value using glm() (R Development

Core Team, 2.13.0, 2011). Depending on whether the response variable was normally distributed or left-shifted, we used either Gaussian distribution with an identity link function or Poisson distribution with a log link function in our models (Table 3.2). Five tests were conducted, setting our alpha at P=0.01 after Bonferroni correction.

Results

Effects of Florivory Treatments. Florivory had no significant effect on any measure of plant growth (MANOVA: F8, 328=1.29, P=0.25), floral defenses (MANOVA:

F2, 136=0.05, P=0.99), floral attractive traits (MANOVA: F2, 86=1.62, P=0.15), or nectar or pollen production (MANOVA: F4, 116=1.00, P=0.41).

Our florivory treatments significantly increased subsequent natural florivory compared to the control (GLIM: df=161, z=-7.706, P<0.0001), but the medium and high florivory treatments were not significantly different from one another (GLIM: df=161, z=0.597, P=0.55; Figure 3.2a). Florivory also significantly reduced leaf herbivory

(GLIM: df=179, z=3.571, P=0.0004) and the number of flower spiders (GLIM: df=179, z=2.844, P=0.004) compared to control plants, with no difference between medium and high florivory treatments (GLIM: df=179, z=-0.706, P=0.48 and df=179, z=-1.211,

P=0.23, respectively; Figure 3.2). Florivory also reduced flower galls (GLIM: df=179,

z=2.013, P=0.04) and pollinator visits (GLIM: df=117, z=2.306, P=0.02), but neither of these results were significant after Bonferroni corrections.

Florivory significantly decreased the number of CH fruits (GLIM: df=197, z=8.568, P<0.0001) and the ratio of CH to CL fruits (GLIM: df=176, z=6.485,

P<0.0001; Figure 3.3). Additionally, the medium florivory treatment had significantly fewer CH fruits (GLIM: df=197, z=-12.268, P<0.0001), fewer CL fruits (GLIM: df=197, z=-3.873, P=0.0001), and a smaller CH to CL fruits ratio (GLIM: df=176, z=-

6.772, P<0.0001) than the high florivory treatment (Figure 3.3). Thus, while any level of florivory reduced reproduction, plants with moderate florivory had significantly less reproduction than plants with high florivory or control plants. There was no effect of florivory on number of CL fruits (GLIM: df=197, z=1.072, P=0.28), seeds per CH or CL fruit (GLIM: df=73, z=-1.605, P=0.11 and df=139, z=0.863, P=0.39, respectively), or mass per CH or CL seed (GLIM: df=73, t=0.125, P=0.90 and df=139, t=1.085, P=0.28, respectively). Additionally, there was no difference between the two levels of damage on number of seeds per CH or CL fruit (GLIM: df=73, z=1.272, P=0.20 and df=139, z=0.863, P=0.39, respectively) or mass per CH or CL seed (GLIM: df=73, t=-0.38,

P=0.71 and df=139, t=-1.406, P=0.16, respectively).

Effects of Floral Traits. Early-season floral traits had many significant relationships with behavior of all insects except for flower bud galls (Table 3.1).

Pollinators and nectar robbers visited taller plants more, and florivores damaged flowers on taller plants more. Nectar thieves visited earlier flowering plants more, and florivores damaged late flowering plants more. Both floral chemical defenses were negatively correlated with floral antagonists; initial floral anthocyanins were correlated with more

pollinator visits and less florivory, while initial floral condensed tannins were correlated with fewer nectar robbers, nectar thieves, and flower spiders, and more leaf herbivory.

Flowers with more red were visited more often by nectar robbers, but had less florivory.

Both nectar thieves and florivores preferred larger flowers. Pollinators and nectar robbers visited plants with higher pollen production more often, while florivores preferred plants with less pollen production.

Many early season floral traits were highly correlated with late season flower traits (Table 3.2). Plants that started out as tall seedlings became taller plants. Similarly, plants that flowered earlier had more total flowers, and flower color was significantly correlated across time. Neither anthocyanins nor condensed tannins were correlated across time, suggesting that they are more strongly influenced by environment or phenology than genotype.

Discussion

Effects of Florivory. Experimental florivory reduced plant reproduction by leading to fewer CH (outcrossed) fruits, but had no effect on CL (selfed) fruit production, number of seeds per fruit, or seed mass (Figure 3.3). Florivory can affect plant reproduction directly, by damaging reproductive organs, or indirectly, by altering floral interactions (McCall and Irwin 2006). In this experiment, the effects on total reproduction may not be due to direct damage to reproductive parts, since we only damaged petals. Instead, there are likely indirect effects through changes in other plant- insect interactions, or changes in resource allocation. For example, experimental florivory increased natural subsequent florivory (Figure 3.2A), which may have caused

damage to reproductive parts. Alternately, there may be a direct cost by removing floral tissue containing nutrients, leaving fewer resources to invest in fruits.

Experimental florivory also reduced the ratio of CH (outcrossed) to CL (selfed) reproduction. This shift towards a greater reliance on self-pollination after florivory is seen in other species (Ashman and Penet 2007, Penet et al. 2009). There have also been several studies showing that I. capensis responds to other antagonisms, including leaf herbivory and competition, by increasing selfing CL reproduction (Steets 2005, Steets et al. 2006a, Steets et al. 2006b). Making flowers less apparent (for example, with small corollas or inserted anthers) may provide resistance to florivores (Ashman et al. 2004,

McCall and Irwin 2006); increased allocation to inconspicuous CL over showy CH flowers could be a mechanism of induced resistance to or tolerance of florivores. It is possible that in plants with a mixed mating system, such as I. capensis, relying on selfing may serve as a mechanism of tolerating antagonists. Compared to leaves and CH flowers, CL flowers require few resources to produce (Waller 1979) and inbreeding depression for most traits in I. capensis is low (Heschel et al. 2005). Therefore, tolerance via increased selfing may be a more effective strategy than investing in chemical defenses against florivory.

Surprisingly, the 30% florivory treatment had a significantly greater impact on fruit production and mating system than the 60% florivory treatment (Figure 3.3).

Another florivory study in I. capensis also found a stronger effect of medium levels of floral damage on subsequent florivory and patterns of defense induction (Boyer et al., in prep). This suggests there may be a damage threshold, with low or intermediate levels of floral damage eliciting a stronger response than high levels of damage. For many

species, there is a damage threshold for leaf herbivory which must be reached before the plant responds with induced resistance or changes in plant growth (Underwood 2000,

2010). In florivory, too, there can be a threshold of damage that must be reached before there are effects on plant reproduction (McCall 2008). However, there are fewer examples where leaf herbivory has the greatest effect at intermediate damage levels (but see Huhta et al. 2000, Underwood 2000, Utsumi et al. 2009), and no studies we know of that have found this pattern with florivory. It is possible that plants maintain partially damaged flowers for reproduction even if the flowers lack the resources for full fruit development, while heavily damaged flowers are aborted and resources reabsorbed to be used in future flowers or fruits. If moderately damaged flowers set few fruits and reabsorbed resources can ameliorate the effects of damage, this may explain the pattern of moderate damage leading to fewer CH fruits.

Experimental florivory affected several other species interactions. Florivory increased natural subsequent florivory and decreased both leaf herbivory and the number of flower spiders on plants. This suggests that floral damage may have consequences beyond direct damage to flowers or even pollinator deterrence by altering the community of subsequent interactions on both flowers and leaves. While there have been several studies finding effects of florivory on pollination (Botto-Mahan and Ojeda-Camacho

2000, Leavitt and Robertson 2006, McCall 2008, Zangerl and Berenbaum 2009, Cardel and Koptur 2010, Sõber et al. 2010, Botto-Mahan et al. 2011), there are few studies testing effects on other plant-insect interactions. In this experiment, not only did florivory affect subsequent flower-using insects, but also affected leaf herbivores. This, combined with the fact that florivory affected plant reproduction (Figure 3.3) and

exhibited preference for some floral traits (Table 3.1) suggests the potential for conflicting selection on plant traits. Additionally, our results suggest that floral herbivory may have as many and as strong indirect effects on communities as leaf herbivore

(Ohgushi 2008).

Surprisingly, our floral damage treatments increased subsequent natural florivory.

This suggests that, rather than inducing resistance in plants, flower damage made plants more attractive or less resistant to florivores. Several leaf herbivory studies have found induced susceptibility, where leaf damage leads to increased leaf herbivory (Karban and

Niiho 1995, Underwood 1998, Kaplan and Denno 2007, Utsumi and Ohgushi 2008). So far, few studies have tested how floral damage affects subsequent florivory, with one study finding that floral damage decreases florivory (Boyer et al., in prep). This study is the first we know of that shows induced susceptibility to florivores after floral damage. It is possible that increased florivore attraction after floral damage can be adaptive, and flowers that are damaged may be maintained as a “bait” to keep florivores from attacking other, healthy flowers. While florivory treatments still had a negative impact on fruit production (Figure 3.3), we applied our treatments to new flowers each time, unlike the pattern of natural florivory. By drawing subsequent floral damage to flowers that are already damaged by making those flowers more attractive to florivores, plants may reduce the strength of the negative impact on reproduction. Additionally, by changing interactions with other plant antagonists, florivores may have complex effects on selection for floral traits via altering the interaction community.

Floral damage had no effect on either of the two floral chemical defenses measured, even though floral damage was a good predictor of future florivory. Optimal defense theory predicts that flowers should be well-defended against damage because of their 89

close association with fitness (Begon et al. 2006, McCall and Irwin 2006, McCall and

Fordyce 2010). It is possible that plants protect themselves from florivory using constitutive defenses instead of induced defenses to protect flowers before damage occurs. Indeed optimal defense theory predicts higher levels of constitutive versus induced defenses in costly tissues (McCall and Irwin 2006, McCall and Fordyce 2010). Although we did not manipulate floral chemical defenses in this experiment, we did see a negative correlation between anthocyanins and florivory (Table 3.1), and anthocyanins have been implicated as defenses against florivores in petunias (Johnson et al. 2008). However, natural levels of florivory are high despite anthocyanins, reaching 90% on some flowers (NLSG, unpublished data). Impatiens capensis may rely on tolerance mechanisms (such as increasing selfing) instead of resistance.

While experimental florivory increased subsequent florivory, we saw induced resistance to other antagonists. Floral damage decreased the amount of leaf herbivory

(Figure 3.2b), suggesting either a systemic change signaled by floral damage that induces vegetative resistance, or that floral damage induces volatiles that deter leaf herbivores.

While previous studies have shown that damage to leaves can induce resistance in flowers (Euler and Baldwin 1996, Wackers and Bezemer 2003, Adler et al. 2006, McCall and Karban 2006), we are unaware of previous work examining how floral damage affects vegetative resistance. All of the florivores on I. capensis are generalists that consume leaves as well as flowers. Thus, if florivory accurately predicts the possibility of leaf damage by the same insects, induced vegetative resistance in response to floral damage could be adaptive (Karban et al. 1999). Alternately, since experimental florivory caused increased natural florivory as well as decreased leaf herbivory, these generalist

insects may have chosen to switch from feeding on leaves to feeding on flowers, either because flowers became more attractive or nutritious, or because leaves became less so.

There were significantly fewer flower spiders on plants with experimental florivory (Figure 3.2c), perhaps because our treatments altered floral symmetry and flower spiders prefer symmetrical flowers (Wignall et al. 2006). Although we only measured flower spiders in response to florivory, this change in spider abundance may have larger community effects on plants with or without florivory. Not only can flower spiders have significant negative impacts on pollinators and plant reproduction, but can also affect seed predators and possibly other plant-insect interactions (Louda 1982).

Florivory may have indirect positive effects on pollination by reducing the number of flower spiders, perhaps leading to greater outcrossing or healthier seeds. However, flower spiders can consume nectar thieves and nectar robbers as well as pollinators on I. capensis (NLSG, personal observation). Since there tended to more nectar larcenists than pollinators, florivory may have indirect negative effects on pollination by reducing flower spider numbers and removing their predation on nectar larcenists. The final outcome of the effects of florivory on flower spiders likely depends on what other insects are interacting with plants and in what densities.

Although our florivory treatments altered flower size and symmetry (Figure 3.1), both of which can significantly affect insect behavior (Lehrer et al. 1995, Møller and

Eriksson 1995, Lara and Ornelas 2001, Kaczorowski et al. 2012), none of the nectar- feeding insects, including pollinators, were significantly affected by flower damage. This suggests that visitors to I. capensis flowers are driven more by rewards such as nectar and pollen, neither of which we manipulated with our damage treatments, than visual cues.

This hypothesis is consistent with a study showing that altered symmetry had no effect on pollinator visitation or plant reproduction in Impatiens pallida, a congener of I. capensis

(Frey et al. 2005).

Effects of Floral Traits. As seen in many previous studies, many insects preferred larger plants with more flowers (Table 3.1): pollinators, nectar robbers, and florivores preferred taller plants, while nectar thieves preferred earlier flowering plants. This can be a function of plant resources, where plants with higher resources grow larger and therefore attract more interactions (Burkle and Irwin 2010). Florivory was higher on plants with a late flowering date more, which is surprising since early flowering date was correlated with more total flowers per plant (Table 3.2). This means there may be conflicting selection on flowering data by florivores and nectar thieves, the outcome of which will depend on their relative densities and effects on plant reproduction.

Florivores and nectar thieves both visited plants with larger flowers more frequently, while pollinators and nectar robbers both preferred plants with greater pollen production

(Table 3.1). Nectar larcenists are often attracted to floral traits such as larger flower size and greater floral rewards (Irwin et al. 2010). Florivores preferred to visit flowers with low pollen production, suggesting either that florivores do not target pollen as a source of food, avoid excess pollen (perhaps due to high defense levels in pollen; see Gronquist et al. 2001, Frölich et al. 2006), or cause plants to produce less pollen. Although most floral traits were measured before treatments began, pollen production was only measured once in the middle of the summer, making it difficult to differentiate between effects of pollen on florivores and vice versa. It is interesting that nectar production was not correlated with any insect choice (Table 3.1). In Impatiens spp., nectar production is continuous

(Rust 1977, Marden 1984), allowing compensation for nectar removal (Temeles and Pan

2002). Therefore, it is possible that insects do not discriminate based on nectar volume because nectar is constantly produced.

This study is one of the most comprehensive examinations of floral defenses, in terms of both induction after florivory and effects on insect preference. We found that the two floral defenses measured (anthocyanins and condensed tannins) were both associated with differences in insect preference. Plants with more floral condensed tannins had fewer nectar robbers, nectar thieves, and flower spiders, suggesting these secondary compounds may act as an effective floral defense. Condensed tannins have been shown to deter leaf herbivores in several studies (reviewed in Barbehenn and

Constabel 2011), but this is the first investigation of their ecological role in floral tissue.

Plants with more floral anthocyanins had less natural florivory and more pollination.

Anthocyanins play many roles in plants. Aside from providing protection against UV damage (Mancinelli 1990, Dixon et al. 2001, Close and Beadle 2003, Karageorgou and

Manetas 2006), anthocyanins act as pigments providing red or blue color (Rausher 2008,

Tanaka et al. 2008). These pigments in petal tissue can significantly affect pollinator preference, often increasing pollinator visitation (Schemske and Bradshaw 1999,

Hoballah et al. 2007) (but see Irwin et al. 2003). In our system, pollinator attraction to anthocyanins is not entirely due to color, since there was no correlation between early season flower redness and pollinator preference (Table 3.1). In addition to UV protection and pigmentation, anthocyanins have also been implicated as a resistance trait against florivores. In Petunia, florivores avoided segments of petal tissue colored by anthocyanins, and insect florivores gained less weight when fed on anthocyanin-rich petal

tissue than white petal tissue (Johnson et al. 2008). Our negative correlation between anthocyanins and florivory suggests that anthocyanins may provide resistance to florivory in I. capensis as well.

Some floral traits were significantly correlated over time, including plant height, flowering, and flower redness (Table 3.2). While it may not be surprising that tall plants stayed tall or that early flowering plants flowered more, these results do suggest that there is little plasticity in flower color. This is supported by the observation that individual plants had a bivariate distribution of flower color, with consistently all low-red or all high-red flowers (NLSG, unpublished data), suggesting a genetic basis to this trait.

Neither of the two floral chemical defenses (anthocyanins or condensed tannins) were correlated across time, indicating floral defenses are more plastic, possibly via induction after damage. While we saw no effect of florivory on floral defenses, there are many other antagonists which could induce resistance in I. capensis flowers. For example, since nectar larcenists are negatively correlated with floral condensed tannins (Table 3.1), plants may induce floral condensed tannins after nectar robbing to protect against future antagonisms (Irwin et al. 2010). Alternately, the levels of chemical defenses in flowers may depend more on phenology or seasonal resource availability.

Conclusions. Florivory significantly reduced plant reproduction and altered mating system expression, leading to a greater proportion of selfed reproduction.

Decreasing allocation to outcrossing reproduction could provide a mechanism of tolerating florivory, or of resistance through reduced floral display. Experimental florivory increased subsequent natural florivory, suggesting that either damaged flowers are more attractive to florivores or that generalist consumers move from leaves to flowers

after florivory damage. Surprisingly, our florivory treatments significantly reduced leaf herbivory, suggesting a systemic response to flower damage that has not been examined previously. Two floral chemical defenses, condensed tannins and anthocyanins, were both negatively correlated with several floral antagonisms, adding to a growing literature identifying floral chemical defenses that confer resistance to floral antagonists. However, we found no evidence that florivores induce either anthocyanins or condensed tannins in flowers. Overall, our results suggest that florivory may shape the community of species that interact with plants, alter interactions such as leaf herbivory that occur outside the realm of flowers, and shape the evolution of mating system.

Tables

Table 3.2. GLIM regression results between early season floral traits on late season floral traits in Impatiens capensis. Bold P values indicate significance after Bonferroni corrections set alpha at P=0.01. Test Test Distribution Statistic P Value Early floral condensed tannins on late Gaussian t=-0.399 0.69 Early floral anthocyanins on late Gaussian t=1.110 0.27 Early flower redness on late Poisson z=49.29 <0.0001 Date of first flower on total flowers produced Poisson z=-38.27 <0.0001 Initial seedling height on final plant height Gaussian t=5.401 <0.0001

Figures

Figure 3.1. Flower damage treatments, showing the amount of petal tissue removed from the lips and throat of flowers in each treatment. Treatments were applied to every fourth flower on each plant in each treatment using dissecting scissors.

12 a 10 A a

6 4

2 NaturalFlorivory(%) 0 Control 30% 60% Damage Damage 7.0 B 6.0 a a 5.0 4.0 3.0 2.0

Leaf Damage (%)DamageLeaf 1.0

0.0

Control 30% 60% Damage Damage

0.3 C a a 0.2

0.1

SpidersFlower(#) 0 Control 30% 60% Damage Damage Figure 3.2. Effect of artificial florivory (no damage, 30% flower tissue removed, or 60% flower tissue removed) on plant antagonists in Impatiens capensis. A) percent floral tissue removed due to subsequent natural florivory. B) Percent leaf damage due to leaf herbivores. C) Number of flower spiders per plant. For each test, we used generalized linear models with a Poisson distributions to test whether the combined damage treatments were different from the control, as well as whether the medium treatment was different from the high treatment. Error bars show standard error. Different color bars indicate damaged plants are significantly different from the control; letters indicate if the two damage treatments are significantly different from each other at alpha = 0.005 (with Bonferroni correction).

5.0 A b 4.0 a 3.0 2.0

(#) FruitsCH 1.0

0.0

Control 30% 60%

Damage Damage

b 8.0 a 7.0 B 6.0 5.0 4.0 3.0

CL(#) Fruits 2.0 1.0 0.0 Control 30% 60% Damage Damage 0.3 C b a 0.2

0.1

ProportionFruitsCH 0 Control 30% 60% Damage Damage Figure 3.3. Effect of artificial florivory (no damage, 30% flower tissue removed, or 60% flower tissue removed) on Impatiens capensis reproduction. “CH” (chasmogamous) is outcrossing reproduction, while “CL” (cleistogamous) is selfed reproduction. A) number of CH fruits. B) number of CL fruits. C) Proportion of all fruits that were outcrossed vs. selfed. For each test, we used generalized linear models with a Poisson distribution to test whether the combined damage treatments were different from the control, as well as whether the medium treatment was different from the high treatment. Error bars show standard error. Different color bars indicate damaged plants are significantly different from the control; letters indicate if the two damage treatments are significantly different from each other at alpha = 0.004 (with Bonferroni correction).

100

CHAPTER 4

CONSEQUENCES OF MULTIPLE FLORAL INTERACTIONS FOR

SUBSEQUENT INTERACTIONS AND PLANT REPRODUCTION

Abstract

Plants often interact simultaneously with multiple floral antagonists and mutualists. Despite a recent interest in studying the community context of multiple interactions, very few studies have tested the effects of multiple floral interactions on subsequent plant-insect interactions, floral traits, or plant reproduction. Additionally, most studies of floral interactions have been pairwise in nature, while floral interactions may have non-additive effects that cannot be predicted from the outcome of each interaction in isolation. We manipulated florivory, nectar robbing, and pollination on

Impatiens capensis to determine interactive effects on floral traits, subsequent plant- insect interactions, and plant reproduction. We found that florivory significantly deterred both pollinators and nectar larcenists, but increased subsequent florivory. Surprisingly, pollinators and nectar larcenists both preferred plants that had previously been pollinated and robbed, even though many other plant species reduce floral attractive traits after successful pollination or nectar robbing. All treatments had significant multi-way interactions on subsequent floral visitors, indicating that the effect of one interaction depends on the presence of other interactions. Additionally, this frequently led to a ranking of interaction importance: for nectar feeders, florivory had a dominant negative effect that eclipsed the secondary positive effects of pollination and nectar robbing. The only floral trait influenced by floral interactions was the production of outcrossing

101

flowers, which was reduced by florivory. This result, combined with shifts from outcrossing to selfing reproduction in the presence of florivores, suggests that I. capensis may tolerate or resist florivory by shifting the mating system towards more selfing.

There were three-way multivariate effects of floral interactions on plant reproduction; florivory tended to reduce outcrossing and increase selfing reproduction, while pollination resulted in greater outcrossing reproduction and decreased selfing reproduction in the absence florivory. Taken together, our results suggest that florivory has stronger effects on I. capensis than pollination or nectar robbing, a result that would not have been obvious by examining pairwise interactions.

Introduction

Plants interact with many types of antagonists. Floral antagonists may have particularly strong effects on plant reproduction because flowers are closely related to reproduction, and yet floral antagonists are studied much less frequently than leaf antagonists (McCall and Irwin 2006, Irwin et al. 2010). While many floral antagonists have individually been shown to reduce plant reproduction in some conditions, plants do not interact with only one type of antagonist at a time (McCall and Irwin 2006, Agrawal et al. 2007, Irwin et al. 2010). The effect of one floral antagonist on the behavior of others is relatively unknown. For example, florivory changes floral symmetry (McCall

2008), volatile emissions (Lucas-Barbosa et al. 2011), nectar guides (Botto-Mahan and

Ojeda-Camacho 2000), rewards (Krupnick et al. 1999), and resistance traits (McCall

2006), all of which could affect the preference of other floral antagonists or mutualists.

Understanding pairwise interactions in isolation is often insufficient to predict fitness

102

outcomes, community composition, coevolution, or shapes of selection in natural systems

(Thompson 2002, Agrawal and Van Zandt 2003, Agrawal et al. 2007). Ecologists now realize the importance of studying plant-animal interactions in a more comprehensive context, integrating the direct and indirect effects of many antagonists and mutualists simultaneously (Thompson 2002, Agrawal et al. 2007). While it is known that the diversity of leaf damage on a plant can structure subsequent whole-plant interactions

(e.g., Van Zandt and Agrawal 2004a, Johnson and Agrawal 2007, Ohgushi 2008, Utsumi et al. 2009), consequences of multiple floral antagonisms on subsequent floral interactions are largely unknown.

Typically, studies that include both pollinators and plant antagonists assume that antagonists affect pollinators but not vice versa (e.g. Bronstein et al. 2003, Ivey and Carr

2005). Many studies have demonstrated that nectar robbers (Irwin and Brody 1998,

Maloof and Inouye 2000, Temeles and Pan 2002, Richardson 2004) and florivores

(Botto-Mahan and Ojeda-Camacho 2000, Malo et al. 2001, Leavitt and Robertson 2006,

McCall 2008, Cardel and Koptur 2010, Sõber et al. 2010, Botto-Mahan et al. 2011,

Cares-Suarez et al. 2011) affect pollinator preference. However, research is lacking on whether pollinators can affect floral antagonists. Pollination can alter plant traits in many ways, including changes in flower color (Weiss 1991, Nuttman and Willmer 2003), floral sex ratio (Sato 2002), shape (vanDoorn 1997), and longevity (vanDoorn 1997, Sato

2002). These changes can often make flowers less attractive to subsequent pollinators

(Weiss 1991, vanDoorn 1997, Sato 2002, Nuttman and Willmer 2003). Since many floral antagonists are attracted to the same traits as pollinators (Temeles and Pan 2002,

McCall and Irwin 2006, Irwin et al. 2010), pollination could also reduce attractiveness to

103

subsequent floral antagonists. Additionally, some plant species can alter resource allocation to future flowers based on pollination quality (Olivieri et al. 1994, Liu et al.

2010, Albert et al. 2011, Canto et al. 2011), which could influence attractiveness to floral antagonists as well as future pollinators.

In addition to attractive traits, flowers also produce chemical defenses, often at higher concentrations than in leaves (Zangerl and Rutledge 1996, Hause et al. 2000,

Strauss et al. 2004, Damle et al. 2005, Frölich et al. 2006, Frölich et al. 2007). Both petals (Euler and Baldwin 1996, Adler et al. 2001, Strauss et al. 2004, Irwin and Adler

2006) and nectar (Adler 2000, Irwin et al. 2004, Adler and Irwin 2005, Adler et al. 2006,

Johnson et al. 2006) can contain the same defenses that frequently deter leaf herbivores, and other flower parts (e.g., pollen, ovules, stigmas) can also contain high levels of defenses (Gronquist et al. 2001, Frölich et al. 2006, Frölich et al. 2007). Floral chemical defenses have been implicated in deterring both florivores (Johnson et al. 2008, Lucas-

Barbosa et al. 2011) and nectar larcenists (Adler and Irwin 2005), but can also deter some pollinators (Adler and Irwin 2005, Gegear et al. 2007, Lucas-Barbosa et al. 2011, Adler and Irwin 2012), causing possible ecological tradeoffs. While a few previous studies have found that floral chemical defenses can be induced after leaf herbivory (Euler and

Baldwin 1996, Wackers and Bezemer 2003, Adler et al. 2006, McCall and Karban 2006), and one study found induced resistance to florivores after floral damage (McCall 2006), there have been no studies testing the induction of chemical defenses after attack by floral antagonists. Understanding how floral mutualists and antagonists alter both attractive and defense floral traits has important implications understanding how these interactions shape the subsequent community of interactions.

104

We conducted a factorial manipulation of florivory, nectar robbing, and pollination to test their combined effects on floral attractive and defense traits, subsequent plant-insect interactions, and plant reproduction in the field. Additionally, by manipulating three floral interactions, we will test for non-additive effects that would not be obvious in pairwise studies. Together, this will give us a comprehensive picture of how floral traits and insects interact to affect plant reproduction.

Methods

Study System. Impatiens capensis Meerb. (Balsaminaceae) is an excellent system to examine multiple interactions. It is an annual found in moist soils in much of North America

(Leck 1979, Eastman 1995). It has a mixed mating system, with selfing cleistogamous (CL) and open-pollinated chasmogamous (CH) flowers; the showy orange CH flowers produce more seeds with better dispersal and survival than seeds from CL flowers (Mitchell-Olds and

Waller 1985, Schmitt et al. 1985, Eastman 1995). CH flowers are heavily reliant on pollinators, mostly Bombus sp. and Apis mellifera, to produce fruits (Rust 1977, Leck 1979,

Eastman 1995, Steets and Ashman 2004). Both flower types produce a capsule fruit containing one to several seeds, which are dispersed explosively as the fruit ripens. The seed bank generally does not last more than one year (Simpson et al. 1985).

Besides pollinators, CH flowers are visited by several floral antagonists. Nectar robbers and nectar thieves both consume nectar without contacting the plant’s reproductive parts (Rust 1979); while nectar robbers pierce the corolla or spur to consume nectar, nectar thieves are simply too small to transfer pollen while entering the corolla opening (Inouye

1980, Irwin et al. 2010). Collectively, nectar robbers and nectar thieves are considered nectar larcenists. Flowers can have very high rates of nectar robbing (up to 80% of flowers) by

105

several insect species, including the plant’s main pollinators (Eastman 1995, Young 2008).

Although some robbers can also be pollinators, flowers are never pollinated during the act of robbing (Rust 1979). Nectar robbers include Vespula maculifrons, Apis mellifera, and some

Bombus species (Rust 1979, Zimmerman and Cook 1985). Nectar thieves include halictid bees, syrphid flies, and ants (Eastman 1995). Generalist herbivores, including Popillia japonica, regularly consume flowers as florivores (0-90% damage, pers. obs.). The specialist galler Schizomyia impatientis attacks flower buds, causing them to form galls instead of flowers and preventing fruit production (Hummel 1956, Eastman 1995).

Chasmogamous flowers have a range of floral attractive and defense traits. Plants vary in number of flowers, flower size, nectar production, pollen production, and flower height (NLSG, pers. obs.). Flower color can vary from entirely yellow (no red spotting) to almost entirely red (extensive red spotting; Boyer et al., in prep). While little has been published on defensive chemistry in I. capensis or its role in mediating species interactions (but see Soper Gorden and Adler, in prep), Impatiens spp. in general contain both anthocyanins and condensed tannins. Anthocyanins are flavonoid pigments that often attract pollinators (Delpech 2000, Koes et al. 2005), but have also been implicated in reducing florivore preference and performance (Johnson et al. 2008). Anthocyanins are present in Impatiens spp. leaves, flowers, and stems, and cause the variable red spots on the lip petals of I. capensis (Aras et al. 2007). Condensed tannins are also present in

Impatiens spp. in general (Waterman et al. 1983) and I. capensis flowers specifically

(Soper Gorden, unpublished data). Although condensed tannins are usually studied as vegetative defenses, they are also found in floral tissue and have the potential to deter florivores (Gautierhion and Maisels 1994, Burggraaf et al. 2008).

106

Plant propagation and treatments. On 6 May 2011, we collected I. capensis seedlings from Hampshire Farm (N 42º 19’ W 72º 31’), transplanted them into 10 cm pots in Fafard #2 mix (Conrad Fafard, Inc, Agawam, MA), and kept them in a greenhouse on the University of Massachusetts campus (N 42º 23’ W 72º 32’) until they were planted in the field. Plants were watered daily and bench locations were randomized once a week. On 15 June 2011, when all plants were ~0.4 m tall, we transplanted 200 plants along the northwest edge of a wet forest at Hampshire Farm where I. capensis occurs naturally. Plants were in four rows of 50, with each plant 1 m from all neighbors. To reduce intraspecific competition, conspecifics within 0.25 m of each plant were removed.

Florivory was manipulated by removing 30% floral tissue from one quarter of flowers using dissecting scissors (average natural florivory; Soper Gorden, unpublished data), while control plants did not receive artificial florivory. Nectar robbing was manipulated by using dissecting scissors to cut a small hole at the base of the flower’s throat (where many nectar robbers puncture the corolla) and using a microcapillary tube to remove nectar, which has successfully simulated nectar robbing in other systems (e.g.

Irwin and Brody 1998, Burkle et al. 2007, Brody et al. 2008). Control plants did not receive artificial nectar robbing. Pollination was manipulated by using a paintbrush to apply mixed pollen from at least three wild plants to stigmas of female-phase flowers, while control plants had no additional pollen added. Florivory and nectar robbing were applied to every fourth flower (the average proportion of flowers naturally damaged by florivores, and within the recorded range of nectar robbing frequencies; Soper Gorden, unpublished data; Eastman 1995, Young 2008), but not the same individual flowers for

107

both treatments; hand pollination was applied to all female-phase flowers each day. All plants also received natural levels of florivory, nectar robbing, and pollination.

Treatments were randomly applied in a fully factorial manner, resulting in eight total treatment combinations, each with 25 plants.

Responses. Plant growth was measured three times during the summer, including plant height, number of nodes and CL flowers, and leaf area (estimated by multiplying the length and width of the two most apical fully expanded leaves). In this species, plant height is highly correlated with aboveground dry biomass (correlation: df=173, r=0.78, r2=0.61, P<0.0001; unpublished data), so biomass was not measured.

For chemical analysis, we collected the first two flowers each plant produced

(before treatments were applied; July 7 through September 1, depending on plant) and two flowers later (August 25 or after all treatments had been applied at least once, whichever came later). These flowers were digitally photographed for flower color analysis, then frozen at -80°C until chemical extractions. Floral anthocyanins were extracted from one early and one late flower, using a modified protocol (Mancinelli 1990,

Aras et al. 2007, Brussland 2007). Briefly, frozen flowers were extracted in acidified methanol (1% HCl v/v) at 4°C in the dark for 48 hours, then measured at 530 nm and 657 nm on a spectrophotometer (Genesys 10S UV-Vis Spectrophotometer, Thermo-

Scientific). Relative anthocyanins content was calculated as A530 – 0.25A657, as per

Mancinelli (1990), then standardized by flower mass. Floral condensed tannins were extracted from one early and one late flower, using an acid butanol method modified from Hagerman (2002). Briefly, frozen flowers were ground and sonicated in 70% acetone. The supernatant was added to acid butanol (5% HCl v/v) and 2% ferric

108

ammonium sulfate in 2N HCl, heated in a boiling water bath for 50 minutes, then measured at 550 nm on a spectrophotometer (Genesys 10S UV-Vis Spectrophotometer,

Thermo-Scientific). Relative condensed tannin content was calculated as A550 standardized by flower mass. For both floral defenses, we subtracted the early season value from the late season value to calculate the change in defense levels over the course of the summer.

All CH flowers produced were counted at least every other day, marking pedicels of counted flowers with Wite-Out Quick Dry (Bic, Clinchy, France) to prevent double counting, giving total CH flower production. Flower size was measured on up to three flowers per plant five times during the summer, including flower length, nectar spur length, lip petal width and height, and corolla opening width and height. These six morphometric measures were condensed into one value using principal components analysis (prcomp() in R; R Development Core Team, 2.13.0, 2011). The first and only significant principal component (PC1) explained 66% of the total variation and was used to represent larger flower size for analysis. Nectar and pollen production were measured on up to three flowers per plant, bagged as buds to prevent pollinator visitation.

Androecia were collected from bagged flowers upon anthesis; if pollen had already been shed in the flower, the androecium was not collected. Androecia were dried for 24 hours at 45°C, suspended in 1ml 70% ethanol, and six subsamples were counted on a hemacytometer under a microscope. Nectar was collected from bagged male-phase flowers using microcapillary tubes to remove pooled nectar from the base of the throat, then the tip of the nectar spur was cut and any additional nectar squeezed into the same tube. We quantified flower color from photographs of flowers used for defense

109

extractions using ImageJ’s threshold and measure features to calculate the percent of lip petals that were red versus yellow-orange in color (ImageJ v.1.43, National Institute of

Health, 2010). Since this was measured on two early and two late flowers, we could also calculate the change in redness over the season.

Pollinators, nectar robbers, and nectar thieves were observed during 15-minute surveys on eight days spread throughout the summer. In each survey, we recorded the number, interaction type (pollinator, nectar robber, or nectar thief), taxon, probe time per flower, and number of flowers probed for each visitor for all plants that had open CH flowers. We also completed three monthly herbivory surveys, during which we counted and identified each herbivore on each plant, allowing measures of herbivore density and richness (to functional group), and estimated leaf damage on four apical fully expanded leaves. Finally, we estimated florivory during six surveys every other week, during which we recorded the average percent floral tissue missing per flower for each plant.

During biweekly surveys, we counted the number of CH and CL fruits on each plant to estimate the total number of fruits produced. Total fruit production can be counted from pedicel scars from dehisced fruits, but the process is extremely time- consuming. Previous work showed that the average number of fruits per day was highly correlated with the total number of fruits produced up to that point (n = 40, r2 = 0.91, P <

0.0001; Soper Gorden and Adler, in review), so average CH and CL fruits per day was used to estimate total fruit production. During these surveys, we also collected mature

CH and CL fruits, and stored them at 4°C until seeds per fruit and average seed mass could be measured. Finally, to test for shifts in mating system, we calculated the proportion of CH versus CL fruits the plant produced.

110

Statistical analyses. To test the effects of our treatments, we conducted seven

MANOVAs on groups of related traits: plant growth (height, leaf size, number of nodes, and CL flowers); floral defenses (change in anthocyanins and condensed tannins); flowers (number of CH flowers and flower size PC); floral rewards (nectar and pollen production); leaf herbivory (leaf damage, density of herbivores, and herbivore richness); fruits (number of CH and CL fruits and proportion of CH versus CL fruits); and seeds

(seeds per CH and CL fruit and average mass of CH and CL seeds). For each test, we included florivory treatment, nectar robbing treatment, pollination treatment, and all two- and three-way interaction terms. Nectar production, leaf damage, and herbivore richness were log transformed; number of CL and CH flowers, herbivore density, number of CH and CL fruits, number of seeds per CL fruit, and average CL seed mass were square root transformed; plant height was X2 transformed. All MANOVAs were conducted in SAS

(v 9.2, SAS Institute, 2008). Individual ANOVAs were investigated when a MANOVA was significant.

A few variables had left-shifted distributions that could not be transformed to make them normal: percent florivory, pollinator visits, nectar robber visits, and nectar thief visits. For these responses, we conducted individual generalized linear models

(GLIMs) using a Poisson distribution with a log link function, and including florivory, nectar robbing, and pollination treatments and all two- and three-way interactions as explanatory variables. GLIMs were run in R using glm() (R Development Core Team,

2.12.0, 2011). Seasonal change in flower redness was extremely non-normal, and did not fit a Poisson distribution. Instead, we ran a GLIM using a Gaussian distribution with an identity link.

111

Results

Our treatments had a significant effect on flowering, but no significant effect on any other measure of plant growth or floral traits (Table 4.1). Supplemented florivory significantly reduced the number of CH flowers plants produced (ANOVA: F2, 150=4.19,

P=0.017; Figure 4.1). There was no effect on CL flower production or any other plant growth or flowering trait.

There were significant effects on all subsequent floral interactions, with many two- and three-way interactions, but there was no significant effect of our treatments on leaf herbivory (Table 4.1). There was a significant three-way interaction on pollinator visits per hour, such that pollination and nectar robbing had little effect on pollinator visits on plants with supplemented florivory, but pollinators preferred plants with either pollination or nectar robbing (but not both) when florivory was not supplemented (Figure

4.2a). Overall, pollinators preferred flowers without supplemented florivory and with supplemented nectar robbing or pollination.

Similarly, there was a significant three-way interaction between florivory, nectar robbing, and pollination on nectar robber visits (Table 4.1). Nectar robbers overwhelmingly preferred plants with natural florivory and nectar robbing but supplemented pollination to all other treatment combinations (Figure 4.2b). Overall, nectar robbers preferred previously pollinated plants (Table 4.1).

Nectar thieves responded differently to combinations of florivory and nectar robbing treatments, and nectar robbing and pollination treatments (Table 4.1). Nectar thieves preferred plants without florivory or nectar robbing over plants with higher levels

112

of either damage (Figure 4.3a), and also preferred plants that had been hand pollinated, especially without nectar robbing (Figure 4.3b). There were also significant main effects of florivory and pollination; florivory reduced nectar thieves, while pollination increased nectar thief visits.

There were two-way interactions between florivory and nectar robbing, and between nectar robbing and pollinators on subsequent florivory (Table 4.1). Florivores preferred plants that had already had florivory damage when there was no added nectar robbing, but that preference disappeared in the presence of nectar robbing (Figure 4.4a).

There was also significantly less florivory on plants with supplemental pollination and natural nectar robbing than other treatment combinations (Figure 4.4b). Overall, plants with supplemented florivory or nectar robbing h ad significant more subsequent florivory, and plants with supplemented pollination had significantly less subsequent florivory.

There were also significant three-way interaction effects between florivory, nectar robbing, and pollination on fruits and seeds (Table 4.1). When individual ANOVAs were tested, no single reproduction response had a significant three-way interaction (F<0.59,

P>0.44 for all), indicating that the effect on reproduction is multivariate. To investigate the relationship more closely, we calculated the standardized canonical coefficients for each response variable in the reproduction MANOVAs to examine how strongly each variable contributed to the significant interaction (Scheiner 2001; Table 4.2). In the fruit number MANOVA, all three responses (number of CH fruits and CL fruits, as well as proportion of CH versus CL fruits) were equally responsible for the significant interaction, with CH fruits having an equal but opposite effect compared to CL fruits and the proportion of CH fruits (Table 4.2). When comparing treatment means, plants

113

produced the most CH fruits and had the highest proportion of CH fruits in the pollination treatment when there was no florivory or nectar robbing, and the least in any treatment combination including florivory (Figure 4.5). Florivory also resulted in more

CL fruits (Figure 4.5b). In the seed trait MANOVA, the strongest effects were due to number of seeds per CH and CL fruit rather than seed mass (Table 4.2). When comparing treatment means, nectar robbing reduced seeds per CH fruit, but hand pollination could rescue this (Figure 4.6). Florivory in general reduced seeds per CH fruit in any combination of pollination or nectar robbing treatment (Figure 4.6a).

However, there were more CL seeds per fruit in plants with florivory, especially when there was also nectar robbing, suggesting plants may shunt resources into CL seeds when

CH flowers are damaged (Figure 4.6b).

Discussion

Manipulating floral interactions had the strongest effects on other floral interactions, with less impact on plant traits. There were significant multi-way interactions between all floral treatments, indicating that both antagonist and mutualist behavior varied depending on combinations of previous floral interactions. Similarly, effects on plant reproduction were complex, and were shaped by non-additive combinations of florivory, nectar robbing, and pollination. These results indicate that the context of previous interactions shapes the effect of both floral antagonisms and mutualisms on floral interactions and plant reproduction.

Pollinators, nectar robbers, and nectar thieves, all of whom consume nectar, preferred plants without florivory in at least some conditions. Several other experiments

114

found reduced pollinator visitation after floral damage, often leading to reduced reproduction (Leavitt and Robertson 2006, Cardel and Koptur 2010, Sõber et al. 2010,

Cares-Suarez et al. 2011) (but see McCall 2008). In this experiment, plants with florivory produced fewer flowers (Figure 4.1) and therefore may have had a smaller display, attracting fewer floral visitors (Schmid-Hempel and Speiser 1988, Brody and

Mitchell 1997, Ishii et al. 2008, Kilkenny and Galloway 2008, Irwin et al. 2010).

Although our treatments had little effect on the many floral traits we measured (including two defenses, three attractive traits, and two rewards), florivores may have altered a floral trait we did not measure, such as nectar composition or volatiles, which can alter attractiveness to nectar feeding insects. Interestingly, a previous study in this system manipulating florivory found no effects on pollinators, nectar robbers, or nectar thieves but did decrease leaf herbivory, which we didn’t see in this experiment (Soper Gorden and Adler, in prep). However, this previous study only manipulated florivory; the fact that we found so many significant non-additive effects in this study suggests that it is important to look at multispecies interactions to elucidate significant effects on plants.

In general, it seems that research is lacking on whether pollinators alter antagonist behavior. Studies that include pollinators, cheaters, and herbivores typically assume that antagonists affect pollinators but not vice versa (e.g. Bronstein et al. 2003, Ivey and Carr

2005). In this study, we found that hand pollination significantly increased subsequent pollinator and nectar larcenist visitation, suggesting that pollinators can influence visitation by floral antagonists. However, the fact that hand pollination increased visitation by nectar feeders is somewhat surprising. In many species, pollination leads to rapid reduction in floral attractiveness, to promote cross pollination with other flowers

115

(vanDoorn 1997). This may include color changes, reduced reward production, changes in volatiles, or changes in floral morphology including wilting or closing (Weiss 1991, vanDoorn 1997, Sato 2002, Nuttman and Willmer 2003). Besides direct effects on floral attractiveness or lifespan, pollinators can also change other floral traits including nectar or pollen availability and the presence of pathogens, all of which typically decrease attractiveness to subsequent pollinators and potentially to floral antagonists (Weiss 1991,

Cnaani et al. 2006, Gurevitch et al. 2006). In some species, including species that, like I. capensis, are protandrous with a short female phase, pollination may have no effect floral attractiveness (vanDoorn 1997). However, we could find no record of supplemental pollination making flowers more attractive to other nectar feeders. Since pollinators had no effect on any of the floral attractive or defense traits we measured (Table 4.1), it is unclear why plants with hand-pollinated flowers were consistently more attractive to subsequent pollinators, nectar robbers, and nectar thieves. Regardless of the cause, however, plants with early pollination were more likely to receive more subsequent pollinator visits, as well as increased visitation by nectar larcenists. Since I. capensis plants are able to continually replace nectar (Rust 1977, Marden 1984, Temeles and Pan

2002), the benefits of increased pollinator visits may outweigh any costs of increased nectar larcenists.

Pollinators visited plants with nectar robbing more, as long as those plants weren’t pollinated as well. The significant three-way interaction shows that on plants without florivory, pollinators preferred plants with either pollination or nectar robbing, but not both. A previous study in I. capensis found no difference in pollen receipt between flowers that had been robbed and those that had not (Temeles and Pan 2002),

116

although they did not record pollinator visitation; this suggests a trend opposite to what we saw here, where pollinators preferred plants that had been robbed. Our pollination treatment did not remove nectar, and our nectar robbing treatment did not remove or add pollen, so it seems unlikely that the preference of pollinators for robbed plants is due to changes in rewards. We measured many traits (including nectar production, flower size, pollen production, number of flowers, flower color, and floral defenses), none of which were affected by our nectar robbing treatment. If floral traits do change in response to nectar robbing, it may be a trait we did not measure, such as volatile production, nectar composition, nectar guides, flower gender, or other traits. For example, a previous study in this system found that nectar robbing reduced the length of the male phase without changing floral lifespan, resulting in flowers that spent significantly more time in the female phase (Temeles and Pan 2002). Gender of flowers can affect pollinator preference (Ashman et al. 2000, Huang et al. 2006, Waelti et al. 2009, de Jong et al.

2011), though in nearly all cases pollinators prefer male or hermaphrodite flowers and nectar robbing increased female phase length. It is possible that piercing the corolla lets nectar scents or other floral volatiles escape at a greater rate, perhaps making flowers more attractive or attractive at a greater distance. Alternately, since we recorded interactions on subsequent flowers as well as treated flowers, the piercing and/or nectar removal may signal the plant to change floral traits of future flowers.

The negative impact of florivory appeared to outweigh any effects of hand pollination or nectar robbing on pollinator preference. On plants with florivory, there were low levels of pollinator visitation regardless of the pollination or nectar robbing treatment (Figure 4.2a). Similarly, there seems to be a ranking of the importance of floral

117

interactions for effects on nectar robbers, which have a strong preference for supplemented pollination that both nectar robbing and florivory can cancel out (Figure

4.2b). This suggests that florivores have stronger effects on plants than nectar robbers or pollinators, and have the potential to significantly alter plant-insect interaction communities. It also suggests that increased florivory may reduce selection by pollinators and nectar robbers, since florivory overwhelms the effects of other floral interactions. This same ranking of floral interaction importance seems to translate to effects on reproduction, with florivory having a dominant effect reducing CH fruit and seed traits that masked the positive effects of pollination. Therefore, the ranking of plant- insect interaction importance may have corresponding effects on plant reproduction, fitness, and even selection. For example, in areas or years when florivory is high, florivores may exert the strongest selection pressure on plants regardless of pollination or nectar robbing. Alternately, when florivory is rare, pollinators and nectar robbers may exert concurrent conflicting selection.

Florivores responded to our treatments differently from nectar feeders. Enhanced florivory increased subsequent natural florivory while reducing pollination, nectar robbing, and nectar thieving. Induced susceptibility to florivory after floral damage in I. capensis is consistent with previous research in this system (Soper Gorden and Adler, in prep), which found evidence that generalist herbivores may switch from feeding on leaves to feeding on flowers after floral damage, perhaps because of changes in floral versus leaf attractiveness. Florivores also preferred plants with enhanced nectar robbing, suggesting that multiple types of floral damage increase attractiveness to florivores.

Simply piercing corollas without removing any floral tissue is enough to change

118

interactions with I. capensis flowers (Temeles and Pan 2002), so it is possible that nectar robbers, by chewing through the corolla, can induce the same floral changes as florivores.

Because none of the traits we measured were affected by any treatment, other traits, such as volatile release, must be responsible for the attraction of florivores. Although plants that had either florivore or nectar robbing damage were attractive to subsequent florivores, plants with both kinds of damage tended to be less attractive (Figure 4.4a). It is possible that too much damage or too many kinds of damage can reach a threshold where flowers are no longer attractive to florivores, perhaps through changes in volatiles, symmetry, or plant resource allocation.

While other floral insects preferred plants in the pollination supplementation treatment, florivores avoided them, especially if the plants had no florivory or nectar robbing damage (Figure 4.4b). This suggests that florivores and nectar feeders may be attracted to different floral traits. Indeed, a previous study in this system found that there was a consistent negative correlation between pollinator and florivore choice for floral traits (Soper Gorden and Adler, in prep). For example, while pollinators preferred redder flowers with higher levels of anthocyanins, florivores preferred more yellow flowers with less anthocyanins (Soper Gorden and Adler, in prep). This is an interesting and unexpected result; florivores are theorized to be attracted to similar traits as pollinators

(McCall and Irwin 2006). Additionally, if pollinators and florivores prefer exclusively different floral traits, one would expect strong selection for traits that both attract pollinators and deter florivores. Yet we still see many plants with florivore-preferred traits, and high levels of florivory. It is possible that resource availability limits the ability of plants to produce exclusively pollinator-preferred traits or that there are other

119

trade-offs limiting the benefits of traits preferred by pollinators. For example, nectar robbers and nectar thieves also both preferred pollinated plants (Figures 2, 3) and may exert conflicting selection on pollinator-preferred traits.

Florivores reduced CH flower production but not CL flower production.

Additionally, while there was no main effect of florivory on any measure of reproduction, there were significant three-way interactions on both fruit and seed traits, and plants with florivory tended to have more CL reproduction and less CH reproduction. This suggests that florivory shifts the mating system towards more selfing (CL) reproduction. We have seen a strong reduction in CH outcrossing but not selfed reproduction previously (Soper

Gorden and Adler, in prep). In fact, I. capensis seems able to use a shift in mating system towards increased selfing as a tolerance mechanism against many antagonists, including leaf herbivores and competitors (Steets 2005, Steets et al. 2006a, Steets et al. 2006b).

Since other systems with mixed mating systems have also shown increased selfing in response to floral antagonists or inadequate pollination (Ivey and Carr 2005, Penet et al.

2009, Albert et al. 2011), increased selfing may be a general defense mechanism; this, in turn, may help explain the evolutionary maintenance of mixed mating systems.

With the exception of florivory reducing CH flower production (Figure 4.1), our manipulations had no effect on any of the attractive or defensive floral traits we measured. This suggests that the floral traits we measured are either relatively constant, or that they are shaped more by phenology, genetics, or resource availability that plant- insect interactions. There is evidence that some traits (plant height, number of flowers, and flower color) are constant within plants across the flowering season, though floral anthocyanins and condensed tannins vary (Soper Gorden and Adler, in prep). In other

120

systems, flower color (Irwin and Strauss 2005, Hoballah et al. 2007, Rausher 2008,

Hopkins and Rausher 2012), reward production (Brandenburg et al. 2009, Bolstad et al.

2010), and flower size (Galen 1996, Parachnowitsch and Kessler 2010, Andersson 2012) can have a genetic basis. Together, this may suggest that floral traits are relatively tightly controlled, possibly because of their importance for plant reproduction. While plasticity can be adaptive in some systems (Sultan 2000), sometimes plasticity can alter traits to such a degree that they become detrimental (DeWitt et al. 1998, Langerhans and DeWitt

2002). For example, it may be adaptive to have high constitutive defenses in flowers, rather than defenses induced after damage (Zangerl and Rutledge 1996, Ohnmeiss and

Baldwin 2000, McCall and Fordyce 2010). Despite the limited effects on floral traits, our treatments had multiple significant effects on all floral insects surveyed, which suggests that indirect effects of floral insects on subsequent plant-insect interactions may play an important role in this system.

Florivory, nectar robbing, and pollination had a multivariate three-way interaction in both MANOVAs involving plant reproduction (Table 4.1). However, none of the univariate analyses were significant, suggesting that reproduction in I. capensis is a multivariate trait that depends on both CH and CL reproduction simultaneously. As measures of CH reproduction increased, measures of CL reproduction tended to decrease at nearly the same rate (Table 4.2), suggesting that plants are able to keep total reproduction relatively constant by shifting the mating system from outcrossing to selfing in times of poor resource availability or high levels antagonisms. The significant three- way interaction indicates that floral interactions have complex and context-dependent

121

impacts on plant reproduction, where the effect of one interaction depends on the presence and density of other insect interactions.

All of our treatments were supplementation treatments that led to total levels still within recorded ranges. Our florivory treatment almost exactly doubled natural florivory levels, as hoped; we damaged 25% of flowers by removing 30% of flower tissue, and natural florivory occurred on 26% of flowers with an average of 28% flower tissue missing. Similarly, our pollination treatment was applied to roughly 20% of flowers, while we saw 25% of flowers naturally visited. Our nectar robbing treatment was the most extreme increase; we robbed 25% of flowers, while naturally only 10% of flowers were robbed. While this may seem like a large increase in nectar robbing levels, it is still well within the natural range (Eastman 1995, Young 2008). Additionally, despite the larger increase in nectar robbing, florivory still had the strongest effects, suggesting that the dominant effects of florivory are both strong and robust.

Conclusions. Our manipulations of floral insects had complex effects on subsequent natural floral interactions, including both floral mutualists and antagonists.

While nectar-feeding insects (including pollinators, nectar robbers, and nectar thieves) preferred plants that had been pollinated but without artificial florivory, florivores preferred the opposite – plants that had already been damaged but not pollinated. This suggests that florivores and nectar feeders have opposite responses to floral traits.

Although we tested many measures of floral attractive and defense traits in response to our treatments, the only significant effect was that florivory reduced CH flower production. This, coupled with the context-dependent reduction in CH reproduction with florivory, suggests that plants may shift towards a more selfing mating system in the presence of florivores. Taken together, our results suggest that some antagonists such as 122

nectar robbers may offer conflicting selection with pollinators for floral traits, while others such as florivores may actually select for the same traits as mutualists, leading to complex selection mosaics. Additionally, florivory seems able to reduce selection pressures by nectar robbers and pollinators. Combined with the multivariate interactive effects of floral insects on plant reproduction, our results exemplify the idea that studying pairwise interactions provides an incomplete picture of interaction networks and patterns of selection. In this system, all significant effects – including those on plant reproduction

– differed depending on which floral interactions were present.

123

Tables

124

Table 4.2. Standardized canonical coefficients of response variables from significant three-way MANOVAs testing the effects of manipulations of floral interactions (florivory, nectar robbing, and pollination) on plant reproduction in Impatiens capensis (see Table 4.1). CH = outcrossing; CL = selfing. Standardized Response Variable Canonical Coefficients Fruit Number MANOVA Number of CH fruits -2.8689 Number of CL fruits 2.3495 Proportion of CH vs CL fruits 2.2619 Seed Trait MANOVA Seeds per CH fruit 1.2703 Seeds per CL fruit -0.7148 CH seed mass -0.1807 CL seed mass 0.1010

125

Figures

200 180 160 140 120 100 80 60

Number of CH CH ofFlowers Number 40 20 0 Control Supplemental Florivory

Figure 4.1. Effect of supplemented florivory on the number of outcrossing (CH) flowers produced by Impatiens capensis plants. Error bars show standard error.

126

127

8 A) Natural Florivory 7 Supplemented Florivory

Hour(#) 6 5 4 3 2 1

Nectar Thief Visits per Visits Thief Nectar 0 Natural Nectar Robbing Supplemented Nectar Robbing

Natural Nectar Robbing 8 B) 7 Supplemented Nectar Robbing 6 5 4 3 2 1

Nectar Thief Visits Per Hour(#) Per Visits Thief Nectar 0 Natural Pollination Supplemented Pollination

Figure 4.3. Effects of florivory, nectar robbing, and pollination supplementation treatments on nectar thief visits, tested using a generalized linear model (GLIM) with a Poisson distribution. A) two-way interaction between supplemented florivory and nectar robbing on nectar thieves. B) two-way interaction between supplemented nectar robbing and pollination on nectar thieves. Error bars show standard error.

128

12 A) Natural Florivory 10 Supplemented Florivory

Florivory (%) Florivory 4

0 Natural Nectar Robbing Supplemented Nectar Robbing

12 B) Natural Nectar Robbing 10 Supplemented Nectar Robbing

Florivory (%) Florivory 4

0 Natural Pollination Supplemented Pollination

Figure 4.4. Effects of florivory, nectar robbing, and pollination supplementation treatments on subsequent natural florivory, tested using a generalized linear model (GLIM) with a Poisson distribution. A) two-way interaction between supplemented florivory and nectar robbing on florivory. B) two-way interaction between supplemented nectar robbing and pollination on florivory. Error bars show standard error.

129

Figure 4.5. Effects of florivory, nectar robbing, and pollination on fruit production in Impatiens capensis. A) Number of CH (outcrossing) fruits. B) Number of CL (selfing) fruits. C) Proportion of total fruits that were CH versus CL. Error bars show standard error.

130

Figure 4.6. Effects of florivory, nectar robbing, and pollination on seed traits in Impatiens capensis. A) Number of seeds per CH (outcrossing) fruit. B) Number of seeds per CL (selfing) fruit. C) Average mass of CH seeds. D) Average mass of CL seeds. Error bars show standard error.

131

BIBLIOGRAPHY

Abrahamson, W. G. and B. J. Hershey. 1977. Resource-allocation and growth of

Impatiens capensis (Balsaminaceae) in 2 habitats. Bulletin of the Torrey

Botanical Club 104:160-164.

Abrahamson, W. G., K. D. McCrea, A. J. Whitwell, and L. A. Vernieri. 1991. The role of

phenolics in goldenrod ball gall resistance and formation. Biochemical

Systematics and Ecology 19:615-622.

Abrahamson, W. G. and A. E. Weis. 1987. Nutritional Ecology of Arthropod Gall

Makers. Pages 235-258 in J. Frank Slansky and J. G. Rodriguez, editors.

Nutritional Ecology of Insects, Mites, Spiders, and Related Invertebrates. John

Wiley & Sons, New York.

Adler, L. S. 2000. The ecological significance of toxic nectar. Oikos 91:409-420.

Adler, L. S. and R. E. Irwin. 2005. Ecological costs and benefits of defenses in nectar.

Ecology 86:2968-2978.

Adler, L. S. and R. E. Irwin. 2012. Nectar alkaloids decrease pollination and female

reproduction in a native plant. Oecologia 168:1033-1041.

Adler, L. S., R. Karban, and S. Y. Strauss. 2001. Direct and indirect effects of alkaloids

on plant fitness via herbivory and pollination. Ecology 82:2032-2044.

Adler, L. S., M. Wink, M. Distl, and A. J. Lentz. 2006. Leaf herbivory and nutrients

increase nectar alkaloids. Ecology Letters 9:960-967.

132

Agrawal, A. A., D. D. Ackerly, F. Adler, A. E. Arnold, C. Caceres, D. F. Doak, E. Post,

P. J. Hudson, J. Maron, K. A. Mooney, M. Power, D. Schemske, J. Stachowicz, S.

Strauss, M. G. Turner, and E. Werner. 2007. Filling key gaps in population and

community ecology. Frontiers in Ecology and the Environment 5:145-152.

Agrawal, A. A. and P. A. Van Zandt. 2003. Ecological play in the coevolutionary theatre:

genetic and environmental determinants of attack by a specialist weevil on

milkweed. Journal of Ecology 91:1049-1059.

Albert, L. P., L. G. Campbell, and K. D. Whitney. 2011. Beyond simple reproductive

assurance: Cleistogamy allows adaptive plastic responses to pollen limitation.

International Journal of Plant Sciences 172:862-869.

Aldea, M., J. G. Hamilton, J. P. Resti, A. R. Zangerl, M. R. Berenbaum, and E. H.

DeLucia. 2005. Indirect effects of insect herbivory on leaf gas exchange in

soybean. Plant Cell and Environment 28:402-411.

Aldrich, J. R., A. Khrimian, and M. J. Camp. 2007. Methyl 2,4,6-decatrienoates attract

stink bugs and tachinid parasitoids. Journal of Chemical Ecology 33:801-815.

Anderson, S. S., K. D. McCrea, W. G. Abrahamson, and L. M. Hartzel. 1989. Host

genotype choice by the ball gallmaker Eurosta solidaginis (Diptera, Tephritidae).

Ecology 70:1048-1054.

Andersson, S. 2012. Does inbreeding promote evolutionary reduction of flower size?

Experimental evidence from Crepis tectorum (Asteraceae). American Journal of

Botany 99:1388-1398.

Andrews, E. S., N. Theis, and L. S. Adler. 2007. Pollinator and herbivore attraction to

Cucurbita floral volatiles. Journal of Chemical Ecology 33:1682-1691.

133

Aras, A., G. Cevahir, S. Yentur, F. Eryilmaz, M. Sarsag, and S. Cag. 2007. Investigation

of anthocyanin localization in various parts of Impatiens balsamina L.

Biotechnology & Biotechnological Equipment 21:69-73.

Aronne, G., M. Giovanetti, M. R. Guarracino, and V. de Micco. 2012. Foraging rules of

flower selection applied by colonies of Apis mellifera: ranking and associations of

floral sources. Functional Ecology 26:1186-1196.

Ashman, T. L. 2002. The role of herbivores in the evolution of separate sexes from

hermaphroditism. Ecology 83:1175-1184.

Ashman, T. L., D. H. Cole, and M. Bradburn. 2004. Sex-differential resistance and

tolerance to herbivory in a gynodioecious wild strawberry. Ecology 85:2550-

2559.

Ashman, T. L. and L. Penet. 2007. Direct and indirect effects of a sex-biased antagonist

on male and female fertility: Consequences for reproductive trait evolution in a

gender-dimorphic plant. American Naturalist 169:595-608.

Ashman, T. L., J. Swetz, and S. Shivittz. 2000. Understanding the basis of pollinator

selectivity in sexually dimorphic Fragaria virginiana. Oikos 90:347-356.

Asikainen, E. and P. Mutikainen. 2005. Preferences of pollinators and herbivores in

gynodioecious Geranium sylvaticum. Annals of Botany 95:879-886.

Augspurger, C. K., J. M. Cheeseman, and C. F. Salk. 2005. Light gains and physiological

capacity of understorey woody plants during phenological avoidance of canopy

shade. Functional Ecology 19:537-546.

134

Bagatto, G. and J. D. Shorthouse. 1994. Seasonal acquisition of mineral nutrients by a

Chalcid gall on lowbush blueberry. Entomologia Experimentalis et Applicata

73:61-66.

Banta, J. A., M. H. H. Stevens, and M. Pigliucci. 2010. A comprehensive test of the

'limiting resources' framework applied to plant tolerance to apical meristem

damage. Oikos 119:359-369.

Baraza, E., J. Gomez, J. Hodar, and R. Zamora. 2004. Herbivory has a greater impact in

shade than in sun: Response of Quercus pyrenaica seedlings to multifactorial

environmental variation. Canadian Journal of Botany-Revue Canadienne De

Botanique 82:357-364.

Barbehenn, R. V. and C. P. Constabel. 2011. Tannins in plant-herbivore interactions.

Phytochemistry 72:1551-1565.

Begon, M., C. R. Townsend, and J. L. Harper. 2006. Ecology: From Ecosystems to

Individuals. 4th edition. Blackwell Publishing, Malden, MA.

Behmer, S. T. and A. Joern. 2008. Coexisting generalist herbivores occupy unique

nutritional feeding niches. Proceedings of the National Academy of Sciences of

the United States of America 105:1977-1982.

Benjamin, F. and R. Winfree. 2011. Pollination services from native bees: What scale

matters? 96th Ecological Society of America Annual Meeting. Ecological Society

of America, Austin, TX.

Bigger, D. S. and M. A. Marvier. 1998. How different would a world without herbivory

be?: A search for generality in ecology. Integrative Biology.

135

Bissuel-Belaygue, C., A. A. Cowan, A. H. Marshall, and J. Wery. 2002. Reproductive

development of white clover (Trifolium repens L.) is not impaired by a moderate

water deficit that reduces vegetative growth: II. Fertilization efficiency and seed

set. Crop Science 42:414-422.

Boggs, C. L. and D. W. Inouye. 2012. A single climate driver has direct and indirect

effects on insect population dynamics. Ecology Letters 15:502-508.

Bolstad, G. H., W. S. Armbruster, C. Pelabon, R. Perez-Barrales, and T. F. Hansen. 2010.

Direct selection at the blossom level on floral reward by pollinators in a natural

population of Dalechampia schottii: Full-disclosure honesty? New Phytologist

188:370-384.

Botto-Mahan, C. and M. Ojeda-Camacho. 2000. The importance of floral damage for

pollinator visitation in Alstoemeria ligtu L. Revista Chilena de Entomologia

26:73-76.

Botto-Mahan, C., P. A. Ramirez, C. G. Ossa, R. Medel, M. Ojeda-Camacho, and A. V.

Gonzalez. 2011. Floral herbivory affects female reproductive success and

pollinator visitation in the perennial herb Alstroemeria ligtu (Alstroemeriaceae).

International Journal of Plant Sciences 172:1130-1136.

Brandenburg, A., A. Dell'Olivo, R. Bshary, and C. Kuhlemeier. 2009. The sweetest thing:

Advances in nectar research. Current Opinion in Plant Biology 12:486-490.

Brody, A. K. and R. E. Irwin. 2012. When resources don't rescue: Flowering phenology

and species interactions affect compensation to herbivory in Ipomopsis aggregata.

Oikos 121:1424-1434.

136

Brody, A. K., R. E. Irwin, M. L. McCutcheon, and E. C. Parsons. 2008. Interactions

between nectar robbers and seed predators mediated by a shared host plant,

Ipomopsis aggregata. Oecologia 155:75-84.

Brody, A. K. and R. J. Mitchell. 1997. Effects of experimental manipulation of

inflorescence size on pollination and pre-dispersal seed predation in the

hummingbird-pollinated plant Ipomopsis aggregata. Oecologia 110:86-93.

Bronstein, J. L. 1994. Conditional outcomes in mutualistic interactions. Trends in

Ecology & Evolution 9:214-217.

Bronstein, J. L., R. Alarcon, and M. Geber. 2006. The evolution of plant-insect

mutualisms. New Phytologist 172:412-428.

Bronstein, J. L., W. G. Wilson, and W. E. Morris. 2003. Ecological dynamics of

mutualist/antagonist communities. American Naturalist 162:S24-S39.

Brussland, J. 2007. Anthocyanin Extraction Protocol. California State University, Long

Beach, CA.

Burggraaf, V., G. Waghorn, S. Woodward, and E. Thom. 2008. Effects of condensed

tannins in white clover flowers on their digestion in vitro. Animal Feed Science

and Technology 142:44-58.

Burkle, L. A. and R. E. Irwin. 2009. The effects of nutrient addition on floral characters

and pollination in two subalpine plants, Ipomopsis aggregata and Linum lewisii.

Plant Ecology 203:83-98.

Burkle, L. A. and R. E. Irwin. 2010. Beyond biomass: Measuring the effects of

community-level nitrogen enrichment on floral traits, pollinator visitation and

plant reproduction. Journal of Ecology 98:705-717.

137

Burkle, L. A., R. E. Irwin, and D. A. Newman. 2007. Predicting the effects of nectar

robbing on plant reproduction: Implications of pollen limitation and plant mating

system. American Journal of Botany 94:1935-1943.

Campbell, D. R., M. Bischoff, J. M. Lord, and A. W. Robertson. 2012. Where have all

the blue flowers gone: Pollinator responses and selection on flower colour in

New Zealand Wahlenbergia albomarginata. Journal of Evolutionary Biology

25:352-364.

Campbell, D. R. and K. J. Halama. 1993. Resource and pollen limitations to lifetime seed

production in a natural plant population. Ecology 74:1043-1051.

Canto, A., C. M. Herrera, I. M. Garcia, R. Perez, and M. Vaz. 2011. Intraplant variation

in nectar traits in Helleborus foetidus (Ranunculaceae) as related to floral phase,

environmental conditions and pollinator exposure. Flora 206:668-675.

Cardel, Y. J. and S. Koptur. 2010. Effects of florivory on the pollination of flowers: An

experimental field study with a perennial plant. International Journal of Plant

Sciences 171:283-292.

Cares-Suarez, R., T. Poch, R. F. Acevedo, I. Acosta-Bravo, C. Pimentel, C. Espinoza, R.

A. Cares, P. Munoz, A. V. Gonzalez, and C. Botto-Mahan. 2011. Do pollinators

respond in a dose-dependent manner to flower herbivory?: An experimental

assessment in Loasa tricolor (Loasaceae). Gayana Botanica 68:176-181.

Carroll, A. B., S. G. Pallardy, and C. Galen. 2001. Drought stress, plant water status, and

floral trait expression in fireweed, Epilobium angustifolium (Onagraceae).

American Journal of Botany 88:438-446.

138

Close, D. C. and C. L. Beadle. 2003. The ecophysiology of foliar anthocyanin. Botanical

Review 69:149-161.

Cnaani, J., J. D. Thomson, and D. R. Papaj. 2006. Flower choice and learning in foraging

bumblebees: Effects of variation in nectar volume and concentration. Ethology

112:278-285.

Coley, P. D. 1986. Costs and benefits of defense by tannins in a neotropical tree.

Oecologia 70:238-241.

Coley, P. D., J. P. Bryant, and F. S. Chapin. 1985. Resource availability and plant

antiherbivore defense. Science 230:895-899.

Conner, J. K. and S. Rush. 1996. Effects of flower size and number on pollinator

visitation to wild radish, Raphanus raphanistrum. Oecologia 105:509-516.

Cornelissen, T. and P. Stiling. 2006. Responses of different herbivore guilds to nutrient

addition and natural enemy exclusion. Ecoscience 13:66-74.

Cox, P. A. 1982. Vertebrate pollination and the maintenance of dioecism in Freycinetia.

American Naturalist 120:65-80.

Crespi, B. J., D. A. Carmean, and T. W. Chapman. 1997. Ecology and evolution of

galling thrips and their allies. Annual Review of Entomology 42:51-71.

Cronk, J. K. and M. S. Fennessy. 2001. Weland Plants: Biology and Ecology. CRC

Press, Boca Raton, FL.

Cuevas-Reyes, P., M. Quesada, C. Siebe, and K. Oyama. 2004. Spatial patterns of

herbivory by gall-forming insects: a test of the soil fertility hypothesis in a

Mexican tropical dry forest. Oikos 107:181-189.

139

Damle, M. S., A. P. Giri, M. N. Sainani, and V. S. Gupta. 2005. Higher accumulation of

proteinase inhibitors in flowers than leaves and fruits as a possible basis for

differential feeding preference of Helicoverpa armigera on tomato (Lycopersicon

esculentum Mill, Cv. Dhanashree). Phytochemistry 66:2659-2667.

Darwin, C. 1862. On the Various Contrivances by Which British and Foreign Orchids are

Fertilized by Insects, and on the Good Effects of Intercrossing. John Murray,

London. de Jong, T. J., P. G. L. Klinkhamer, A. Shmida, and F. Thuijsman. 2011. On the

evolution of protandry and the distinction between preference and rank order in

pollinator visitation. Evolutionary Ecology Research 13:307-314. de Souza, R. A., R. Nessim, J. C. Santos, and G. W. Fernandes. 2006. Influence of Apion

sp. (Brentidae, Apioninae) stem-galls on induced resistance and leaf area of

Diospyros hispida (Ebenaceae). Revista Brasileira de Entomologia 50:433-435. de Waal, C., S. C. H. Barrett, and B. Anderson. 2012. The effect of mammalian herbivory

on inflorescence architecture in Ornithophilous babiana (Iridaceae): Implications

for the evolution of a bird perch. American Journal of Botany 99:1096-1103.

Delaney, K. J., F. J. Haile, R. K. D. Peterson, and L. G. Higley. 2008. Impairment of leaf

photosynthesis after insect herbivory or mechanical injury on common milkweed,

Asclepias syriaca. Environmental Entomology 37:1332-1343.

Delpech, R. 2000. The importance of red pigments to plant life: experiments with

anthocyanins. Journal of Biological Education 34:206-210.

DeWitt, T. J., A. Sih, and D. S. Wilson. 1998. Costs and limits of phenotypic plasticity.

Trends in Ecology & Evolution 13:77-81.

140

Dicke, M. and I. T. Baldwin. 2010. The evolutionary context for herbivore-induced plant

volatiles: Beyond the ‘cry for help’. Trends in Plant Science 15:167-175.

Dixon, P., C. Weinig, and J. Schmitt. 2001. Susceptibility to UV damage in Impatiens

capensis (Balsaminaceae): Testing for opportunity costs to shade-avoidance and

population differentiation. American Journal of Botany 88:1401-1408.

Du, D. L., J. A. Winsor, M. Smith, A. Denicco, and A. G. Stephenson. 2008. Resistance

and tolerance to herbivory changes with inbreeding and ontogeny in a wild gourd

(Cucurbitaceae). American Journal of Botany 95:84-92.

Dudt, J. F. and D. J. Shure. 1994. The influence of light and nutrients on foliar phenolics

and insect herbivory. Ecology 75:86-98.

Eastman, J. 1995. Jewelweeds. Pages 91-95 The book of swamp and bog: trees, shrubs,

and wildflowers of eastern freshwater wetlands. Stackpole Books,

Mechanicsburg, PA.

Euler, M. and I. T. Baldwin. 1996. The chemistry of defense and apparency in the

corollas of Nicotiana attenuata. Oecologia 107:102-112.

Fay, P. A. and D. C. Hartnett. 1991. Constraints on growth and allocation patterns of

Silphium integrifolium (Asteraceae) caused by a cynipid gall wasp. Oecologia

88:243-250.

Fay, P. A., D. C. Hartnett, and A. K. Knapp. 1996. Plant tolerance of gall-insect attack

and gall-insect performance. Ecology 77:521-534.

Fenster, C. B., W. S. Armbruster, P. Wilson, M. R. Dudash, and J. D. Thomson. 2004.

Pollination syndromes and floral specialization. Annual Review of Ecology

Evolution and Systematics 35:375-403.

141

Ferrari, M. J., D. L. Du, J. A. Winsor, and A. G. Stephenson. 2007. Inbreeding depression

of plant quality reduces incidence of an insect-borne pathogen in a wild gourd.

International Journal of Plant Sciences 168:603-610.

Ferrari, M. J., A. G. Stephenson, M. C. Mescher, and C. M. De Moraes. 2006. Inbreeding

effects on blossom volatiles in Cucurbita pepdo subsp texana (Cucurbitaceae).

American Journal of Botany 93:1768-1774.

Foss, L. K. and L. K. Rieske. 2004. Stem galls affect oak foliage with potential

consequences for herbivory. Ecological Entomology 29:273-280.

Freeman, R. S., A. K. Brody, and C. D. Neefus. 2003. Flowering phenology and

compensation for herbivory in Ipomopsis aggregata. Oecologia 136:394-401.

Frey, F. M. 2004. Opposing natural selection from herbivores and pathogens may

maintain floral-color variation in Claytonia virginica (Portulacaceae). Evolution

58:2426-2437.

Frey, F. M., R. Davis, and L. F. Delph. 2005. Manipulation of floral symmetry does not

affect seed production in Impatiens pallida. International Journal of Plant

Sciences 166:659-662.

Frölich, C., T. Hartmann, and D. Ober. 2006. Tissue distribution and biosynthesis of 1,2-

saturated pyrrolizidine alkaloids in Phalaenopsis hybrids (Orchidaceae).

Phytochemistry 67:1493-1502.

Frölich, C., D. Ober, and T. Hartmann. 2007. Tissue distribution, core blosynthesis and

diversification of pyrrolizidine alkaloids of the lycopsamine type in three

Boraginaceae species. Phytochemistry 68:1026-1037.

142

Frost, S. W. 1924. A Study of the Leaf-mining Diptera of North America. Cornell

University, Ithaca, NY.

Gagné, R. J. 1989. The Plant-Feeding Gall Midges of North America. Comstock

Publishing Associates, Ithaca, NY.

Galen, C. 1996. Rates of floral evolution: Adaptation to bumblebee pollination in an

alpine wildflower, Polemonium viscosum. Evolution 50:120-125.

Galen, C. and J. Cuba. 2001. Down the tube: Pollinators, predators, and the evolution of

flower shape in the alpine skypilot, Polemonium viscosum. Evolution 55:1963-

1971.

Galen, C., R. Kaczorowski, S. L. Todd, J. Geib, and R. A. Raguso. 2011. Dosage-

dependent impacts of a floral volatile compound on pollinators, larcenists, and the

potential for floral evolution in the alpine skypilot Polemonium viscosum.

American Naturalist 177:258-272.

Gautierhion, A. and F. Maisels. 1994. Mutualism between a leguminous tree and large

African monkeys as pollinators. Behavioral Ecology and Sociobiology 34:203-

210.

Gegear, R. J., J. S. Manson, and J. D. Thomson. 2007. Ecological context influences

pollinator deterrence by alkaloids in floral nectar. Ecology Letters 10:375-382.

Gomulkiewicz, R., S. L. Nuismer, and J. N. Thompson. 2003. Coevolution in variable

mutualisms. American Naturalist 162:S80-S93.

143

Gronquist, M., A. Bezzerides, A. Attygalle, J. Meinwald, M. Eisner, and T. Eisner. 2001.

Attractive and defensive functions of the ultraviolet pigments of a flower

(Hypericum calycinum). Proceedings of the National Academy of Sciences of the

United States of America 98:13745-13750.

Gullen, P. J. and P. S. Cranston. 2005. The Insects: An Outline of Entomology. 3rd

edition. Blackwell Publishing, Malden, MA.

Gurevitch, J., S. M. Scheiner, and G. A. Fox. 2006. The Ecology of Plants. 2nd edition.

Sinauer Associates, Sunderland, MA.

Hagerman, A. E. 2002. The Tannin Handbook. Miami University.

Hargreaves, A. L., L. D. Harder, and S. D. Johnson. 2009. Consumptive emasculation:

The ecological and evolutionary consequences of pollen theft. Biological Reviews

84:259-276.

Hartley, S. E. 1998. The chemical composition of plant galls: are levels of nutrients and

secondary compounds controlled by the gall-former? Oecologia 113:492-501.

Hause, B., I. Stenzel, O. Miersch, H. Maucher, R. Kramell, J. Ziegler, and C.

Wasternack. 2000. Tissue-specific oxylipin signature of tomato flowers: Allene

oxide cyclase is highly expressed in distinct flower organs and vascular bundles.

Plant Journal 24:113-126.

Heath, K. D. and P. Tiffin. 2007. Context dependence in the coevolution of plant and

rhizobial mutualists. Proceedings of the Royal Society B-Biological Sciences

274:1905-1912.

Hegland, S. J., A. Nielsen, A. Lázaro, A.-L. Bjerknes, and Ø. Totland. 2009. How does

climate warming affect plant-pollinator interactions? Ecology Letters 12:184-195.

144

Heil, M., A. Hilpert, B. Fiala, and K. E. Linsenmair. 2001. Nutrient availability and

indirect (biotic) defence in a Malaysian ant-plant. Oecologia 126:404-408.

Held, D. W. and D. A. Potter. 2004. Floral affinity and benefits of dietary mixing with

flowers for a polyphagous scarab, Popillia japonica Newman. Oecologia

140:312-320.

Heschel, M. S., N. Hausmann, and J. Schmitt. 2005. Testing for stress-dependent

inbreeding depression in Impatiens capensis (Balsaminaceae). American Journal

of Botany 92:1322-1329.

Heschel, M. S., J. R. Stinchcombe, K. E. Holsinger, and J. Schmitt. 2004. Natural

selection on light response curve parameters in the herbaceous annual, Impatiens

capensis. Oecologia 139:487-494.

Hoballah, M. E., T. Gubitz, J. Stuurman, L. Broger, M. Barone, T. Mandel, A.

Dell'Olivo, M. Arnold, and C. Kuhlemeier. 2007. Single gene-mediated shift in

pollinator attraction in Petunia. Plant Cell 19:779-790.

Hopkins, R. and M. D. Rausher. 2012. Pollinator-mediated selection on flower color

allele drives reinforcement. Science 335:1090-1092.

Hopkins, W. G. and N. P. A. Hüner. 2004. Introduction to Plant Physiology. 3rd edition.

John Wiley & Sons, Hoboken, NJ.

Huang, S. Q., L. L. Tang, J. F. Sun, and Y. Lu. 2006. Pollinator response to female and

male floral display in a monoecious species and its implications for the evolution

of floral dimorphism. New Phytologist 171:417-424.

Huberty, A. F. and R. F. Denno. 2004. Plant water stress and its consequences for

herbivorous insects: A new synthesis. Ecology 85:1383-1398.

145

Huhta, A. P., T. Lennartsson, J. Tuomi, P. Rautio, and K. Laine. 2000. Tolerance of

Gentianella campestris in relation to damage intensity: An interplay between

apical dominance and herbivory. Evolutionary Ecology 14:373-392.

Hummel, C. M. 1956. The biology, ecology, and description of the jewelweed flower-bud

midge Schizomyia impatientis (Osten Saken). Cornell University.

Hurlbert, S. H. 1971. The nonconcept of species diversity: A critique and alternative

parameters. Ecology 52:577-586.

Ingersoll, C. M., R. A. Niesenbaum, C. E. Weigle, and J. H. Lehman. 2010. Total

phenolics and individual phenolic acids vary with light environment in Lindera

benzoin. Botany-Botanique 88:1007-1010.

Inouye, D. W. 1980. The terminology of floral larceny. Ecology 61:1251-1253.

Irwin, R. E. 2009. Realized tolerance to nectar robbing: Compensation to floral enemies

in Ipomopsis aggregata. Annals of Botany 103:1425-1433.

Irwin, R. E. and L. S. Adler. 2006. Correlations among traits associated with herbivore

resistance and pollination: Implications for pollination and nectar robbing in a

distylous plant. American Journal of Botany 93:64-72.

Irwin, R. E., L. S. Adler, and A. A. Agrawal. 2004. Community and evolutionary ecology

of nectar. Ecology 85:1477-1478.

Irwin, R. E. and A. K. Brody. 1998. Nectar robbing in Ipomopsis aggregata: effects on

pollinator behavior and plant fitness. Oecologia 116:519-527.

Irwin, R. E., J. L. Bronstein, J. S. Manson, and L. Richardson. 2010. Nectar robbing:

Ecological and evolutionary perspectives. Pages 271-292 Annual Review of

Ecology, Evolution, and Systematics, Vol 41.

146

Irwin, R. E., C. Galen, J. J. Rabenold, R. Kaczorowski, and M. L. McCutcheon. 2008.

Mechanisms of tolerance to floral larceny in two wildflower species. Ecology

89:3093-3104.

Irwin, R. E. and J. E. Maloof. 2002. Variation in nectar robbing over time, space, and

species. Oecologia 133:525-533.

Irwin, R. E. and S. Y. Strauss. 2005. Flower color microevolution in wild radish:

Evolutionary response to pollinator-mediated selection. American Naturalist

165:225-237.

Irwin, R. E., S. Y. Strauss, S. Storz, A. Emerson, and G. Guibert. 2003. The role of

herbivores in the maintenance of a flower color polymorphism in wild radish.

Ecology 84:1733-1743.

Ishii, H. S., Y. Hirabayashi, and G. Kudo. 2008. Combined effects of inflorescence

architecture, display size, plant density and empty flowers on bumble bee

behaviour: experimental study with artificial inflorescences. Oecologia 156:341-

350.

Ivey, C. T. and D. E. Carr. 2005. Effects of herbivory and inbreeding on the pollinators

and mating system of Mimulus guttatus (Phrymaceae). American Journal of

Botany 92:1641-1649.

Johnson, E. T., M. A. Berhow, and P. F. Dowd. 2008. Colored and white sectors from

star-patterned petunia flowers display differential resistance to corn earworm and

cabbage looper larvae. Journal of Chemical Ecology 34:757-765.

147

Johnson, M. T. J. and A. A. Agrawal. 2007. Covariation and composition of arthropod

species across plant genotypes of evening primrose (Oenothera biennis). Oikos

116:941-956.

Johnson, S. D., A. L. Hargreaves, and M. Brown. 2006. Dark, bitter-tasting nectar

functions as a filter of flower visitors in a bird-pollinated plant. Ecology 87:2709-

2716.

Jones, M. E., M. E. Fenn, and T. D. Paine. 2011. The effect of nitrogen additions on

bracken fern and its insect herbivores at sites with high and low atmospheric

pollution. Arthropod-Plant Interactions 5:163-173.

Kaczorowski, R. L., A. R. Seliger, A. C. Gaskett, S. K. Wigsten, and R. A. Raguso. 2012.

Corolla shape vs. size in flower choice by a nocturnal hawkmoth pollinator.

Functional Ecology 26:577-587.

Kaplan, I. and R. F. Denno. 2007. Interspecific interactions in phytophagous insects

revisited: a quantitative assessment of competition theory. Ecology Letters

10:977-994.

Kaplan, I., S. Sardanelli, B. J. Rehill, and R. F. Denno. 2011. Toward a mechanistic

understanding of competition in vascular-feeding herbivores: An empirical test of

the sink competition hypothesis. Oecologia 166:627-636.

Karageorgou, P. and Y. Manetas. 2006. The importance of being red when young:

anthocyanins and the protection of young leaves of Quercus coccifera from insect

herbivory and excess light. Tree Physiology 26:613-621.

148

Karban, R., A. A. Agrawal, J. S. Thaler, and L. S. Adler. 1999. Induced plant responses

and information content about risk of herbivory. Trends in Ecology & Evolution

14:443-447.

Karban, R. and C. Niiho. 1995. Induced resistance and susceptibility to herbivory: Plant

memory and altered plant development. Ecology 76:1220-1225.

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.

Katjiua, M. L. J. and D. Ward. 2006. Resistance and tolerance of Terminalia sericea trees

to simulated herbivore damage under different soil nutrient and moisture

conditions. Journal of Chemical Ecology 32:1431-1443.

Kawano, S., M. Odaki, R. Yamaoka, M. Ooa-Tanabe, M. Takeuchi, and N. Kawano.

1995. Pollination biology of Oenothera (Onagraceae): The interplay between

floral UV-absorbancy patterns and floral volatiles as signals to nocturnal insects.

Plant Species Biology 10:31-38.

Kehl, A., S. Dotterl, G. Aas, and G. Rambold. 2010. Is flower scent influencing host

plant selection of leaf-galling sawflies (Hymenoptera, Tenthredinidae) on

willows? Chemoecology 20:215-221.

Kennedy, P. G. and K. G. Peay. 2007. Different soil moisture conditions change the

outcome of the ectomycorrhizal symbiosis between Rhizopogon species and Pinus

muricata. Plant and Soil 291:155-165.

Kersch, M. F. and C. R. Fonseca. 2005. Abiotic factors and the conditional outcome of an

ant-plant mutualism. Ecology 86:2117-2126.

149

Kessler, A. and I. T. Baldwin. 2001. Defensive function of herbivore-induced plant

volatile emissions in nature. Science 291:2141-2144.

Kessler, A. and R. Halitschke. 2009. Testing the potential for conflicting selection on

floral chemical traits by pollinators and herbivores: Predictions and case study.

Functional Ecology 23:901-912.

Kessler, D., C. Diezel, D. G. Clark, T. A. Colquhoun, and I. T. Baldwin. 2012. Petunia

flowers solve the defence/apparency dilemma of pollinator attraction by

deploying complex floral blends. Ecology Letters:n/a-n/a.

Kilkenny, F. F. and L. F. Galloway. 2008. Reproductive success in varying light

environments: direct and indirect effects of light on plants and pollinators.

Oecologia 155:247-255.

Klein, A. M., B. E. Vaissiere, J. H. Cane, I. Steffan-Dewenter, S. A. Cunningham, C.

Kremen, and T. Tscharntke. 2007. Importance of pollinators in changing

landscapes for world crops. Proceedings of the Royal Society B-Biological

Sciences 274:303-313.

Koes, R., W. Verweij, and F. Quattrocchio. 2005. Flavonoids: a colorful model for the

regulation and evolution of biochemical pathways. Trends in Plant Science

10:236-242.

Kohyani, P. T., B. Bossuyt, D. Bonte, and M. Hoffmann. 2009. Differential herbivory

tolerance of dominant and subordinate plant species along gradients of nutrient

availability and competition. Plant Ecology 201:611-619.

150

Koslow, J. M. and K. Clay. 2007. The mixed mating system of Impatiens capensis and

infection by a foliar rust pathogen: Patterns of resistance and fitness

consequences. Evolution 61:2643-2654.

Krupnick, G. A. and A. E. Weis. 1999. The effect of floral herbivory on male and female

reproductive success in Isomeris arborea. Ecology 80:135-149.

Krupnick, G. A., A. E. Weis, and D. R. Campbell. 1999. The consequences of floral

herbivory for pollinator service to Isomeris arborea. Ecology 80:125-134.

Langerhans, R. B. and T. J. DeWitt. 2002. Plasticity constrained: over-generalized

induction cues cause maladaptive phenotypes. Evolutionary Ecology Research

4:857-870.

Lanza, J., G. C. Smith, S. Sack, and A. Cash. 1995. Variation in nectar volume and

composition of Impatiens capensis at the individual, plant, and population levels.

Oecologia 102:113-119.

Lara, C. and J. F. Ornelas. 2001. Preferential nectar robbing of flowers with long

corollas: experimental studies of two hummingbird species visiting three plant

species. Oecologia 128:263-273.

Larson, K. C. 1998. The impact of two gall-forming arthropods on the photosynthetic

rates of their hosts. Oecologia 115:161-166.

Larson, K. C. and T. G. Whitham. 1991. Manipulation of food resources by a gall-

forming aphid: The physiology of sink-source interactions. Oecologia 88:15-21.

Larson, K. C. and T. G. Whitham. 1997. Competition between gall aphids and natural

plant sinks: Plant architecture affects resistance to galling. Oecologia 109:575-

582.

151

Leavitt, H. and I. C. Robertson. 2006. Petal herbivory by chrysomelid beetles

(Phyllotreta sp.) is detrimental to pollination and seed production in Lepidium

papilliferum (Brassicaceae). Ecological Entomology 31:657-660.

Lechowicz, M. J., D. J. Schoen, and G. Bell. 1988. Environmental correlates of habitat

distribution and fitness components in Impatiens capensis and Impatiens pallida.

Journal of Ecology 76:1043-1054.

Leck, M. A. 1979. Germination behavior of Impatiens capensis Meerb. (Balsaminaceae).

Bartonia 46:1-14.

Lee, D. W. 1985. Duplicating foliage shade for research on plant development.

HortScience 20:116-118.

Leege, L. M. 2006. The relationship between psyllid leaf galls and redbay (Persea

borbonia) fitness traits in sun and shade. Plant Ecology 184:203-212.

Leege, L. M. and L. M. Wolfe. 2002. Do floral herbivores respond to variation in flower

characteristics in Gelsemium sempervirens (Loganiaceae), a distylous vine?

American Journal of Botany 89:1270-1274.

Lehrer, M., G. A. Horridge, S. W. Zhang, and R. Gadagkar. 1995. Shape vision in bees:

Innate preference for flower-like patterns. Philosophical Transactions of the Royal

Society of London Series B-Biological Sciences 347:123-137.

Liu, F., Y. Y. Liao, W. Li, J. M. Chen, Q. F. Wang, and T. J. Motley. 2010. The effect of

pollination on resource allocation among sexual reproduction, clonal

reproduction, and vegetative growth in Sagittaria potamogetifolia (Alismataceae).

Ecological Research 25:495-499.

152

Lively, C. M., S. G. Johnson, L. F. Delph, and K. Clay. 1995. Thinning reduces the effect

of rust infection on jewelweed (Impatiens capensis). Ecology 76:1859-1862.

Losey, J. E. and M. Vaughan. 2006. The economic value of ecological services provided

by insects. Bioscience 56:311-323.

Louda, S. M. 1982. Inflorescence spiders: A cost-benefit analysis for the host plant,

Haplopappus venetus Blake (Asteraceae). Oecologia 55:185-191.

Lu, Y., Y. B. Luo, and S. Q. Huang. 2012. Effects of soil moisture and floral herbivory

on sexual expression in a gynodioecious orchid. Journal of Systematics and

Evolution 50:454-459.

Lucas-Barbosa, D., J. J. A. van Loon, and M. Dicke. 2011. The effects of herbivore-

induced plant volatiles on interactions between plants and flower-visiting insects.

Phytochemistry 72:1647-1654.

Luo, S. X., S. M. Chaw, D. X. Zhang, and S. S. Renner. 2010. Flower heating following

anthesis and the evolution of gall midge pollination in Schisandraceae. American

Journal of Botany 97:1220-1228.

Malerba, R. and J. Nattero. 2012. Pollinator response to flower color polymorphism and

floral display in a plant with a single-locus floral color polymorphism:

consequences for plant reproduction. Ecological Research 27:377-385.

Maliakal, S. K., K. McDonnell, S. A. Dudley, and J. Schmitt. 1999. Effects of red to far-

red ratio and plant density on biomass allocation and gas exchange in Impatiens

capensis. International Journal of Plant Sciences 160:723-733.

153

Malo, J. E., J. Leirana-Alcocer, and V. Parra-Tabla. 2001. Population fragmentation,

florivory, and the effects of flower morphology alterations on the pollination

success of Myrmecophila tibicinis (Orchidaceae). Biotropica 33:529-534.

Maloof, J. E. and D. W. Inouye. 2000. Are nectar robbers cheaters or mutualists?

Ecology 81:2651-2661.

Mancinelli, A. L. 1990. Interaction between light quality and light quantity in the

photoregulation of anthocyanin production. Plant Physiology 92:1191-1195.

Marden, J. H. 1984. Intrapopulation variation in nectar secretion in Impatiens capensis.

Oecologia 63:418-422.

Marshall, C. B., G. Avila-Sakar, and E. G. Reekie. 2008. Effects of nutrient and CO2

availability on tolerance to herbivory in Brassica rapa. Plant Ecology 196:1-13.

Martin, N. H. 2004. Flower size preferences of the honeybee (Apis mellifera) foraging on

Mimulus guttatus (Scrophulariaceae). Evolutionary Ecology Research 6:777-782.

Mathews, S. and D. Tremonte. 2012. Tests of the link between functional innovation and

positive selection at phytochrome A: The phylogenetic distribution of far-red

high-irradiance responses in seedling development. International Journal of Plant

Sciences 173:662-672.

Mattson, W. J. 1980. Herbivory in relation to plant nitrogen-content. Annual Review of

Ecology and Systematics 11:119-161.

McCall, A. C. 2006. Natural and artificial floral damage induces resistance in Nemophila

menziesii (Hydrophyllaceae) flowers. Oikos 112:660-666.

McCall, A. C. 2008. Florivory affects pollinator visitation and female fitness in

Nemophila menziesii. Oecologia 155:729-737.

154

McCall, A. C. and J. A. Fordyce. 2010. Can optimal defence theory be used to predict the

distribution of plant chemical defences? Journal of Ecology 98:985-992.

McCall, A. C. and R. E. Irwin. 2006. Florivory: the intersection of pollination and

herbivory. Ecology Letters 9:1351-1365.

McCall, A. C. and R. Karban. 2006. Induced defense in Nicotiana attenuata (Solanaceae)

fruit and flowers. Oecologia 146:566-571.

McCrea, K. D. and W. G. Abrahamson. 1987. Variation in herbivore infestation:

Historical vs. genetic factors. Ecology 68:822-827.

Mitchell-Olds, T. and D. M. Waller. 1985. Relative performance of selfed and outcrossed

progeny in Impatiens capensis. Evolution 39:533-544.

Mitchellolds, T. and D. M. Waller. 1985. Relative performance of selfed and outcrossed

progeny in Impatiens capensis. Evolution 39:533-544.

Møller, A. P. and M. Eriksson. 1995. Pollinator preference for symmetrical flowers and

sexual selection in plants. Oikos 73:15-22.

Mothershead, K. a. R. J. M. 2000. Fitness impacts of herbivory through indirect effects

on plant-pollinator interactions in Oenothera macrocarpa. Ecology 81:30-40.

Mougi, A. and M. Kondoh. 2012. Diversity of interaction types and ecological

community stability. Science 337:349-351.

Muñoz, A., C. Celedon-Neghme, L. A. Cavieres, and M. T. K. Arroyo. 2005. Bottom-up

effects of nutrient availability on flower production, pollinator visitation, and seed

output in a high-Andean shrub. Oecologia 143:126-135.

155

Muth, N. Z., E. C. Kluger, J. H. Levy, M. J. Edwards, and R. A. Niesenbaum. 2008.

Increased per capita herbivory in the shade: Necessity, feedback, or luxury

consumption? Ecoscience 15:182-188.

Nabity, P. D., J. A. Zavala, and E. H. DeLucia. 2009. Indirect suppression of

photosynthesis on individual leaves by arthropod herbivory. Annals of Botany

103:655-663.

National Weather Service. 2012. Observed Weather for Worcester MA. National Oceanic

and Atmospheric Administration.

Navarro, L. 2001. Reproductive biology and effect of nectar robbing on fruit production

in Macleania bullata (Ericaceae). Plant Ecology 152:59-65.

Neff, M. M. and J. Chory. 1998. Genetic interactions between phytochrome A,

phytochrome B, and cryptochrome 1 during Arabidopsis development. Plant

Physiology 118:27-36.

Nuttman, C. and P. Willmer. 2003. How does insect visitation trigger floral colour

change? Ecological Entomology 28:467-474.

Ohgushi, T. 2008. Herbivore-induced indirect interaction webs on terrestrial plants: the

importance of non-trophic, indirect, and facilitative interactions. Entomologia

Experimentalis et Applicata 128:217-229.

Ohnmeiss, T. E. and I. T. Baldwin. 2000. Optimal Defense Theory predicts the ontogeny

of an induced nicotine defense. Ecology 81:1765-1783.

Olivieri, I., D. Couvet, and M. Slatkin. 1994. Allocation of reproductive effort in

perennial plants under pollen limitation. American Naturalist 144:373-394.

156

Ollerstam, O. and S. Larsson. 2003. Salicylic acid mediates resistance in the willow Salix

viminalis against the gall midge Dasineura marginemtorquens. Journal of

Chemical Ecology 29:163-174.

Ollerton, J., R. Winfree, and S. Tarrant. 2011. How many flowering plants are pollinated

by animals? Oikos 120:321-326.

Onoda, Y., F. Schieving, and N. P. R. Anten. 2008. Effects of light and nutrient

availability on leaf mechanical properties of Plantago major: A conceptual

approach. Annals of Botany 101:727-736.

Ort, D. R. 2001. When there is too much light. Plant Physiology 125:29-32.

Parachnowitsch, A. L. and A. Kessler. 2010. Pollinators exert natural selection on flower

size and floral display in Penstemon digitalis. New Phytologist 188:393-402.

Pascual-Alvarado, E., P. Cuevas-Reyes, M. Quesada, and K. Oyama. 2008. Interactions

between galling insects and leaf-feeding insects: the role of plant phenolic

compounds and their possible interference with herbivores. Journal of Tropical

Ecology 24:329-336.

Penet, L., C. L. Collin, and T. L. Ashman. 2009. Florivory increases selfing: An

experimental study in the wild strawberry, Fragaria virginiana. Plant Biology

11:38-45.

Petanidou, T. and E. Smets. 1996. Does temperature stress induce nectar secretion in

Mediterranean plants? New Phytologist 133:513-518.

Petanidou, T., A. Van Laere, W. N. Ellis, and E. Smets. 2006. What shapes amino acid

and sugar composition in Mediterranean floral nectars? Oikos 115:155-169.

157

Poelman, E. H., C. Broekgaarden, J. J. A. Van Loon, and M. Dicke. 2008. Early season

herbivore differentially affects plant defence responses to subsequently colonizing

herbivores and their abundance in the field. Molecular Ecology 17:3352-3365.

Prada, M., O. J. Marini-Filho, and P. W. Price. 1995. Insects in flower heads of Aspilia

foliacea (Asteraceae) after a fire in a central Brazilian savanna: Evidence for the

plant vigor hypothesis. Biotropica 27:513-518.

Price, P. W. 1991. The plant vigor hypothesis and herbivore attack. Oikos 62:244-251.

Rausher, M. D. 2008. Evolutionary transitions in floral color. International Journal of

Plant Sciences 169:7-21.

Richardson, S. C. 2004. Are nectar-robbers mutualists or antagonists? Oecologia

139:246-254.

Rodriguez-Rajo, F. J., J. Mendez, and V. Jato. 2005. Factors affecting pollination ecology

of Quercus anemophilous species in north-west Spain. Botanical Journal of the

Linnean Society 149:283-297.

Rossi, A. M. and P. Stiling. 1998. The interactions of plant clone and abiotic factors on a

gall-making midge. Oecologia 116:170-176.

Rust, R. W. 1977. Pollination in Impatiens capensis and Impatiens pallida

(Balsaminaceae). Bulletin of the Torrey Botanical Club 104:361-367.

Rust, R. W. 1979. Pollination of Impatiens capensis: Pollinators and nectar robbers.

Journal of the Kansas Entomological Society 52:297-308.

Sakai, S. 2002. A review of brood-site pollination mutualism: plants providing breeding

sites for their pollinators. Journal of Plant Research 115:161-168.

158

Sato, H. 2002. The role of autonomous self-pollination in floral longevity in varieties of

Impatiens hypophylla (Balsaminaceae). American Journal of Botany 89:263-269.

Scheiner, S. M. 2001. MANOVA: Multiple Response Variables and Multispecies

Interactions. Pages 99-115 in J. Gurevitch and S. M. Scheiner, editors. Design and

Analysis of Ecological Experiments. Oxford University Press, Cary, NC.

Schemske, D. W. 1978. Evolution of reproductive characteristics in Impatiens

(Balsaminaceae) - Significance of cleistogamy and chasmogamy. Ecology

59:596-613.

Schemske, D. W. and H. D. Bradshaw. 1999. Pollinator preference and the evolution of

floral traits in monkeyflowers (Mimulus). Proceedings of the National Academy

of Sciences of the United States of America 96:11910-11915.

Schmid-Hempel, P. and B. Speiser. 1988. Effects of inflorescence size on pollination in

Epilobium angustifolium. Oikos 53:98-104.

Schmitt, J. 1993. Reaction norms of morphological and life-history traits to light

availability in Impatiens capensis. Evolution 47:1654-1668.

Schmitt, J., D. Ehrhardt, and D. Swartz. 1985. Differential dispersal of self-fertilized and

outcrossed progeny in jewelweed (Impatiens capensis). American Naturalist

126:570-575.

Schowalter, T. D., D. C. Lightfoot, and W. G. Whitford. 1999. Diversity of arthropod

responses to host-plant water stress in a desert ecosystem in southern New

Mexico. American Midland Naturalist 142:281-290.

Schultz, B. B. 1992. Insect herbivores as potential causes of mortality and adaptation in

gallforming insects. Oecologia 90:297-299.

159

Schulz, K. E., M. Smith, and Y. J. Wu. 1993. Gas-exchange of Impatiens pallida Nutt.

(Balsaminaceae) in relation to wilting under high light. American Journal of

Botany 80:361-368.

Seibert, T. F. 1993. A nectar-secreting gall wasp and ant mutualism: Selection and

counter-selection shaping gall wasp phenology, fecundity, and persistence.

Ecological Entomology 18:247-253.

Shorthouse, J. D. and O. Rohfritsch. 1992. Biology of Insect-Induced Galls. Oxford

University Press, New York.

Siemann, E. 1998. Experimental tests of effects of plant productivity and diversity on

grassland arthropod diversity. Ecology 79:2057-2070.

Simpson, R. L., M. A. Leck, and V. T. Parker. 1985. The comparative ecology of

Impatiens capensis Meerb (Balsaminaceae) in central New Jersey. Bulletin of the

Torrey Botanical Club 112:295-311.

Sõber, V., M. Moora, and T. Teder. 2010. Florivores decrease pollinator visitation in a

self-incompatible plant. Basic and Applied Ecology 11:669-675.

Steets, J. A. 2005. Antagonists and mixed mating: Consequences for the demography of

Impatiens capensis (Balsaminaceae). University of Pittsburgh.

Steets, J. A. and T. L. Ashman. 2004. Herbivory alters the expression of a mixed-mating

system. American Journal of Botany 91:1046-1051.

Steets, J. A., J. L. Hamrick, and T. L. Ashman. 2006a. Consequences of vegetative

herbivory for maintenance of intermediate outcrossing in an annual plant.

Ecology 87:2717-2727.

160

Steets, J. A., R. Salla, and T. L. Ashman. 2006b. Herbivory and competition interact to

affect reproductive traits and mating system expression in Impatiens capensis.

American Naturalist 167:591-600.

Stiling, P. and D. C. Moon. 2005. Quality or quantity: the direct and indirect effects of

host plants on herbivores and their natural enemies. Oecologia 142:413-420.

Stiling, P. and A. M. Rossi. 1997. Experimental manipulations of top-down and bottom-

up factors in a tri-trophic system. Ecology 78:1602-1606.

Stinchcombe, J. R. and M. D. Rausher. 2002. The evolution of tolerance to deer

herbivory: modifications caused by the abundance of insect herbivores.

Proceedings of the Royal Society of London Series B-Biological Sciences

269:1241-1246.

Stone, G. N. and K. Schönrogge. 2003. The adaptive significance of insect gall

morphology. Trends in Ecology & Evolution 18:512-522.

Strauss, S. Y. and R. E. Irwin. 2004. Ecological and evolutionary consequences of

multispecies plant-animal interactions. Annual Review of Ecology Evolution and

Systematics 35:435-466.

Strauss, S. Y., R. E. Irwin, and V. M. Lambrix. 2004. Optimal defence theory and flower

petal colour predict variation in the secondary chemistry of wild radish. Journal of

Ecology 92:132-141.

Sultan, S. E. 2000. Phenotypic plasticity for plant development, function and life history.

Trends in Plant Science 5:537-542.

Sultan, S. E. 2010. Plant developmental responses to the environment: Eco-devo insights.

Current Opinion in Plant Biology 13:96-101.

161

Sun, Y., J. Ding, and M. J. Frye. 2010. Effects of resource availability on tolerance of

herbivory in the invasive Alternanthera philoxeroides and the native

Alternanthera sessilis. Weed Research 50:527-536.

Suwa, T. and H. Maherali. 2008. Influence of nutrient availability on the mechanisms of

tolerance to herbivory in an annual grass, Avena barbata (Poaceae). American

Journal of Botany 95:434-440.

Tanaka, Y., N. Sasaki, and A. Ohmiya. 2008. Biosynthesis of plant pigments:

Anthocyanins, betalains and carotenoids. Plant Journal 54:733-749.

Temeles, E. J. and I. L. Pan. 2002. Effect of nectar robbery on phase duration, nectar

volume, and pollination in a protandrous plant. International Journal of Plant

Sciences 163:803-808.

Thompson, J. N. 2002. Plant-animal Interactions: Future Directions. Pages 236-247 in C.

M. Herrera and O. Pellmyr, editors. Plant-animal Interactions: An Evolutionary

Approach. Blackwell Science Ltd, Malden, MA.

Thompson, J. N. and B. M. Cunningham. 2002. Geographic structure and dynamics of

coevolutionary selection. Nature 417:735-738.

Tooker, J. F. and C. M. De Moraes. 2007. Feeding by Hessian fly [Mayetiola destructor

(Say)] larvae does not induce plant indirect defenses. Ecological Entomology

32:153-161.

Underwood, N. 2000. Density dependence in induced plant resistance to herbivore

damage: threshold, strength and genetic variation. Oikos 89:295-300.

Underwood, N. 2010. Density dependence in insect performance within individual plants:

Induced resistance to Spodoptera exigua in tomato. Oikos 119:1993-1999.

162

Underwood, N. C. 1998. The timing of induced resistance and induced susceptibility in

the soybean-Mexican bean beetle system. Oecologia 114:376-381.

Utsumi, S., M. Nakamura, and T. Ohgushi. 2009. Community consequences of

herbivore-induced bottom-up trophic cascades: the importance of resource

heterogeneity. Journal of Animal Ecology 78:953-963.

Utsumi, S. and T. Ohgushi. 2008. Host plant variation in plant-mediated indirect effects:

moth boring-induced susceptibility of willows to a specialist leaf beetle.

Ecological Entomology 33:250-260.

Van Zandt, P. A. and A. A. Agrawal. 2004a. Community-wide impacts of herbivore-

induced plant responses in milkweed (Asclepias syriaca). Ecology 85:2616-2629.

Van Zandt, P. A. and A. A. Agrawal. 2004b. Specificity of induced plant responses to

specialist herbivores of the common milkweed Asclepias syriaca. Oikos 104:401-

409. vanDoorn, W. G. 1997. Effects of pollination on floral attraction and longevity. Journal

of Experimental Botany 48:1615-1622.

Verhoeven, J. T. A., W. Koerselman, and A. F. M. Meuleman. 1996. Nitrogen- or

phosphorus-limited growth in herbaceous, wet vegetation: Relations with

atmospheric inputs and management regimes. Trends in Ecology & Evolution

11:494-497.

Villarreal, A. G. and C. E. Freeman. 1990. Effects of temperature and water-stress on

some floral nectar characteristics in Ipomopsis longiflora (Polemoniaceae) under

controlled conditions. Botanical Gazette 151:5-9.

163

von Wettberg, E. J. and J. Schmitt. 2005. Physiological mechanism of population

differentiation in shade-avoidance responses between woodland and clearing

genotypes of Impatiens capensis. American Journal of Botany 92:868-874. von Wettberg, E. J. B., J. R. Stinchcombe, and J. Schmitt. 2012. Early developmental

responses to seedling environment modulate later plasticity to light spectral

quality. Plos One 7:e34121.

Wackers, F. L. and T. M. Bezemer. 2003. Root herbivory induces an above-ground

indirect defence. Ecology Letters 6:9-12.

Waelti, M. O., P. A. Page, A. Widmer, and F. P. Schiestl. 2009. How to be an attractive

male: floral dimorphism and attractiveness to pollinators in a dioecious plant.

BMC Evolutionary Biology 9.

Walisch, T. J., G. Colling, M. Poncelet, and D. Matthies. 2012. Effects of inbreeding and

interpopulation crosses on the performance and plasticity of two generations of

offspring of a declining grassland plant. American Journal of Botany 99:1300-

1313.

Waller, D. M. 1979. The relative costs of self- and cross-fertilized seeds in Impatiens

capensis (Balsaminaceae). American Journal of Botany 66:313-320.

Waller, D. M. 1980. Environmental determinants of outcrossing in Impatiens capensis

(Balsaminaceae). Evolution 34:747-761.

Waller, D. M. 1985. The genesis of size hierarchies in seedling populations of Impatiens

capensis Meerb. New Phytologist 100:243-260.

Walling, L. L. 2000. The myriad plant responses to herbivores. Journal of Plant Growth

Regulation 19:195-216.

164

Walters, B. B. and E. W. Stiles. 1996. Effect of canopy gaps and flower patch size on

pollinator visitation of Impatiens capensis. Bulletin of the Torrey Botanical Club

123:184-188.

Waser, N. M. and M. V. Price. 1981. Pollinator choice and stabilizing selection for

flower color in Delphinium nelsonii. Evolution 35:376-390.

Washitani, I., Y. Okayama, K. Sato, H. Takahashi, and T. Ohgushi. 1996. Spatial

variation in female fertility related to interactions with flower consumers and

pathogens in a forest metapopulation of Primula sieboldii. Researches on

Population Ecology 38:249-256.

Waterman, P. G., G. M. Choo, A. L. Vedder, and D. Watts. 1983. Digestibility,

Digestion-Inhibitors and Nutrients of Herbaceous Foliage and Green Stems from

an African Montane Flora and Comparison with Other Tropical Flora. Oecologia

60:244-249.

Weinig, C., K. A. Gravuer, N. C. Kane, and J. Schmitt. 2004. Testing adaptive plasticity

to UV: Costs and benefits of stem elongation and light-induced phenolics.

Evolution 58:2645-2656.

Weiss, M. R. 1991. Floral color changes as cues for pollinators. Nature 354:227-229.

White, T. C. R. 1984. The abundance of invertebrate herbivores in relation to the

availability of nitrogen in stressed food plants. Oecologia 63:90-105.

Whitehead, S. R. and K. Poveda. 2011. Herbivore-induced changes in fruit–frugivore

interactions. Journal of Ecology 99:964-969.

Wignall, A. E., A. M. Heiling, K. Cheng, and M. E. Herberstein. 2006. Flower symmetry

preferences in honeybees and their crab spider predators. Ethology 112:510-518.

165

Wise, M. J. and W. G. Abrahamson. 2008. Applying the Limiting Resource Model to

plant tolerance of apical meristem damage. American Naturalist 172:635-647.

Wise, M. J., J. J. Cummins, and C. De Young. 2008. Compensation for floral herbivory

in Solanum carolinense: Identifying mechanisms of tolerance. Evolutionary

Ecology 22:19-37.

Wolfe, L. M. 1997. Differential flower herbivory and gall formation on males and

females of Neea psychotrioides, a dioecious tree. Biotropica 29:169-174.

Wu, Z. T., P. Dijkstra, G. W. Koch, J. Penuelas, and B. A. Hungate. 2011. Responses of

terrestrial ecosystems to temperature and precipitation change: A meta-analysis of

experimental manipulation. Global Change Biology 17:927-942.

Young, H. J. 2008. Selection on spur shape in Impatiens capensis. Oecologia 156:535-

543.

Yukawa, J. 2000. Synchronization of gallers with host plant phenology. Population

Ecology 42:105-113.

Zangerl, A. R. and M. R. Berenbaum. 2009. Effects of florivory on floral volatile

emissions and pollination success in the wild parsnip. Arthropod-Plant

Interactions 3:181-191.

Zangerl, A. R. and C. E. Rutledge. 1996. The probability of attack and patterns of

constitutive and induced defense: A test of optimal defense theory. American

Naturalist 147:599-608.

Zimmerman, M. and S. Cook. 1985. Pollinator foraging, experimental nectar-robbing and

plant fitness in Impatiens capensis. American Midland Naturalist 113:84-91.

166