The Ecological Consequences and Adaptive Function

Home , Bombus impatiens, Gelsemine, Gelsemium elegans

THE ECOLOGICAL CONSEQUENCES AND ADAPTIVE FUNCTION

OF NECTAR SECONDARY METABOLITES

Jessamyn Sara Manson

A thesis submitted in conformity with the requirements

for the degree of Doctor of Philosophy

Graduate Department of Ecology and Evolutionary Biology

University of Toronto

The Ecological Consequences and Adaptive Function of Nectar Secondary Metabolites

Jessamyn Sara Manson

Doctor of Philosophy

Department of Ecology and Evolutionary Biology

University of Toronto

2009

ABSTRACT

Plants are under selection to simultaneously attract pollinators while deterring herbivores. This dilemma can lead to tradeoffs in floral traits, which are traditionally thought to be optimized for pollinators. My dissertation addresses the ecological costs and putative functional significance of nectar secondary metabolites, a paradoxical but widespread phenomenon in the angiosperms. I investigate this issue from the pollinator’s perspective using a series of controlled laboratory investigations focused primarily on the bumble bee Bombus impatiens and the nectar alkaloid gelsemine, from Gelsemium sempervirens . I begin by demonstrating that nectar enriched with the alkaloid gelsemine significantly deters visits from bumble bees at a range of natural alkaloid concentrations.

However, this aversion can be mitigated by increasing the sucrose concentration such that the alkaloid-rich nectar is more rewarding than its alkaloid-free counterpart. I then demonstrate that the consumption of gelsemine-rich nectar can inhibit oocyte development and protein utilization in bees, but that this effect is limited to bees of suboptimal condition. Continuous consumption of the nectar alkaloid gelsemine also leads to a reduction in the pathogen load of bumble bees infected with Crithidia bombi ,

but direct interactions between the pathogen and the alkaloid have no impact on infection intensity. Gelsemine also fails to inhibit floral yeast growth, suggesting that nectar alkaloids may not be universally antimicrobial. Finally, I demonstrate that gross nectar cardenolides from the genus Asclepias are strongly correlated with gross leaf cardenolides and that the majority of individual cardenolides found in nectar are a subset of those identified in leaves. This pattern suggests that nectar cardenolides are a consequence of defense for Asclepias ; however, they may not be a costly corollary because bumble bees show an overall preference for nectar cardenolides at mean concentrations. Altogether, my dissertation provides a new perspective on the role of chemical defenses against herbivores in plant-pollinator interactions.

iii

ACKNOWLEDGEMENTS

Writing a dissertation is like raising a child – it takes a village! First of all, I’d like to thank my supervisor, James Thomson, who willingly welcomed this stray graduate student back in 2004. James understood that my interests lay in the field of chemical ecology and encouraged me to pursue research questions beyond the expertise of the lab.

I am extraordinarily grateful to James for letting me take so many risks and teaching me to be stubbornly independent. I would also like to thank to Spencer Barrett and Peter

Kotanen, the intrepid members of my supervisory committee, for their discussion and advice during my doctorate. Because of the integrative nature of my work, I have had a lot of help over the years. A huge thank you to three of my collaborators, Marc-André

Lachance, Mario Vallejo-Marìn and Anurag Agrawal, who have allowed my research to grow in exciting new directions. I am very appreciative of all the time and energy that others have invested so that I could complete my chemical analyses; thank you to Sergio

Rasmann, Rayko Halitchke, John Arnason and especially to Ammar Saleem, who spent many hours at the HPLC with me. Thanks to Bruce Hall and Andrew Petrie for watering my army of useless plants every time I was out of town. And thanks to Lynn Adler and

Rebecca Irwin for suggesting that I work on a plant that shares my name!

The members of the Thomson lab have been a wonderful source of guidance, support and collaboration. Thank you to Robert Gegear, without whom I would not know the joys of the flight cage, to Michael Otterstatter, who rekindled my passion for pathogens and to James Burns, Jessica Forrest, Nathan Muchhala, Jane Ogilvie and

Alison Parker for being great friends and mentors. I have forged strong bonds with many

members of the graduate community. I’d especially like to recognize Kate Edwards,

Heather Coiner, Danielle Way, Danielle Marcos, Marc Johnson, Patrick Vogan, Sarah

Yakimowski, Jannice Friedman, Brechann McGoey, Anna Simonsen and Brandon

Campitelli for their friendship, humour and compassion throughout the course of my degree.

I have also had an immense about of support from my family, particularly my parents. Thank you mom for teaching me to overcome obstacles with a smile on my face, thank you pa for instilling in me such a strong work ethic, and thank you dad for always telling me to go outside! Finally, this would not have been possible without the love and encouragement of my husband, Chris; your strength and generosity amaze me!

I would also like to acknowledge several presses for permitting me to include previously published work in my thesis. Chapter two (R.J. Gegear, J.S. Manson and J.D.

Thomson. 2007. Ecological context influences pollinator deterrence by alkaloids in floral nectar. Ecology Letters 10: 375-382) and chapter three (J.S. Manson and J.D. Thomson.

2009. Post-ingestive effects of nectar alkaloids depend on dominance status of bumble bees. Ecological Entomology 34: 421-426) were reproduced with permission from

Wiley-Blackwell Publishing. Chapter four (J.S. Manson, M.C. Otterstatter and J.D.

Thomson. In Press. Consumption of a nectar alkaloid reduces pathogen load in bumble bees. Oecologia DOI: 10.1007/s00442-009-1431-9) and appendix one (J.S. Manson,

M.A. Lachance and J.D. Thomson. 2007. Candida gelsemii sp. nov., a yeast of the

Metschnikowiaceae clade isolated from nectar of the poisonous Carolina Jessamine.

Antonie von Leeuwenhoeck 92: 37-42) were reproduced with permission from Springer

Science and Business Media.

TABLE OF CONTENTS

ABSTRACT………………………………………………………………………...ii

ACKNOWLEDGEMENTS………………………………………………………... iv

TABLE OF CONTENTS…………………………………………………………... vi

LIST OF TABLES…………………………………………………………………. x

LIST OF FIGURES………………………………………………………………... xi

LIST OF APPENDICES…………………………………………………………... xiii

CHAPTER ONE – Introduction…………………………………………………… 1

A brief history of nectar secondary metabolites…………………………… 2

A spotlight on alkaloids……………………………………………………. 6

Principal experimental systems……………………………………………. 7

How toxic is “toxic” nectar? Addressing the ecological consequences

and adaptive functions of nectar secondary metabolites……………………9

CHAPTER TWO – Ecological context influences pollinator deterrence by

alkaloids in floral nectar……………………………………………………………12

Abstract…………………………………………………………………….. 12

Introduction………………………………………………………………… 13

Methods……………………………………………………………………. 16

Bees and flowers …………………………………………………………… 16

Experimental procedure …………………………………………………….17

Data analysis ………………………………………………………………. 18

Results……………………………………………………………………... 20

Flower preference ………………………………………………………….. 20

Foraging proficiency ………………………………………………………. 21

Discussion………………………………………………………………….. 21

Acknowledgements………………………………………………………… 26

CHAPTER THREE – Post-ingestive effects of nectar alkaloids depend on dominance status of bumble bees………………………………………………….. 32

Abstract……………………………………………………………………. 32

Introduction………………………………………………………………… 33

Methods……………………………………………………………………. 36

Oocyte development ………………………………………………………... 36

Haemolymph carbohydrates ……………………………………………….. 39

Results……………………………………………………………………... 39

Protein metabolism ………………………………………………………… 39

Carbohydrate concentrations ……………………………………………… 40

Discussion………………………………………………………………….. 40

Acknowledgements………………………………………………………… 45

CHAPTER FOUR – Consumption of a nectar alkaloid reduces pathogen load in bumble bees……………………………………………………………………... 50

Abstract…………………………………………………………………….. 50

Introduction………………………………………………………………… 51

Methods……………………………………………………………………. 54

Statistical analysis …………………………………………………………. 57

Results……………………………………………………………………... 58

Discussion…………………………………………………………………. 60

vii

Acknowledgements………………………………………………………… 65

CHAPTER FIVE – Cardenolide concentrations of nectar, leaves and flowers:

A comparative study across Asclepias series Incarnatae…………………………... 72

Abstract…………………………………………………………………….. 72

Introduction………………………………………………………………… 73

Methods……………………………………………………………………. 77

Study system ………………………………………………………………... 77

Quantifying cardenolides …………………………………………………... 77

Pollination biology ………………………………………………………….80

Statistical analysis …………………………………………………………. 83

Quantitative cardenolide analysis …………………………………………. 83

Qualitative cardenolide analysis …………………………………………... 84

Behaviour analysis …………………………………………………………. 86

Results……………………………………………………………………....87

Quantitative cardenolide analysis …………………………………………. 87

Qualitative cardenolide analysis …………………………………………... 88

Behaviour analysis …………………………………………………………. 90

Discussion………………………………………………………………….. 92

Acknowledgements………………………………………………………… 101

CHAPTER SIX – CONCLUDING DISCUSSION………………………………... 108

Ecological context is crucial……………………………………………….. 109

A subtle effect is still an effect…………………………………………….. 110

There is no such thing as a general adaptive hypothesis…………………... 111

viii

Consequence of defense is often assumed but rarely tested……………….. 112

Costs and benefits are not always obvious………………………………… 114

REFERENCES.……………………………………………………………………. 116

LIST OF TABLES

Table 2.1. Description of reward conditions for each behavioural assay…….……. 27

Table 2.2. Results of t-tests evaluating pollinator preference…………………...... 28

Table 2.3. Generalized linear model results for foraging proficiency……………... 29

Table 3.1. The effect of gelsemine on oocyte length and width …………………... 46

Table 3.2. Daily pollen consumption in microcolonies……………………………. 47

Table 4.1. Mixed model statistics describing the effects of continuous gelsemine

consumption on Crithidia bombi infections……………………………….. 66

Table 4.2. Mixed model statistics describing the direct effects of gelsemine on Crithidia

bombi infections……………………………………………………………. 67

Table A.1. Summary of yeasts recovered from the nectar of flowers in

Statesboro, Georgia……………………………………………………….... 152

Table A.2. Growth responses of Candida gelsemii that exhibit variation…………..153

LIST OF FIGURES

Figure 2.1. Choice behaviour of individual bees over 80 consecutive visits………. 30

Figure 3.1. Mean oocyte size in dominant and subordinate bees at three

concentrations of gelsemine-rich nectar…………………………………… 48

Figure 3.2. Mean carbohydrate concentrations in bee haemeolymph 24 hours

after consuming gelsemine-rich nectar at three concentrations…………… 49

Figure 4.1. Diagram of experimental design for Crithidia bombi assays …………. 68

Figure 4.2. Effect of continuous alkaloid consumption on Crithidia bombi

infections in bumble bees………………………………………………….. 69

Figure 4.3. Relationship between Crithidia bombi infection intensity and

bumble bee body size………………………………………………………. 70

Figure 4.4. Effect of exposing Crithidia bombi cells to alkaloids prior to

bumble bee inoculation…………………………………………………….. 71

Figure 5.1. Average total cardenolide concentrations for the nectar, leaves and

flowers of twelve species of Asclepias series Incarnatae…………………...102

Figure 5.2. Correlations between total cardenolide concentrations in Asclepias

nectar and leaves …………………………………………………………... 103

Figure 5.3. NMDS two-dimensional ordination of individual cardenolide

concentrations of Asclepias nectar and leaves…………………………….. 104

Figure 5.4. Correlation between intensity of Crithidia bombi infections in bees

and their preference for blue flowers………………………………………. 105

Figure 5.5. Boxplots of average visit length and mean foraging rate by bees to

flowers varying in nectar cardenolide concentration………………………. 106

Figure 6.1. Path diagrams summarizing interactions between secondary

metabolites, pollinators, herbivores and plant fitness……………………… 115

Figure A.1. Phylogram of Candida gelsemii and its closest relatives……………... 154

Figure A.2. Differentiation of Candida gelsemii into “pulcherrima” cells…………155

xii

LIST OF APPENDICES

APPENDIX ONE – Candida gelsemii sp. nov., a yeast of the Metschnikowiaceae clade isolated from nectar of the poisonous Carolina jessamine…………………... 144

Abstract…………………………………………………………………….. 144

Introduction………………………………………………………………… 145

Methods……………………………………………………………………. 146

Results and Discussion…………………………………………………...... 147

Species delineation, phylogenetic placement and phenotypic

variability ………………………………………………………………….. 147

Ecology …………………………………………………………………...... 148

Description of Candida gelsemii Lachance sp. nov………………………... 150

Acknowledgements………………………………………………………… 151

APPENDIX TWO – Concentrations of cardenolides from the nectar, leaves and flowers of twelve species of Asclepias series Incarnatae………………………...... 156

APPENDIX THREE – Raw visit data from behavioural assays testing the effect of nectar cardenolides on bumble bee preference………………………………….. 160

xiii CHAPTER ONE

Introduction

For outcrossing flowering plants, maximizing fitness frequently entails attracting effective pollinators while deterring herbivores. Traits that are critical to fulfilling these two mandates have been studied largely in isolation, but there is a growing body of evidence indicating that many of these characteristics are intrinsically linked and therefore under selection from both pollinators and herbivores (Armbruster 1997, Strauss

1997, Strauss et al. 1999, Strauss and Whittall 2006, Bronstein et al. 2007). In many cases, selection by these two groups is antagonistic, forcing adaptive compromises ( sensu

Strauss and Whittall 2006) between reproduction and defense.

Tradeoffs to accommodate pollination and herbivory are particularly evident when they affect floral traits. Several studies have shown shifts in flower shape (Gomez

2003), size (Ashman et al. 2004) and flowering phenology (Pilson 2000) in response to selection from herbivores that reduce the frequency, efficiency or likelihood of pollination. In plants such as Raphanus sativus and Ipomoea purpurea , flower petal colour is directly associated with the production of anthocyanins, a secondary compound that chemically defends plants from herbivory (Fineblum and Rausher 1997, Irwin et al.

2003). Although tortoise beetle larvae have lower survivorship on the anthocyanin-rich purple flowers of I. purpurea , the rarer white (anthocyanin-free) morph exports more pollen per capita, but has lower visitation rates than its purple counterpart (Simms and

Bucher 1996). In the case of I. purpurea , being chemically defended does not unequivocally lead to a cost in terms of pollination services. However, the effects of

1 2 secondary metabolites on pollination services may be more insidious; compounds such as alkaloids, phenolics and glycosides have been detected in floral nectar (Adler 2000).

Could this so-called “toxic” nectar represent another tradeoff between pollinator attraction and herbivore defense?

My dissertation investigates the ecological consequences of nectar secondary metabolites. More specifically, I examine the effects of nectar alkaloids and nectar cardenolides on bumble bee behaviour and physiology. I also research adaptive functions for nectar alkaloids, focusing on their putative antimicrobial properties. Finally, I investigate whether cardenolides in milkweed nectar are likely to have particular adaptive roles in nectar, or are present systemically as a byproduct of the chemical defense of foliage against herbivores. Taken together, my research provides novel insight into the impact of herbivory on plant-pollinator interactions from the pollinator’s perspective and is an important contribution to the field of plant-animal interactions as a whole.

A brief history of nectar secondary metabolites

Floral nectar is produced by plants to reward pollinators and has two principal ingredients, water and sugar. In the mid-twentieth century, however, biologists began finding that nectar was not just a simple syrup and that it contained a myriad of other components in small concentrations (Baker and Baker 1983). Some of these additional nectar constituents may increase the value of floral nectar as a reward. Amino acids, which are found nearly universally in nectar (Baker and Baker 1973, Baker 1977, Baker and Baker 1983), can make nectar more palatable for pollinators (Kim and Smith 2000) and are particularly important to adult lepidoptera that have no other dietary source for protein-building material (Baker and Baker 1973). Nectar may also contain lipids (Baker

3 and Baker 1975, Baker 1977), proteins (Baker and Baker 1975, Thornburg et al. 2003) and antioxidants (Baker and Baker 1973), all of which may help to attract and retain pollinators. However, many plants have nectar constituents that are unpalatable, repellant, and sometimes poisonous to floral visitors. One of the most frequently observed “toxic” nectar components are secondary metabolites, compounds that are produced to defend plants against herbivores and include alkaloids, phenolics, tannins and glycosides.

The first comprehensive work on nectar secondary metabolites was completed by

Baker and Baker in the 1970s. The Bakers adapted a series of simple colourimetric assays to test for the presence of alkaloids (Baker and Baker 1975) and phenolics (Baker

1977) in the field. The surveys that they performed remain unrivalled in their breadth, characterizing the nectar chemistry of over 1000 plant species in environments as diverse as California, Costa Rica (Baker and Baker 1975) and the Rocky Mountains of Colorado

(Baker and Baker 1982). These studies detected nectar alkaloids in 0-12% and nectar phenolics in 30-49% of samples, depending on the latitude or altitude of the site. Nectar alkaloids and phenolics were most prevalent in the Costa Rican lowlands, and nectar alkaloids were entirely absent from the alpine tundra of Colorado (Baker and Baker

1982).

The Bakers’ work on nectar constituents was motivated by their interest in plant- pollinator coevolution, and they made several predictions regarding the role of nectar secondary metabolites in plant-animal interactions. They were particularly intrigued by the observation that nectar alkaloids were absent from plants pollinated by lepidoptera but present in plants pollinated by bees (Baker and Baker 1975). The Bakers reasoned

4 that nectar alkaloids may deter lepidoptera, which are inconstant pollinators, while bees, which exhibit strong floral constancy, are somehow resistant to the compounds; this rationale became the basis for the pollinator fidelity hypothesis (see below). A few years later, Rhoades and Bergdahl (1981) built upon this idea, suggesting that nectar secondary metabolites are another means by which plants encourage specialist pollinators and discourage ineffective floral visitors. While neither the Bakers nor Rhoades and Bergdahl followed up their hypotheses for the function of “toxic” nectar with empirical evidence, other researchers began testing their ideas. Stephenson, for example, found that nectar robbers were deterred by iridoid glycosides while valid pollinators were unaffected

(1981, 1982). Masters determined that specialists will visit nectar with pyrrolizidine alkaloids but generalists will not (1991); however, his results have been contested

(Landolt and Lenczewski 1993).

In 2000, Adler published the authoritative review on “toxic” nectar in which she summarized all substantiated accounts of nectar secondary metabolites. Adler collected

33 reports of nectar that was toxic or aversive for animals, nectar where secondary metabolites had been isolated and nectar with both identified compounds and deleterious effects. The studies, which date back to 1933, find evidence for “toxic” nectar in at least

21 angiosperm families and specify the presence of alkaloids, phenolics, iridoid glycosides and non-protein amino acids, along with some assorted rare compounds.

Animals affected by the noxious nectar include honey bees, ants, butterflies and humans, while the responses range from undetectable to lethal. However, the greatest strength of

Adler’s review is her synopsis of the hypotheses for the functional significance of this strange but widespread phenomenon.

The most compelling aspect of the nectar secondary metabolite question is the why: Why does nectar have toxic constituents? To date, five potential reasons have been proposed, all of which were succinctly summarized by Adler and which I will touch on only briefly here. The pollinator fidelity hypothesis, proposed by Baker and Baker

(1975) but fleshed out by Rhoades and Bergdahl (1981), stipulates that nectar secondary metabolites may encourage specialist pollinators while discouraging generalists by promoting secondary metabolite tolerance in more frequent visitors. The nectar robber hypothesis was suggested by both Janzen (1977) and Baker (1978) and states that chemical defenses in nectar could defend plants from floral visitors that take resources but do not pollinate. Hagler and Buchmann (1993) were the first to hypothesize that nectar secondary metabolites could have an antimicrobial function, thereby keeping nectar palatable for pollinators. The drunken pollinator hypothesis, from Ehlers and

Olesen (1997), is a tangential proposal suggesting that the ethanol produced from floral yeasts can improve pollen transfer because intoxicated pollinators spend less time grooming. This hypothesis can be expanded to include a “nectar as narcotic” hypothesis, in which consuming very small doses of secondary metabolites such as caffeine and nicotine could cause both intoxication and addiction (Singaravelan et al. 2005). Finally,

Adler herself (2000) suggests that secondary metabolites in nectar occur as a consequence of a plant’s chemical defense system. This last hypothesis implies a non- adaptive and potentially deleterious function for “toxic” nectar, but is not mutually exclusive from the previous hypotheses. In fact, it is likely that nectar secondary metabolites originated as a sort of chemical overflow, but may have gained new, nectar- specific functions through “exaptation” (Armbruster 1997).

Since Adler’s review, interest in the field of “toxic” nectar, and the implications of chemical defenses on plant mutualisms in general, has increased dramatically. Nectar secondary metabolites have been identified in ten new plants, including almonds

(London-Shafir et al. 2003) and avocados (Afik et al. 2006). Work on the impact of putatively “toxic” nectar on pollinators has expanded to include effects on solitary bees

(Elliott et al. 2008) and birds (Tadmor-Melamed et al. 2004, Johnson et al. 2006). Those focusing on the honey bee are now looking at dose-dependent effects of nectar secondary metabolites (Singaravelan et al. 2005) and mechanisms for tolerating toxic compounds

(Liu et al. 2005). Studies are addressing the hypothesized adaptive functions of noxious compounds in nectar; there is now evidence from one system to suggest that nectar alkaloids are ineffective at deterring nectar robbers (Adler and Irwin 2005), while recent work demonstrating how floral yeasts degrade nectar quality (Herrera et al. 2008) has prompted interest in exploring the antimicrobial properties of nectar secondary metabolites. In addition, researchers are considering the role of secondary metabolites as components (either directly or indirectly) of other floral traits like pollen (Praz et al.

2008) and floral scent (Raguso 2008). Perhaps most importantly, authors no longer discount the influence of herbivory and herbivore protection on nectar, nectarivorous animals and plant-pollinator interactions as a whole.

A spotlight on alkaloids

Alkaloids, which represent the largest and most diverse group of secondary metabolites, are loosely defined by a cyclic structure and the presence of nitrogen in a negative oxidation state (Hartmann 1992). Infamously toxic to vertebrates and invertebrates (think strychnine and atropine), alkaloids can affect neuroreceptors,

7 structurally disrupt DNA, induce apoptosis or inhibit critical enzymes (Wink and

Schimmer 1999). However, alkaloids also have favorable properties at appropriate doses and can be antimicrobial (e.g. quinine; Cowan 1999), therapeutic (e.g., morphine), or can act as stimulants (e.g., caffeine). Alkaloids are found in 20% of flowering plants and are considered a highly effective constitutive defense against herbivores (Hartmann 1992).

The first circumstantial evidence for alkaloids in nectar came from a study on

Sophora microphylla in 1972 (Clinch et al.). Baker and Baker (1983) subsequently detected nectar alkaloids in 8% of the species they tested (75 of 910 species). Attention quickly turned to the effects of nectar alkaloids on flower visitors; studies suggest that responses vary depending on alkaloid concentration and type of visitor. Clinch et al.

(1972) found that honey bees died as a result of ingesting as little as 10 µL of S. micorphylla nectar, while Singaravalen (2005) suggested that low concentrations of nicotine and caffeine may cause addiction to nectar in honey bees. Nectarivorous birds that consume nectar rich in nicotine are less able to assimilate sucrose (Tadmor-Melamed et al. 2004), while specialist butterflies may (Masters 1991) or may not (Landolt and

Lenczewski 1993) find nectar alkaloids unattractive. Finally, high concentrations of nectar in alkaloids reduce the transfer rate of a pollen analogue (Adler and Irwin 2005), suggesting that nectar alkaloids can reduce plant fitness.

Principal experimental systems

The majority of my thesis focuses on the relationship between a single plant,

Gelsemium sempervirens , and one of its key pollinators, Bombus impatiens . When my research questions required a plant system amenable to comparative studies, I had to look

8 beyond the tiny Gelsemium genus to the larger, more diverse genus Asclepias . Here I describe the relevant natural history of my two principal experimental systems.

Gelsemium sempervirens , the Carolina jessamine, is a high-climbing or trailing woody vine distributed throughout the southeastern United States and into Central

America (Radford et al. 1968, Ornduff 1970). The genus Gelsemium , part of the

Loganiaceae, contains only three species, two of which are native to North America, while the third, G. elegans is found in Asia (Ornduff 1970, Wyatt et al. 1993). G. sempervirens grows abundantly on the margins of tilled fields, along fence lines and in ditches along the road. It flowers from mid-January until early April, depending on longitude (Leege and Wolfe 2002, Pascarella 2007), but does not produce mature seeds until late fall. G. sempervirens flowers are 1.7 – 3.3 cm in length, vivid yellow, tubular and very fragrant. The flowers are distylous, with familiar pin and thrum morphology, and are largely self-incompatible (Ornduff 1970) . G. sempervirens is pollinated by the solitary bees Osmia lignaria (orchard mason bee) and Habropoda laboriosa (blueberry bee), the social bees Apis mellifera (honey bee), Bombus impatiens and Bombus bimaculatus (bumble bees) and the nectar robber Xylocopa virginica (carpenter bee)

(Ornduff 1970, Adler and Irwin 2005, Pascarella 2007). Floral visitors differ in their pollen transfer ability, with Bombus spp. being among the most efficient (Adler and Irwin

2006).

Gelsemium sempervirens is heavily defended from herbivory by indole alkaloids, with the plant’s roots and floral nectar being particularly noxious (Hardin and Arena

1969, Burrow and Tyrl 2001). The most abundant alkaloid, gelsemine, is found in high concentrations in flowers (Blaw et al. 1979) and is the principal alkaloid detected in floral

9 nectar (Adler and Irwin 2005), with gelsemine concentrations in nectar ranging from 5.8 to 246.1 ng/ µL (Adler and Irwin 2005; Manson, unpublished data). G. sempervirens is highly toxic to mammals: the ingestion of five flowers rendered a young child semi- comatose (Blaw et al. 1979), while livestock and chickens are also reportedly at risk

(Burrow and Tyrl 2001). The plant’s effect on invertebrates is less clear, with reports of

G. sempervirens nectar poisoning honey bees being unsubstantiated (Hardin and Arena

1969). However, G. sempervirens nectar artificially enriched with gelsemine did reduce the length and number of flower visits made by several pollinators, including bumble bees (Adler and Irwin 2005).

The bumble bee Bombus impatiens is native to eastern North America, with a range from Ontario to Maine and south to Florida (Kearns and Thomson 2001). B. impatiens has a moderate tongue length (5.8 +/- 0.8 mm; Harder and Barrett 1993), which allows it to be a generalist pollinator on a large number of plants, including several species known to have nectar secondary metabolites, such as milkweeds, rhododendrons and Gelsemium sempervirens (Manson, personal observation). The foraging habits of bumble bees, including B. impatiens , have been well-documented, and bumble bees have been recognized as an excellent model system for behavioural assays (Heinrich 1979,

Real 1991, Chittka et al. 1999, Chittka and Thomson 2001). In addition, B. impatiens is commercially available as a pollinator of greenhouse crops.

How toxic is “toxic” nectar? Addressing the ecological consequences and adaptive functions of nectar secondary metabolites

My dissertation contributes to a growing body of literature addressing the ecological consequences of chemical defense on plant-insect mutualisms. I have focused

10 primarily on the effects of a single nectar secondary metabolite, gelsemine, on a single pollinator, Bombus impatiens . I broaden my scope to nectar cardenolides in the genus

Asclepias , adding a phylogenetic framework to a larger comparative study. My work addresses many of the key questions raised by my peers and represents the most comprehensive treatment of the impact of nectar secondary metabolites on bumble bees to date.

Previous studies have reported that nectar alkaloids can be aversive, however a rigorous examination of their impact on pollinator behaviour was lacking. In chapter two, I describe the first laboratory study to address the effects of the nectar alkaloid gelsemine on bumble bee behaviour under a range of ecologically relevant nectar conditions. I trained workers to forage on artificial flower arrays with distinctly coloured flowers that contrasted in alkaloid or sucrose concentration, or both, and examined the effects of gelsemine on preference, foraging rate and flower handling time. In this controlled environment, I also evaluated whether pollinators learn to avoid nectar alkaloids over time.

In chapter three, I explore whether bees experience a physiological cost due to nectar alkaloid consumption. This study, which took advantage of a unique developmental process to induce queen-like behaviour in worker bees, focused on how the constant consumption of gelsemine-rich nectar affected oocyte development and carbohydrate metabolism in bees. These two traits relate to protein utilization and energetics, respectively, and are therefore good metrics for sublethal costs of nectar alkaloid ingestion. This is one of only a handful of studies dealing with post- consumptive effects of nectar secondary metabolites.

One of the most interesting issues surrounding nectar secondary metabolites is whether they have an adaptive function. The hypothesis that nectar alkaloids might be antimicrobial is addressed in chapter four and appendix one. In appendix one I report the discovery of a new floral yeast species found in Gelsemium sempervirens and its susceptibility to gelsemine. In chapter four, I take the antimicrobial hypothesis further and evaluate whether gelsemine-rich nectar can reduce infections of the gut pathogen

Crithidia bombi in bumble bees. This chapter turns the adaptive hypothesis on its head by suggesting that consuming nectar alkaloids might be beneficial for pollinators.

Chapter five used high-performance liquid chromatography to survey cardenolides from twelve species of the genus Asclepias . This survey looks at correlations between cardenolide concentrations in nectar, leaves and flowers. I paid particular attention to differences in the identity and chemical polarity of the individual cardenolides found in nectar and leaves. I then used these data to evaluate whether nectar cardenolides are a consequence of a plant’s systemic chemical defenses. This represents the first study to identify and quantify nectar cardenolides and only the third to compare nectar secondary metabolites to secondary metabolites in other plant parts.

Finally, chapter six is a general discussion that synthesizes my results and identifies how the findings contribute to our understanding of plant-pollinator interactions. In particular, I discuss themes that bridge my chapters such as the importance of ecological context, the functional significance of nectar alkaloids and nectar cardenolides and the subtle costs of nectar secondary metabolites on pollinators and pollination.

CHAPTER TWO

Ecological context influences pollinator deterrence by alkaloids in floral nectar

Robert J. Gegear*, Jessamyn S. Manson*, and James D. Thomson

*equally contributing authors

R. J. Gegear and I contributed equally to the design and execution of this project and the writing of the manuscript. J.D. Thomson provided substantial comments on the experiments and the paper, which is published in Ecology Letters, 2007, 10: 375-382 .

Abstract

Secondary metabolites may benefit plants by deterring herbivores, but the presence of these defensive chemicals in floral nectar may also deter beneficial pollinators. This tradeoff between sexual reproduction and defense has received minimal study. We determined whether the pollinator-deterring effects of a nectar alkaloid found in the perennial vine Gelsemium sempervirens depend on ecological context (i.e., the availability of alternative nectar sources) by monitoring the behavioural response of captive bumble bees ( Bombus impatiens , an important pollinator of G. sempervirens in nature) to nectar alkaloids in several ecologically-relevant scenarios. Although alkaloids in floral nectar tended to deter visitation by bumble bees, the magnitude of that effect depended greatly on the availability and nectar properties of alternative flowers.

Ecological context should thus be considered when assessing ecological costs of plant

12 13 defense in terms of pollination services. We consider adaptive strategies that would enable plants to minimize pollinator deterrence due to defensive compounds in flowers.

Introduction

The interaction between plants and their animal pollinators has been a significant force in the evolution of floral characters. However, plants also simultaneously interact with other types of animal visitors, such as herbivores, which may affect pollination and ultimately influence the evolution of floral traits. Although herbivory can influence pollination through direct damage to reproductive tissues (e.g. pistils and stamens, Leege and Wolfe 2002) or floral characters used to attract pollinators (e.g. corolla characters,

Strauss et al. 1996, for further review see McCall and Irwin 2006), it can also influence pollination through more subtle, but nonetheless ecologically significant, mechanisms

(Agrawal et al. 1999, Strauss et al. 1999, Strauss et al. 2002). For instance, pollination services to plants may be reduced when plant characters used to defend against herbivores are linked to characters used to attract pollinators (Simms and Bucher 1996,

Strauss 1997). This integration of attractive and defensive traits presents plants with a potential fitness tradeoff between the benefits of pollinator attraction and the costs of reduced herbivore defense. Consequently, many pollination- and herbivory-related traits that have traditionally been considered the result of selection by pollinators or herbivores alone may actually be an evolutionary compromise between the contrasting selection pressures exerted by both plant interactors together (Herrera et al. 2002). Despite the important implications of interactions between pollination and herbivory for the ecology and evolution of plant characters, there is little known about how pollinator visitation is influenced by the concurrent presence of plant attractive and defensive traits. Here, we

14 describe a series of controlled experiments that were designed to examine the effects of plant defensive compounds in flowers on the attractiveness of plants to pollinators.

From the pollinator’s perspective, the attractiveness of plants is determined primarily by the perceived amount of beneficial compounds such as carbohydrates and amino acids contained in floral nectar (Proctor et al. 1996). Paradoxically, floral nectar of some plant species also contains secondary metabolites such as phenols and alkaloids

(Baker and Baker 1983) that occur in leaves, stems, and roots to defend against attack by herbivores and microorganisms (Berenbaum 1995). Indeed, secondary metabolites have been reported in the floral nectar of at least twenty-one angiosperm families (Adler

2000), indicating that this phenomenon is widespread. Although the presence of secondary metabolites in floral nectar may provide reproductive benefits to plants in some special cases (see Rhoades and Bergdahl 1981, Adler 2000 for reviews of hypotheses), it is predicted to have detrimental effects if pollinators are deterred from visiting flowers (Strauss et al. 1999, Adler and Irwin 2005). Although we are testing adaptive hypotheses in other work, here we implicitly assume that defense chemicals occur in nectar as an unavoidable byproduct of their production in other tissues; we therefore consider the deterrence of pollinators as an ecological cost of defense (Strauss et al. 1999, 2002).

Because the foraging decisions of pollinators are contingent on current floral conditions and past floral experiences, the ecological cost of defense to plants with secondary compounds in floral nectar likely depend on the ecological context in which pollinators interact with flowers. For example, ecological costs may be reduced if secondary compounds are present in nectar for short periods or increased if pollinators

15 have the option of visiting alternative plants with no secondary compounds in floral nectar. How these costs vary with ecological context has not been rigorously examined, presumably because of the difficulty in manipulating floral environments and tracking the behaviour of individual pollinators under natural conditions.

Based on the pollination ecology of Gelsemium sempervirens (L.), we devised laboratory choice experiments in which we used one of its major floral visitors and pollinators, the bumble bee Bombus impatiens Cresson (Ornduff 1970; Manson unpublished data) , as a model system to investigate how ecological context influences the effects of secondary compounds in floral nectar on pollinator choice behaviour and foraging proficiency (flowers visited per minute and flower-handling time). Gelsemium sempervirens is an obligate outcrosser and secretes the commercially available alkaloid gelsemine in floral nectar. Previous work has shown that bumble bees ( Bombus bimaculatus , which is closely related to B. impatiens ) spend less time on G. sempervirens flowers and visit fewer flowers per plant when natural gelsemine concentrations are increased in nectar (Adler and Irwin 2005, 2006), suggesting that nectar gelsemine imposes an ecological cost on plants by altering visitation by bumble bees. We determined whether the behavioural response of bumble bees to gelsemine-rich floral nectar, and thus the ecological cost of defense to plants, depends on ecological context by monitoring the choice behaviour, foraging rate and flower-handling time of freely foraging bees on artificial floral arrays that simulated the following ecologically-relevant scenarios: (1) G. sempervirens co-occurring and co-flowering with an equally rewarding plant species without alkaloids in floral nectar, (2) G. sempervirens co-occurring and co- flowering with less rewarding plant species without alkaloids in floral nectar, and (3) A

16 population of G. sempervirens in which plants have either a low or high level of gelsemine in floral nectar. By comparing the behaviour of bees foraging under these conditions, we were not only able to assess ecological costs of gelsemine in floral nectar to G. sempervirens in terms of potential bumble bee pollination services, but provide new perspectives on the adaptive significance of traits commonly observed in plant species with secondary compounds in flowers.

Methods

Bees and flowers

Colonies of Bombus impatiens Cresson, each with 30-50 workers, were obtained from Biobest Canada (Leamington, Ontario). Nest boxes were connected to a 2.2m x

2.2m x 2.4m flight cage by a gated tube so that we could control the number of bees entering the flight cage. Prior to experiments, colonies were allowed to collect 30% w/w sugar solution from feeders located in the center of the flight cage. Colonies were supplied with pollen ad libitum . Workers that made regular foraging trips between the colony and feeders were individually marked with coloured liquid paper.

Artificial flowers were constructed by removing the lids from 30 yellow and 30 blue 1.5 mL microcentrifuge tubes and adding 4.2 cm circles of blue and yellow construction paper respectively around the mouth of the tubes. Yellow flowers resembled the tubular yellow flowers of Gelsemium sempervirens . A blue-yellow colour dimorphism was used to make it easier for bees to discriminate flowers based on gelsemine and sucrose content of nectar. To access the test solution (hereafter referred to as ‘nectar’), bees had to land on the surface of the paper corolla and crawl to the bottom of the tube, much as they do in real G. sempervirens flowers. Flowers were presented to

17 bees by embedding them upright in a 1.26 x 0.79 x 0.032 m styrofoam board covered in green paper. Flowers were positioned in a 67.5 cm by 56.0 cm grid so that bees had an equidistant choice of each flower type upon departing any flower on the array (with the exception of the two outer columns). Our artificial array was designed in this manner so that we could create realistic floral environments for bees while controlling for the availability, distribution and nectar properties of flowers.

The principal alkaloid found in the nectar of Gelsemium sempervirens is gelsemine (Irwin and Adler 2006), with natural concentrations ranging across populations from 5.8 ng/µL to 246.1 ng/µL (Adler and Irwin 2005). Sucrose concentrations in G. sempervirens nectar reportedly range from 11 to 62% under natural conditions (Leege and Wolfe 2002, Adler and Irwin 2005; Manson, personal observation). For our behavioral assays, nectar containing both gelsemine and sucrose was created by adding gelsemine hydrochloride (ChromaDex, Santa Ana, CA) to aqueous sucrose solutions

(either 30% or 50% w/w sucrose) until all gelsemine was dissolved (gelsemine concentrations were 0, 5, 50, and 125 ng/µL sucrose solution; Table 2.1). Solutions were refrigerated at 4°C when not in use and replaced every 3-5 days. For brevity, we refer to the sucrose concentration of nectar as either S30 or S50 and the gelsemine concentration of nectar as either G0, G5, G50, or G125.

Experimental procedure

We determined the behavioural response of bumble bee foragers to gelsemine in nectar under the three ecological scenarios described above. Gelsemine and sucrose concentrations used in each assay were selected based on values reported for G. sempervirens under natural conditions (Table 2.1).

Marked bees were trained by allowing them to forage freely on an array of each flower type (i.e., each nectar condition) in succession for three foraging trips. This procedure ensured that bees had experienced the nectar condition associated with each flower colour prior to testing. The flower colour associated with each nectar condition was randomized among bees to control for the possibility that floral preference was influenced directly by colour. Immediately following training, bees were individually presented with a mixed array containing 30 flowers of each type, and we videotaped at least 80 flower visits for later analysis. Flowers were filled with 3µL of nectar and refilled quickly after being drained by bees. Flowers were replaced between bees. After testing, bees were freeze-killed, and body size was estimated by measuring the length of the radial cell on the right forewing (Harder 1982).

Data analysis

For each assay, we determined whether bees overall had a preference for flowers with lower levels of gelsemine on the mixed array (i.e. visitation frequency was non- random with respect to alkaloid level) by using a two-tailed one-sample t-test to compare the mean proportion of visits to flowers with the lower concentration of nectar gelsemine to the proportion of visits expected given the abundance of both flower types on the mixed array (0.5 in all cases). Proportions were arcsin-transformed so that they conformed more closely to a normal distribution. We then examined how the flower- choice behaviour of individual bees changed as they gained foraging experience on the mixed array by dividing the first 80 flower visits for each bee tested into four blocks of

20 consecutive visits. For each block, we determined whether individuals had a preference for one of the available flower types by using a G-test of independence (Sokal

19 and Rohlf 1995) to compare the observed frequency of visits to low gelsemine flowers to the frequency of visits expected given random flower selection (10 visits). An observed visit frequency of 15 or greater indicated that the bee had a preference for low gelsemine flowers whereas a visit frequency of 5 or less indicated a preference for high gelsemine flowers. We then tested for changes in flower-choice behaviour of bees over time by using a repeated measures ANOVA to compare visit frequency to low gelsemine flowers among the four blocks, followed by Tukey’s multiple comparison test.

Because previous work has suggested that nectar alkaloids, including gelsemine, may affect plant fitness by altering the behaviour of pollinators on flowers rather than through pollinator deterrence (Strauss et al. 1999, Adler and Irwin 2005), we also examined the effect of gelsemine on bee foraging proficiency. Here, we assess foraging proficiency by measuring foraging rate (number of flowers visited per minute) and flower-handling time (total time in seconds that the bee spends on a flower), which are two components of bumble bee behavior that may affect how they collect and deposit pollen and thus provide another measure of how nectar gelsemine may influence plant reproductive success though behavioral alterations to bees. We used a generalized linear model with radial cell length as a covariate (Proc Genmod; SAS version 8) to compare foraging rates and flower-handling times of individuals that showed a preference (i.e. visitation frequency was significantly biased in favour of the flower type) for S30G0 flowers (n=20) and those that showed a preference for S30G50 flowers (n=10). Measures of foraging proficiency were calculated based on 10 consecutive flower visits taken randomly between visits 50 and 70 and results are reported as likelihood ratio statistics

(G) (SAS Institute1999). All foraging proficiency measures were log-transformed to meet the assumptions of normality and equal variance.

Results

Flower preference

Bumble bee choice behaviour was significantly influenced by the concentration of gelsemine in the nectar of available flower types. Although bees as a group readily collected floral nectar containing gelsemine on monotypic arrays during training, they had a strong preference for nectar with equal sucrose rewards but no gelsemine (Assays

1A-C) or lower levels of gelsemine (Assay 3 - S30G50 vs S30G125) on mixed arrays

(Table 2.2). Bees showed no overall nectar preference when the sucrose concentration of nectar with alkaloids was increased relative to an alkaloid-free nectar alternative (Assay 2

- S30G0 vs S50G50; Table 2.1). In fact, there was a significant decrease in the proportion of visits to flowers with no nectar gelsemine between Assays 1B and 2 (t=6.37, df=16, p<0.001), indicating that an increase in sucrose concentration of floral nectar containing gelsemine relative to alkaloid-free nectar alternatives increased its attractiveness to bees.

At the individual level, there was a considerable amount of variation in flower- choice behaviour of bees on the mixed array over time (Fig. 2.1). With the exception of

Assay 2, most individuals had a strong preference for flowers with nectar containing sucrose only or low levels of gelsemine in the final visit block (percentage of bees with a preference for flowers with no or low nectar gelsemine in the final visit block was: 92.3%

(Assay 1A), 63.6% (Assay 1B), 77.8% (Assay 1C), 36.4% (Assay 2), and 77.8% (Assay

3). Interestingly, a small percentage of bees showed a preference for flowers with higher nectar gelsemine concentrations in the final visit block (7.7% (Assay 1A), 9.1% (Assay

1B), 11.1% (Assay 1C), 45.5% (Assay 2)). The mean proportion of visits to the flower type on the mixed array with the lower nectar gelsemine concentration differed significantly among visit blocks for Assay 1a (F 3,12 =12.32, P<0.0001), 1b (F 3,10 =7.09,

P=0.001), 1c (F 3,8 =6.98, P=0.0015, and 3 (F 3.8 =6.61, P=0.0021), but not for Assay 2

(F 3,10 =0.542, P=0.657). Pairwise comparisons showed that the proportion of flower visits to low gelsemine flowers significantly increased between block 1 and 3 and block 1 and 4 for Assays 1A-C and Assay 3 and also between block 2 and 3 and block 2 and 4 for

Assay 1A, indicating that bees tended to sample both flower types on the mixed array prior to developing a preference for the flower type with the lower level of nectar gelsemine.

Foraging proficiency

Visitation to gelsemine-rich flowers had no significant effect on foraging rate or mean handling time (Table 2.3). Bee size, determined from radial cell length, did positively correlate with foraging rate (G 1,27 =5.01, P=0.03), but did not correlate with mean handling time. There was no interaction between preference and size for any of the three foraging efficiency measures, so the interaction term was removed from the model.

Discussion

Our study supports the hypothesis that defensive compounds in floral nectar impose an ecological cost on plants in the form of reduced pollinator visitation, and demonstrates for the first time that such ecological costs to plants depend heavily on the ecological context in which pollinators make foraging decisions. Bumble bees readily foraged on monotypic arrays of alkaloid-rich flowers regardless of nectar alkaloid level,

22 but quickly developed a strong aversion to them when flowers with lower levels of nectar alkaloids were made available. These results suggest that nectar alkaloids would only be a significant ecological cost to Gelsemium sempervirens plants when they must compete for pollination services with alternate plants that have lower levels of nectar alkaloids.

Interestingly, G. sempervirens flowers very early in the spring (Pascarella, 2007), perhaps because pollinator response to alkaloids in floral nectar was an important selective pressure on flowering time. Nectar quality has been postulated to be a significant factor in the evolution of flowering phenology (e.g. Mosquin 1971, Heinrich 1975, Brody

1997), with plant species of low nectar quality evolving earlier bloom times to escape competition for pollinators with plant species of high nectar quality. Early bloom time may thus be one adaptive mechanism, or ‘counteradaptation’ (Strauss et al. 1999), in plants to mitigate the loss of pollination services due to nectar alkaloids.

Although most bees quickly learned to associate alkaloid concentration with flower-colour cues, and avoided alkaloid-rich flowers when equally rewarding alkaloid- free alternatives were available, the deterrent effect of the alkaloid was offset by higher sugar concentrations. Thus, bees acted as if they were balancing economic gains (sugar collection) against palatability (alkaloid concentration). Previous work on feeding behaviour in herbivorous insects has shown that carbohydrates can counteract the deterrent effects of many plant secondary compounds, including alkaloids, and do so through a variety of complex physiological response mechanisms (Dethier 1982, Mitchell and Sutcliffe 1984, Mitchell 1987, Dethier and Bowdan 1992, Shields and Mitchell 1995,

Glendinning et al. 2000). Similar mechanisms may mediate the behavioral response of bumble bees, and other insect pollinators, to nectar alkaloids. For instance, the

23 unpleasant taste of the alkaloids may be ‘masked’ by higher sucrose concentrations

(Glendinning 2002), in the same way that a person can make a bitter food like chocolate more palatable by adding sugar. Alternatively, alkaloids may interfere with the ability of sucrose-sensitive receptor cells in the peripheral taste system to detect the correct sucrose concentration of nectar (a process called sensory inhibition; Mitchell & Sutcliffe 1984).

In this view, higher sucrose concentrations are required for alkaloid-rich nectar to be perceived as a profitable resource. Regardless of the behavioural mechanisms involved, the combined effect of nectar alkaloids and sugars on floral attractiveness has important implications for our understanding of how pollinators assess nectar quality and make adaptive foraging decisions. For example, pollinators performing a “behavioral titration”

(Kotler and Blaustein 1995, Webster and Dill 2006) of sucrose and alkaloid uptake would alter how they allocate foraging effort to available plant species.

The reduced deterrent effect of alkaloids caused by a relative increase in nectar sucrose concentration suggests that plants with nectar alkaloids would minimize the loss of pollination services to co-flowering plants with lower alkaloid levels by increasing the caloric content of nectar. At present, there is little information on the relationship between secondary metabolites and caloric content of nectar because past studies have either held caloric content and secondary metabolite content constant (Stephenson 1982,

Masters 1991, Hagler and Buchmann 1993, Landolt and Lenczewski 1993, Singaravelan et al. 2005), or not compared caloric content and secondary metabolite concentration among plant species available to pollinators (Stephenson 1981, Adler and Irwin 2005).

We predict that plants with more alkaloids in floral nectar will contain higher caloric rewards (concentration and possibly volume) relative to other available plants with lower

24 nectar alkaloid levels. This hypothesis for nectar alkaloids is akin to the “nutrient/toxin titration” model proposed for the presence of toxins in fruit (Cipollini and Levey 1997b).

Our results indicate that ecological costs of alkaloids in floral nectar to G. sempervirens are due to a reduction in the quantity (number of individuals visiting plants) and not the quality (individual behaviour on flowers) of floral visitation by pollinators.

Bees that collected gelsemine-rich nectar spent the same amount of time on flowers and visited the same number of flowers per minute as bees that collected gelsemine-free nectar, indicating that there would be no reproductive cost to G. sempervirens in terms of a reduction in the ability of pollinators to remove and deposit pollen. In contrast to our results, Adler and Irwin (2005) found that increasing gelsemine concentrations in flowers of natural G. sempervirens populations had no effect on initial attraction of pollinators to plants, but reduced the amount of time that they spent per flower and the number of flowers visited per plant. One likely explanation for the discrepancy between our results and those of Adler and Irwin is that bees in our study visited flowers with different levels of gelsemine in sucrose rewards for extended periods of time and were thus able to learn the association between floral cue (colour in our case) and nectar properties. Indeed, many bees in our study entered and quickly departed from flowers with gelsemine-rich nectar while learning to associate flower colour with reward condition (see Fig. 2.1), suggesting that ecological costs to plants increase with pollinator foraging experience.

This point underscores the importance of identifying and tracking individual pollinators over ecologically relevant periods during field studies aimed at determining the effects of pollinator behaviour on plant fitness.

Plasticity in the behavioural response of pollinators to secondary compounds in floral nectar has important implications for the evolution of plant defenses against herbivory. Optimal defense theory (ODT) predicts that constitutive defenses in valuable tissues should be more advantageous to plants than induced defenses because such tissues are protected prior to damage by herbivores (Rhoades 1979, McCall and Karban 2006).

Our study suggests that predictions of ODT for plants with defense compounds in flowers need to incorporate potential tradeoffs between benefits of herbivore resistance and costs of pollinator deterrence. Based on the behavioural response of bumble bees to nectar alkaloids observed in our study, we expect constitutive defenses in floral tissues and nectar (i.e. reproductive traits) of outcrossed plants to be favoured only when pollinators have few other floral resources available or caloric rewards compensate for reduced nectar palatability. In contrast, we expect induced defenses in flowers to be advantageous when there is strong competition for pollinators, since floral attractiveness would be reduced for short time periods subsequent to damage by herbivores. Moreover, induced defenses would reduce the floral attractiveness of a small subset of plants in the population (assuming that levels of herbivory are low), thereby decreasing the likelihood that pollinators will learn to discriminate against all plants with a floral signal similar to that of the defended plant (i.e. other plants in the population). Thus, inducible defenses can benefit plants under many ecological conditions by allowing them to mount a strong defense against herbivores while minimizing ecological costs due to pollinator deterrence. Although there is growing evidence for induced defenses in flowers (e.g.

Adler et al. 2006, McCall 2006, McCall and Karban 2006), the costs and benefits of induced versus constitutive defense strategies has not been considered in terms of

26 pollinator deterrence. In future field experiments, we plan to determine how plant defense strategies affect pollination services, and thus plant fitness, in different ecological contexts.

Acknowledgements

We would like to thank J. Forrest, J. Thaler, and four anonymous reviewers for helpful comments on the manuscript. This study was funded by an NSERC grant to JDT.

Table 2.1. Descriptions of the reward conditions used in each behavioural assay.

Sucrose and gelsemine concentrations used in each assay were selected based on their ecological significance. Concerning Assay 3, note that gelsemine production by

Gelsemium sempervirens has yet to be characterized as either induced or constitutive.

Nectar Condition

Assay Ecological Significance Gelsemine-poor Gelsemine-rich

30% w/w sucrose 50 ng/µL is in the middle range of natural 30% w/w 1A with 50 ng/µL concentrations of gelsemine in G. sempervirens sucrose gelsemine nectar (Adler and Irwin 2005)

50% sucrose with Simulates high nectar concentrations found in the 1B 50% sucrose 50 ng/µL gelsemine field (Adler and Irwin 2005)

5 ng/µL is the lowest concentration of gelsemine 30% sucrose with 1C 30% sucrose found in G. sempervirens nectar (Adler and Irwin 5 ng/µL gelsemine 2005)

50% sucrose with Presents bees with a trade-off between palatability 2 30% sucrose 50 ng/µL gelsemine and economic gain

Simulates a possible induced chemical response to 30% sucrose 30% sucrose with herbivory that increases nectar alkaloid 3 with 50 ng/µL 125 ng/µL concentrations or natural variation in nectar gelsemine gelsemine alkaloid concentrations

Table 2.2. Results of one-sample t-test for Assays 1-3. For each assay, the mean proportion of visits to the flower type with the lower level of nectar alkaloids on the mixed array was compared to the proportion of visits to flowers with lower nectar alkaloids expected given the abundance of both flower types on the mixed array (0.5 in all cases). Means are given +/- SE.

Mean proportion of visits to Assay t-value df P low gelsemine flowers

1A 0.86 +/- 0.04 7.932 12 <0.0001

1B 0.76 +/- 0.09 2.54 10 0.029

1C 0.84 +/- 0.07 4.89 8 0.0012

2 0.50 +/- 0.10 -0.149 10 0.88

3 0.82 +/- 0.04 7.04 8 0.0001

Table 2.3. Generalized linear model results for bee foraging rate (flowers visited per minute), and flower-handling time (in seconds). Data are from foragers in Assays 1A, 1C and 3 that had a significant preference for either sucrose-only (n=20) or sucrose plus gelsemine (n=10) nectar. All values are mean +/- SE and df=1, 27 for each analysis.

Foraging Proficiency Sucrose Only Sucrose and Gelsemine G P Measure

Foraging Rate (flowers/minute) 9.38 +/- 0.36 8.91 +/- 0.29 1.82 0.18

Flower- Handling Time 4.76 +/- 0.6 5.77 +/- 0.42 3.2 0.07 (seconds)

1A: S30G0 vs S30G50 2: S30G0 vs S50G50 100 100 75 75

50 50 25 25 0 0

1B: S50G0 vs S50G50 3: S30G50 vs S30G125 100 100 75 75 50 50 25 25 0 0

1C: S30G0 vs S30G5 100 75 50

Percent of visits toalkaloid low flowers 25 0 1 2 3 4

Visit Block

Figure 2.1. (previous page) Choice behaviour of individual bees over four consecutive blocks of 20 flower visits. Figure panel numbers correspond to assay numbers from

Table 2.1. Changes are displayed in percent preference for flowers with low or no gelsemine in nectar, as indicated by the treatment in bold text, and each line represents the preference trajectory of a single foraging bee. Dashed reference lines demarcate zones of significance as determined by G-test values, with percent of visits to flowers with low or no gelsemine in nectar above 71.6% indicating that the bee had a preference for flowers with no or low levels of gelsemine in nectar and below 28.4% indicating that the bee had a preference for flowers with gelsemine in nectar.

CHAPTER THREE

Post-ingestive effects of nectar alkaloids depend on dominance status of bumble bees

Jessamyn S. Manson and James D. Thomson

This project was designed in collaboration with J. D. Thomson. I completed the experiments and wrote the paper with assistance from J. D. Thomson, which appears in Ecological Entomology, 2009, 34: 421-426.

Abstract

Secondary metabolites have acute or chronic post-ingestive effects on animals, ranging from death to growth inhibition to reduced nutrient assimilation. Although characterized as toxic, the nectar of Gelsemium sempervirens is not lethal to pollinators, even when the concentration of the nectar alkaloid gelsemine is very high. However, little is known about the sublethal costs of nectar alkaloids. Using a microcolony assay and paired worker bumble bees we measured the effects of artificial nectar containing gelsemine on oocyte development. Oocytes are a sensitive indicator of protein utilization and general metabolic processes. We also calculated carbohydrate concentrations in the haemolymph to examine energetic costs of gelsemine consumption. High concentrations of gelsemine significantly reduced mean oocyte width in subordinate bees, while dominant bees showed only a trend towards oocyte inhibition. Gelsemine consumption did not reduce carbohydrate concentrations in haemolymph. The cost of ingesting gelsemine may be due to direct toxicity of alkaloids or may be an expense associated with detoxifying gelsemine. Detoxification of alkaloids can require reallocation of resources

32 33 away from essential metabolic functions like reproduction. The risks associated with nectar alkaloid consumption are tied to both the social and nutritional status of the bee.

Introduction

Plant secondary metabolites are believed to have evolved as chemical defenses against herbivorous animals (Whittaker and Feeny 1971, Janzen 1973, Feeny 1992,

Berenbaum 1995). Acute toxicity, resulting in death, is reported in many of the major secondary metabolite families (e.g., alkaloids, phenolics, glycosides, Rosenthal and

Berenbaum 1991). Although lower concentrations of a secondary metabolite may not be lethal, they can reduce the overall health and fitness of an animal (chronic toxicity,

Berenbaum et al. 1986).

Such chronic effects are subtle and highly variable. Typically, bioassays of growth, development or reproduction are used. Zangerl and Berenbaum (1993) showed decreased growth of parsnip webworm ( Depressaria pastinacella ) larvae when fed on wild parsnip ( Pastinaca sativa ) umbels with high furanocoumarin levels. Similarly, winter moth caterpillars ( Operophtera brumata ) fed on oak leaves with high tannin concentrations had reduced larval and pupal weights, along with reduced adult emergence

(Feeny 1970). Tannic acid caused developmental malformations in tent caterpillar

(Malacosoma disstria ) pupae (Karowe 1989), while phenolic glycoside concentrations were negatively correlated with fecundity in gypsy moth, Lymantria dispar (Osier et al.

2000).

Nearly all studies on plant chemical defense focus on secondary metabolites in the shoots or roots; however, these compounds are also paradoxically found in floral nectar.

Although the functional significance of nectar secondary metabolites is not fully

34 understood (but see Adler 2000 for a review of hypotheses), studies do suggest that this so-called “toxic” nectar can have deleterious consequences for nectar-collecting floral visitors. Honey bees have died after consuming artificial nectar containing very low concentrations of alkaloids and glycosides (Detzel and Wink 1993); in other cases, ingesting “toxic” nectar has less severe consequences. For example, Palestine sunbirds consuming nectar containing pyridine alkaloids were less able to assimilate sugar from their diet (Tadmor-Melamed et al. 2004). Secondary metabolites in nectar, particularly alkaloids and phenolics, have been shown to deter pollinators and reduce number of flower visits (Adler and Irwin 2005, Singaravelan et al. 2005, Johnson et al. 2006, Gegear et al. 2007). Despite multiple reports of the distastefulness of “toxic” nectar, few studies have correlated behavioural responses with effects of nectar secondary metabolite consumption on floral visitors.

The chemical arsenal of Gelsemium sempervirens L. (Carolina jessamine) includes the indole alkaloid gelsemine, a compound found in the roots, shoots, flowers and floral nectar of the plant. G. sempervirens is a perennial vine native to the southeastern United States; its fragrant yellow flowers open in the early spring and attract several flower visitors (Adler and Irwin 2005, Pascarella 2007; Manson, personal observation), including bumble bee queens and workers ( Bombus bimaculatus, B. impatiens ), honey bees ( Apis mellifera ), carpenter bees ( Xylocopa virginica) and solitary bees ( Osmia lignaria, Habropoda laboriosa ). The consequences of ingesting flowers or leaves from Gelsemium spp. are severe for mammals and include psychosis, respiratory failure, severe convulsions and death (Blaw et al. 1979, Ott 1998, Rujjanawate et al.

2003, Fung et al. 2007). In contrast, adult bumble bees exposed to high levels of

35 gelsemine experience no acute effects, even when gelsemine concentrations are 20 times higher than natural levels (Manson, personal observation). Similarly, Elliott et al. (2008) reported no effect of gelsemine on the number or survivorship of offspring produced by the megachilid solitary bee, Osmia lignaria . However, laboratory assays indicate that bees prefer to feed on sucrose-only nectar rather than a solution of gelsemine and sucrose

(Gegear et al. 2007), while enriching G. sempervirens nectar with gelsemine deterred visitors in nature (Adler and Irwin 2005), implying that there are consequences to the ingestion of alkaloid-rich nectar.

Given the absence of acute effects, we tested for sublethal costs of nectar alkaloids by feeding bumble bee workers ( Bombus impatiens Cresson) artificial nectar containing gelsemine and measuring development of their oocytes. Oocyte development provides a good bioassay because it is a defined metabolic challenge that can be induced by pairing worker bees without a queen (Cnaani et al. 2002, Cnaani et al. 2007).

Furthermore, it is a complex and costly physiological process which we predict to be sensitive to toxins for several reasons. First, oocyte development is highly correlated with protein utilization in worker bees (Lin and Winston 1998, Pernal and Currie 2000).

Although there are several ways for insects to cope with secondary metabolites, a common mechanism is to detoxify these compounds into less hazardous ones (Slansky

1992). Detoxification requires the production of specific enzymes from dietary proteins.

If this mechanism is used by bumble bee workers, protein could be re-allocated towards the construction of enzymes and away from oocyte production. Second, alkaloid processing requires energy, so carbohydrates used for normal metabolic processes may be redirected to alkaloid metabolism, leaving less energy for the formation of reproductive

36 structures. In addition, alkaloids may directly interfere with the absorption of nutrients by inhibiting digestive enzymes or forming nutrient-allelochemical complexes (Slansky

1992). We therefore hypothesize that metabolic costs associated with nectar alkaloid consumption will result in reduced oocyte development. We also evaluated whether alkaloid processing directly reduces available carbohydrates by measuring carbohydrate levels in bee haemeolymph 24 hours after ingestion. We discuss our findings with a focus on possible mechanisms for alkaloid tolerance in pollinators.

Methods

Oocyte development

Oocyte development in Bombus spp. depends on social circumstances. If multiple workers are kept in queenless colonies, one of them frequently assumes a queen-like role, becoming the dominant worker and developing oocytes (Cnaani et al. 2002, Cnaani et al.

2007). When two bumble bee workers interact in a queenless colony, the dominant worker will develop its ovaries at an optimal rate while suppressing the ovary development rate of the subordinate worker. We took advantage of this developmental strategy, building “microcolonies” from worker bees to assess how gelsemine consumption affects ovary development under optimal and suboptimal conditions.

We obtained pupal clumps of Bombus impatiens from Biobest Canada Ltd.

(Leamington, ON). Bumble bee microcolonies were composed of two unfed callow workers (<24 hours old). We created a size dichotomy in each container in an effort to enhance differences between dominant and subordinate bees, as previous work suggests that larger bees are more likely to be dominant (Ayasse et al. 1995). Each pair of bees was housed in a closed 500 mL clear plastic food container, lined with paper to absorb

37 feces, and equipped with holes for ventilation (along the sides) and nectar access (on the base). This container was nested in a second food container, which held a small petri dish of artificial nectar, made accessible by a cotton wick that led up to the holes on the base of the first container. This arrangement reduced spilling and contamination of the nectar by preventing direct contact between the nectar and the bees. Individuals were divided evenly among treatments so that we had a total of 28, 29, and 28 pairs of bees in control, moderate and high gelsemine treatments, respectively, after three replicates, which were run at three separate dates.

We used a 30% w/w solution of sucrose as artificial nectar, which fell well within the range of natural sugar levels reported in Gelsemium sempervirens flowers (Leege and

Wolfe 2002, Adler and Irwin 2005), to which we added gelsemine hydrochloride

(hereafter referred to as gelsemine; Chromadex, Santa Ana, CA). We used three diet treatments: control, composed of sucrose only; moderate gelsemine, a solution of sucrose plus 50 ng/ µL gelsemine; and high gelsemine, a solution of sucrose plus 250 ng/ µL gelsemine. The two gelsemine treatments simulate the mean and maximum concentrations, respectively, of gelsemine found in the nectar of natural Gelsemium sempervirens populations (Adler and Irwin 2005). Bees avoid nectar of both of these concentrations if control nectar is available (Gegear et al. 2007). We supplied 1.5 mL of artificial nectar daily, as well as commercially available pollen ad libitum . We provided new pollen every day; pollen lumps were weighed before and after they were provided to a container to determine daily pollen consumption by the pair of bees. Microcolonies were maintained for six days under controlled environmental conditions (23-27ºC in the dark, except for during feeding), which is the estimated time needed for B. impatiens

38 oocytes to mature (Cnaani et al. 2002, Cnaani et al. 2007). After six days, we froze the bees and dissected them in distilled water to determine ovary development. Using a scaled ocular, we measured the length and width of the largest oocyte in each of the two paired ovaries using the average of the two in our analyses. Development in the two tended to be symmetrical. We also recorded the length of the radial cell in the front right wing as a proxy for bee size (Harder 1982).

We compared oocyte length and width between treatments with ANCOVA, using radial cell length as a covariate. We chose to analyze length and width separately to pinpoint the effects of gelsemine consumption on each of these size parameters. Previous work on oocyte development has used either a subjective size ‘score’ (Pernal and Currie

2000) or measured length alone (Bloch and Hefetz 1999, Cnaani et al. 2007), which may have overlooked possible variation in oocyte width. Dominance was assigned to the bee within each pair with the larger oocytes, estimated as length times width, and we analyzed dominant and subordinate bees separately. We pooled data from the three experimental replicates, as the data did not significantly differ between replicates. We removed four pairs of bees from the analysis because one of the pair died before the experiment was completed, changing the social environment of the remaining bee. These pairs were spread across treatments. We also removed two subordinate bee outliers with extremely small oocytes.

To assess possible differences in protein intake between treatments, we analyzed daily pollen consumption by microcolonies using a repeated measures ANOVA. When necessary, data was transformed to meet assumptions for normality and homogeneity of variance.

Haemolymph carbohydrates

We removed pupal clumps from individual commercial colonies and isolated unfed callow bees (<24 hours old) in individual vials. We provided bees with 500 µL of one of the three treatments: control (30% w/w sucrose), moderate gelsemine (50 ng/ µL gelsemine in 30% sucrose) or high gelsemine (250 ng/ µL gelsemine in 30% sucrose).

After 24 hours, when nearly all the nectar was consumed, we refrigerated the bees and then decapitated them. We took haemolymph samples by separating the ventral terga with forceps and gently inserting a 5 µL microcapillary tube into the lower abdominal cavity.

We estimated the volume of each haemolymph sample and stored them individually in 1 mL of 80% ethanol. We analyzed carbohydrates in 18, 17, and 17 individuals in the control, moderate and high gelsemine treatments, respectively.

We calculated carbohydrate concentrations, expressed as micrograms of trehalose equivalents per microlitre of haemolymph, using the anthrone method (modified from

Siegert 1987). Because the carbohydrate data were not normally distributed, we tested for differences in carbohydrate concentration between gelsemine treatments using a non- parametric Kruskal-Wallis test.

All statistical analyses were performed in R (version 2.6.0).

Results

Protein metabolism

Nectar gelsemine concentration did not affect oocyte length of dominant or subordinate bees (Table 3.1, Fig. 3.1). However, high levels of gelsemine did significantly reduce oocyte width in subordinate bees (post-hoc Tukey tests using the multcomp package in R, Fig. 3.1), while there was a trend towards smaller widths in

40 oocytes of dominant bees in the high gelsemine treatment. An unexpected element of the experiment was that dominance was not reliably predicted based on bee size; that is to say, the largest bees did not always have the largest oocytes. However, there was still a positive relationship between oocyte size (LxW) and radial cell length (R 2=0.2, F=42.36, df=162, P<0.001). Therefore, radial cell was kept in the analyses and did contribute to the explanatory power of each model.

The reduction in oocyte size was not correlated with reduced protein intake, as pollen consumption did not differ between treatments (Table 3.2). Pollen consumption did vary significantly between days within each treatment, with bees eating the most on the second day of the assay, followed by a decline in appetite and a resurgence in pollen consumption by day 6. There was no interaction between treatment and day.

Carbohydrate concentrations

Gelsemine did not affect the concentration of carbohydrates found in bee haemolymph (Kruskal-Wallis test, χ2 =3.01, df=2, P=0.22). Although the data are highly variable (Figure 3.2), removing outliers did not reveal differences between treatments

(analysis not shown).

Discussion

The nectar alkaloid gelsemine significantly inhibits oocyte development in subordinate bees, but is only marginally effective at reducing oocyte size in dominant bees. This effect was detectable at ecologically relevant concentrations, suggesting that ingestion of nectar alkaloids can incur a cost to pollinators. The severity of this cost, however, appears to depend on the condition of the bumble bee and the concentration of

41 the alkaloid. Overall, the mean concentration of gelsemine found in nature may be largely innocuous to healthy bees. Under suboptimal circumstances, however, the ingestion of nectar alkaloids might be chronically deleterious to pollinators.

Sublethal effects of nectar alkaloids on bumble bees may arise in various ways.

First, the alkaloids may not be toxic enough to kill bees, but they may be toxic enough to compromise nutrient absorption, alter neurohormonal processes, or damage internal organs (see Slansky 1992 for review). These outcomes may lead to protein excretion or increased protein investment in immune function, reducing oocyte size. Another explanation for inhibited oocyte development is that detoxifying alkaloids is costly.

Detoxification of secondary metabolites is a common process whereby compounds are metabolized into less toxic components, and it is often accompanied by rapid excretion.

This process requires protein to build detoxification enzymes and energy to process the deconstruction of the secondary metabolites; it is therefore assumed to be metabolically expensive. The assumption of costliness is supported by the observation that many detoxification mechanisms are induced only after the consumption of a secondary metabolite. Inducibility is interpreted as an energy-saving strategy (Berenbaum and

Zangerl 1994). Although this explanation is attractive, empirical evidence is divided on whether detoxification is a significant expense. Detoxification of alkaloids is reported to reduce digestive efficiency in the southern armyworm, Spodoptera eridania (Cresswell et al. 1992), while parsnip webworms ( Depressaria pastinacella ) shunt energy away from growth to metabolize furanocoumarins (Berenbaum and Zangerl 1994). In contrast, alkaloid detoxification in Helicoverpa zea required a negligible amount of energy compared to that spent on regular metabolic activity (Neal 1987). Evidence of metabolic

42 costs due to “toxic” nectar ingestion are sparse; we know that Palestine sunbirds experienced reduced sucrose assimilation after consuming nectar containing the alkaloids nicotine and anabasine (Tadmor-Melamed et al. 2004), but whether this was the result of reallocation to alkaloid detoxification was not identified. In our study we did not find that gelsemine affected carbohydrate levels (see Fig. 3.2). However, carbohydrate levels in insect haemolymph are reportedly highly variable and may lack the resolution necessary to detect the energetic costs associated with alkaloid detoxification (Thompson

2003).

Secondary metabolites have reduced protein utilization in previous studies on both vertebrates and invertebrates (reviewed in Duffey and Stout 1996). Oocyte size is a sensitive measure of protein utilization (Duchateau and Velthuis 1989, Lin and Winston

1998, Pernal and Currie 2000) and smaller oocytes were reported in worker bumble bees infected with the gut pathogen Crithidia bombi , suggesting that oocyte size can indeed be an indicator of poor health (Shykoff and Schmid-Hempel 1991). We must therefore conclude that protein metabolism and, consequently, fecundity in dominant bees is only modestly affected by ingested nectar alkaloids. This conclusion is supported by work done on Osmia lignaria (Elliott et al. 2008), which found that gelsemine did not reduce the fecundity of healthy solitary bees. The significant reduction in the oocyte size of subordinate bees in the high gelsemine treatment suggests nectar alkaloids may incur a cost to protein metabolism in individuals of suboptimal condition. Whether the inhibition of oocyte development due to gelsemine results in extended ovary development time or smaller offspring is unknown, but both outcomes could affect fitness.

The microcolony assay was designed to test the direct effects of gelsemine on dominant bees because previous studies have shown a predictable response on the ovary development of the dominant worker under different social and nutritional environments

(Duchateau and Velthuis 1989, Cnaani et al. 2002, Cnaani et al. 2007). The role of the subordinate bees in the microcolonies was simply to fulfill the social conditions required for optimal ovary development in their dominant counterparts. However, the significant treatment response by the subordinate bees, which experienced suboptimal conditions for oocyte development, is an unexpected but important result. The response of subordinate bees to the consumption of gelsemine is a complex effect that may involve both metabolism and behaviour. Previous studies on worker oocyte development have reported that bees exert dominance, in part, by monopolizing the pollen ball (Cnaani et al.

2007). This behaviour reduces the subordinate bee’s access to protein, which likely explains why all subordinate bees have smaller oocytes. In addition to obtaining less dietary protein to support oocyte development, these food-stressed bees may experience heightened costs of detoxification. In fact, Wahl and Ulm (1983) demonstrated that the cost of metabolizing pesticides increased when honey bee pollen intake was reduced.

Compensatory pollen feeding by the dominant bees in the high-gelsemine treatment might further reduce the amount available to subordinates in those treatments, either directly through consumption by the dominants, or indirectly because dominant bees spend more time at the pollen ball and guard it more stringently. The possibility of competition for access to pollen could also explain why larger bees tended to fare better

(significant effect of radial cell length Table 3.2). We found no effects of gelsemine on pollen consumption (Table 3.2), but those data include pollen consumption by both bees.

We cannot determine whether the allocation of pollen to dominants and subordinates may have differed among treatments. Furthermore, the hygroscopic nature of pollen, coupled with the necessity of using fresh weights, renders the pollen consumption data only approximate.

The effects of nectar alkaloids on pollinators must be interpreted with natural plant-insect interactions in mind. In this study, we found that the inhibition of oocyte development due to the consumption of gelsemine was related to a pollinator’s condition.

Despite its acute toxicity to mammals, gelsemine seems to be distasteful but largely harmless to bees, except if they are consuming the highest natural concentrations, have little other food to dilute the toxic effects, or are metabolically challenged. These circumstances might apply to bumble bee queens foraging on the early spring flowers of

G. sempervirens. Queens that have recently emerged from hibernation are developing their ovaries to begin nestmaking, and often have few other nectar and pollen resources to choose from. They may be ingesting substantial amounts of gelsemine-rich nectar while highly food-stressed and therefore vulnerable to the deleterious consequences of gelsemine. Even a slight sublethal effect of nectar alkaloids may present a subtle but significant impediment to pollinator fitness. Whether that impediment offsets the positive value of the nectar sugars obtained would depend on the dietary choices available. The role that nectar secondary metabolites play in plant-pollinator communities will therefore be shaped by the composition of each community, and future work needs to move beyond the interactions of a single plant and pollinator to include more complex, community-level interactions. Although “toxic” nectar may be less

45 severe than its name suggests, its deleterious effects still have the potential for widespread consequences to pollinators and the plants they visit.

Acknowledgements

We would like to thank Kate Edwards and Tamryn Ah-Long for their assistance, and two anonymous reviewers for comments on the manuscript. This research was funded by

NSERC.

Table 3.1. The effect of zero, moderate (50 ng/µL) and high (250 ng/µL) gelsemine on oocyte length and width, on dominant and subordinate bees. All analyses are ANCOVAs with Type III SS and radial cell length (a proxy for bee size) as a covariate. Significant effects of the gelsemine treatment are in bold.

Oocyte Length Oocyte Width

Source

df SS F P df SS F P

Dominant Bees

Treatment 2 0.01 0.39 0.68 2 0.01 1.60 0.21

Radial Cell 1 0.09 4.98 0.03 1 0.02 4.87 0.03

Subordinate Bees

Treatment 2 0.04 1.57 0.22 2 0.03 4.80 0.01

Radial Cell 1 0.12 10.11 <0.001 1 0.02 6.92 0.01

Table 3.2. Daily pollen consumption for six day microcolony assay, compared between the three gelsemine treatments using a repeated measures ANOVA. Note that because bees were raised in pairs, each measurement represents pollen consumed for one dominant and one subordinate bee.

Source df SS F P

Between Subjects

Treatment 2 8.96 1.53 0.29

Day 1 0.41 0.14 0.72

Within Subjects

Treatment 2 7.48 3.74 0.25

Day 5 181.4 13.36 <0.001

Treatment x Day 10 20.10 0.74 0.69

Figure 3.1. Mean oocyte size, plotted as length against width, in dominant and subordinate bees fed 0 ng/µL (closed circles), 50 ng/µL (open circles) or 250 ng/µL

(closed inverted triangle) gelsemine. Dominant bees have larger oocytes and clump together in the upper right corner of the graph, while the smaller oocytes of subordinate bees fall in the lower left corner. The graph indicates the SE of both length and width measurements.

Figure 3.2. Median carbohydrate concentrations in haemolymph 24 hours after bees consumed artificial nectar with 0 ng/µL, 50 ng/µL, or 250 ng/µL gelsemine.

Carbohydrates are expressed as micrograms of trehalose equivalents per microlitre haemolymph. The boxes represent the 25 th and 75 th percentiles, the whiskers indicate roughly two standard deviations and filled circles represent outliers.

CHAPTER FOUR

Consumption of a nectar alkaloid reduces pathogen load in bumble bees

Jessamyn S. Manson*, Michael C. Otterstatter* and James D. Thomson

*equally contributing authors

M.C. Otterstatter and I contributed equally to the design and execution of this project and the writing of the manuscript. J.D. Thomson provided comments on the paper, which is in press at Oecologia.

Abstract

Floral nectar is produced to attract pollinators, but frequently contains secondary metabolites that are often distasteful and toxic. One hypothesized function of this so- called “toxic” nectar is that it has antimicrobial properties, which could prevent nectar from spoiling. Microbicidal nectar may also benefit insect pollinators by reducing the infection intensity of pathogens. We tested whether gelsemine, a nectar alkaloid of the bee-pollinated plant Gelsemium sempervirens , could reduce pathogen loads in bumble bees infected with the gut protozoan Crithidia bombi . In our first laboratory experiment, artificially infected bees consumed a daily diet of gelsemine post-infection to simulate continuous ingestion of alkaloid-rich nectar. In the second experiment, bees were inoculated with C. bombi cells that were pre-exposed to gelsemine, simulating the direct effects of nectar alkaloids on pathogen cells that are transmitted at flowers. Gelsemine significantly reduced the fecal intensity of C. bombi seven days after infection when it was consumed continuously by infected bees, whereas direct exposure of the pathogen to

50 51 gelsemine showed a non-significant trend towards reduced infection. Lighter pathogen loads may relieve bees from the behavioural impairments associated with the infection, thereby improving their foraging efficiency. If the collection of nectar secondary metabolites by pollinators is done as a means of self-medication, pollinators may selectively maintain this unusual trait in natural plant populations.

Introduction

At first glance, the function of plant-produced secondary metabolites seems straightforward: these distasteful and often toxic compounds defend plants from herbivores (Rosenthal and Berenbaum 1991). However, identical secondary metabolites are also found in the floral nectar of plants, which is paradoxical given that floral nectar is usually interpreted as attractive, not deterrent, to pollinators. Secondary metabolites, including tannins, phenols, alkaloids and terpenes, have been found in floral nectar across

21 angiosperm families (Adler 2000). The prevalence and diversity of secondary compounds across the angiosperms suggests that this so-called “toxic” nectar has some adaptive function for plants. Hypothesized functions of secondary metabolites in nectar include deterrence of nectar robbers, increased constancy of effective pollinators, or protection against deleterious microbes (see Rhoades and Bergdahl 1981, Adler 2000 for full review). Of these, the antimicrobial hypothesis is perhaps the most general because microbes are ubiquitous and nectar is an otherwise ideal medium to support a variety of microorganisms that can harm plants and deter pollinators. Although many secondary metabolites have microbicidal properties (Cowan 1999), and diverse microorganisms often occur in floral nectar (Ehlers and Olesen 1997, Golonka 2002, Brysch-Herzberg

2004), few studies have tested whether nectar secondary metabolites actually suppresses microbes (but see Manson et al. 2007).

For plants that rely on animal vectors for pollination, microbial contamination of nectar can be costly. Spoilage due to yeast, fungi, and bacteria can affect nectar palatability, leading to changes in pollinator behaviour (Ehlers and Olesen 1997) and potentially reducing pollen transfer. Microbes may also affect pollination by disrupting processes like pollen tube growth (Kevan et al. 1989) or stigmatic receptivity (Buban and

Orosz-Kovacs 2003). In some cases, the pollinators themselves cause nectar contamination by transporting plant venereal diseases that destructively infect floral structures, reducing seed set through damaged pollen and ovules (Buban and Orosz-

Kovacs 2003, Antonovics 2005). Allocating secondary metabolites to nectar might reduce microbial growth, thereby protecting plants from pathogens and maintaining the quality of their rewards.

Antimicrobial properties of this so-called “toxic” nectar may also benefit nectar- collecting pollinators. Gathering nectar exposes pollinators to a variety of pathogenic microorganisms that can reduce their survival and foraging efficiency. For example, bumble bees in Europe and North America frequently carry the intestinal protozoan

Crithidia bombi (Lipa and Triggiani 1988, Schmid-Hempel 2001, Colla et al. 2006), which elevates their mortality rate under food stress (Brown et al. 2000) and impairs their associative learning, flower handling, and foraging efficiency (Gegear et al. 2005,

Otterstatter et al. 2005, Gegear et al. 2006). Horizontal transmission of C. bombi occurs at flowers, when infected bees deposit ‘free-living’ pathogen cells that are subsequently ingested by susceptible foragers (Durrer and Schmid-Hempel 1994). Since C. bombi is

53 known to occur in the nectar of wild flowers (Durrer and Schmid-Hempel 1994), nectar secondary metabolites may influence the survival and infectivity, and consequently the transmission, of this pathogen. Furthermore, for gut pathogens such as C. bombi , host diet can significantly effect the severity of infection by altering immunocompetence, metabolic processes, or by limiting nutrient availability for the parasite (Wink and Theile

2002, Logan et al. 2005, Cory and Hoover 2006). Hence, in flowers or in the guts of flower visitors, “toxic” nectar may benefit pollinators via antimicrobial action.

The nectar of the Carolina jessamine ( Gelsemium sempervirens L.) contains the indole alkaloid gelsemine, a secondary metabolite that is highly toxic to vertebrates

(Blaw et al. 1979). Gelsemine appears to have little effect on the fitness or physiology of bees (Elliott et al. 2008, Manson and Thomson 2009) and no effect on non-pathogenic floral yeasts (Manson et al. 2007). Although gelsemine-rich nectar can be distasteful and deterrent to pollinators (Adler and Irwin 2005, Gegear et al. 2007), G. sempervirens flowers consistently attract a number of floral visitors in nature, including bumble bees.

In the present study, we examined the putative antimicrobial properties of the nectar alkaloid gelsemine on the bumble bee pathogen Crithidia bombi. First, we asked, does consumption of alkaloid-rich nectar by bumble bees reduce the severity of intestinal infections by C. bombi ? Second, given that this pathogen is naturally transmitted at flowers, we asked, does alkaloid-rich nectar directly reduce the infectivity of C. bombi cells? We then discuss the ecological impact of nectar alkaloids on pollinator-pathogen dynamics and the potential for pollinators to selectively maintain secondary metabolites as a natural microbicide.

Methods

Our experimental protocol is illustrated in Figure 4.1. In both experiments, we exposed the pathogen C. bombi to either a gelsemine solution (Alkaloid) or plain sucrose solution (Control) and then compared the intensity of developing infections in inoculated bumble bees ( Bombus impatiens Cresson). In Experiment 1, bees were first inoculated with C. bombi and then fed on a daily diet of Alkaloid or Control solution. In Experiment

2, we exposed C. bombi cells to Alkaloid or Control solutions for varying durations before inoculating bees.

We made artificial alkaloid-rich ‘nectar’ by mixing gelsemine hydrochloride

(purchased from Chromadex, Irvine, CA, hereafter referred to as gelsemine) into a 30% w/w aqueous sucrose solution. The concentration of gelsemine was 250 ng/ µL , which represents the highest naturally occurring concentration of gelsemine reported (Adler and

Irwin 2005). Alkaloid solutions were refrigerated at 4°C when not in use and stored for up to two days, although they were usually prepared immediately before use.

We prepared pathogen inocula from the gut tracts of four ‘donor’ B. impatiens workers from each of five hives infected by C. bombi (provided by a commercial rearing company). Following the general protocol of Otterstatter and Thomson (2006), gut tracts were excised and crushed in a microcentrifuge tube containing 300 µL of distilled water.

The mixture was allowed to settle at room temperature for three hours, after which the supernatant was removed and mixed thoroughly. Supernatants were diluted to the appropriate density of C. bombi cells (Neubauer haemocytometer counts) and sucrose was added to a concentration of 30%. In each of the two experiments (described below), we used 20 new donor bees from five new hives; thus, within experiments, all bees

55 received the same cocktail of C. bombi strains (genotypes), but between experiments, inocula may have contained different pathogen genotypes.

We obtained susceptible ‘recipient’ B. impatiens workers from pupal clumps originating from commercially reared hives (same supplier as above). Previous studies have found that Crithidia infections are not acquired until workers emerge (Otterstatter and Thomson 2007), making new workers naïve to Crithidia regardless of the infection status of the source colony. Newly emerging (< 24 hours old) worker bees were placed in containers according to their hive of origin and given 30% sucrose solution and pollen ad libitum . After two days, workers were starved overnight, weighed (± 0.1 mg), and then arbitrarily assigned to an experimental group. We ensured that each of the Alkaloid and Control groups in both experiments contained recipient bees from at least three hives, in roughly equal numbers.

In Experiment 1, ‘Continuous Exposure’, bees inoculated with C. bombi were allowed to feed daily on gelsemine, simulating the continual ingestion of nectar constituents by an infected foraging bee. Each bee was initially fed a 2 µL drop containing 10 4 C. bombi in 30% sucrose solution and we monitored individuals until the entire drop was consumed. This dose falls within the range of C. bombi cells shed in the faeces of infected bees in previous studies (Schmid-Hempel and Schmid-Hempel 1993,

Logan et al. 2005), and therefore simulates cells available for transmission to naïve individuals. Bees were reared in individual 15 mL vials and received either a 0.5 mL solution of 250 ng/ µL gelsemine in 30% sucrose (Alkaloid bees, n = 35) or 0.5 mL of

30% sucrose only (Control bees, n=35) along with a pollen lump daily for 10 days.

In Experiment 2, ‘Delayed Exposure’, C. bombi was exposed to gelsemine for various durations prior to host ingestion, simulating direct exposure of the pathogen to nectar in a flower. We placed 10 4 C. bombi (in 2 µL of 30% sucrose solution) into each of 60 microcentrifuge tubes: 30 of these contained 8 µL of a 250 ng/ µL solution of gelsemine in 30% sucrose (Alkaloid), and 30 contained 8 µL of 30% sucrose only

(Control). In the ‘Immediate’ group, we fed the Alkaloid and Control pathogen mixtures to recipient bees immediately; each bee was housed individually and received only one dose, yielding 10 Alkaloid bees and 10 Control bees. In the ‘1 hr Delay’ and ‘2 hr Delay’ groups, we left the Alkaloid and Control pathogen mixtures at room temperature (~ 21-

24°C) under fluorescent lighting for one and two hours, respectively, before feeding them to recipient bees (as before, 10 Alkaloid bees and 10 Control bees per group). We selected these two delay treatments to simulate the time delay between the deposition of

C. bombi cells by infected bees and the next flower visit by a naïve bee. In this experiment, we compensated for evaporative water loss by starting with more dilute sucrose solutions that evaporated to a concentration of 30% sucrose after one or two hours (dilutions calculated from a preliminary study). Following the inoculation with C. bombi , bees were kept in individual vials and given 0.5 mL of 30% sucrose solution and a fresh pollen lump daily.

In both experiments, we quantified infection intensities of all bees at day 7 and day 10 post-inoculation, as these days delimit the period in which pathogen load is saturated (Schmid-Hempel and Schmid-Hempel 1993, Otterstatter and Thomson 2006)

On day 7, all bees were transferred to clean vials without food and left until they defecated. The density of C. bombi in each bee’s feces was determined with a

57 haemocytometer. On day 10, all bees were sacrificed and the total density of C. bombi in their gut tracts was determined with a haemocytometer following Otterstatter and

Thomson (2006).

Statistical analysis

Our final sample sizes were lower than the original design due to mortality from unknown causes (9% of bees died before day 10; subsequent examinations did not reveal unusually intense C. bombi infections), missing fecal samples (17% of bees did not produce enough feces for analysis on day 7 post-inoculation), and the failure of certain bees to develop an infection (10% of bees remained uninfected throughout the experiment). We excluded all of these bees from further analyses. Likelihood ratio ( G) tests showed that, in each case, the proportion of ‘excluded bees’ did not differ between

Control and Alkaloid groups (P > 0.20 in all cases), suggesting that these were not serious sources of bias. In total, we analyzed the infection intensities of 76 bees for day 7

(fecal counts), and 102 bees for day 10 (gut counts).

Given their differing designs, Experiments 1 and 2 were analyzed separately. In both cases, we used multiple regression analysis, with repeated measures on bees (to account for the non-independence of observations on the same individual at day 7 and day 10), to determine whether or not gelsemine reduced the intensity of gut infections. In order to directly compare a bee’s intensity of infection at day 7 (measured as C. bombi cells / µL host feces) and day 10 ( C. bombi cells / µL of gut fluid) post-inoculation, fecal counts were converted to estimated gut counts using the linear regression, gut count =

2 -6.3455 + 0.6955 x feces count (F1,41 = 268.93, P < 0.001, R = 0.93) based on data in

Otterstatter and Thomson (2007). We treated pathogen counts (square-root transformed

58 to satisfy the standard assumption of normally-distributed errors) as our dependent variable, and whether or not C. bombi was exposed to gelsemine (‘Alkaloid’ or ‘Control’ group), time (day 7 or day 10 post-inoculation), and bee body mass, as explanatory factors in our analyses. In Experiment 2, we also included ‘Delay’ as an explanatory factor, i.e., the duration that C. bombi was exposed to gelsemine prior to host inoculation

(no delay, 1 hr delay, 2 hr delay). Preliminary analyses showed that infection intensity in

Control and Alkaloid groups did not meet the standard assumption of homoscedasticity

(F-test for equal variance, day 7: F = 3.96, P = 0.004; day 10: F = 2.07, P = 0.067); we therefore used a heterogeneous variance model (Proc MIXED, SAS Institute 2006) to account for this deviation. For both experiments, we began with a saturated model and removed non-significant effects via backward stepwise elimination. Akaike information criterion (AIC) values were used to compare candidate models; ultimately, the model with the lowest AIC value was chosen as the best fit to the data. We used linear contrasts

(t-tests) in our regression models to compare the average infection intensities of Control and Alkaloid bees at day 7 and day 10. Finally, we used Kolmogorov-Smirnov (K-S) two-sample tests to compare the distributions of infection intensities between Control and

Alkaloid bees; P-values for K-S tests were computed using Monte Carlo estimation (Proc

NPAR1WAY, SAS Institute 2006).

Results

In Experiment 1, an alkaloid-rich diet reduced the intensity of C. bombi infections in bumble bees. Our regression analysis revealed significant main effects of gelsemine

(Alkaloid or Control diet), time since inoculation, and bee body size on infection intensity

(Table 4.1). At 7 days post-inoculation, bees receiving dietary gelsemine had infections

59 that were, on average, 2.2 times less intense than bees receiving the control diet (t = 2.45, df = 36, P = 0.019; Fig. 4.2). Indeed, gelsemine completely prevented heavy infections in bees by day 7: whereas infections in the Control group ranged from 0 – 51500 cells/ µL, the most intense infection in the Alkaloid group was only 5300 cells/ µL (significantly different distributions of infection intensity, K-S test: D = 0.42, P = 0.02). Infection intensities increased significantly from day 7 to day 10, and this effect did not differ between Control and Alkaloid groups (non-significant ‘Gelsemine x Time’ effect, Table

4.1). At 10 days post-inoculation, although average infection intensities were similar in

Alkaloid and Control groups (t = 0.96, df = 36, P = 0.342; Fig. 4.2), the distribution of infection intensities was skewed to significantly lighter infections among bees receiving gelsemine compared to bees receiving the control diet (K-S test: D = 0.35, P = 0.02). For example, while Alkaloid and Control bees exhibited similar ranges of infection intensity at day 10 (150 – 39 188 cells/ µL vs. 150 – 35 250 cells/ µL, respectively), the median infection intensity of Alkaloid bees was less than half that of Control bees (4775 cells/ µL vs. 10850 cells/ µL, respectively). Overall, in Experiment 1, larger bodied bees developed lighter infections than small bees, independently of alkaloid treatment (Table 4.1; Fig.

4.3).

Given that a continuous diet of gelsemine reduced infection intensity in bumble bees, we asked in Experiment 2 if exposing C. bombi cells to gelsemine prior to host inoculation would also reduce infections. Exposing C. bombi inocula to gelsemine did not have a clear effect, however. Gelsemine did not significantly reduce average infection intensity (non-significant ‘Gelsemine’ effect, Table 4.2), nor did the distribution of infection intensities differ between Control and Alkaloid groups for any of the

60 treatments (K-S tests: P > 0.60 in all cases). Average infection intensity increased significantly over time (from day 7 to day 10) when C. bombi was fed to bees immediately (t = 3.40, df = 41, P = 0.002; Fig. 4.4a), but this effect decreased when pathogen cells sat for one hour before inoculation (t = 1.91, df = 41, P = 0.064; Fig. 4.4b), and disappeared when pathogen cells sat for two hours before inoculation (t = 1.60, df = 41, P = 0.118; Fig. 4.4c) (significant ‘Delay x Time’ interaction, Table 4.2). There was no indication that gelsemine affected this variation in infection intensity over time in any of the three experimental treatments (non-significant ‘Gelsemine x Delay x Time’ interaction, Table 4.2).

Discussion

Insect pollinators regularly feed from flowers that contain alkaloid-rich nectar but the consequences of such nectar for pollinators and plants remain unclear (Adler 2000).

Our results demonstrate for the first time that artificial nectar containing a naturally occurring nectar alkaloid reduces the severity of gut infections in pollinators. Bumble bees ( Bombus impatiens ) inoculated with the intestinal parasite Crithidia bombi developed less intense infections when feeding on the alkaloid gelsemine for several days

(Fig. 4.2). In particular, the distribution of infections differed substantially between treatments, with most gelsemine-consuming bees experiencing far lighter infections than the control bees. However, the infectivity of C. bombi inocula was unaffected when pathogen cells were exposed to gelsemine outside of the host. These results suggest that alkaloid-rich nectar can act as a microbicide against a protozoan pathogen of pollinators when ingested, but does not directly interfere with pathogen viability. Given that

C. bombi is deposited at flowers by infected foragers, and can spread between bees via

61 contaminated floral nectar, alkaloid-rich nectar could have substantial effects on the transmission of this pathogen both within the hive and through bumble bee populations.

Our experiments were conducted under laboratory conditions, which facilitated the careful manipulation of both alkaloid and pathogen. However, the artificiality of the lab may also have limited certain aspects of our study. Natural floral nectar is rarely as simple as the artificial nectar that we used; thus, our experiment may have eliminated some of the subtle interactions between nectar components and Crithidia bombi .

Similarly, bees may not forage on a single nectar source continuously for ten days, as we simulated in Experiment 1. Nevertheless, Gelsemium sempervirens flowers very early in the spring (Pascarella 2007) and thus represents one of the few nectar sources for early emerging bumble bees. Finally, by isolating bees in individual vials, we may have disrupted important aspects of infection dynamics that naturally occur within hives, such as the exchange of pathogen cells and strains among nestmates.

The effects of plant secondary metabolites on host-pathogen interaction are understudied and poorly understood (Price et al. 1980, Cory and Hoover 2006). Plant- derived alkaloids appear to have anti-protozoal properties that are effective against human parasites, such as Trypanosoma brucei rhodesiense , the causative agent African sleeping sickness (Freiburghaus et al. 1996). In bumble bees, the plant-derived alkaloid gelsemine appears to have similar anti-protozoal effects on C. bombi , another trypanosome parasite. Although the underlying mechanism is not yet clear, it may be that when a host’s gut contains substantial concentrations of alkaloids, C. bombi cells suffer reduced growth and replication because of costs associated with alkaloid tolerance.

Similar reductions in pathogen proliferation have been reported for secondary metabolite-

62 tolerant plant pathogens (Vanetten et al. 2001). Alternatively, the consumption of alkaloids might alter the host’s gut environment, making it less hospitable for pathogen cells. Logan et al. (2005) proposed this mechanism after pollen consumption altered the rate at which C. bombi populations increase within hosts, perhaps by affecting their adherence to the gut wall. Consumption of alkaloids may also increase gut pH, which could be deleterious to pathogen cells (Stiles and Paschke 1980). Finally, an alkaloid- rich diet may increase a bee’s excretion rate, effectively ‘flushing’ C. bombi cells from the gut wall. Indeed, animals that consume secondary metabolites often deal with the inherent toxicity through rapid excretion (Wink and Theile 2002, Despres et al. 2007), and alkaloid-rich nectar in particular has been shown to increase excretion rates in a nectarivorous bird (Tadmor-Melamed et al. 2004). Gelsemine does not, however, seem to hinder a bee’s immunocompetence towards C. bombi , since this would result in a pattern opposite to what we observed, i.e., higher levels of infection in the gelsemine-consuming bees.

The impact of nectar secondary metabolites on pathogens could be ecologically significant both to bumble bees and the plants they pollinate. Although C. bombi is often considered a benign pathogen (Schmid-Hempel 1998), it renders foragers less able to provide food for their colonies. For example, infected workers have reduced foraging rates, a decreased capacity to learn floral cues, and difficulty manipulating complex flowers (Gegear et al. 2005, Otterstatter et al. 2005, Gegear et al. 2006). The severity of these impairments increases with infection intensity, so although dietary gelsemine does not appear to cure C. bombi infections, it could curtail the adverse effects of the pathogen on host behaviour by reducing infection intensity. Bumble bee queens might derive the

63 greatest benefit from “toxic” nectar. In the spring, queens that emerge from hibernation harbouring C. bombi are less likely to found a colony than healthy queens (Brown et al.

2003). It is possible that a gelsemine-rich diet would suppress an infected queen’s pathogen load to the extent that she could establish a viable colony. In the south-eastern

United States, Bombus impatiens and B. bimaculatus queens often collect alkaloid-rich nectar from Gelsemium sempervirens in the spring (Manson, personal observation); whether or not these queens receive a ‘medicinal’ benefit from this nectar is an important topic for further study. The medicinal properties of “toxic” nectar might also have consequences for plant communities, as parasitic infections are known to alter pollen collection (Schmid-Hempel and Schmid-Hempel 1991) and plant species choice

(Schmid-Hempel and Stauffer 1998) in bumble bees.

Our demonstration that gelsemine can mitigate infections raises the possibility that infected bees might actively self-medicate. There is mounting evidence that infected insects alter their foraging strategies in order to fight pathogens. Some insects adjust basic nutrient intake to improve their overall immune response (Lee et al. 2006), whereas others actively seek compounds that have antimicrobial properties. The active collection of non-nutritive secondary metabolites, or ‘pharmacophagy’ (Boppre et al. 2005), is often associated with a significant shift in diet. For example, parasitoid-infested Platyprepia virginalis caterpillars preferentially consume alkaloid-rich hemlock instead of lupine, their primary host plant, in field choice experiments (Karban and English-Loeb 1997).

Parasitized Grammia geneura caterpillars also choose a mixed diet of plants rich in secondary metabolites rather than a nutrient-rich, but toxin-poor, single-plant diet (Singer et al. 2004). In fact, Singer et al. (2009) elegantly demonstrated that G. geneura self-

64 medicate with pyrrolizidine alkaloids to reduce parasite infections and increase caterpillar survival, despite the fact that the alkaloid reduces fitness in unparasitized individuals. The preferential consumption of secondary metabolites in parasitized G. geneura is caused by an increase in the firing rates of the animal’s taste receptors, which results in increased consumption of pyrrolizidine alkaloids (Bernays and Singer 2005), although the generality of this mechanism is unknown. Several social insect species, including wood ants and honey bees, are known to collect antimicrobial resins to prevent microbe growth within their hives (Konig 1988, Marcucci 1995, Christe et al. 2003, Chapuisat et al.

2007). In the current study, we did not allow bees to choose their diet, so we were unable to test for a gelsemine preference amongst infected individuals. We know of no reports of self-medication by pollinators; the possibility warrants study.

Plants experience multidirectional selection on secondary metabolite concentrations. Strong chemical defenses that reduce herbivory may also reduce pollinator attraction, unless secondary compounds confer a fitness benefit to either plants or pollinators via reduced pathogen loads. Indeed, Price et al. (1980) proposed that the very multifunctional nature of plant defenses may shape the concentration of plant secondary metabolites. If pollinators benefit from, and even seek out, nectar rich in secondary metabolites, selection on plants to decrease alkaloid compounds in nectar may be minimal, and potentially even countered by stabilizing selection from pollinators

(Clayton and Wolfe 1993). Few studies have attempted to tease apart the various forces that select for or against “toxic” nectar in plants (but see Irwin et al. 2004); however, the potential impact of this unusual trait on both plant and pollinator fitness suggests that it merits further investigation.

Acknowledgements

We would like to thank Nathan Muchhala and Mario Vallejo-Marín for comments on the manuscript. This study was supported by grants from the Natural Sciences and

Engineering Council (NSERC). All experiments complied with current laws of Canada.

Table 4.1. Experiment 1: Mixed model statistics describing the effect of an alkaloid-rich diet on the intensity of Crithidia bombi infections in bumble bees. ‘Bee’ was included in each model as a repeated factor to account for the non-independence of sequential observations on individuals. Bees were inoculated with pathogen cells and then fed a daily diet of either alkaloid or control solution (‘Gelsemine’). Pathogen counts were done at 7 and 10 days post-inoculation (‘Time’). Numerator and denominator degrees of freedom (df) are shown for each explanatory factor.

Explanatory Factor F df P

Gelsemine 4.65 1,57 0.035

Time 32.97 1,36 <0.001

Bee Body Size 7.61 1,57 0.008

Gelsemine x Time 0.88 1,36 0.36

Table 4.2. Experiment 2: Mixed model statistics describing the effect of exposing

Crithidia bombi cells to gelsemine for varying durations prior to bumble bee inoculation on infection intensity. ‘Bee’ was included in each model as a repeated factor to account for the non-independence of sequential observations on individuals. Pathogen inocula were mixed with either an alkaloid or control solution (‘Gelsemine’) and fed to bees immediately, or after either a 1 hr or 2 hr delay (‘Delay’). Pathogen counts were done at

7 and 10 days post-inoculation (‘Time’). Numerator and denominator degrees of freedom

(df) are shown for each explanatory factor.

Explanatory Factor F df P

Gelsemine 0.11 1,51 0.74

Delay 10.48 2,51 <0.001

Time 3.74 1,41 0.06

Gelsemine x Delay 0.1 2,51 0.91

Gelsemine x Time 1.1 1,41 0.3

Delay x Time 6.15 2,41 0.005

Gelsemine x Delay x Time 1.06 2,41 0.36

Bee body size was non-significant and excluded from the final model ( F = 1.53, df = 1,51, P = 0.22)

Figure 4.1. Diagram of the experimental design. Throughout, ‘Alkaloid’ refers to a

30% sucrose solution containing 250 ng/ µL of the alkaloid gelsemine and ‘Control’ refers to a 30% sucrose-only solution. In Experiment 1, bees harbouring the gut pathogen

Crithidia bombi were fed a daily diet of Alkaloid or Control solution, whereas in

Experiment 2, bees were fed C.bombi cells that had bee pre-exposed to Alkaloid or

Control solutions for varying durations.

Figure 4.2. Experiment 1: Effect of an alkaloid-rich diet on the intensity of Crithidia bombi infections in bumble bees. Bees were inoculated with a standard dose of pathogen cells and then fed a daily diet of either a gelsemine or control solution. The lower and upper edges of each box indicate the 25 th and 75 th percentiles, respectively, the solid and dashed lines within a box indicate the median and mean values, respectively. Error bars, where visible above and below a box, indicate the 90th and 10th percentiles, respectively.

Infection intensities have been square-root transformed.

Figure 4.3. Relationship between intensity of infection by Crithidia bombi and bumble bee body size in Experiment 1. Each point represents a bee’s infection intensity (adjusted value, from the statistical model shown in Table 4.1) at 7 or 10 days post-inoculation.

Figure 4.4. Experiment 2: Effect of exposing Crithidia bombi cells to the alkaloid gelsemine for varying durations prior to bumble bee inoculation. Boxes are as described in Figure 4.2. Pathogen inocula were mixed with either a gelsemine or control solution and fed to bees (a) immediately, (b) after a 1 hr delay, or (c) after a 2 hr delay.

CHAPTER FIVE

Cardenolide concentrations of nectar, leaves and flowers:

A comparative study across Asclepias series Incarnatae

Jessamyn S. Manson, Sergio Rasmann, Rayko Halitschke, James D. Thomson

and Anurag A. Agrawal

This project was designed in collaboration with A.A. Agrawal. I completed the chemical analyses with the assistance of S. Rasmann and R. Halitschke, while I executed the behavioural assays alone. I received substantial statistical assistance from both A.A. Agrawal and J.D. Thomson, and J.D. Thomson also assisted in writing this chapter.

Abstract

Secondary metabolites are usually thought to deter insects, so their occurrence in floral nectar is surprising. These compounds may have been selected for particular roles in nectar, or they may occur there non-adaptively as a passive consequence of systemic chemical defense of other tissues. Finding the same relative abundances of the same compounds in nectar and leaves would support the “consequence-of-defense” hypothesis, whereas finding unique profiles of compounds in nectar would suggest some adaptive role. We quantified cardenolides in the nectar, leaves and flowers of twelve species from the Asclepias series Incarnatae to test this “consequence-of-defense” hypothesis. We collected samples from a large greenhouse collection and used HPLC to determine cardenolide concentrations. Using the nectar data, we developed behaviour assays to

72 73 assess the ecological consequences of a representative cardenolides on a bumble bee, an important Asclepias pollinator. We found that nectar, leaf and flower cardenolides varied substantially among Asclepias species, but within species there were positive correlations in cardenolide concentration between the plant components. Nectar had fewer individual cardenolides than leaves and, within species, nectar compounds were generally a subset of leaf compounds, supporting our non-adaptive hypothesis. A multivariate analysis of the concentrations of individual cardenolides also found that leaves and nectar of the same species were similar and that this similarity was independent of evolutionary history. Surprisingly, bumble bees had an overall preference for nectar with average concentrations of the cardenolide digoxin, suggesting that nectar cardenolides may contribute to rewarding and retaining pollinators. With some exceptions, our data generally support the consequence-of-defense hypothesis, but imply that “toxic” nectar may not deter pollinating bees. This study is the first to identify and quantify cardenolides in Asclepias nectar and to explore their effects on pollinators.

Introduction

To defend themselves against herbivores, many plants produce noxious compounds that are distasteful, deterrent and often deleterious (Rosenthal and

Berenbaum 1991). Although defensive secondary metabolites have been identified and quantified in the leaves of a vast number of plant species, plants rarely limit the distribution to of these compounds to leaves. Secondary metabolites are also common in roots, stems and flowers (Van der Putten et al. 2001, McCall and Irwin 2006), where defense against herbivores remains a plausible function, but their function in other plant parts such as nectar remains uncertain (Adler 2000). An accurate characterization of

74 plant chemical defense requires a whole-plant perspective that compares concentrations of secondary metabolites between plant “components”, by which we include both organs and exudates such as floral nectar. Given that resources are finite, differences in the allocation of secondary metabolites between plant components should relate to both the relative importance and vulnerability of that structure to herbivory (McKey 1974).

Although this proposition has been addressed by recent work comparing above- and below-ground chemical defenses (Van der Putten et al. 2001, Bezemer and van Dam

2005, Rasmann et al. In Press), substantially less work has been conducted on floral structures and floral rewards. Reports of positive correlations between the concentrations of secondary metabolites in leaves and floral nectar (Adler et al. 2006) provide the first empirical evidence that so-called “toxic” nectar may be due to the systemic production of defensive chemicals.

The presence of secondary metabolites in nectar is a widespread if paradoxical phenomenon, reported in at least 21 angiosperm families (Adler 2000). When investigators have assessed the consequences for animal pollinators of imbibing these

“toxic” nectars, the effects range from putative nicotine and caffeine addiction in honey bee colonies (Singaravelan et al. 2005) to lethal poisoning of honey bee workers after the consumption of a mere 10 µL of nectar from Sophora microphylla (Clinch et al. 1972).

Adaptive hypotheses for nectar-specific functional roles for secondary metabolites are largely derived from the herbivory literature and include deterring inefficient pollinators and nectar robbers (Rhoades and Bergdahl 1981, Adler 2000). Alternatively, secondary metabolites in nectar may arise from leakage of compounds into the nectary during vascular transport or from the systemic production of phytochemicals, making their

75 presence in nectar an undesirable consequence of defense ( sensu Adler 2000). Finding correlations between the concentration and identity of compounds in nectar and other plant components would support this less adaptive explanation, whereas finding compounds unique to nectar would suggest that nectar secondary metabolites have functional significance (e.g. phylogenetic constraint vs. adaptive function; Rhoades and

Bergdahl 1981, Strauss et al. 1999, Adler et al. 2006). The best plant system for evaluating these functional hypotheses would therefore possess defense chemicals that have been identified and quantified across a number of plant structures as well as a collection of pollinators that vary in efficiency and may thus select for or against nectar secondary metabolites.

Milkweeds ( Asclepias spp.) are a classic system for studying plant chemical defenses and their effects on animals. Milkweeds have evolved a number of strategies to protect themselves against a suite of generalist and specialist herbivores, including both physical and chemical defenses (Agrawal and Fishbein 2006). The most potent secondary metabolites found in milkweeds are cardenolides, bitter-tasting compounds that elicit aversive or emetic responses in both vertebrates (Brower et al. 1968) and invertebrates (Dussourd and Hoyle 2000). Cardenolides affect animals by inhibiting sodium pumps, thereby altering sodium and potassium transportation in cells (Malcolm

1991). Cardenolides are often considered a “qualitative” defense, because they are poisonous to generalists at very low doses, yet they can be tolerated and even co-opted by specialists (Feeny 1976, Agrawal and Fishbein 2006). Herbivore damage can increase cardenolide concentrations in some species of Asclepias , but not in others (Rasmann et al. In Press).

The concentrations of cardenolides in leaves vary significantly across the genus

Asclepias (Agrawal and Fishbein 2006, Agrawal et al. 2008) and can also vary among plant components such as the leaves, roots, pith and epidermis within the same plant

(Nelson et al. 1981, Fordyce and Malcolm 2000, Rasmann et al. In Press). Milkweeds are pollinated by a diverse collection of generalists such as lepidoptera, honey bees and bumble bees (Wyatt and Broyles 1994), but there are substantial differences in pollinator efficiency between guilds (Fishbein and Venable 1996, Kephart and Theiss 2004). Given that Asclepias nectar can apparently be toxic to bees (Pryce-Jones 1942), the milkweeds are a good candidate for examining consequences of secondary metabolites in nectar.

In this study, we use the genus Asclepias to explore the functional significance of nectar cardenolides. We quantified cardenolides in the nectar, leaves and flowers of twelve species from the series Incarnatae, a monophyletic group that varies in a number of key traits, including leaf cardenolide concentration (Agrawal et al. 2008). We then used these data to evaluate the “consequence-of-defense” hypothesis. We interpret similarity of cardenolide occurrence in different plant components as supporting a non- adaptive hypothesis such as leakage or a systemically controlled chemical defense strategy; differences in cardenolide profiles among plant components may indicate compartmentalized regulation of cardenolide production, with the possibility of such regulation being adaptive. We propose that similarities in individual cardenolides across plant components are more informative than similarities in gross cardenolides, as overlaps in cardenolide identity between leaves and nectar support a shared chemical production site, while correlations in gross cardenolides may mask significant differences in cardenolide composition. Thus, we specifically asked: 1) Are there positive

77 correlations between cardenolide concentrations in different plant components?, 2) Do cardenolides in nectar have an independent source or are they instead a subset of the cardenolides found in leaves?, 3) Are differences in the concentrations of cardenolides between leaves and nectar constrained by evolutionary history in the series Incarnatae, and 4) What are the ecological consequences of nectar cardenolides for bumble bees, a common pollinator of Asclepias ?

Methods

Study system

The genus Asclepias is a monophyletic group (Agrawal and Fishbein 2008) composed of about 135 species found in the Americas (Woodson 1954, Agrawal and

Fishbein 2008, Fishbein et al. In Press). We selected species from the series Incarnatae, as this is a monophyletic group with relatively well-resolved phylogenetic relationships

(Agrawal and Fishbein 2008) and a significant variation in leaf cardenolide concentrations among species (Agrawal et al. 2008). The plants were part of two independently established permanent greenhouse collections at Cornell University.

Quantifying cardenolides

We collected samples to analyze constitutive cardenolide concentrations on five occasions between July 2007 and August 2008. The frequency and timing of nectar collections depended on the flowering time of the different species; we collected nectar from species that flowered abundantly over several sampling periods, whereas we were only able to collect nectar from rarely flowering species on a single sampling day. We also collected whole mature leaves from all twelve species. On the four occasions when

78 we did this, we sampled the leaves after taking nectar samples in order to avoid inducing cardenolide production in nectar or leaves by the leaf removal. We also collected whole flowers but were unable to gather samples from all twelve species; this was, in part, due to the fact that flowers were often damaged during nectar collection and were therefore not appropriate for evaluating constitutive cardenolide concentrations.

We extracted nectar from flowers using 5 µL graduated microcapillary tubes primarily, although several species (e.g., Asclepias nivea ) produced nectar so copiously that we used a 200 µL microcapillary tubes for sample collection. We took every precaution to ensure that we caused no damage to the nectary, as this could induce cardenolides; fortunately, damage can be visually detected in Asclepias as it induces the exudation of latex. On the few occasions where damage did occur, samples were discarded. Because most plants produce very little nectar per flower, we pooled nectar samples to improve our chances of detecting trace amounts of cardenolides. Samples were pooled within days and individual plants, with each sample containing between 20 and 230 µL of nectar, representing tens to hundreds of individual flowers. After pooling we had between 1 and 8 individual samples per species. Nectar was added to 500 µL of

70% ethanol (following Bluthgen et al. 2004) and stored at -80°C prior to analysis. We harvested leaves and flowers from a subset of plants also used for nectar sampling and froze them immediately at -80°C.

We used high performance liquid chromatography (HPLC) to quantify cardenolides in nectar, leaf and flower samples, adapting a protocol from Zehnder and

Hunter (2007). We prepared nectar samples for extraction by drying down all water and ethanol from the stored samples using a rotary evaporator (Labconco). We extracted the

79 residuum with 1 mL of 100% methanol and added 10 µL of a 0.2 g/L solution of the cardenolide digitoxin as an internal standard, which allowed for the direct comparison of unknown cardenolide peaks to a known cardenolide of predetermined quantity. We let the samples shake gently for 24 hours (cardenolides are generally stable at room temperature; Malcolm 1991) and then spun them down in the centrifuge for 30 minutes at

14000 rpm and 15°C, removing the supernatant. In some cases, preliminary analysis of samples had very low cardenolide yield, so where necessary, we pooled samples yet again to increase our power to detect cardenolides. For leaf and flower samples, we ground wet tissue with liquid nitrogen, and added 10 µL of 2 g/L digitoxin to each sample, which was around 100 mg of tissue (wet weight). We extracted the samples with

1 mL of 100% methanol using the FastPrep® homogenization system (MP Biomedicals) to rapidly lyse cells set at 5.0 m/s for 30 seconds. Flower and leaf samples were spun down as before, but we did not further concentrate the samples.

Cardenolides were analyzed on an Agilent 1100 series HPLC. We injected 15 µL of the extract. Compounds were separated on a Gemini C18 reversed phase column

(3 µm, 150 x 4.6 mm, Phenomenex, Torrance, CA) using the following solvent gradient

(solvent A: 0.25% phosphoric acid in water; solvent B: acetonitrile): 0-5 min 20% B, 20 min 70% B, 20-25 min 70% B, 30 min 95% B, 30-35 min 95 % B at a flow rate of 0.7 mL/min. UV absorption spectra were recorded from 200 to 400 nm and cardenolides were quantified by integrating the peak area at 218 nm. Cardenolides are identifiable from other compounds by a single symmetrical peak that absorbs at 218 nm (Zehnder and

Hunter 2007), and we confirmed this by checking the shape and absorption of the peak in control samples containing only digitoxin. The amounts of cardenolides present were

80 then calculated relative to the peak area of the digitoxin internal standard, of which a known concentration was added to the sample during extraction. This estimation assumes that all cardenolide molecules have the same absorption characteristics as digitoxin. Final cardenolide concentration estimates were calculated as nanograms of cardenolides per microlitre of nectar or per microgram of fresh tissue collected.

Pollination biology

Milkweeds have a unique and complex pollination system: their pollen grains are contained and transported in discrete pollinia. Pollinia become attached to the legs of flower visitors and must then be oriented and inserted correctly into the stigmatic chamber of the recipient flower to achieve pollination (Wyatt and Broyles 1994). Pollen tubes travel down from the surface of pollinium, through the nectary to the ovary below for successful fertilization to occur (Wyatt and Broyles 1994). Milkweed pollination requires an animal pollinator, but despite their mechanically elaborate pollination system, all milkweeds that have been studied to date are visited by a suite of generalist flower visitors, frequently including bumble bees (Wyatt and Broyles 1994).

Nectar secondary metabolites can substantially reduce pollinator visitation frequency in choice experiments against flowers containing sugar alone (Gegear et al.

2007), but can also reduce pollinator proficiency by increasing the length of time a pollinator spends on a flower (Adler and Irwin 2005). While the number of visits clearly correlates with reproductive success, the length of a floral visit can be tied to the number of pollen grains removed and subsequently donated to the next flower (Thomson 1986).

Whether visit length affects pollinia removal and deposition rates in Asclepias spp. remains untested (Morse 1982), but we can presume that longer visits should increase a

81 bees’ probability of collecting pollinia and inserting them in subsequence flowers. We evaluated the effect of nectar cardenolides on both aspects of bumble bee behaviour by using artificial nectar enriched with a cardenolide and testing pollinator preference on artificial flower arrays. We chose bumble bees because they are generalists, important natural pollinators of milkweeds and were commercially available to us. We created an artificial nectar solution by mixing 30% w/w sucrose with the cardenolide digoxin (92%

HPLC grade, Sigma).

Digoxin, a cardenolide found in Digitalis spp., is a bitter cardenolide (Malcolm

1991) that causes 50% mortality at a concentration of 0.5% in honey bees (Detzel and

Wink 1993). The concentration range of cardenolides we observed was 0-109 ng/ µL across the twelve species found in the series Incarnatae, with a mean nectar cardenolide concentration of 11.5 ng/ µL (see Fig 5.1 and appendix two). With these data for reference, we created artificial nectar concentrations of 0, 10 and 50 ng/ µL digoxin, simulating control, mean and moderate nectar cardenolide concentrations, respectively.

Our base artificial nectar was a 30% w/w sucrose solution, a reasonable simulation of natural Asclepias nectar. We mixed nectar solutions every two days, refrigerating unused portions at 4°C for no more than 24 hours.

We evaluated pollinator preference and flower-handling proficiency using methods reported by Gegear et al. (2007). In short, marked worker bees were trained to associate artificial flower colour (either blue or yellow) with one of two nectar conditions

(see descriptions below) by foraging freely on alternating monotypic arrays of each flower type. The association between flower colour and nectar condition was randomized among bees to control for any potential bias due to innate colour preferences.

Immediately following training, individual bees foraged on an array with 30 flowers of each type for at least 80 flower visits. We filled flowers with 2 µL of nectar and refilled each flower immediately after it was drained by a bee. We replaced flowers between bees. All foraging bouts were videotaped for later analysis using JWatcher Video

Version 1.0 (Blumstein and Daniel 2007).

We completed three separate experiments to evaluate pollinator preference. In the first, bees chose between control nectar (30% w/w sucrose only) and nectar simulating the mean nectar cardenolide concentration found in the greenhouse populations of

Asclepias (10 ng/µL digoxin in 30% sucrose). In the second experiment, bees chose between control nectar and nectar with a higher cardenolide concentration (50 ng/µL digoxin); this concentration simulates the upper limit of cardenolides found in all species save Asclepias pumila , which had an extremely high cardenolide concentration (see appendix two). Finally, we evaluated whether bees reacted differently to the two concentrations (10 ng/µL vs. 50 ng/µL digoxin). Training periods were consistent between experiments except for the differences in nectar treatments. Individual bees were used only once and then dissected for pathogen analysis.

Bumble bee foraging behaviour can be significantly affected by the gut protozoan

Crithidia bombi (Gegear et al. 2005, Otterstatter et al. 2005), but the ingestion of nectar secondary metabolites can reduce pathogen loads in bumble bee workers (Manson et al.

In Press). To evaluate whether bumble bees infected with C. bombi preferred to forage on “toxic” nectar as a means of self-medication, we evaluated each bee’s pathogen load after each foraging assay. Each bee was refrigerated for 48 hours and then its gut was excised and ground with 100 µL of distilled water in a microcentrifuge tube. We allowed

83 samples to settle for 24 hours at 4°C, then we removed the supernatant and counted C. bombi cells using an improved Newbauer haemocytometer (Otterstatter and Thomson

2006). As an additional covariate potentially affecting foraging, we also estimated the bee’s body size by measuring the length of the radial cell on the right forewing (Harder

1982).

Statistical analysis

Quantitative cardenolide analysis

Because many cardenolides did not appear in all samples, the data set contains many zeros. The distribution of estimated concentrations could not be made normal via data transformation, so we chose non-parametric analyses. Nectar was quantified by volume during collection, but we converted these values to mass to facilitate comparisons across plant components. To do this, we determined the mass of 2 µL of a 60% sucrose solution in the lab (representing the average sucrose concentration of our nectar samples) and converted microlitres to micrograms by a factor of 1.057 (data not shown). Prior to any statistical analyses, we calculated average concentrations as follows: for average total concentration, we summed all cardenolides in a sample and divided that by the mass of the sample, giving us a total concentration in ng/ µg and then we summed the concentrations of all samples of that plant component for each species and divided by the number of samples collected to achieve an average total concentration (see appendix two for raw data); for average concentration of individual cardenolides, we divided each compound by the sample mass to get a ng/ µg measure, then averaged the concentration of each compound across all samples within a species (see Fig. 5.1).

We began our analyses with a general comparison of the concentrations of average total nectar cardenolides between the twelve species as well as between plant components within each species using Kruskal-Wallis tests. This analysis was followed by a sign test to examine consistent directional differences between nectar, leaves and flowers, regardless of species identity. We then tested for positive correlations in total cardenolide concentrations between leaves, nectar and flowers of the same species using

Kendall’s rank correlation.

Qualitative cardenolide analysis

Similarities in the suite of individual cardenolides found in different species and plant structures can indicate whether cardenolides have a shared site of production. Here we focused our analysis on differences between nectar and leaves only, as we have the most complete data for these two plant components.

We compared the number of individual cardenolides in leaves and nectar between all twelve species using a generalized linear model with a Poisson distribution. We then examined differences in chemical polarity, a characteristic that affects the mobility and absorbency of compounds. Highly polar cardenolides are poorly absorbed by animals and less mobile within plants, while less polar cardenolides are absorbed quickly and are highly mobile (Malcolm 1991); because reverse phase HPLC filters compounds based on chemical polarity, retention time is a reasonable surrogate for that polarity, with highly polar compounds having a short retention time and less polar compounds having longer retention times. We examined the average retention time of samples with detectable cardenolides, comparing differences across species and between plant parts. To calculate the average retention time for a sample containing multiple compounds, we weighted the

85 retention times of individual compounds by their proportion of the sample’s total cardenolide concentration (Fordyce and Malcolm 2000). We analyzed the data using a generalized linear model with a Gaussian distribution. We also used a sign test to check for directionality of changes in weighted averages between leaf and nectar samples.

Comparing the concentrations of individual cardenolides across species and plant parts required a multivariate approach. Because of fundamental differences in the physical construction of nectar and leaves, we converted raw cardenolide concentrations into relative concentrations, or the proportion that the cardenolide contributed to the total cardenolide concentration of the sample, thus reducing spurious associations driven by absolute concentration differences in plant components. We used a two-dimensional non-metric multidimensional scaling (NMDS) ordination to order nectar and leaf tissues from each species (entities) by the similarity of their cardenolide profiles (attributes).

The ordination used a Bray-Curtis dissimilarity index and was conducted in R’s vegan package (Oksanen 2009). We were required to remove samples that had no detectable cardenolides, as their positions in an ordination are undefined; these samples included the leaves and nectar of Asclepias angustifolia and A. fascicularis , but also the nectar samples of A. curassavica , A. incarnate ssp. pulchra and A. texana , leaving some of the leaf samples in the ordination unpaired. We examined contributions of each cardenolide to the ordination using the envfit function in the R package vegan (Oksanen 2009) followed by Bonferonni corrections to account for the multiple pairwise comparisons.

We performed Mantel tests to examine correlations between chemical composition and phylogenic relationships. We did this by comparing our chemical dissimilarity (distance) matrix, which identifies whether plant components within a

86 species have more compounds in common than plant components in different species, to a phylogenetic distance matrix, which was constructed from pairwise distances based on the molecular branch lengths from a pruned phylogenetic tree of the series Incarnatae (see

Agrawal and Fishbein 2008, Fishbein et al. In Press for complete phylogeny). We analyzed nectar and leaf data separately to prevent spurious correlations between leaves and nectar from the same species, which have a phylogenetic distance of zero. We performed the analysis using the R package ape (Paradis 2009).

Behaviour analysis

We assessed pollinator preference for flowers with or without nectar cardenolides using a G-test to evaluate the total goodness of fit of all bees within an assay, testing for an expected visit frequency of 0.5 or random. We also examined whether the visitation data from each bee was homogeneous across an assay. In addition, we tested for correlations between Crithidia bombi infection intensity and pollinator preference.

We then examined foraging proficiency by comparing average flower handling time and foraging rate (visits per minute) between treatments. To capture data that demonstrated consistent foraging behaviour on each flower type, we calculated the average length of a flower visit from ten consecutive visits to the same flower type found between visits 45 and 75. We analyzed the extracted data using a generalized linear model, with radial cell length and pathogen load as covariates. Data were transformed to meet assumptions of normality when necessary. Analyses were performed in R (version

2.9.0).

Results

Quantitative cardenolide analysis

The concentration of average total cardenolides per microgram of Asclepias nectar, calculated as the sum of total cardenolide concentrations within a species divided by the number of samples, differed significantly among species (χ2=38.8, df=11,

P<0.001; Fig. 5.1), with five of the twelve species having no constitutive cardenolides in their nectar at all. Species that had quantifiable nectar cardenolides showed a 30-fold difference between the lowest and highest total cardenolide concentrations. A within- species non-parametric analysis revealed that all but two species demonstrated no difference in average total cardenolide concentrations between plant components

(analysis not shown), while the two species that did differ significantly, Asclepias curassavica and Asclepias angustifolia , had no detectable nectar cardenolides.

The cardenolide concentration for nectar samples (12.17 +/- 3.22 ng/ µg) regardless of species, was more than twelve-fold higher than the average cardenolide concentration in leaves (0.54 +/- 0.18 ng/ µg) or flowers (0.76 +/- 0.41 ng/ µg).

Nonetheless, it is difficult to quantitatively compare these values due to the inherent physical differences between nectar and leaf tissue. Sign tests comparing nectar cardenolides to leaf and flower cardenolides found that these differences were not consistent between species (leaf: s=7, n=11, P=0.34; flower: s=2, n=6, P=0.69). It should also be noted that cardenolide concentrations were entirely independent of initial mass

(analysis not shown), such that gross cardenolide mass (ng) in each leaf, nectar and leaf sample is not correlated with the amount of material collected.

There was a significant positive correlation between average total cardenolide concentrations in nectar and those in leaves in the series Incarnatae (n=12, τ =0.43,

P=0.03). Although this relationship is maintained between the average concentration of cardenolides in leaves and flowers (n= 6, τ =0.73, P=0.03), there is a poor correlation between nectar cardenolides and flower cardenolides (n=6, τ =0.15, P=0.34), likely due to the limited flower sample size.

Qualitative cardenolide analysis

The number of individual cardenolides differed significantly across species

(F 11,74 = 12.07, P<0.01) and we found a significant interaction between plant part and species (F 10,74 = 7.69, P<0.01). Furthermore, within-species analyses using Wilcoxon rank-sum tests showed that this pattern was driven by two sister species, Asclepias curassavica and Asclepias nivea (Agrawal and Fishbein 2008), which have significantly fewer individual cardenolides in their nectar relative to their leaves (see appendix two).

The identities of cardenolides in leaves and nectar were somewhat different. In all, we identified thirty individual cardenolides (see appendix two). There was a single cardenolide that was unique to Asclepias nectar: both A. pumila and A. perennis had a detectable peak at 14.1 minutes that was found only in their nectar samples. Conversely, nearly one third (9 of 30) of the compounds found in leaves were not found in nectar samples. The compounds that were missing from nectar were scattered throughout the range of cardenolide retention times. The majority of individual cardenolides were found in both leaf and nectar samples, although seven compounds were found in only a single species, five of which originated from A. perennis .

The average weighted retention time, which estimates polarity, also differed across species (F 9, 39 =16.59, P<0.01). While there was no difference in retention time between leaves and nectar (F 1, 38 =0.18, P=0.68), there was a significant species by plant component interaction in the model (F 6, 39 =5.92, P<0.01). A sign test found no directional differences in average weighted retention time between leaf and nectar samples (n=12, s=6, P=0.75).

Our NMDS ordination analysis (Fig. 5.3) revealed that identity and concentration of the cardenolides in nectar tended to match those of leaves, with the ordination rejecting the notion that nectar cardenolides are unique or that their proportions are independently controlled. Five of the seven species with leaf and nectar data have consistent directionality, with nectar samples closer to the y-axis and leaf samples close to the x- axis; this trend was driven by the reduction in number of individual compounds found in nectar samples relative to leaf samples. A sixth species, Asclepias pumila , had the opposite relationship, with its nectar and leaf samples sitting closer to the x- and y-axis, respectively; this reversed trend was because both plant components have the same number of compounds and nectar cardenolides were therefore not a subset of leaf cardenolides in this case (see appendix two). The three points that fall near the edge of the ordination are isolated by the fact that each species has only a single detectable cardenolide. This single compound makes these samples much less complex than the rest, but also means that each compound represents 100% of their relative cardenolides, which is what drives the points to the ordination’s edge. Interestingly, the leaf samples from Asclepias mexicana and its paired nectar sample differ in their single cardenolide, but the compounds were not unique to the species.

Our Mantel tests indicated that there was no autocorrelation between chemical differences and phylogenetic distances in nectar (z=12.12, P=0.67), or in leaves (z=27.82,

P=0.41), suggesting that the spatial associations in the ordination, which summarize cardenolide similarity, are independent of the species’ evolutionary history. We also found that no individual cardenolide had a significant effect on the organization of the ordination (P>0.05 for correlations of all compounds with both axes of the ordination after Bonferonni correction).

Behaviour analysis

Bees had different responses to the two different nectar cardenolide concentrations. In assay 1, bees exhibited a preference for nectar with 10 ng/ µL digoxin rather than the sucrose-only control (G=45.49, df=1, P<0.01). Bees in the second assay were generally indifferent to nectar with 50 ng/ µL digoxin (G=0.87, df=1, P=0.35), while bees in the third assay preferred to forage on nectar with 10 ng/ µL digoxin significantly more than nectar with 50 ng/ µL digoxin (G=23.19, df=1, P<0.01). All three assays had significant levels of heterogeneity between bees (assay 1 G=670.6, df=11; assay 2

G=843.2, df=10; assay 3 G=551.4, df=9, all P<0.01; see appendix three), indicating that individual bees had variable preferences. The key result from these behavioural assays is that low levels of cardenolides were attractive to bees, while intermediate cardenolide levels neither attracted nor deterred foragers.

Despite these general trends in pollinator preference, individual bees had very different responses to nectar cardenolides (see appendix three). In the first assay, three of the twelve individual bees randomly visited flowers (bees 8, 9 and 11), while three bees primarily visited flowers with sucrose-only nectar (bees 2, 3, 6), and six bees preferred

91 nectar with 10 ng/ µL digoxin (bees 1, 4, 5, 7, 10, 12). Bees in the second assay had only a single bee that foraged without preference (bee 21), while the remaining ten individuals were evenly split between strong preferences for sucrose-only flowers and strong preferences for flowers with moderate cardenolides. In assay 3, half the bees chose the lower cardenolide concentration, while four bees chose higher cardenolides and a single bee foraged without preference. We found no correlation between the proportion of sucrose-only visits and intensity of Crithidia bombi infection detected in workers (n= 29,

τ=-0.02, P=0.55).

Further analysis revealed that bees in all three treatments had a significant overall preference for blue flowers (G=23.09 for assay 1, G=225.4 for assay 2, G=300.3 for assay 3, all df=1, P<0.01; see appendix three). However, we still saw significant heterogeneity for this preference within the data (assay 1 G=693.0, df=11; assay 2

G=618.8 df=10; assay 3 G=261.9, df=9, all P<0.01). The majority of all individual bees, twenty-one of thirty-three, preferred to forage on blue (appendix three). This apparent blue bias may be associated with the colour of the first flower visited, which was blue

95%, 91% and 100% of the time for assays 1, 2 and 3, respectively. Interestingly, the intensity of C. bombi in bees did significantly correlate with the proportion of visits to blue flowers, (n= 29, τ=0.22, P=0.04). This pattern was driven by one particular bee with a very high infection level and strong affinity for blue flower. However, removing this bee from the analysis still shows a suggestive, if non-significant, trend (n= 28, τ=0.18,

P=0.08, Fig. 5.4).

The consumption of cardenolides had no effect on bumble bee foraging proficiency. There was no difference in average visit length between bees foraging on

92 control or cardenolide-enriched nectar of either average or moderate concentration

(F 2,24 =1.79, P=0.19; Fig. 5.5a). We also found no difference in the number of visits per minute between the three treatments (F 2,24 =0.14, P=0.87, Fig. 5.5b). Neither visit length nor visitation rate was affected by either bee size or pathogen load.

Discussion

Nectar cardenolides in the Asclepias series Incarnatae vary in concentration, identity, number and chemical polarity; the variation across species is substantially higher than between nectar and other plant components within species. More importantly, the comparative data generally support the consequence-of-defense hypothesis for nectar.

Positive correlations between the cardenolide concentrations of different plant components suggest that the presence of nectar cardenolides is likely a consequence of systemic chemical defense. Nectar appears to contain a simpler cocktail of secondary metabolites than foliage, and the compounds are generally a subset of those found in leaves (see appendix two). With one exception, the individual cardenolides found in nectar can also be found in leaves, refuting the suggestion that nectar has evolved an independent chemical arsenal. Finally, we found that bumble bees preferred to forage on flowers with average cardenolide concentrations, which could imply that pollinator- mediated selection could reinforce or stabilize the concentrations of cardenolides in nectar. Our study, the first to detect and quantify nectar cardenolides, suggests that the chemical complexity of Asclepias nectar may be a byproduct of systemic phytochemical defenses.

The relationship between cardenolide concentrations in leaves and in other plant components is an understudied area, but recent interest in the feedback between above-

93 and below-ground herbivory has provided some insight. Rasmann et al. (In Press) found a significant correlation in total constitutive cardenolide levels between the roots and leaves of twelve Asclepias species from the series Incarnatae (ten of which are shared between this study and our own), a pattern which complements our findings in nectar.

Taken together, the correlations between total cardenolide concentrations in Asclepias shoots, nectar and roots suggest that chemical defense concentrations in milkweeds are systemically linked. This is further supported by evidence that total leaf cardenolides also correlate with flavonoids, another foliar defensive chemical (Agrawal et al. 2009).

In contrast, leaf cardenolides are not correlated with physical defensive traits such as trichome number and leaf thickness, refuting the concept of a complex plant defense syndrome (sensu Agrawal and Fishbein 2006). Because our study is only the second to compare nectar secondary metabolites to foliar secondary metabolites, we cannot infer a general trend towards correlated chemical composition across these plant parts.

However, Adler et al.’s study on alkaloids in Nicotiana tabacum (2006) also found a significant positive correlation between leaf and nectar secondary metabolites. Despite the implications of these correlations, analyses that focus on total cardenolide concentrations may mask differences in the concentrations of individual cardenolides, differences which are critical to understanding cardenolide allocation within a plant.

Similarities in the identity and concentration of individual cardenolides found in Asclepias nectar and leaves tend to support the consequence-of-defense hypothesis.

Regardless of species, of the twenty-one individual cardenolides found in nectar, twenty are also found in leaves, so nectar cardenolides in general appear to be a subset of leaf cardenolides and not a unique suite of compounds. However, when we compare leaf and

94 nectar cardenolides within species, we see that individual nectar cardenolides have matching compounds in leaves only 63% of the time, on average, while the remaining nectar compounds are not unique to Asclepias in general but may be unique to that species (see appendix two). In other words, although the majority of nectar compounds are nested within leaf compounds, there is still a substantial proportion that, while not novel, are absent from the leaves of a species but present in its nectar. These cardenolides, found in the nectar but absent from the leaves, can represent up to 40% of a species’ total nectar cardenolide concentration. A simple nested relationship is further refuted by the single unique nectar cardenolide (14.1 min), which is found in the nectar of two species but has not been identified in leaves or flowers and may therefore have a novel source and function. One explanation for this discrepancy between nectar and leaves could be that the “missing” compounds are in very low concentrations in leaves and are therefore not detected during analysis. Although the majority of nectar cardenolides are a subset of leaf cardenolides, the inconsistencies in this putatively nested relationship suggest that the systemic defense hypothesis does not explain every feature of cardenolide distribution within a plant.

In contrast, our multivariate analysis indicates that the concentrations of individual cardenolides in nectar and leaves from the same Asclepias species have more in common than individual cardenolides of nectar from different species, reaffirming that nectar cardenolides are largely a subset of leaf cardenolides. The lack of a phylogenetic signal in the data suggests that the similarities in chemistry across species are independent of evolutionary history. Two previous studies comparing total cardenolide concentrations across species have found no phylogenetic conservatism in Asclepias leaf

95 cardenolides (Agrawal et al. 2008, Rasmann et al. In Press), while another did detect a phylogenetic signal in total leaf cardenolides (Agrawal et al. 2009); all three studies used substantially more species than ours and the species used in the comparisons were similar but did not overlap completely. In addition, the study by Rasmann et al. (In Press) did not find a phylogenetic signal in the total leaf or total root cardenolides from twelve species within the series Incarnatae (ten of which overlap with our own study), although the authors suggest this may be due the limited number of species examined. Overall, similarities in cardenolide composition between leaves and nectar seem to provide some evidence for a phylogenetically independent systemic chemical defense strategy.

Nearly a third of leaf cardenolides were undetected in nectar, but the reason for their absence was not clear. Previous work comparing the chemistry of different

Asclepias plant parts has found that the chemical polarity of cardenolides can differ significantly between leaves and roots (Nelson et al. 1981), and between leaves and stems

(Fordyce and Malcolm 2000), which led to us to question whether the cardenolides missing from nectar might all have similar chemical polarities. The chemical polarity of a cardenolide is linked to the compound’s rate of absorption in an animal’s gut: compounds with low polarity are bitter-tasting and quickly absorbed after consumption, causing rapid emetic responses but also potential systemic toxicity, whereas highly polar compounds are harder to taste and are poorly absorbed, but can become cumulatively toxic over time (Malcolm 1991). In addition, chemical polarity seems to be related to the mobility or transportability of a compound within a plant, with compounds of low polarity being much more portable than compounds of high polarity (Malcolm 1991).

Given these characteristics, we might expect that the absence of compounds with high

96 polarity from nectar would support the hypothesis that secondary metabolites leak into nectar as a consequence of defense. However, if nectar secondary metabolites function to deter inefficient pollinators, we would also expect that highly polar, and therefore less toxic, cardenolides would be excluded from nectar. Unfortunately for this conjecture, we can infer from the both the raw data (appendix two) and the weighted average retention times that excluded compounds have a range of polarities. In addition, there is no directionality in weighted average retention time, suggesting that there is no trend of an increased proportion of cardenolides with low polarity in nectar. Finally, we did not find that the absent compounds were of particularly low concentration in leaves, refuting the possibility that missing compounds were simply too dilute to detect in nectar. We must therefore conclude that neither chemical polarity nor reduced concentration explain the exclusion of certain cardenolides from nectar.

The effects of nectar secondary metabolites on pollinator behaviour are inconsistent, with some studies reporting that bees have a strong aversive reaction to compounds such as alkaloids (Adler and Irwin 2005, Gegear et al. 2007), phenolics

(Hagler and Buchmann 1993) and cyanogenic glycosides (London-Shafir et al. 2003), and others showing indifference or preference for secondary metabolites such as alkaloids

(Singaravelan et al. 2005). Although Pryce-Jones (1942) observed that Asclepias nectar was toxic to bees, this is the first study to rigorously evaluate how pollinators respond to a cardenolide that is believed to be in the nectar of Digitalis spp. (Detzel and Wink 1993).

Our analysis of bumble bee preference suggests that bumble bees find average concentrations of nectar cardenolides more attractive than nectar containing only sugar, and are indifferent to higher concentrations of nectar cardenolides. Although pollinator

97 attraction to nectar secondary metabolites is not without precedent (Singaravelan et al.

2005, Liu et al. 2007), the result was surprising in light of previous research that demonstrates cardenolide toxicity in other animals (Malcolm 1991).

Despite their reputed toxicity, secondary metabolites may act as a post- consumptive foraging cue, creating an association between the taste of the secondary compound and the quality of the reward (Cipollini and Levey 1997a). This association is, however, unlikely under our experimental conditions, which provided rewards of equal caloric value and a clear pre-consumption colour cue to identify each nectar treatment. We must therefore consider that cardenolides at low concentrations are actually appealing to bumble bees; to our knowledge this is the first case of cardenolides acting as a feeding stimulant to a generalist forager. In this study, bumble bees preferred nectar with low cardenolides to nectar with high cardenolides or sucrose-only nectar, which could suggest a dose-dependent association. Artificial nectar with low concentrations of caffeine or nicotine are preferred by free-foraging honey bees over a sugar-only solution (Singaravelan et al. 2005), whereas nectar with average concentrations of either alkaloid deterred foragers. However, it is unclear whether the preference for low-dose nicotine or caffeine is driven by a simple cue association or by a dependence on these compounds, which are known to be addictive. There is currently no evidence to suggest that cardenolides have addictive properties, although there are many examples of cardenolide preferences in specialist herbivores (Malcolm 1991).

Despite these intriguing results, it should be noted that our behaviour assays may not accurately reflect bumble bee foraging decisions in natural populations of Asclepias .

The cardenolide concentrations we used reflect the constitutive concentrations found in

98 our greenhouse grown collections, but concentrations of nectar cardenolides may be significantly higher in natural Asclepias populations, where plants are under selection by herbivores, and where concentrations may be elevated through induction or evaporation.

In addition, while many of our Asclepias nectar samples contain a number of different cardenolides, we used only a single compound in our artificial nectar. We chose to use digoxin to synthesize Asclepias nectar because it is a commercially available botanical cardenolide that has lethal post-consumptive consequences for honey bees (Detzel and

Wink 1993). Digoxin also has low polarity and is therefore quite bitter and more rapidly absorbed than its more polar counterparts (Malcolm 1991). We predicted that these characteristics would enhance the likelihood of both aversive and emetic behaviour in bees, which was clearly not the case. It is difficult to predict how nectar with a more complex chemical profile might affect foraging behaviour, and the issue merits further investigation.

The unexpected preference for blue flowers makes interpreting the results of the behavioural assays more complicated. This trend may be driven simply by the overwhelming preference for the first flower visited on the array to be blue (true in 94% of the assays, despite randomizing the colour of the closest flower to the hive), which may have been driven by an innate preference for blue (Heinrich et al. 1977, Gumbert

2000). Previous studies suggest that bees rapidly associate colour with reward (e.g.

Heinrich et al. 1977, Dukas and Real 1991) and that bees with foraging experience rely on learned colour cues over innate colour bias (Gegear and Laverty 2004); this does not appear to be the case in our study. Interestingly, we found that the infection intensity of

Crithidia bombi correlated significantly with the proportion of visits to blue flowers (Fig.

5.4). C. bombi is known to impair associative learning in foraging bees (Otterstatter et al.

2005, Gegear et al. 2006) and we hypothesize that, as C. bombi infection intensity increased, bees became unable to associate floral colour with nectar treatment during training, rendering the training period irrelevant. We may further hypothesize that this innate preference for blue may have motivated bees to make their first foraging visit on the mixed array to a blue artificial flower. Bees may then have chosen to become constant visitors on blue flowers because they were adequately rewarding. Although the blue bias complicates our understanding of the results, it actually reinforces the fact that the nectar cardenolide treatments are not particularly deleterious for pollinators, as digoxin at moderate concentrations did not discourage visitation to blue flowers.

Pre-existing infections of the protozoan pathogen Crithidia bombi within our experimental colony gave us the opportunity to simultaneously test another adaptive hypothesis of “toxic” nectar, that of its putative antimicrobial properties (Adler 2000).

Secondary metabolites are hypothesized to have antimicrobial activity to reduce degredation of nectar quality by floral yeasts (Hagler and Buchmann 1993). However, these antimicrobial properties may also provide benefits for foragers if they can reduce pathogen loads (Manson et al. In Press). C. bombi is a protozoan pathogen that affects pollinator foraging (Gegear et al. 2005, Otterstatter et al. 2005) and can reduce colony fitness (Imhoof and Schmid-Hempel 1999). Previous work has found that leaf cardenolides from A. curassavica can reduce infections of the protozoan pathogen

Ophryocystis elektroscirrha in monarch butterfly caterpillars (de Roode et al. 2008). In addition, the nectar alkaloid gelsemine has been shown to reduce C. bombi loads in bumble bees at natural concentrations (Manson et al. In Press). In our study, we assessed

100 standing pathogen loads only at the end of preference trials, so we cannot evaluate whether cardenolides reduced C. bombi infections. It also seems unlikely that the consumption of cardenolides for tens of minutes might be effective at mitigating infection. Instead, we postulated that if cardenolides are therapeutic for ill bumble bees, infected individuals may demonstrate a preference for cardenolide-rich nectar indicative of self-medication (Singer et al. 2009). We found that this was not the case, nor was there any indication that the greater the infection intensity, the more visits made to cardenolide- rich flowers. However, given the tantalizing possibility of pollinators actively mitigating pathogen levels with cardenolides, the issue warrants further investigation.

Finally, we consciously decided to sample plants free from herbivore damage in order to restrict our analysis to constitutive cardenolides. However, herbivore-induced chemical defenses can significantly alter the distribution and concentration of secondary metabolites in a plant (Karban and Baldwin 1997). This is particularly true in Asclepias , where differences in the concentration of leaf cardenolides between species can jump from five-fold to forty-five fold after damage by monarch butterfly larvae (Rasmann et al.

In Press). In contrast, the same study found that foliar damage did not increase cardenolide concentrations in roots (Rasmann et al. In Press), demonstrating that induction may not be a systemic response in Asclepias . While foliar herbivory in

Nicotiana tobacum increased nectar alkaloids by 33% (Adler et al. 2006), no studies have addressed the immediate consequences of herbivory on nectar cardenolide concentrations. If nectar cardenolides are inducible, examining the concentrations of individual compounds before and after damage could provide concrete evidence for the

101 consequence-of-defense hypothesis and may even illuminate the underlying mechanism for the presence of secondary metabolites in nectar.

Many plants must balance the costs of simultaneously attracting pollinators and deterring herbivores, and strategies that accommodate both factors may result in suboptimal reproduction or defense. Our chemical survey of nectar cardenolides and their relationship with leaf and flower cardenolides suggests that the presence of secondary metabolites in nectar is likely a consequence of a systemic chemical defense strategy. Determining the allocation and distribution of plant secondary metabolites, along with their subsequent effects on both antagonists and mutualists, requires a wider lens than most studies have employed.

Acknowledgements

We would like to thank A. Parachnowitsch, S. DeLeon, M. Stasny, S. Campbell, A.

Erwin and A. Hastings for their support during cardenolide collection and analysis.

Cardenolide analyses were conducted in the Cornell Chemical Ecology Core Facility, with support from Paul Feeny, New Life Sciences Initiative, College of Agriculture and

Life Sciences, Center for a Sustainable Future, Boyce Thompson Institute, and

Departments of Ecology & Evolutionary Biology, Neurobiology & Behavior,

Entomology, Plant Biology, and Horticulture. The chapter was vastly improved by C.

Parsons. This research was supported by NSERC of Canada and US NSF-DEB

0447550.

102

120

100

20 CardenolideConcentration (ng/µg)

0 ang bar bol can cur fas inc mex niv per pum tex

Asclepias species

Figure 5.1. Average total cardenolide concentration for nectar, leaf and flower samples of twelve species from the Asclepias series Incarnatae: Asclepias angustifolia, A. barjoniifolia, A. boliviensis, A. candida, A. curassavica , A. fascicularis, A. incarnata pulchra, A. mexicana, A. nivea , A. perennis , A. pumila and A. texana . Black bars represent nectar, light grey represent leaves and dark grey represent flowers, all with standard error bars. Flower samples were analyzed for the underlined species, while those not underlined represent species for which flower samples were not collected.

Nectar concentrations have been converted to mass equivalents for comparative purposes

(see methods). All other missing peaks represent a sample concentration of zero. For exact cardenolide concentrations, consult appendix two.

103

3.5 g)

µ 3.0

2.5

2.0

1.5

1.0

0.5 Nectar Cardenolide Concentration (ng/ Nectar Cardenolide Concentration

0.0

0 10 20 30 40 50 60 Leaf Cardenolide Concentration (ng/µg)

Figure 5.2. Correlation between leaf and nectar cardenolide concentrations. Each point represents paired concentration data for total cardenolide concentration of leaf and nectar samples from a single plant. Note that one extreme outlier, Asclepias pumila (nectar concentration: 110 ng/ µg, leaf concentration: 0.17 ng/ µg), has been omitted from the plot for visual effect but is included in the analyses.

104

MEX 3

inc

1 Dim2 BOL BAR mex NIV 0 cur niv bar per bol

PUM PER

pum can −1 tex

CAN

−1 0 1 2 3 Dim1

Figure 5.3. NMDS two-dimensional ordination of the individual carndenolide concentrations of Asclepias nectar and leaves. Nectar samples are denoted by upper case letters, while lower case denotes leaf samples (see Fig. 5.1 for species abbreviations).

Ellipses highlight nectar and leaves from the same species. Leaves and nectar with no detectable cardenolides could not be included in the ordination (see Methods for more information).

105

1.0

0.8

0.6

0.4

Proportion of Visits to Blue Flowers 0.2

10000 20000 30000 40000 50000 60000 Crithidia (cells/µL)

Figure 5.4. Correlation between intensity of Crithidia bombi infections in bees and their preference for blue flowers (n=28; outlier excluded in plot).

106

10 Average Visit Length (s) Average

0 10 50 Nectar Cardenolide Concentration (ng/µL)

Figure 5.5a. Boxplots of average length of ten consecutive flowers visits to flowers with no (0 ng/µL, n=6 bees), low (10 ng/µL, n=11 bees) or moderate (50 ng/µL, n=8 bees) cardenolide concentrations. The boxes represent the 25 th and 75 th percentiles, the whiskers are roughly two standard deviations and open circles represent outliers.

107

6 Foraging Rate (visits/min) Foraging 4

0 10 50 Nectar Cardenolide Concentration (ng/µL)

Figure 5.5b. Boxplots of visitation rate to flowers with no (0 ng/µL, n=6 bees), low (10 ng/µL, n=11 bees) or moderate (50 ng/µL, n=8 bees) cardenolide concentrations.

The rate was calculated using ten consecutive visits to each nectar type and therefore estimates visitation rate during a period of foraging constancy. Again, the boxes are the

25 th and 75 th percentiles, the whiskers represent approximately two standard deviations and the open circle represents an outlier.

CHAPTER SIX

Concluding Discussion

The study of nectar secondary metabolites has emerged as an important intersection between plant-herbivore and plant-pollinator studies; recognizing that floral traits are shaped by selection from both mutualists and antagonists, researchers have begun to integrate theory and methods from these historically separated fields. My dissertation contributes to this new area of study by examining the consequences of nectar secondary metabolites from a pollinator’s perspective. Specifically, I am broadening our understanding of how pollinators respond both behaviourally and physiologically to nectar alkaloids and nectar cardenolides. I also provide the first evidence to support an antimicrobial function for nectar secondary metabolites and the first survey of nectar cardenolides. In addition, my findings support the hypothesis that nectar secondary metabolites are a byproduct of a plant’s defenses against herbivores.

The relationships between plant defense, pollination and herbivory can be summarized using path diagrams to illustrate the direct and indirect effects of secondary metabolites on plant fitness. Secondary metabolites mitigate herbivore damage and therefore have a direct positive effect on plant fitness, but the direct effects of these defensive chemicals are context dependent when one considers how secondary metabolites affect pollinators (Fig 6.1a). The direction of these undefined arrows could be shaped by both the identity of the secondary compound and the type of pollinator (e.g. generalist or specialist, robber or legitimate pollinator); for example, results from chapters two and three suggest that nectar alkaloids have a negative effect on pollinators,

108 109 while chapter five found that nectar cardenolides have a positive effect on pollinators.

When the path diagram is expanded to include the role of pollinator pathogens on plant fitness (Fig 6.1b), the relationships depicted in the diagram become more defined. Based on results from chapter four, I can predict that secondary metabolites will have a direct positive effect on plant fitness mediated indirectly by their negative effect on pollinator pathogens. While the empirical data necessary to determine the strength of each effect has yet to be collected, these diagrams provide a framework for further evaluations of chemically-mediated plant-herbivore-pollinator interactions.

My research addresses the costs and benefits of nectar secondary metabolites for pollinators in a number of different ways. Here I synthesize my results and suggest their broader implications.

Ecological context is crucial

Secondary metabolites are known to be toxic, unpalatable and aversive, yet pollinators often forage on nectar containing these compounds. In chapter two of my dissertation, I show that although the nectar alkaloid gelsemine deters bees at even the lowest concentration tested, this response is mitigated when alkaloid-rich nectar has higher sucrose concentration than alkaloid-free control nectar. This result not only suggests a potential mechanism that plants may use to overcome the deterrent properties of nectar alkaloids, but it also highlights the importance of context for bumble bee foraging choices. Animals make foraging decisions by weighing their options and selecting food that provides the most reward at the least cost (Sih 1980). Nectar with secondary metabolites can fit this description if: a) it provides a better source of calories for foragers, b) it is larger in volume or c) other nectar options are scarce. Several studies

110 have found that “toxic” nectar collection by honey bees is associated with a reduction in alternative nectar availability (London-Shafir et al. 2003, Tan et al. 2007), and I have observed the same patterns with bumble bees in the lab. We might therefore expect that plants with nectar secondary metabolites may be under selection by pollinators to bloom either early or late in the season, times when other floral resources are limited.

Gelsemium sempervirens , which flowers from January to April (Pascarella 2007), meets this expectation; whether there is a broader pattern between plants with nectar secondary metabolites and flower phenology is currently unknown. Pollinators that choose to forage on “toxic” nectar do not make this decision blindly, and the effects of nectar secondary metabolites on pollinator preference cannot be interpreted without considering the ecological circumstances surrounding the animal’s decision.

A subtle effect is still an effect

Despite the dramatic reports of pollinator mortality due to nectar secondary metabolites (Baker and Baker 1975, Rhoades and Bergdahl 1981, Adler 2000), the post- consumptive effects of gelsemine on pollinators are less apparent but no less important.

In chapter three, I found that subordinate bees developed smaller oocytes after consuming nectar enriched with high concentrations of gelsemine. This sublethal effect of nectar alkaloids suggests that gelsemine can inhibit protein utilization and is potentially costly to reproduction. It is important to recognize that this effect was detected only when a preliminary study prompted me to measure the width of oocytes. Previous work on

Osmia lignaria revealed no detectable cost of consuming gelsemine at very high doses during several developmental stages (Elliott et al. 2008). Given that costs in my study were detectable only in bumble bees with limited access to food, I would suggest that

111 costs may have been masked by the generous provisions supplied to each O. lignaria by

Elliott et al. (2008) and that fitness effects could exist but are simply difficult to detect.

In contrast, the results of chapter four suggest that consuming nectar alkaloids could have a subtle positive effect on bumble bees. Continuous ingestion of gelsemine- enriched nectar significantly reduces the number of Crithidia bombi cells in foragers’ guts. C. bombi is not a lethal pathogen; instead, it impairs worker foraging efficiency

(Gegear et al. 2005, Otterstatter et al. 2005, Gegear et al. 2006), an effect that can subsequently reduce colony fitness. Because the behavioural costs of a C. bombi infection are correlated with infection intensity, we can infer that gelsemine consumption should mitigate these costs and subsequently improve overall colony fitness. Detecting improvements at the colony level is challenging due to natural variation in hive fitness but does not negate that these improvements occur. The long-term costs and benefits of nectar alkaloid consumption for pollinators are far from resolved, but future studies should not discount the importance of an effect based on magnitude alone.

There is no such thing as a general adaptive hypothesis

In her review paper, Adler (2000) hypothesized that an antimicrobial function would be the most general adaptive mechanism to explain nectar secondary metabolites.

Rather than being directed at a particular type of floral visitors, this hypothesis is based on the ubiquity of microbes in nature and the fact that nectar is an excellent medium for microbial growth. I first tested this hypothesis on floral nectar yeasts, which degrade nectar and reduce its palatability for pollinators (Herrera et al. 2008). In appendix one, I demonstrate that a number of floral yeasts from a variety of different plants were

112 unaffected when grown on gelsemine-enriched media, indicating that nectar alkaloids do not confer “anti-yeast” properties to floral nectar. In chapter four, I report that nectar alkaloids can have deleterious effects on the microbe Crithidia bombi , a bumble bee pathogen that is transferred from infected to naïve individuals at the flower (Durrer and

Schmid-Hempel 1994). This experiment did not find a direct effect of nectar alkaloids on

C. bombi , but rather suggested that ingesting gelsemine can indirectly reduce the number of C. bombi cells in the bees’ guts. These results call into question the generality of the antimicrobial hypothesis for two reasons: first, the fact that some microbes are affected while others aren’t indicates that the nectar alkaloid gelsemine is not universally antimicrobial; second, I expected to see microbial inhibition in floral yeasts, which are ubiquitous in flowers, but not inhibition of a pathogen that affects only bumble bees. In other words, the more general microbe, which is predicted to reduce pollination indiscriminately through nectar degradation, is immune to nectar alkaloids, while the microbe that is deleterious to only a single genus of pollinators is inhibited by the alkaloid. There is significant diversity in the characteristics of secondary metabolites found in nectar, and the impact of these compounds on biotic interactions with microbes and flower visitors is highly variable. It is probably simplistic to predict a general mechanism to explain the presence of secondary compounds in nectar.

Consequence of defense is often assumed but rarely tested

Nectar secondary metabolites may or may not have an adaptive function, but it is frequently suggested that they appear in nectar as a consequence of a plant’s need to chemically defend itself (Adler 2000, Strauss and Whittall 2006). There are at least two mechanisms that could explain nectar secondary metabolites as a byproduct of foliar

113 defense: first, secondary metabolites might leak into nectar during vascular transportation from the site of production to the site of defense; second, if production of secondary metabolites is controlled systemically but compounds are synthesized locally in the nectary, a systemic augmentation in defense could result in the production of nectar secondary metabolites. Both of these mechanisms would result in a correlation between secondary metabolite concentrations in nectar and in leaves, but characterizing the distribution of chemical defenses is difficult and therefore understudied. To my knowledge, only two previous studies have addressed this issue, both finding positive correlations between the total alkaloid concentrations of leaves and nectar in plants from across the genus Nicotiana (Adler et al. 2006; Adler unpublished). In chapter five, I report positive correlations between total cardenolide concentrations in the nectar and leaves of the genus Asclepias series Incarnatae, providing support for the consequence- of-defense hypothesis. The relationship between individual cardenolides in nectar and those in leaves is not nearly as clear. Most nectar cardenolides represent a subset of those found in leaves, which suggests vascular leakage; however, some of the cardenolides isolated in nectar did not match those found in leaves, suggesting that at least some cardenolides may be produced or regulated independently in nectar and leaves.

Evaluating the consequence-of-defense hypothesis is fundamental to the study of nectar secondary metabolites because it establishes the link between herbivore defenses and nectar toxicity. Future studies should compare chemical composition across plant components in a number of different species and focus on investigating the physical mechanism responsible for secondary metabolites in nectar.

114

Costs and benefits are not always obvious

My dissertation would have been substantially easier if the consumption of nectar secondary metabolites killed bumble bees. Unfortunately, detecting the costs or benefits of this paradoxical phenomenon is not that simple. In chapter three, detecting any sublethal cost of nectar alkaloids required measuring a very small trait in a certain subset of bees that could have easily been overlooked. The reduction in pathogen load found in chapter four would have been impossible to detect through the anticipated improvements of this reduction on pollinator foraging efficiency. The overall preference for nectar cardenolides of average concentration in chapter five was not immediately obvious because of the significantly heterogeneous behaviour of individual bees and was only detected after careful statistical analyses. Finally, even though bumble bees have a strong aversion to gelsemine-rich nectar in the lab, they readily collect nectar from Gelsemium sempervirens in the field, suggesting that even significant behavioural patterns can be obscured depending on the conditions of observation. Nectar secondary metabolites may not always be lethal, but they may nevertheless have ecologically important effects on pollinators which need to be taken into consideration when interpreting plant-pollinator interactions.

Although often described as a simple mutualism, the relationship between plants and their pollinators is anything but simple. In recognizing that this relationship is shaped by both mutualists and antagonists, my dissertation enriches study of pollination ecology and, more broadly, the study of plant-animal interactions as a whole.

115

Figure 6.1. Path diagrams of the relationships between secondary metabolites, herbivores, pollinators and plant fitness, both without (A) and with (B) the inclusion of pollinator pathogens. Solid lines represent positive effects while dashed lines represent negative effects and both solid and dashed suggest a context-dependent outcome. See text for explanation of models.

REFERENCES

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. 2006. Comparison of pollen transfer dynamics by multiple

floral visitors: experiments with pollen and fluorescent dye. Annals of Botany

97:141-150.

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

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

Afik, O., A. Dag, Z. Kerem, and S. Shafir. 2006. Analyses of avocado ( Persea

americana ) nectar properties and their perception by honey bees ( Apis mellifera ).

Journal of Chemical Ecology 32:1949-1963.

Agrawal, A. A., and M. Fishbein. 2006. Plant defense syndromes. Ecology 87:S132-

S149.

Agrawal, A. A., and M. Fishbein. 2008. Phylogenetic escalation and decline of plant

defense strategies. Proceedings of the National Academy of Sciences of the

United States of America 105:10057-10060.

Agrawal, A. A., M. J. Lajeunesse, and M. Fishbein. 2008. Evolution of latex and its

constituent defensive chemistry in milkweeds ( Asclepias ): a phylogenetic test of

plant defense escalation. Entomologia Experimentalis Et Applicata 128:126-138.

116 117

Agrawal, A. A., J. P. Salminen, and M. Fishbein. 2009. Phylogenetic trends in phenolic

metabolism of milkweeds ( Asclepias ): evidence for escalation. Evolution 63:663-

673.

Agrawal, A. A., S. Y. Strauss, and M. J. Stout. 1999. Costs of induced responses and

tolerance to herbivory in male and female fitness components of wild radish.

Evolution 53:1093-1104.

Antonovics, J. 2005. Plant venereal diseases: insights from a messy metaphor. New

Phytologist 165:71-80.

Armbruster, W. S. 1997. Exaptations link evolution of plant-herbivore and plant-

pollinator interactions: a phylogenetic inquiry. Ecology 78:1661-1672.

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.

Ayasse, M., T. Marlovits, J. Tengo, T. Taghizadeh, and W. Francke. 1995. Are there

pheromonal dominance signals in the bumblebee Bombus hypnorum L

(Hymenoptera, Apidae). Apidologie 26:163-180.

Baker, H. G. 1977. Non-sugar chemical constituents of nectar. Apidologie 8:349-356.

Baker, H. G. 1978. Chemical aspects of the pollination biology of woody plants in the

tropics. Pages 57-82 in P. B. Thomlinson and M. H. Zimmerman, editors.

Tropical trees as living systems. Cambridge University Press, Cambridge.

118

Baker, H. G., and I. Baker. 1973. Amino acids in nectar and their evolutionary

significance. Nature 241:543-545.

Baker, H. G., and I. Baker. 1975. Studies of nectar-constitution and pollinator-plant

coevolution. Pages 100-140 in L. E. Gilbert and P. H. Raven, editors. Coevolution

of animals and plants. University of Texas Press, Austin.

Baker, H. G., and I. Baker. 1982. Chemical constituents of nectar in relation to

pollination mechanisms and phylogeny. University of Chicago Press, Chicago.

Baker, H. G., and I. Baker. 1983. A brief historical review of the chemistry of floral

nectar. Pages 126-152 in B. Bentley and T. Elias, editors. The Biology of

Nectaries. Columbia University Press, New York.

Berenbaum, M. R. 1995. The chemistry of defense - theory and practice. Proceedings of

the National Academy of Sciences of the United States of America 92:2-8.

Berenbaum, M. R., and A. R. Zangerl. 1994. Costs of inducible defense - protein

limitation, growth, and detoxification in parsnip webworms. Ecology 75:2311-

2317.

Berenbaum, M. R., A. R. Zangerl, and J. K. Nitao. 1986. Constraints on chemical

coevolution - wild parsnips and the parsnip webworm. Evolution 40:1215-1228.

Bernays, E. A., and M. S. Singer. 2005. Taste alteration and endoparasites. Nature

436:476-476.

119

Bezemer, T. M., and N. M. van Dam. 2005. Linking aboveground and belowground

interactions via induced plant defenses. Trends in Ecology & Evolution 20:617-

624.

Blaw, M. E., M. A. Adkisson, D. Levin, J. C. Garriott, and R. S. A. Tindall. 1979.

Poisoning with Carolina jessamine ( Gelsemium sempervirens [L] Ait). Journal of

Pediatrics 94:998-1001.

Bloch, G., and A. Hefetz. 1999. Regulation of reproduction by dominant workers in

bumblebee ( Bombus terrestris ) queenright colonies. Behavioral Ecology and

Sociobiology 45:125-135.

Blumstein, D. T., and J. C. Daniel. 2007. Quantifying behaviour the JWatcher way.

Sinauer Associates, Inc., Sunderland.

Bluthgen, N., G. Gottsberger, and K. Fiedler. 2004. Sugar and amino acid composition of

ant-attended nectar and honeydew sources from an Australian rainforest. Austral

Ecology 29:418-429.

Boppre, M., S. M. Colegate, and J. A. Edgar. 2005. Pyrrolizidine alkaloids of Echium

vulgare honey found in pure pollen. Journal of Agricultural and Food Chemistry

53:594-600.

Brody, A. K. 1997. Effects of pollinators, herbivores, and seed predators on flowering

phenology. Ecology 78:1624-1631.

120

Bronstein, J. L., T. E. Huxman, and G. Davidowitz. 2007. Plant-mediated effects linking

herbivory and pollination. Pages 75-103 in T. Ohgushi, T. P. Craig, and P. W.

Price, editors. Ecological communities: plant mediation in indirect interaction

webs. Cambridge University Press, Cambridge.

Brower, L. P., W. N. Ryerson, L. Coppinger, and S. C. Glazier. 1968. Ecological

chemistry and palatability spectrum. Science 161:1349-1351.

Brown, M. J. F., R. Loosli, and P. Schmid-Hempel. 2000. Condition-dependent

expression of virulence in a trypanosome infecting bumblebees. Oikos 91:421-

427.

Brown, M. J. F., R. Schmid-Hempel, and P. Schmid-Hempel. 2003. Strong context-

dependent virulence in a host-parasite system: reconciling genetic evidence with

theory. Journal of Animal Ecology 72:994-1002.

Brysch-Herzberg, M. 2004. Ecology of yeasts in plant-bumblebee mutualism in Central

Europe. Fems Microbiology Ecology 50:87-100.

Buban, T., and Z. Orosz-Kovacs. 2003. The nectary as the primary site of infection by

Erwinia amylovora (Burr.) Winslow et al.: a mini review. Plant Systematics and

Evolution 238:183-194.

Burrow, G. E., and R. J. Tyrl. 2001. Toxic plants of North America, 1st edition. Iowa

State University Press, Ames.

121

Chapuisat, M., A. Oppliger, P. Magliano, and P. Christe. 2007. Wood ants use resin to

protect themselves against pathogens. Proceedings of the Royal Society B-

Biological Sciences 274:2013-2017.

Chittka, L., and J. D. Thomson. 2001. Cognitive ecology of pollination: animal behaviour

and floral evolution. Cambridge University Press, Cambridge.

Chittka, L., J. D. Thomson, and N. M. Waser. 1999. Flower constancy, insect

psychology, and plant evolution. Naturwissenschaften 86:361-377.

Christe, P., A. Oppliger, F. Bancala, G. Castella, and M. Chapuisat. 2003. Evidence for

collective medication in ants. Ecology Letters 6:19-22.

Cipollini, M. L., and D. J. Levey. 1997a. Secondary metabolites of fleshy vertebrate-

dispersed fruits: adaptive hypotheses and implications for seed dispersal.

American Naturalist 150:346-372.

Cipollini, M. L., and D. J. Levey. 1997b. Why are some fruits toxic? Glycoalkaloids in

Solanum and fruit choice by vertebrates. Ecology 78:782-798.

Clayton, D. H., and N. D. Wolfe. 1993. The adaptive significance of self-medication.

Trends in Ecology & Evolution 8:60-63.

Clinch, P. G., I. W. Forster, and Palmerjo.T. 1972. Effect on honey bees of nectar from

yellow kowhai ( Sophora microphylla Ait). New Zealand Journal of Agricultural

Research 15:194-&.

122

Cnaani, J., R. Schmid-Hempel, and J. O. Schmidt. 2002. Colony development, larval

development and worker reproduction in Bombus impatiens Cresson. Insectes

Sociaux 49:164-170.

Cnaani, J., A. Wong, and J. D. Thomson. 2007. Effect of group size on ovarian

development in bumblebee workers (Hymenoptera : apidae : Bombus ).

Entomologia Generalis 29:305-314.

Colla, S. R., M. C. Otterstatter, R. J. Gegear, and J. D. Thomson. 2006. Plight of the

bumble bee: pathogen spillover from commercial to wild populations. Biological

Conservation 129:461-467.

Cory, J. S., and K. Hoover. 2006. Plant-mediated effects in insect-pathogen interactions.

Trends in Ecology & Evolution 21:278-286.

Cowan, M. M. 1999. Plant products as antimicrobial agents. Clinical Microbiology

Reviews 12:564-582.

Cresswell, J. E., S. Z. Merritt, and M. M. Martin. 1992. The effect of dietary nicotine on

the allocation of assimilated food to energy metabolism and growth in fourth

instar larvae of the southern armyworm, Spodoptera eridania (Lepidoptera,

Noctuidae). Oecologia 89:449-453. de Roode, J. C., A. B. Pedersen, M. D. Hunter, and S. Altizer. 2008. Host plant species

affects virulence in monarch butterfly parasites. Journal of Animal Ecology

77:120-126.

123

Despres, L., J. P. David, and C. Gallet. 2007. The evolutionary ecology of insect

resistance to plant chemicals. Trends in Ecology & Evolution 22:298-307.

Dethier, V. G. 1982. Mechanism of host-plant recognition. Entomologia Experimentalis

Et Applicata 31:49-56.

Dethier, V. G., and E. Bowdan. 1992. Effects of alkaloids on feeding by Phormia regina

confirm the critical role of sensory inhibition. Physiological Entomology 17:325-

330.

Detzel, A., and M. Wink. 1993. Attraction, deterrence or intoxication of bees ( Apis

mellifera ) by plant allelochemicals. Chemoecology 4:8-18.

Duchateau, M. J., and H. H. W. Velthuis. 1989. Ovarian development and egg-laying in

workers of Bombus terrestris . Entomologia Experimentalis Et Applicata 51:199-

213.

Duffey, S. S., and M. J. Stout. 1996. Antinutritive and toxic components of plant defense

against insects. Archives of Insect Biochemistry and Physiology 32:3-37.

Dukas, R., and L. A. Real. 1991. Learning foraging tasks by bees - a comparison between

social and solitary species. Animal Behaviour 42:269-276.

Durrer, S., and P. Schmid-Hempel. 1994. Shared use of flowers leads to horizontal

pathogen transmission. Proceedings of the Royal Society of London Series B-

Biological Sciences 258:299-302.

124

Dussourd, D. E., and A. M. Hoyle. 2000. Poisoned plusiines: toxicity of milkweed latex

and cardenolides to some generalist caterpillars. Chemoecology 10:11-16.

Ehlers, B. K., and J. M. Olesen. 1997. The fruit-wasp route to toxic nectar in Epipactis

orchids? Flora 192:223-229.

Elliott, S. E., R. E. Irwin, L. S. Adler, and N. M. Williams. 2008. The nectar alkaloid,

gelsemine, does not affect offspring performance of a native solitary bee, Osmia

lignaria (Megachilidae). Ecological Entomology 33:298-304.

Feeny, P. 1970. Seasonal changes in oak leaf tannins and nutrients as a cause of spring

feeding by winter moth caterpillars. Ecology 51:565-581.

Feeny, P. 1992. The evolution of chemical ecology: contributions from the study of

herbivorous insects, Vol. II: Ecological and evolutionary processes. Pages 1-44 in

G. A. Rosenthal and M. R. Berenbaum, editors. Herbivores: their interactions with

plant secondary metabolites. Academic Press, Inc., San Diego.

Feeny, P. P. 1976. Plant apparency and chemical defense. Pages 1-40 in J. W. Wallace

and R. L. Mansell, editors. Biochemical interactions between plants and insects.

Plenum, New York.

Fineblum, W. L., and M. D. Rausher. 1997. Do floral pigmentation genes also influence

resistance to enemies? The W locus in Ipomoea purpurea . Ecology 78:1646-1654.

125

Fishbein, M., D. Chuba, C. Ellison, R. Mason-Gamer, and S. P. Lynch. In Press.

Phylogenetic relationships of Asclepias (Apocynaceae) estimated from non-

coding cpDNA sequences. Systematic Botany.

Fishbein, M., and D. L. Venable. 1996. Diversity and temporal change in the effective

pollinators of Asclepias tuberosa . Ecology 77:1061-1073.

Fordyce, J. A., and S. B. Malcolm. 2000. Specialist weevil, Rhyssomatus lineaticollis ,

does not spatially avoid cardenolide defenses of common milkweed by

ovipositing into pith tissue. Journal of Chemical Ecology 26:2857-2874.

Freiburghaus, F., R. Kaminsky, M. H. H. Nkunya, and R. Brun. 1996. Evaluation of

African medicinal plants for their in vitro trypanocidal activity. Journal of

Ethnopharmacology 55:1-11.

Fung, H. T., K. K. Lam, S. K. Lam, O. F. Wong, and S. K. Kam. 2007. Two cases of

Gelsemium elegans Benth. poisoning. Hong Kong Journal of Emergency

Medicine 14:221-224.

Gegear, R. J., and T. M. Laverty. 2004. Effect of a colour dimorphism on the flower

constancy of honey bees and bumble bees. Canadian Journal of Zoology-Revue

Canadienne De Zoologie 82:587-593.

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.

126

Gegear, R. J., M. C. Otterstatter, and J. D. Thomson. 2005. Does parasitic infection

impair the ability of bumblebees to learn flower-handling techniques? Animal

Behaviour 70:209-215.

Gegear, R. J., M. C. Otterstatter, and J. D. Thomson. 2006. Bumble-bee foragers infected

by a gut parasite have an impaired ability to utilize floral information.

Proceedings of the Royal Society B-Biological Sciences 273:1073-1078.

Gimenez-Jurado, G., C. P. Kurtzman, W. T. Starmer, and I. Spencer-Martins. 2003.

Metschnikowia vanudenii sp nov and Metschnikowia lachancei sp nov., from

flowers and associated insects in North America. International Journal of

Systematic and Evolutionary Microbiology 53:1665-1670.

Glendinning, J. I. 2002. How do herbivorous insects cope with noxious secondary plant

compounds in their diet? Entomologia Experimentalis Et Applicata 104:15-25.

Glendinning, J. I., N. M. Nelson, and E. A. Bernays. 2000. How do inositol and glucose

modulate feeding in Manduca sexta caterpillars? Journal of Experimental Biology

203:1299-1315.

Golonka, A. M. 2002. Nectar-inhabiting microorganisms and the dioecious plant species

Silene latifolia . PhD Thesis. Duke University, Durham.

Gomez, J. M. 2003. Herbivory reduces the strength of pollinator-mediated selection in

the Mediterranean herb Erysimum mediohispanicum : consequences for plant

specialization. American Naturalist 162:242-256.

127

Gumbert, A. 2000. Color choices by bumble bees ( Bombus terrestris ): innate preferences

and generalization after learning. Behavioral Ecology and Sociobiology 48:36-43.

Hagler, J. R., and S. L. Buchmann. 1993. Honey bee (Hymenoptera, Apidae) foraging

responses to phenolic-rich nectars. Journal of the Kansas Entomological Society

66:223-230.

Harder, L. D. 1982. Measurement and estimation of functional proboscis length in

bumblebees (Hymenoptera, Apidae). Canadian Journal of Zoology-Revue

Canadienne De Zoologie 60:1073-1079.

Harder, L. D., and S. C. H. Barrett. 1993. Pollen removal from tristylous Pontederia

cordata - effects of anther position and pollinator specialization. Ecology

74:1059-1072.

Hardin, J. W., and J. M. Arena. 1969. Human poisoning from native and cultivated

plants. Duke University Press, Durham.

Hartmann, T. 1992. Alkaloids. Pages 79-121 in G. A. Rosenthal and M. R. Berebaum,

editors. Herbivores: their interactions with secondary plant metabolites, Vol. I:

The chemical participants. Academic Press, Inc., San Diego.

Heinrich, B. 1975. Bee flowers - hypothesis on flower variety and blooming times.

Evolution 29:325-334.

Heinrich, B. 1979. Bumblebee economics. Harvard University Press, Cambridge.

128

Heinrich, B., P. R. Mudge, and P. G. Deringis. 1977. Laboratory analysis of flower

constancy in foraging bumblebees - Bombus ternarius and Bombus terricola .

Behavioral Ecology and Sociobiology 2:247-265.

Herrera, C. M., I. M. Garcia, and R. Perez. 2008. Invisible floral larcenies: microbial

communities degrade floral nectar of bumble bee-pollinated plants. Ecology

89:2369-2376.

Herrera, C. M., M. Medrano, P. J. Rey, A. M. Sanchez-Lafuente, M. B. Garcia, J.

Guitian, and A. J. Manzaneda. 2002. Interaction of pollinators and herbivores on

plant fitness suggests a pathway for correlated evolution of mutualism- and

antagonism-related traits. Proceedings of the National Academy of Sciences of the

United States of America 99:16823-16828.

Imhoof, B., and P. Schmid-Hempel. 1999. Colony success of the bumble bee, Bombus

terrestris , in relation to infections by two protozoan parasites, Crithidia bombi

and Nosema bombi . Insectes Sociaux 46:233-238.

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. K. Brody. 2004. The dual role of floral traits: pollinator

attraction and plant defense. Ecology 85:1503-1511.

129

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.

Janzen, D. H. 1973. Community structure of secondary compounds in plants. Pure and

Applied Chemistry 34:529-538.

Janzen, D. H. 1977. Why don't ants visit flowers? Biotropica 9:252-252.

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.

Karban, R., and I. T. Baldwin. 1997. Induced responses to herbivory. University of

Chigaco Press, Chigaco.

Karban, R., and G. English-Loeb. 1997. Tachinid parasitoids affect host plant choice by

caterpillars to increase caterpillar survival. Ecology 78:603-611.

Karowe, D. N. 1989. Differential effect of tannic acid on 2 tree-feeding lepidoptera -

implications for theories of plant anti-herbivore chemistry. Oecologia 80:507-512.

Kearns, C. A., and J. D. Thomson. 2001. The natural history of bumblebees: a

sourcebook for investigations. University Press of Colorado, Boulder.

Kephart, S., and K. Theiss. 2004. Pollinator-mediated isolation in sympatric milkweeds

(Asclepias ): do floral morphology and insect behavior influence species

boundaries? New Phytologist 161:265-277.

130

Kevan, P. G., D. Eisikowitch, and B. Rathwell. 1989. The role of nectar in the

germination of pollen in Asclepias syriaca -L. Botanical Gazette 150:266-270.

Kim, Y. S., and B. H. Smith. 2000. Effect of an amino acid on feeding preferences and

learning behavior in the honey bee, Apis mellifera . Journal of Insect Physiology

46:793-801.

Konig, B. 1988. The honeybee as pharmacophorus insect. Entomologia Generalis 14:145-

148.

Kotler, B. P., and L. Blaustein. 1995. Titrating food and safety in a heterogeneous

environment: when are the risky and safe patches of equal value? Oikos 74:251-

258.

Kurtzman, C. P., and C. J. Robnett. 1998. Identification and phylogeny of ascomycetous

yeasts from analysis of nuclear large subunit (26S) ribosomal DNA partial

sequences. Antonie Van Leeuwenhoek International Journal of General and

Molecular Microbiology 73:331-371.

Lachance, M. A. 1987. Approaches to yeast identification. Pages 33-51 in D. R. Berry, I.

Russell, and G. G. Stewart, editors. Yeast biotechnology. Allen & Unwin,

London.

Lachance, M. A., W. T. Starmer, C. A. Rosa, J. M. Bowles, J. S. F. Barker, and D. H.

Janzen. 2001. Biogeography of the yeasts of ephemeral flowers and their insects.

FEMs Yeast Research 1:1-8.

131

Landolt, P. J., and B. Lenczewski. 1993. Lack of evidence for the toxic nectar hypothesis

- a plant alkaloid did not deter nectar feeding by lepidoptera. Florida Entomologist

76:556-566.

Lee, K. P., J. S. Cory, K. Wilson, D. Raubenheimer, and S. J. Simpson. 2006. Flexible

diet choice offsets protein costs of pathogen resistance in a caterpillar.

Proceedings of the Royal Society B-Biological Sciences 273:823-829.

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.

Lin, H. R., and M. L. Winston. 1998. The role of nutrition and temperature in the ovarian

development of the worker honey bee ( Apis mellifera ). Canadian Entomologist

130:883-891.

Lipa, J. J., and O. Triggiani. 1988. Crithidia bombi sp n. a flagellated parasite of a

bumblebee Bombus terrestris L. (Hymenoptera, Apidae). Acta Protozoologica

27:287-290.

Liu, F., J. Chen, J. Chai, X. Zhang, X. Bai, D. He, and D. W. Roubik. 2007. Adaptive

functions of defensive plant phenolics and a non-linear bee response to nectar

components. Functional Ecology 21:96-100.

Liu, F. L., J. Z. He, and W. J. Fu. 2005. Highly controlled nest homeostasis of honey bees

helps deactivate phenolics in nectar. Naturwissenschaften 92:297-299.

132

Logan, A., M. X. Ruiz-Gonzalez, and M. J. F. Brown. 2005. The impact of host

starvation on parasite development and population dynamics in an intestinal

trypanosome parasite of bumble bees. Parasitology 130:637-642.

London-Shafir, I., S. Shafir, and D. Eisikowitch. 2003. Amygdalin in almond nectar and

pollen - facts and possible roles. Plant Systematics and Evolution 238:87-95.

Malcolm, S. B. 1991. Cardenolide-mediated interaction between plants and herbivores.

Pages 252-296 in G. A. Rosenthal and M. Berebaum, editors. Herbivores: their

interactions with secondary plant metabolites, Vol. II: Ecological and

evolutionary processes. Academic Press, San Diego.

Manson, J. S., M. A. Lachance, and J. D. Thomson. 2007. Candida gelsemii sp. nov., a

yeast of the Metschnikowiaceae clade isolated from nectar of the poisonous

Carolina jessamine. Antonie Van Leeuwenhoek International Journal of General

and Molecular Microbiology 92:37-42.

Manson, J. S., M. C. Otterstatter, and J. D. Thomson. In Press. Consumption of a nectar

alkaloid reduceds pathogen load in bumble bees. Oecologia.

Manson, J. S., and J. D. Thomson. 2009. Post-ingestive effects of nectar alkaloids depend

on dominance status of bumble bees. Ecological Entomology 34:421-426.

Marcucci, M. C. 1995. Propolis - chemical composition, biological properties and

therapeutic activity. Apidologie 26:83-99.

133

Masters, A. R. 1991. Dual role of pyrrolizidine alkaloids in nectar. Journal of Chemical

Ecology 17:195-205.

McCall, A. C. 2006. Natural and artificial floral damage induces resistance in Nemophila

menziesii (Hydrophyllaceae) flowers. Oikos 112:660-666.

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.

McKey, D. 1974. Adaptive patterns in alkaloid physiology. American Naturalist 108:305-

320.

Miller, M. W., and H. J. Phaff. 1998. Metschnikowia Kamienski. Pages 256-267 in C. P.

Kurtzman and J. W. Fell, editors. The yeasts, a taxonomic study. Elsevier,

Amsterdam.

Mitchell, B. K. 1987. Interactions of alkaloids with galeal chemosensory cells of

Colorado potato beetle. Journal of Chemical Ecology 13:2009-2022.

Mitchell, B. K., and J. F. Sutcliffe. 1984. Sensory inhibition as a mechanism of feeding

deterrence - effects of 3 alkaloids on leaf beetle feeding. Physiological

Entomology 9:57-64.

Morse, D. H. 1982. The turnover of milkweed pollinia on bumble bees, and implications

for outcrossing. Oecologia 53:187-196.

134

Mosquin, T. 1971. Competition for pollinators as a stimulus for evolution of flowering

time. Oikos 22:398-402.

Neal, J. J. 1987. Metabolic costs of mixed-function oxidase induction in Heliothis zea .

Entomologia Experimentalis Et Applicata 43:175-179.

Nelson, C. J., J. N. Seiber, and L. P. Brower. 1981. Seasonal and intraplant variation of

cardenolide content in the California milkweed, Asclepias eriocarpa , and

implications for plant defense. Journal of Chemical Ecology 7:981-1010.

Oksanen, J. 2009. vegan: R functions for vegetation ecologists. URL:

http://cc.oulu.fi/~jarioksa/softhelp/vegan.html.

Ornduff, R. 1970. Systematics and breeding system of Gelsemium (Loganiaceae). Journal

of the Arnold Arboretum 51:1-17.

Osier, T. L., S. Y. Hwang, and R. L. Lindroth. 2000. Effects of phytochemical variation

in quaking aspen Populus tremuloides clones on gypsy moth Lymantria dispar

performance in the field and laboratory. Ecological Entomology 25:197-207.

Ott, J. 1998. The Delphic Bee: bees and toxic honeys as pointers to psychoactive and

other medicinal plants. Economic Botany 52:260-266.

Otterstatter, M. C., R. J. Gegear, S. R. Colla, and J. D. Thomson. 2005. Effects of

parasitic mites and protozoa on the flower constancy and foraging rate of bumble

bees. Behavioral Ecology and Sociobiology 58:383-389.

135

Otterstatter, M. C., and J. D. Thomson. 2006. Within-host dynamics of an intestinal

pathogen of bumble bees. Parasitology 133:749-761.

Otterstatter, M. C., and J. D. Thomson. 2007. Contact networks and transmission of an

intestinal pathogen in bumble bee ( Bombus impatiens ) colonies. Oecologia

154:411-421.

Paradis, E. 2009. ape: analyses of phylogenetics and evolution. URL:

http://ape.mpl.ird.fr/.

Pascarella, J. B. 2007. Mechanisms of prezygotic reproductive isolation between two

sympatric species, Gelsemium rankinii and G. sempervirens (Gelsemiaceae), in

the southeastern United States. American Journal of Botany 94:468-476.

Pernal, S. F., and R. W. Currie. 2000. Pollen quality of fresh and 1-year-old single pollen

diets for worker honey bees ( Apis mellifera L.). Apidologie 31:387-409.

Pilson, D. 2000. Herbivory and natural selection on flowering phenology in wild

sunflower, Helianthus annus . Oecologia 122:72-82.

Praz, C. J., A. Mueller, and S. Dorn. 2008. Specialized bees fail to develop on non-host

pollen: do plants chemically protect their pollen? Ecology 89:795-804.

Price, P. W., C. E. Bouton, P. Gross, B. A. McPheron, J. N. Thompson, and A. E. Weis.

1980. Interactions among 3 trophic levels - influence of plants on interactions

between insect herbivores and natural enemies. Annual Review of Ecology and

Systematics 11:41-65.

136

Proctor, M., P. Yeo, and A. Lack. 1996. The natural history of pollination. Timber Press,

Portland.

Pryce-Jones, J. 1942. Some problems associated with nectar, pollen and honey.

Proceedings of the Linnean Society of London 155:129-174.

R Development Core Team. 2009. R: a language and environment for statistical

computing. Foundation for Statistical Computing, Vienna; URL: http://ww.R-

project.org.

Radford, A. E., H. E. Ahles, and C. R. Bell. 1968. Manual of the vascular flora of the

Carolinas. University of North Carolina Press, Chapel Hill.

Raguso, R. A. 2008. Wake up and smell the roses: the ecology and evolution of floral

scent. Annual Review of Ecology Evolution and Systematics 39:549-569.

Rasmann, S., A. A. Agrawal, S. C. Cook, and A. C. Erwin. In Press. Cardenolides,

induced responses, and interactions between above and belowground herbivores

in the milkweeds ( Asclepias spp.). Ecology.

Real, L. A. 1991. Animal choice behavior and the evolution of cognitive architecture.

Science 253:980-986.

Rhoades, D. F. 1979. Evolution of plant chemical defense against herbivores. in G. A.

Rosenthal and D. H. Janzen, editors. Herbivores: their interaction with secondary

plant metabolites. Academic Press, San Diego.

137

Rhoades, D. F., and J. C. Bergdahl. 1981. Adaptive significance of toxic nectar.

American Naturalist 117:798-803.

Rosa, C. A., M. A. Lachance, J. O. C. Silva, A. C. P. Teixeira, M. M. Marini, Y.

Antonini, and R. P. Martins. 2003. Yeast communities associated with stingless

bees. FEMs Yeast Research 4:271-275.

Rosenthal, G. A., and M. R. Berenbaum. 1991. Herbivores: Their interactions with

secondary plant metabolites, Vol. I: The chemical participants, 2nd edition.

Academic Press, San Diego.

Rujjanawate, C., D. Kanjanapothi, and A. Panthong. 2003. Pharmacological effect and

toxicity of alkaloids from Gelsemium elegans Benth. Journal of

Ethnopharmacology 89:91-95.

SAS Institute. 1999. SAS user's guide. SAS Institute, Cary.

SAS Institute. 2006. SAS/STAT 9.1 user's guide. SAS Institute, Cary.

Schmid-Hempel, P. 1998. Parasites in social insects. Princeton University Press,

Princeton.

Schmid-Hempel, P. 2001. On the evolutionary ecology of host-parasite interactions:

addressing the question with regard to bumblebees and their parasites.

Naturwissenschaften 88:147-158.

138

Schmid-Hempel, P., and R. Schmid-Hempel. 1993. Transmission of a pathogen in

Bombus terrestris , with a note on division-of-labor in social insects. Behavioral

Ecology and Sociobiology 33:319-327.

Schmid-Hempel, P., and H. P. Stauffer. 1998. Parasites and flower choice of bumblebees.

Animal Behaviour 55:819-825.

Schmid-Hempel, R., and P. Schmid-Hempel. 1991. Endoparasitic flies, pollen-collection

by bumblebees and a potential host-parasite conflict. Oecologia 87:227-232.

Shields, V. D. C., and B. K. Mitchell. 1995. The effect of phagostimulant mixtures on

deterrent receptor(s) in 2 crucifer-feeding lepidopterous species. Philosophical

Transactions of the Royal Society of London Series B-Biological Sciences

347:459-464.

Shykoff, J. A., and P. Schmid-Hempel. 1991. Genetic relatedness and eusociality -

parasite-mediated selection on the genetic composition of groups. Behavioral

Ecology and Sociobiology 28:371-376.

Siegert, K. J. 1987. Carbohydrate metabolism in Manduca sexta during late larval

development. Journal of Insect Physiology 33:421-427.

Sih, A. 1980. Optimal behavior - can foragers balance 2 conflicting demands. Science

210:1041-1043.

Simms, E. L., and M. A. Bucher. 1996. Pleiotropic effects of flower-color intensity on

herbivore performance on Ipomoea purpurea . Evolution 50:957-963.

139

Singaravelan, N., G. Nee'man, M. Inbar, and I. Izhaki. 2005. Feeding responses of free-

flying honeybees to secondary compounds mimicking floral nectars. Journal of

Chemical Ecology 31:2791-2804.

Singer, M. S., Y. Carriere, C. Theuring, and T. Hartmann. 2004. Disentangling food

quality from resistance against parasitoids: diet choice by a generalist caterpillar.

American Naturalist 164:423-429.

Singer, M. S., K. C. Mace, and E. A. Bernays. 2009. Self-medication as adaptive

plasticity: increased ingestion of plant toxins by parasitized caterpillars. PLoS

ONE 4:e4796.

Slansky, F. 1992. Allelochemical-nutrient interactions in herbivore nutrient ecology, Vol.

II: Ecological and evolutionary processes. Pages 135-176 in G. A. Rosenthal and

M. R. Berenbaum, editors. Herbivores: their interactions with secondary plant

metabolites. Academic Press, San Diego.

Sokal, R. R., and F. J. Rohlf. 1995. Biometry, 3rd edition. W.H. Freeman and Company,

New York.

Stephenson, A. G. 1981. Toxic nectar deters nectar thieves of Catalpa speciosa .

American Midland Naturalist 105:381-383.

Stephenson, A. G. 1982. Iridoid glycosides in the nectar of Catalpa speciosa are

unpalatable to nectar thieves. Journal of Chemical Ecology 8:1025-1034.

140

Stiles, B., and J. D. Paschke. 1980. Midgut pH in different instars of 3 Aedes mosquito

species and the relation between pH and susceptibility of larvae to a nuclear

polyhedrosis virus. Journal of Invertebrate Pathology 35:58-64.

Strauss, S. Y. 1997. Floral characters link herbivores, pollinators, and plant fitness.

Ecology 78:1640-1645.

Strauss, S. Y., J. K. Conner, and S. L. Rush. 1996. Foliar herbivory affects floral

characters and plant attractiveness to pollinators: implications for male and female

plant fitness. American Naturalist 147:1098-1107.

Strauss, S. Y., J. A. Rudgers, J. A. Lau, and R. E. Irwin. 2002. Direct and ecological costs

of resistance to herbivory. Trends in Ecology & Evolution 17:278-285.

Strauss, S. Y., D. H. Siemens, M. B. Decher, and T. Mitchell-Olds. 1999. Ecological

costs of plant resistance to herbivores in the currency of pollination. Evolution

53:1105-1113.

Strauss, S. Y., and J. B. Whittall. 2006. Non-pollinator agents of selection on floral traits.

Pages 120-138 in L. D. Harder and S. C. H. Barrett, editors. Ecology and

evolution of flowers. Oxford University Press, Oxford.

Tadmor-Melamed, H., S. Markman, A. Arieli, M. Distl, M. Wink, and I. Izhaki. 2004.

Limited ability of Palestine sunbirds Nectarinia osea to cope with pyridine

alkaloids in nectar of tree tobacco Nicotiana glauca . Functional Ecology 18:844-

850.

141

Tan, K., Y. H. Guo, S. W. Nicolson, S. E. Radloff, Q. S. Song, and H. R. Hepburn. 2007.

Honeybee ( Apis cerana ) foraging responses to the toxic honey of Tripterygium

hypoglaucum (Celastraceae): changing threshold of nectar acceptability. Journal

of Chemical Ecology 33:2209-2217.

Thompson, S. N. 2003. Trehalose - The insect 'blood' sugar. Pages 205-285 in Advances

in Insect Physiology, Vol 31. Academic Press, San Diego.

Thomson, J. D. 1986. Pollen transport and deposition by bumble bees in Erythronium -

influences of floral nectar and bee grooming. Journal of Ecology 74:329-341.

Thornburg, R. W., C. Carter, A. Powell, R. Mittler, L. Rizhsky, and H. T. Horner. 2003.

A major function of the tobacco floral nectary is defense against microbial attack.

Plant Systematics and Evolution 238:211-218.

Van der Putten, W. H., L. E. M. Vet, J. A. Harvey, and F. L. Wackers. 2001. Linking

above- and belowground multitrophic interactions of plants, herbivores,

pathogens, and their antagonists. Trends in Ecology & Evolution 16:547-554.

Vanetten, H., E. Temporini, and C. Wasmann. 2001. Phytoalexin (and phytoanticipin)

tolerance as a virulence trait: why is it not required by all pathogens?

Physiological and Molecular Plant Pathology 59:83-93.

Wahl, O., and K. Ulm. 1983. Influence of pollen feeding and physiological condition on

pesticide sensitivity of the honey bee Apis mellifera Carnica. Oecologia 59:106-

128.

142

Webster, S. J., and L. M. Dill. 2006. The energetic equivalence of changing salinity and

temperature to juvenile salmon. Functional Ecology 20:621-629.

Whittaker, R. H., and P. P. Feeny. 1971. Allelochemics - chemical interactions between

species. Science 171:757-770.

Wink, M., and O. Schimmer. 1999. Modes of action of defensive secondary metabolites.

Pages 17-133 in M. Wink, editor. Functions of plant secondary metabolites and

their exploitation in biotechnology, Annual Plant Reviews Vol. III. Wiley-

Blackwell.

Wink, M., and V. Theile. 2002. Alkaloid tolerance in Manduca sexta and

phylogenetically related sphingids (Lepidoptera : Sphingidae). Chemoecology

12:29-46.

Woodson, R. E. 1954. The North American species of Asclepias L. Annals of the

Missouri Botanical Garden 41:1-211.

Wyatt, R., and S. B. Broyles. 1994. Ecology and evolution of reproduction in milkweeds.

Annual Review of Ecology and Systematics 25:423-441.

Wyatt, R., S. B. Broyles, J. L. Hamrick, and A. Stoneburner. 1993. Systematic

relationships within Gelsemium (Loganiaceae) - evidence from isozymes and

cladistics. Systematic Botany 18:345-355.

143

Yarrow, D. 1998. Methods for the isolation and identification of yeasts. Pages 77-100 in

C. P. Kurtzman and J. W. Fell, editors. The yeasts, a taxonomic study. Elsevier,

Amsterdam.

Zangerl, A. R., and M. R. Berenbaum. 1993. Plant chemistry, insect adaptations to plant

chemistry, and host plant utilization patterns. Ecology 74:47-54.

Zehnder, C. B., and M. D. Hunter. 2007. Interspecific variation within the genus

Asclepias in response to herbivory by a phloem-feeding insect herbivore. Journal

of Chemical Ecology 33:2044-2053.

Zhang, Z., S. Schwartz, L. Wagner, and W. Miller. 2000. A greedy algorithm for aligning

DNA sequences. Journal of Computational Biology 7:203-214.

APPENDIX ONE

Candida gelsemii sp. nov., a yeast of the Metshnikowiaceae clade isolated from

nectar of the poisonous Carolina jessamine

Jessamyn S. Manson, Marc-André Lachance and James D. Thomson

This project was conceived by me with guidance from both M.A. Lachance and J.D. Thomson. The work was done in collaboration with M.A. Lachance, who also co-wrote the manuscript, now published in

Antonie van Leeuwenhoeck, 2007, 92: 37-42.

Abstract

A new yeast species, Candida gelsemii , is described to accommodate three isolates recovered in Georgia, USA, from the toxic nectar of the Carolina jessamine

(Gelsemium sempervirens ). The species resembles other members of the

Metschnikowiaceae clade that have been recovered from nectar, but differs in a number of morphological and physiological characteristics. Analysis of rDNA sequences places the new species well into the clade, but in a basal position with respect to a group of

Metschnikowia and Candida species known to occur in association with nectars, bees, as well as marine invertebrates. The type is strain UWOPS 06-24.1 T (CBS 10509 T,

NRRL Y-48212 T).

144 145

Introduction

Floral nectars often contain yeasts that are vectored by pollinating and non- pollinating insects (Lachance et al. 2001, Rosa et al. 2003, Brysch-Herzberg 2004). The role of the yeasts in this ecosystem is poorly understood, but it is clear that the composition of the yeast community is highly dependent on the types of insects involved.

Whereas nitidulid beetles carry yeasts with affinities in the genera Metschnikowia ,

Kodamaea , and Wickerhamiella , bees tend to vector other yeasts related to the genera

Metschnikowia and Starmerella . Even within the Metschnikowiaceae clade, the species associated with nitidulids and those associated with bees are not the same. It was therefore of interest to examine the nectar of the Carolina jessamine ( Gelsemium sempervirens ). This perennial vine is endemic to the southeastern United States. It is also a distylous species, making it an obligate outcrosser. It blooms in the early spring, producing a multitude of fragrant tubular yellow flowers which attract a diverse range of pollinators including eusocial, solitary, and nectar-robbing bees (Ornduff 1970, Adler and

Irwin 2005, 2006). Most importantly, the nectar of this plant contains gelsemine, an alkaloid that is highly toxic to vertebrates (Burrow and Tyrl 2001). The alkaloid, presumed to be a deterrent for herbivores, is reported to deter pollinators and nectar robbers at natural concentrations (Adler and Irwin 2005). Adaptive hypotheses for the presence of alkaloids in floral nectar include providing the nectar with antimicrobial properties to prevent the proliferation of organisms such as floral yeasts (Adler 2000).

In the course of determining whether the yeast community of the nectar is also affected by gelsemine, we sampled nectars from G. sempervirens and also some

146 sympatric azaleas that appeared to have a similar array of bees foraging for nectar. At the time of collecting, both plant species were visited by queens of the native bumble bees

Bombus impatiens and B. bimaculatus , as well as introduced honey bees ( Apis mellifera ).

In the process, strains of Metschnikowia reukaufii were recovered from azalea nectar, whereas the jessamine nectars yielded strains of Candida rancoensis as well as a new asexual species with metschnikowiaceous affinities, which we now describe as Candida gelsemii .

Methods

Nectar samples were collected on March 30 th and 31 st 2006 in four sites located near the campus of Georgia Southern University in Statesboro, Georgia. Other isolation details are given in Table A.1. In each case, approximately 2 µL of nectar was placed onto a plate of YM agar supplemented with 100 mg/L chloramphenicol and the nectar was streak-diluted with a sterile loop. Mould colonies were removed periodically with a knife. The plates were returned to the laboratory (UWO) and colonies were picked for purification and identification by rDNA partial sequencing (Kurtzman and Robnett 1998).

Sequence editing, alignment, and analysis were conducted with DNAMAN version 4.1.

The sequences were queried against the GenBank database, using the Megablast algorithm of Zhang et al. (2000). Strain characterization followed standard methods

(Yarrow 1998). Growth responses were determined by replica plating as detailed by

Lachance (1987). Replica plating was also used to evaluate the effect of gelsemine on yeast growth. The potential effect of gelsemine on yeasts was also assesses by agar diffusion. An ethanol solution of gelsemine was added to sterile discs of Whatman 3mm paper so that each disk contained 35, 3.5, and 0.35 µg, respectively. The air-dried discs

147 were then placed individually on the surface of YM agar plates inoculated with dilute yeast suspensions. The plates were examined periodically for evidence of inhibition zones.

Results and Discussion

Species delineation, phylogenetic placement and phenotypic variability

The three isolates of Candida gelsemii were similar but not identical in sequences, morphology, and growth responses. The ribosomal internal transcribed spacer (rDNA

ITS) sequences of strains 06-17.1 and 06-24.1 differed by a single indel and that of strain

06-11.1 differed from the other two by 8 substitutions and one or two indels. However, the nearly identical large subunit rDNA D1/D2 regions (strain 06-11.1 differs by a single substitution in the D1 domain) and a comparison of morphology and physiology within the greater context of other related yeast species supports assigning the isolates to a single species. Extensive attempts to obtain evidence of a sexual cycle were not successful, and so a biological species concept cannot be applied at present. A Megablast search of the

GenBank database using the D1/D2 domains identified the nearest known relative as

Metschnikowia bicuspidata , with a divergence of 57 substitutions. Figure A.1 shows that the new species occupies a somewhat basal position with respect to that group, which contains a number of species that are often isolated from nectars or other plant components, as well as a small subclade of species thought to be parasitic on certain aquatic invertebrates ( i.e. , Metschnikowia bicuspidata and allies; Miller & Phaff 1998).

Sisterhood of Candida gelsemii and Candida rancoensis is not well supported by the data. However, addition of any other species to the sequence analysis did not alter the

148 presumed monophyly of the ingroup species included in Fig. A.1 ( Metschnikowia pulcherrima was the outgroup).

Considering the narrow distribution of the known isolates of Candida gelsemii , their phenotypic variation is significant. As shown in Fig. A.2, our attempts to obtain ascus formation on dilute V8 agar, although unsuccessful, demonstrated that the three strains are distinct with respect to cell size and propensity to differentiate into chlamydospores that are often referred to as “pulcherrima cells”, in reference to the resting, pre-ascal cells formed by Metschnikowia pulcherrima (Miller and Phaff 1998).

The large lipid globules seen in strain 06-24.1 are most typical of this. The formation of bilobate lipid globules by some of the cells is not typical, however. The variation observed at the physiological level, detailed in Table A.2, was not obviously correlated with the extent of divergence in ITS sequences.

Physiologically, Candida gelsemii superficially resembles phylogenetic congeners such as Metschnikowia lachancei and M. vanudenii , and to a lesser extent M. bicuspidata and M. gruesii , most of which have been isolated from floral nectars (Miller and Phaff

1998, Gimenez-Jurado et al. 2003). M. bicuspidata is of marine origin. The most important differences from other nectar isolates were the weak or negative growth at

30ºC or in the presence of 50% glucose, and the rather weak fermentation. Such differences would not constitute strong key characters for identification.

Ecology

The presence of a highly potent toxic alkaloid in the nectar of Gelsemium sempervirens might constitute a selective factor that favours the presence of resistant yeast species over the more frequently isolated nectar species such as Metschnikowia

149 reukaufii . To test the hypothesis that gelsemine might be such a niche determinant, we tested its effect on the growth of the yeasts listed in Table A.1 as well as others, using two approaches. Two concentrations of synthetic gelsemine, 100ng/µL and 250ng/µL, were added to YM agar and used in the replica plate series used to characterize the isolates as well as others that came from nectar of a tropical palm. These concentrations simulate levels of gelsemine that occur in the nectar of natural G. sempervirens populations and are known to deter several species of flower visitors, including the bumble bee Bombus impatiens , an important pollinator of G. sempervirens (Adler and

Irwin 2005, 2006, Manson, personal observation). To ensure that these concentrations were neither insufficient nor excessive, disks impregnated with gelsemine were also applied to lawns of yeast on agar. The species tested included M. pulcherrima , M. reukaufii , Debaryomyces melissophilus , and Starmerella bombicola . In all cases, no significant effect was detected, indicating that neither yeasts from jessamine nectar nor those from the nectar of other plants experienced reduced growth due to the presence of gelsemine. The mechanism of tolerance to this alkaloid is unknown but appears to be generalized within a broad range of yeasts and suggests that predictions of the toxicity of nectar alkaloids to microbial communities may be incorrect (Adler 2000). We propose that perhaps the yeasts found in jessamine nectar may instead act as a detoxifying agent selectively carried by visitors to that plant as a co-evolved adaptation. We hope to test that hypothesis in the future.

150

Description of Candida gelsemii Lachance sp. nov.

In 2% glucose 0.5% yeast extract after 3 days, the cells are ovoid, occur singly or in bud-mother cell pairs, and measure 3-7 x 5-10 µm. Neither a ring nor a pellicle is formed. On agar media, the colonies are low-convex, slightly umbonate with an entire margin. The surface is glabrous and can be papillate or pitted. On Dalmau plates with

YCB agar supplemented with 0.01% ammonium sulfate, after 2 weeks, a few chains of undifferentiated cells may be formed after two weeks. The cultures were examined individually or mixed in pairs on YCB agar with 0.01% ammonium sulfate and dilute

(1/20) V8 agar. Mating or ascus formation were not observed. Resting cells with conspicuous lipid globules may be formed on dilute V8 (Fig. A.2). Glucose is fermented weakly. Glucose, sucrose, maltose, melezitose, cellobiose (variable or slow), salicin

(variable or weak), glycerol (weak), glucitol (weak), gluconic acid, glucono-∆-lactone

(weak), N-acetyl glucosamine, and hexadecane (weak) are assimilated, but not inulin, raffinose, melibiose, galactose (sometimes weak), lactose, trehalose (sometimes weak), α- methyl-D-glucoside, sorbose, rhamnose, xylose, L-arabinose, D-arabinose, ribose, methanol, 1-propanol, 2-propanol, 1-butanol, erythritol, ribitol, xylitol, galactitol, mannitol, inositol, lactic acid, succinic acid (sometimes weak), citric acid, malic acid, 2- keto-gluconic acid, or glucosamine (occasionally weak). Ethylamine, L-lysine, and cadaverine are utilized as sole nitrogen sources, but not nitrate or nitrite. Growth in vitamin-free medium negative. Growth in amino acid-free medium positive. Growth at

4ºC weak, at 24ºC positive, at 30ºC negative or weak. Gelatin hydrolysis weak. Casein hydrolysis positive. Tween 80 hydrolysis positive. Acid production on chalk agar negative. Growth in YM agar with 10% NaCl positive or slow; 15% negative. Growth in

151

50% W/W glucose, 1% yeast extract agar negative. Growth in the presence of 10 mg/L cycloheximide negative. Growth in the presence of 75 mg/L CTAB positive. Starch production negative. Diazonium Blue B reaction negative. The habitat is nectar of

Gelsemium sempervirens in Georgia, USA. The type culture is strain UWOPS 06-24.1 T isolated from nectar. The type was deposited in the culture collection of the

Centraalbureau voor Schimmelculture, Utrecht, the Netherlands (CBS 10509, =

NRRL Y-48212 T). Mycobank 46490.

Etymology: gel.se’mi.i, L. gen. sing. neut. n., gelsemii, of Gelsemium , referring to the plant from which the isolates were obtained.

Acknowledgements

This work was funded by grants from the Natural Science and Engineering Research

Council of Canada (MAL and JDT). We thank Sheila Colla and Erin Willis for their assistance during field collections, and Lissa Leege for directing us towards our field sites.

152

Table A.1. Summary of yeasts recovered from nectar of flowers obtained in the vicinity

of Georgia Southern University campus in Statesboro, Georgia.

Yeast Species Strain Plant Species ( N) Locality

Candida gelsemii 06-11.1 25.5 km east of campus

06-17.1 25 km east of campus

Gelsemium sempervirens (34) 06-24.1 1.6 km east of campus Candida rancoensis 06-22.1

06-26.1

Metschnikowia reukaufii 06-29.1 Horticultural azalea (6) University campus 06-32.1

153

Table A.2. Growth responses that exhibit variation among the three known strains of

Candida gelsemii . Growth is scored from weakest to strongest response as follows: - = no growth, w = weak growth, s= slow growth, + = successful growth.

Strain Growth Test

06-11.1 06-17.1 06-24.1 T

Galactose w - -

Trehalose - - w

Cellobiose s - s

Salicin w - w

Succinic Acid - w -

Glucosamine w - -

30 ºC w - -

10% NaCl s + s

154

64 Candida gelsemii UWOPS 06-24.106 -24.1 T (DQ988046) 100 Candida gelsemii UWOPS 06-17.1 (DQ988047) Candida gelsemii UWOPS 06-11.1 (DQ988045) Candida rancoensis CBS 8174 T (AJ508580) T 78 Metschnikowia zobellii NRRL Y-5387 (U44823) Metschnikowia bicuspidata NRRL YB-4993 NT (U44822) 60 80 Metschnikowia australis NRRL Y-17414 T (U76526) 79 Metschnikowia krissii NRRL Y-5389 T (U45735) Metschnikowia viticola NCAIM Y.01705 T (AY626892) 52 Metschnikowia reukaufii NRRL Y-7112 T (U44825) 98 Metschnikowia koreensis KCTC 7828 T (AF257272) Metschnikowia vanudenii PYCC 4650 T (AF017404) Metschnikowia lachancei PYCC 4605 T (AY080995) Metschnikowia gruessii NRRL Y-17809 T (U45737) Metschnikowia pulcherrima NRRL Y-7111 T (U45736) 0.05

Figure A.1. Phylogram of Candida gelsemii and closest relatives based on a neighbour- joining analysis (K2P transform) of D1/D2 LSU rDNA sequences. Bootstrap values

(1000 pseudoreplications) of 50% or greater are shown. Strain numbers and sequence accession numbers are given. The superscripts identify type (T) and neotype strains

(NT).

155

Figure A.2. Candida gelsemii after 1 month on dilute V8 agar (1/20) at 18ºC. Strains 06-11.1 (a), 06-17.1 (b), and 06-24.1 (c), showing various degrees of differentiation into “pulcherrima cells”. Asci were not formed on these and other sporulation media.

APPENDIX TWO

Complete list of the concentration of cardenolides from the nectar, leaves and flowers of twelve species of Asclepias from the series Incarnatae, as described in chapter five. Total average concentrations represent the sum of the total cardenolide concentration of each sample within a species divided by the number of samples, with all concentrations on a per microgram basis. The average concentrations of all individual compounds, which are identified by their unique retention times, are also listed. Note that because of variation in the concentration of individual cardenolides within samples, the sum of individual cardenolide concentrations differs from the average total cardenolide concentration. I’ve indicated several trends throughout appendix two: first, columns with borders represent compounds never found in nectar; second, dark grey cells represent nectar cardenolides that are not matched with a cardenolide in the leaf of that species (see chapter five

Discussion); third, the column highlighted in light grey represents the single compound found only in nectar.

(appendix two continued on the next three pages)

156 157

Plant Total Avg Avg Concentration for Individual Cardenolides (ng/ µg) Species Part (ng/ µg) n 12.5 13.1 13.5 13.7 14.1 14.3 14.5 14.7 14.9 15.3 15.7 Asclepias angustifolia nectar 0 5 0 0 0 0 0 0 0 0 0 0 0 Asclepias angustifolia leaf 0 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias angustifolia flower 0.06 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias barjoniifolia nectar 12.26 3 0 0 0 0 0 0 0 0 0 0 0 Asclepias barjoniifolia leaf 0.37 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias boliviensis nectar 4.83 6 0 0 0 0 0 0 0 0 0 2.21 0 Asclepias boliviensis leaf 0.45 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias candida nectar 11.27 3 0 0 0 0 0 0 0 0 0 0 0 Asclepias candida leaf 0.21 2 0 0 0 0 0 0 0 0.05 0 0 0 Asclepias curassavica nectar 0 8 0 0 0 0 0 0 0 0 0 0 0 Asclepias curassavica leaf 1.96 2 0.08 0 0 0 0 0 0 0.26 0 0.11 0 Asclepias curassavica flower 2.71 2 0 0.46 0 0.15 0 0 0 0.49 0 0 0 Asclepias fascicularis nectar 0 7 0 0 0 0 0 0 0 0 0 0 0 Asclepias fascicularis leaf 0 1 0 0 0 0 0 0 0 0 0 0 0 Asclepias incarnata pulchra nectar 0 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias incarnata pulchra leaf 0.12 1 0 0 0 0 0 0 0 0 0 0 0 Asclepias mexicana nectar 3.55 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias mexicana leaf 0.02 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias nivea nectar 32.81 7 0 0 0 0 0 0 0 0.54 0 0 0.3 Asclepias nivea leaf 0.62 2 0 0 0 0 0 0 0 0 0 0 0.08 Asclepias nivea flower 0.79 2 0 0.07 0 0 0 0 0 0 0.1 0 0.07 Asclepias perennis nectar 38.10 4 0 0 0 3.13 2.11 0 0 0 10.6 1.73 0 Asclepias perennis leaf 2.23 2 0.14 0 0.03 0.68 0 0.06 0.44 0 0.23 0.03 0 Asclepias perennis flower 0.79 2 0 0 0 0 0 0 0.14 0 0.31 0 0 Asclepias pumila nectar 109.97 1 0 8.4 0 0 15.9 0 0 0 35.9 0 0 Asclepias pumila leaf 0.17 2 0 0 0 0 0 0 0 0 0.08 0 0 Asclepias pumila flower 0.20 2 0 0 0 0 0 0 0 0 0 0 0 Asclepias texana nectar 0 4 0 0 0 0 0 0 0 0 0 0 0 Asclepias texana leaf 0.06 3 0 0 0 0 0 0 0 0 0 0 0 Asclepias texana flower 0.04 2 0 0 0 0 0 0 0 0 0 0 0

158

Plant Avg Concentration (ng/ µg) Species Part 15.9 16.1 16.3 16.5 16.7 16.9 17.2 17.5 17.8 18.4 18.6 19 19.3 19.5 19.8 Asclepias angustifolia nectar 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias angustifolia leaf 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias angustifolia flower 0 0 0 0 0.04 0 0 0 0 0 0 0 0 0.01 0 Asclepias barjoniifolia nectar 0 0 0 0 0 4.47 0 0 0 0 3.18 0 0 0 0 Asclepias barjoniifolia leaf 0 0 0 0.09 0 0 0 0 0.07 0 0.09 0 0 0 0 Asclepias boliviensis nectar 0 0 0 0 0 0.35 0 0.16 0 0 0.96 0 0 0 2.43 Asclepias boliviensis leaf 0.09 0 0 0 0 0 0 0.11 0 0.01 0.03 0.07 0 0 0.13 Asclepias candida nectar 2.78 0 0 0 0 0 4.15 0 2.55 0 0 1.95 0 0 0 Asclepias candida leaf 0 0 0 0 0.04 0 0.02 0 0.02 0 0 0.05 0 0.01 0.02 Asclepias curassavica nectar 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias curassavica leaf 0 0.14 0 0.74 0 0 0 0.05 0.11 0 0.09 0 0 0.05 0 Asclepias curassavica flower 0 0.3 0 0.29 0 0.27 0 0.19 0.03 0 0.14 0 0 0.19 0 Asclepias fascicularis nectar 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias fascicularis leaf 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias incarnata pulchra nectar 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias incarnata pulchra leaf 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias mexicana nectar 0 0 0 0 0 0 0 0 0 0 0 0 7.51 0 0 Asclepias mexicana leaf 0 0 0 0 0 0 0 0 0 0.05 0 0 0 0 0 Asclepias nivea nectar 0 4.48 0 0.42 0 1.81 0 1.39 0 0 13.4 0.52 0.41 0 0 Asclepias nivea leaf 0 0.07 0 0.1 0.03 0 0 0.06 0.07 0 0.06 0.03 0.01 0.02 0 Asclepias nivea flower 0 0 0 0.13 0 0 0 0.02 0 0.01 0.18 0.02 0 0.06 0 Asclepias perennis nectar 0 0 5.54 0 0 0 0 1.84 5.64 0 10.2 0 0 0 0 Asclepias perennis leaf 0 0 0.1 0.08 0 0 0 0.06 0.11 0 0.08 0 0 0.02 0 Asclepias perennis flower 0 0.06 0 0.03 0 0 0.02 0.09 0 0 0.08 0 0 0.03 0 Asclepias pumila nectar 0 0 0 0 0 0 0 12.9 6.28 0 32.8 0 0 0 0 Asclepias pumila leaf 0 0.01 0 0 0 0 0 0.03 0.02 0 0.03 0 0 0 0 Asclepias pumila flower 0 0 0 0 0 0 0 0.12 0 0 0.05 0 0 0.02 0 Asclepias texana nectar 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 Asclepias texana leaf 0 0 0 0 0 0 0 0 0.11 0 0.06 0 0 0 0 Asclepias texana flower 0 0 0 0 0.06 0 0 0 0 0 0 0 0 0 0

159

Plant Avg Concentration (ng/ µg) Species Part 20 20.3 20.7 20.9 21.3 Asclepias angustifolia nectar 0 0 0 0 0 Asclepias angustifolia leaf 0 0 0 0 0 Asclepias angustifolia flower 0 0 0 0.01 0 Asclepias barjoniifolia nectar 2.45 0 0 2.86 0 Asclepias barjoniifolia leaf 0.06 0 0 0.06 0 Asclepias boliviensis nectar 0 0 0 0 0 Asclepias boliviensis leaf 0 0 0 0.01 0 Asclepias candida nectar 0 0 0 0 0 Asclepias candida leaf 0 0 0 0 0 Asclepias curassavica nectar 0 0 0 0 0 Asclepias curassavica leaf 0.19 0 0 0.14 0 Asclepias curassavica flower 0.09 0.04 0 0.06 0 Asclepias fascicularis nectar 0 0 0 0 0 Asclepias fascicularis leaf 0 0 0 0 0 Asclepias incarnata pulchra nectar 0 0 0 0 0 Asclepias incarnata pulchra leaf 0 0 0 0 0.12 Asclepias mexicana nectar 0 0 0 0 0 Asclepias mexicana leaf 0 0 0 0 0 Asclepias nivea nectar 2.08 0 0 16.4 0 Asclepias nivea leaf 0.02 0.01 0 0.06 0 Asclepias nivea flower 0 20.3 0.01 0.04 0 Asclepias perennis nectar 0 0 0 9.76 0 Asclepias perennis leaf 0.02 0.01 0 0.12 0 Asclepias perennis flower 0 0.02 0 0.02 0 Asclepias pumila nectar 0 0 0 4.07 0 Asclepias pumila leaf 0 0.01 0 0 0 Asclepias pumila flower 0 0.01 0 0 0 Asclepias texana nectar 0 0 0 0 0 Asclepias texana leaf 0 0 0 0 0 Asclepias texana flower 0 0 0 0 0

APPENDIX THREE

Raw visit data from the three behavioural assays testing pollinator preference for the nectar cardenolide digitoxin. Data represents the

number of visits to flowers of each treatment type. The grey cells indicate which nectar treatment was associated with blue flowers for each bee.

Assay 1 Assay 2 Assay 3 Bee 10 ng/µl digitoxin 30% sucrose Bee 50 ng/µl digitoxin 30% sucrose Bee 50 ng/µl digitoxin 10 ng/µl digitoxin 1 80 57 13 100 17 24 1 102 2 24 142 14 3 84 25 51 41 3 47 102 15 6 94 26 108 7 4 95 9 16 51 148 27 44 9 5 99 7 17 104 3 28 41 94 6 20 127 18 87 18 29 17 92 7 80 24 19 2 106 30 4 101 8 88 60 20 104 1 31 70 29 9 65 66 21 60 79 32 38 67 10 143 7 22 18 98 33 66 1 11 91 77 23 114 35 12 121 3

160