The Chemical Mediation of Multi-Trophic Interactions: Testing

THE CHEMICAL MEDIATION OF MULTI-TROPHIC INTERACTIONS: TESTING

COMPONENTS OF THE TRI-TROPHIC INTERACTIONS HYPOTHESIS

by CAITLIN ANNE KELLY B.S. Lafayette College, 2007 M.S. Villanova University, 2009

A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirement for the degree of Doctor of Philosophy Department of Ecology and Evolutionary Biology 2016

This thesis for the Doctor of Philosophy degree by

Caitlin Anne Kelly

has been approved for the department of Ecology and Evolutionary Biology

______M. Deane Bowers

______Michael Breed

______Rebecca Safran

______Timothy Seastedt

______Shannon Murphy

The final copy of this thesis has been examined by the signatories, and we find that both the

content and the form meet acceptable presentation standards of scholarly work in the

abovementioned discipline. iii

Kelly, Caitlin (Ph.D., Ecology and Evolutionary Biology)

The chemical mediation of multi-trophic interactions: Testing components of the tri-trophic interactions hypothesis

Thesis directed by Professor M. Deane Bowers

ABSTRACT

Although there are multiple hypotheses regarding the top-down and bottom-up controls of herbivore populations, the Tri-Trophic Interactions (TTI) Hypothesis is the first to make predictions regarding the simultaneous effects of plant quality, herbivore diet breadth, and natural enemies on herbivore performance. My dissertation research tests several predictions of the TTI hypothesis to determine how plant secondary metabolites, herbivore diet breadth

(specialist or generalist) and natural enemies interactively determine the performance of insect herbivores. I study plants in the genus Penstemon (Plantaginaceae), which are chemically defended by bitter iridoid glycosides (IGs), and their associated lepidopteran herbivores. First, I examined the bi-trophic interactions between two different Penstemon species and generalist and specialist lepidopterans. The fitness of the generalist species varied greatly between the two plants, whereas the specialist performed consistently well on both host plant species. Second, I determined the level of variation in Penstemon IGs to which herbivores are naturally exposed. I focused on three potential sources of variation: inter-annual (across years), location (among geographically distinct populations) and plant tissue type. Finally, I examined the role of IGs in tri-trophic interactions by examining the ability of IGs to protect a specialist caterpillar from predators and parasitoids. Ant bioassays indicated that even low amounts of IGs are effect deterrents against invertebrate predators. However, measurements of phenoloxidase activity, an

iv important component of the insect immune response cascade, suggested that caterpillars reared on certain diets were more vulnerable to parasitoid attack. This research provides important contributions to the recently proposed TTI hypothesis and increases our understanding of the chemical mediation of multi-trophic interactions.

ACKNOWLEDGEMENTS

First and foremost, I would like to thank my committee members who have supported me these last several years. I would especially like to recognize my advisor, Deane Bowers, who never gave up on me, even when I wanted to give up on myself. Michael Breed was my mentor first and my cheerleader second. I am grateful to have been a part of the Animal Behavior team, but more importantly, I am grateful for his unwavering faith in me, which saw me through the toughest of times. I would also like to acknowledge my other committee members, Becca Safran,

Tim Seastedt and Shannon Murphy, for all of their time, insight and assistance.

I owe many thanks to several other CU faculty members, EBIO staff, students, and colleagues who have supported me in many different, and often unexpected, ways. My research would not have been possible without my many, fantastic undergraduate assistants: Lauren

Bradley, Hadley Hanson, Jason Hong, Quinn Langsfeld and many others. I am also grateful for the Colorado Parks and Wildlife for allowing me to use Crescent Meadows as my main field site for six consecutive summers.

I greatly appreciate the support provided by the Department of Ecology and Evolutionary

Biology in the form of fellowships and teaching assistant ships. Funding for research presented in this thesis was provided by the Boulder County Nature Association, John W. Marr Fund in

Plant Ecology, the Department of Ecology and Evolutionary Biology at CU, the Hazel Schmoll

Research Fellowship in Colorado Botany, the William H. Burt award provided through the

University of Colorado Museum of Natural History, and the National Science Foundation.

Finally, I would like to acknowledge my family and friends, for never doubting me and providing endless emotional support. I especially thank Zach, for his statistical wisdom and our loving relationship, and my besties, for being the highlight this journey.

TABLE OF CONTENTS

CHAPTER 1. INTRODUCTION 1.1 Research Rationale and Goals ...... 1 1.2 Plant Secondary Metabolites and Herbivore Preference and Performance ...... 2 1.3 Plant Secondary Metabolites and Herbivore-Natural Enemy Interactions ...... 4 1.4 The Tri-Trophic Interactions Hypothesis ...... 6 1.5 Study System Penstemon glaber ...... 8 Penstemon virgatus ...... 9 Euphydryas anicia ...... 10 1.6 Research Overview ...... 10

CHAPTER 2. PREFERENCE AND PERFORMANCE OF GENERALIST AND SPECIALIST HERBIVORES ON CHEMICALLY DEFENDED HOST

ABSTRACT ...... 13 INTRODUCTION ...... 14 METHODS ...... 18 RESULTS ...... 24 DISCUSSION ...... 30

CHAPTER 3. THE UNPREDICTABLE PENSTEMON: VARIATION IN DEFENSIVE CHEMISTRY ACROSS YEARS, POPULATIONS, AND TISSUES

ABSTRACT ...... 34 INTRODUCTION ...... 34 METHODS ...... 38 RESULTS ...... 41 DISCUSSION ...... 48

CHAPTER 4. HOST PLANT IRIDOID GLYCOSIDES MEDIATE HERBIVORE INTERACTIONS WITH NATURAL ENEMIES

ABSTRACT ...... 52 INTRODUCTION ...... 53 METHODS ...... 56 RESULTS ...... 62 DISCUSSION ...... 68

vii

CHAPTER 5. CONCLUSIONS 5.1 Principle Findings ...... 73 5.2 Hypotheses in Multi-Trophic Interactions ...... 76 5.3 Tests of the TTI Hypothesis ...... 79 5.4 Synthesis ...... 82 5.5 Future Directions and Closing Remarks ...... 83

LITERATURE CITED ...... 86

viii

LIST OF TABLES

Table 3.1 The six Penstemon virgatus sampling populations in Colorado, USA...... 43

Table 3.2 Summary of three-way, repeated measures ANOVAs that tested the effects of tissue type, population and year and their interactions on the iridoid glycoside content of Penstemon virgatus. The iridoid glycosides catalpol and scutellarioside were measured as a concentration (% dry weight)...... 44

Table 3.3 Summary of two-way ANOVAs comparing IG content for each tissue type across six Colorado populations and two sampling years. The IGs catalpol and scutellarioside were measured as a concentration (% dry weight)...... 45

Table 5.1 A brief summary of the main hypotheses or goals addressed in each research chapter and the corresponding conclusions...... 74

LIST OF FIGURES

Figure 1.1 From Mooney et al. 2012. The tri-trophic interactions (TTI) hypothesis combines the physiological efficiency, enemy-free space and slow-growth/high-mortality hypotheses to make predictions regarding the roles of host plant quality, diet breadth and natural enemies in herbivore regulation ...... 7

Figure 1.2 Chemical structures of (a) catalpol and (b) scutellarioside ...... 9

Figure 2.1 The iridoid glycoside (IG) composition of Penstemon glaber and P. virgatus leaves from plants reared in the common garden. Means ± s.e...... 27

Figure 2.2 Leaf consumption by herbivores on Penstemon glaber and P. virgatus plants in the common garden. A representative picture of a typical plant after the experiment is below the matching bar. Means ± s.e...... 28

Figure 2.3 Larval performance: The relative growth rates of (a) Estigmene acrea and (b) Euphydryas anicia when reared exclusively on either Penstemon glaber or P. virgatus and survivorship by day six for (c) Es. acrea and (d) Eu. anicia reared on each plant diet. Means ± s.e ...... 29

Figure 3.1 The mean concentration of (a) scutellarioside and (b) catalpol between sampling years for each population and tissue type (leaves, flowers and stems). In each population, lines connect the 2011 means to the 2013 means for each tissue type. Means ± SE. Population abbreviations, in order of increasing elevation: CM – Crescent Meadows; CWD – Calwood Education Center; LRTH – Lumpy Ridge Trail Head; CL – Cub Lake; CHLK – Chalk Lake Campground; MIH – Michigan Hill...... 46

Figure 3.2 Leaf consumption by herbivores on Penstemon virgatus across six Colorado populations: Crescent Meadows (CM), Calwood Education Center (CWD), Michigan Hill (MIH), Cub Lake (CL), Lumpy Ridge Trailhead (LRTH) and Chalk Lake Campground (CHLK). Means ± s.e...... 47

Figure 4.1 The iridoid glycoside (IG) content of (a) Penstemon glaber (n = 25) and P. virgatus (n = 30) leaves and (b) Euphydryas anicia larvae reared on either P. glaber (n = 30 groups of five siblings) or P. virgatus (n = 35 groups of five siblings). Means ± SE...... 63

Figure 4.2 Ant bioassays of the following pairwise comparisons (a) control solution and Penstemon glaber reared caterpillar solution (36 mg/ml), (b) control solution and P. virgatus reared caterpillar solution (34 mg/ml) (c) P. glaber reared caterpillar solution and solution and P. virgatus reared caterpillar solution. Bars are mean ± SE ...... 65

Figure 4.3 Ant bioassays comparing (a) catalpol and (b) catalpol and scutellarioside. Bars are mean ± SE time spent feeding (in seconds) per feeding bout by foraging individuals of Formica pallidefulva...... 66

Figure 4.4 Phenoloxidase activities of larvae reared on either Penstemon glaber (n = 27) or P. virgatus (n = 27). Means ± SE...... 67

Figure 5.1 Adapted from Mooney et al. 2012. The components of the tri-trophic interactions (TTI) hypothesis examined in Chapter 2. This chapter tested the physiological efficiency (PE) hypothesis (blue circle), which posits variation in host plant quality will result in larger differences in generalist herbivore performance than specialist performance (blue square) ...... 74

Figure 5.2 Adapted from Mooney et al. 2012. Chapter 3 focused variation in host plant quality (highlighted in green), which is one of three major components of the TTI hypothesis...... 75

Figure 5.3 Adapted from Mooney et al. 2012. Chapter 4 examined the role of host plant quality in mediating specialist herbivore interactions with natural enemies (highlighted in purple)...... 76

Figure 5.4 A) A model demonstrating the linear nature of traditional trophic models; B) The tri- trophic model from Mooney et al. (2012). Note its use of overlapping circles instead of straight lines...... 78

Figure 5.5 The number of primary publications that cite Mooney et al. 2012 by publication year. None of these publications explicitly test all components of the TTI hypothesis...... 82

CHAPTER 1

INTRODUCTION

1.1 Research Rationale and Goals

Over the past half century, there have been several hypotheses proposed about factors important in the evolution and maintenance of specialized plant-insect interactions (Dethier

1954; Ehrlich and Raven 1964; Krieger et al. 1971; Bernays and Graham 1988; Futuyma and

Moreno 1988; Farrell 1998), yet no single explanation has prevailed. Agrawal (2011) recently proposed that the evolution of specificity in plant responses to herbivores is a “final frontier” in the field of plant-herbivore interactions. Price et al. (1980) combined multiple theories regarding the “top-down” and “bottom-up” controls of herbivores by promoting the idea of tri-trophic interactions among plants, herbivores and their natural enemies. Since then, many hypotheses in the field of tri-trophic interactions have addressed the effects of plant quality, herbivore diet breadth and natural enemies on herbivore performance. However, it was not until recently that all three of these factors were combined into a single hypothesis. The tri-trophic interaction (TTI) hypothesis integrates three existing hypotheses (physiological efficiency, enemy-free space, and slow-growth/high-mortality) to form predictions regarding how all three of these factors regulate herbivore populations (Mooney et al. 2012). My dissertation research tests several of the predictions of the TTI hypothesis to determine how plant quality, herbivore diet breadth and natural enemies interactively determine herbivore performance.

Although plant quality can refer to a multitude of factors, my research specifically examines the role of plant secondary metabolites (also called secondary compounds, natural products, or allelochemicals). These compounds can act as defenses against plant enemies, including herbivores and pathogens (Fraenkel 1953, Odum and Pinkerton 1955, Ehrlich and

Raven 1964, Whittaker and Feeny 1971); however, they may also attract enemies by providing oviposition and feeding cues to specialized herbivores (Da Costa and Jones 1971, Raybould and

Moyes 2001, Macel and Vrieling 2003, Nieminen et al. 2003). Additionally, these compounds have been shown to attract the natural enemies of herbivores (Turlings et al. 1990, Dicke and van

Loon 2000), and thus play a vital role in mediating multi-trophic interactions. However, in only a few systems do we have an understanding of how these compounds affect multiple trophic levels and many questions remain regarding their role in the evolution of herbivore specialization.

The goal of my dissertation research was to provide a better understanding of the natural variation in plant secondary metabolites and the consequences of this variation for higher trophic levels. I used plants of the genus Penstemon (Plantaginaceae) and their associated lepidopteran herbivores to test the following predictions of the TTI hypothesis (1) plant quality will have stronger effects (i.e. more variation in performance) on generalist herbivores than on specialist herbivores and (2) specialist herbivore performance suffers on low quality plants and makes them more susceptible to attack by natural enemies.

1.2 Plant Secondary Metabolites and Herbivore Preference and Performance

Based on diet breadth, herbivorous insects can be broadly categorized as either specialists or generalists. Although definitions vary, one that is commonly used defines specialist herbivores as those with a narrow diet breadth, typically feeding on fewer than three plant families (Bernays and Chapman 1994). Specialists often feed on plants containing particular secondary metabolites, and may use these compounds as feeding or oviposition cues (Da Costa and Jones 1971,

Raybould and Moyes 2001, Macel and Vrieling 2003, Nieminen et al. 2003). Some specialist herbivores have the ability to sequester these compounds for their own defense (e.g. Dyer and

Bowers 1996, Theodoratus and Bowers 1999, Nishida 2002, Opitz and Mueller 2009). For

3 example, the catalpa sphinx (Ceratomia catalpae; Boisduval; Lepidoptera: Sphingidae) sequesters IGs from the catalpa tree (Catalpa spp. Bignoniaceae; Bowers 2003) and many milkweed (Asclepias spp., Apocynaceae) feeders sequester cardenolides (e.g. Malcolm and

Brower 1989, Dobler et al. 1998). Generalist herbivores are traditionally defined by a wide diet breadth and feed across many (greater than three) plant families (Fraenkel 1959; Bernays and

Chapman 1994). While generalists are often deterred or negatively affected by secondary metabolites, numerous studies have shown that specialists have evolved effective countermeasures to the chemical defenses of their hosts (e.g. Siemens and Mitchell-Olds 1996,

Berenbaum and Zangerl 1998).

There are two points in the lepidopteran lifecycle where host plant choice occurs, the larval stage (feeding) and the adult stage (oviposition by females), and plant secondary metabolites may be important in choices at both these life stages. Though females are presented with a suite of external stimuli (e.g. visual and olfactory cues) to assist with host plant choice, plant secondary metabolites, as detected by a female upon landing, often determine if a female will oviposit (e.g. Wiklund 1981, Thompson and Pellmyr 1991, Bernays and Chapman 1994,

Renwick and Chew 1994,). The acceptance of a host plant by contact chemoreceptive recognition may be governed by either the presence of stimulating signals (typically the compounds on which the species specializes; e.g. Macel and Vrieling 2003) or by the absence of deterrents (e.g. Renwick and Radke 1988). The presence of particular secondary metabolites may be especially important for species capable of sequestering compounds for defense against enemies, as the compounds are then potentially used by the insect for their own defense.

However, the presence of particular secondary metabolites is not sufficient to guarantee offspring survival, and in some cases the compounds can even be detrimental to offspring

(reviewed in Thompson 1988, Bowers 1992). Host plant nutritional content and ability to provide enemy free space are also important components of both larval and female choice

(Fraenkel 1953, Jeffries and Lawton 1984, Awmack and Leather 2002, Gripenberg et al. 2010).

Females’ choice of a host plant often reflects the suitability of the plant for larval survival

(Papaj and Prokopy 1989). In the checkerspot Euphydryas editha (Nymphalidae), for example, it has been demonstrated that females prefer to oviposit on a host plant that ensures the highest fitness for their offspring, who later demonstrate preference for that same host (Singer et al.

1988). Yet, female choice and larval choice do not always correspond (e.g. Wiklund 1974). As larvae are not sessile, they are also capable of making distinct host plant decisions. When parent and offspring choose different host plants, who is making the “correct” choice (i.e. the one which best ensures survival)? A recent meta-analysis found support for positive preference- performance relationships for herbivorous insects in a bitrophic setting (Gripenberg et al. 2010).

In general, offspring survive better on plant hosts preferred by their mother, and females lay more eggs on hosts conducive to better offspring performance. Yet, several studies report cases where female preference and performance appear uncoupled (e.g. Rausher 1979; Valladares and

Lawton 1991; Underwood 1994; Fritz et al. 2000; Faria and Fernandes 2001). Though several factors influence preference-performance relationships, diet breadth may be a more reliable predictor of the strength of these relationships (Gripenberg et al. 2010).

1.3 Plant Secondary Metabolites and Herbivore-Natural Enemy Interactions

Ehrlich and Raven (1964) hypothesized that diet specialization is the result of herbivore coevolution with plant secondary metabolites, and that many herbivores are physiologically constrained to host plants with similar chemistry. Bernays and Graham (1988) challenged this

5 hypothesis when they proposed that predators are actually the driving force constricting the diet of herbivores. According to the hypothesis proposed by Bernays and Graham (1988), predators are more likely to select generalist herbivores because they are less likely to be chemically defended and may be easier to find. Yet the role of secondary metabolites in mediating herbivore interactions with natural enemies is complex and likely depends on the type of natural enemy.

Numerous studies have shown that secondary metabolites negatively affect predators because certain herbivores can sequester harmful toxins (e.g. Hey et al. 1990, Bowers 1992, Dyer 1995,

Dyer and Bowers 1996, Nishida 2002, Ode 2006, Petschenka and Agrawal 2016). In contrast, secondary metabolites have been shown to have a positive effect on predators by signaling herbivore location via plant volatiles (Turlings et al. 1990, Dicke and van Loon 2000).

Parasitoids are one of the greatest sources of mortality for lepidopteran larvae (Hawkins et al. 1997) and their unique life cycle dictates that they interact with prey differently than predators. One may predict that parasitoids developing within a chemically defended host would suffer negative developmental effects; indeed, several studies have shown that sequestered toxins do adversely affect parasitoids (Barbosa et al. 1986, Sime 2002). However, parasitoids may benefit from preying on herbivores feeding on plants with secondary metabolites and thus show a preference for chemically defended hosts (Dyer and Gentry 1999, Gentry and Dyer 2002,

Sznajder and Harvey 2003, Zvereva and Rank 2003). Parasitoid larvae may be better protected from predators if they develop within a chemically defended host; the sequestered compounds may provide enemy-free space inside the insect host (Gentry and Dyer 2002). Furthermore, insect hosts may become immunocompromised by diverting energy away from immune function in order to metabolize secondary metabolites (Schmid-Hempel and Ebert 2003), thus rendering them more vulnerable to parasitism.

The insect immune response is one of the most effective defenses against parasitoids and can also functions as a predictor of parasitism (Smilanich et al. 2009b). Encapsulation is one such immune response whereby specialized immune cells (hemocytes) build layers of cells around a foreign body, which will eventually become melanized (see Gillespie et al. 1997,

Strand 2008). The melanization process relies on rapidly activated cascades of enzymes to catalyze the initial steps in the production of melanin (Sugumaran 2002; Siva-Jothy et al. 2005,

González-Santoyo and Córdoba-Aguilar 2012). Phenoloxidase (PO) is an enzyme known for its role in the melanization of cuticle, wound healing, and immune responses of insects (Cory and

Hoover 2006; González-Santoyo and Córdoba-Aguilar 2011). However, such immune responses can be costly in terms of both host growth and reproduction (Ahmed et al. 2002, Freitak et al.

2003, Cory and Hoover 2006, McKean et al. 2008, Honkavaara et al. 2009, Ardia et al. 2012).

For example, oxidized phenolics can covalently bind to proteins, reducing the nutritive quality of plant protein for herbivores, thereby reducing the growth of the herbivore (Cory and Hoover

2006). Therefore, herbivores, particularly those that sequester secondary metabolites, face a trade-off between defense from predators and protection from parasitoids (Smilanich et al.

2009a)

1.4 The Tri-Trophic Interactions Hypothesis

The most influential hypotheses in plant-herbivore and herbivore-predator interactions consider the interactive effects of plant quality, herbivore diet breadth, and natural enemies on herbivore performance. However, these hypotheses typically only address two of these three factors in different pairwise combinations. The recently proposed Tri-Trophic Interactions (TTI) hypothesis is the first to integrate existing hypotheses to propose predictions regarding the

The Tri-Trophic Interactions Hypothesis 7

FigureFigure 1. 1.1 Predictions From Mooney of the tri-trophic et al. 2012. interactions The tri-trophic (TTI) hypothesis interactions for (TTI) the interactive hypothesis effects of natural enemies, host-plant quality andcombines diet breadth the physiological on herbivores. efficiency,Three well-studied enemy- hypothesesfree space – and the physiologicalslow-growth/high efficiency- (PE), enemy free space (EFS) hypotheses, and slow- growth/high-mortality (SGHM) – each address unique, pairwise combinations of these factors. The physiological efficiency (PE) hypothesis predicts specialistsmortality should hypotheses outperform to make generalists predictions on shared regarding host plants the (e.g. roles a.b), of and host that plant generalists quality, should be more sensitive to variation in host-plant qualitydiet breadth than specialists and natural (e.g. a–cenemies,b–d). Thein herbivore Enemy Free re Spacegulation. (EFS) hypothesisOne of the predicts main natural enemies should have a stronger effect on dietary specialistspredictions than is generalists that specialists (e.g. a–e ,performb–f). The better Slow-Growth/High-Mortality than generalists, on (SGHM) their hypothesis host plant, predicts and low host-plant quality enhances the effects of natural enemies (e.g. b–f,d–h). The TTI hypothesis offers novel predictions for the three-way interaction among these factors: Dietary specialists (as comparedthis difference to generalists) is exaggerated are predicted in the to escape presence natural of enemiesenemies. and be competitively dominant due to faster growth rates, and such differences should be greater on low quality (as compared to high quality) host plants. Such non-additive dynamics imply that predictions for the PE, EFS, and SGHM hypotheses are contingent upon the third, discounted factor. Natural enemies should mediate the predictions of the PE hypothesis, such that simultaneousthe differential effects effects of of all host-plant three factors quality on on specialists the regulation and generalists of herbivore is greater in thepopulations presence of natural(Mooney enemies et al. (e–g %f–h) than in the absence of natural enemies (a–c,b–d). Host-plant quality should mediate the predictions of the EFS hypothesis, such that the differential effects of natural enemies on specialist and generalist herbivores is greater on low-quality host plants (c–g%d–h) than on high-quality host plants (a–e,b–f). 2012).Herbivore The TTI diet hypothesis breadth should integrates mediate the three predictions existing of the hypotheses: SGHM hypothesis, physiological such that SGHM efficiency, dynamics areenemy stronger- for dietary generalist (b– d%f–h) than specialist herbivores (a–c,e–g). doi:10.1371/journal.pone.0034403.g001 free space, and slow-growth/high-mortality (Figure 1.1). Current hypotheses The PE hypothesis also offers three predictions for the TheThe physiological efficiency efficiency (PE) hypothesis(PE) hypothesis states that posits dietary that dietaryinteractive specialist effects ofs host-plant are better quality and herbivore diet breadth. specialists are better adapted than generalists at physiologically First, variation in host-plant quality should have stronger effects on dietary generalists than on better-adapted dietary specialists [28]. adaptedutilizing than theirgeneralists host plants at asphysiologically food [1]. As a result, utilizi specialistsng their should host plants as food. Therefore, have superior physiological performance (e.g., more efficient As depicted in Fig. 1, the PE hypothesis predicts a–c,b–d (PE resource assimilation and faster growth rates) than generalists on effects without natural enemies) and e–g,f–h (PE effects with specialiststheir sharedshould host demonstrate plants [24,25]. superior As depicted physiological in Fig. 1, performance the PE natural as compared enemies). However,to generalists the PE on hypothesis does not offer hypothesis predicts a.b, c.d, e.f, and g.h. This central predictions for the relative magnitude of PE effects between the presence and absence of natural enemies. Past studies support this their sharedprediction host of plants the PE (Scriber hypothesis and has Feeny found support 1979, inMoran some (e.g.1986). The enemy-free space hypothesis [26]) but not all (e.g. [24]) studies (reviewed by [27]). first prediction, showing that toxic forms of plant secondary compounds have larger effects on the performance of generalist (EFS) hypothesizes that specialist herbivores are better adapted than generalists at using their Table 1. Descriptions of three long-standing hypotheses for plant-herbivore and herbivore-predator interactions and their relation host plantsto the for tri-trophic defense interactionsfrom natural hypothesis. enemies due to their improved crypsis or ability to sequester

plant secondary compounds (Bernays 1998). The slow-growth/high-mortality hypothesisPredictions under tri-trophic interactions Original hypothesis hypothesis (SGHM)Name proposes that low-quality Factors considered host plants Predictionsslow the development of herbivores and thus Physiological efficiency Diet breadth, plant Specialists are better adapted than generalists at The benefits of specialization for performance are increase their vulnerability toquality natural enemies usingby extend shared plantsing astime food spend (a.b, c. ind, ethe.f, larvalgreater stage in the(Price presence of natural enemies (e–g%f–h) and g.h) and variation in host-plant quality than absence of natural enemies (a–c,b–d) should have stronger effects on generalists than et al. 1980, Moran and Hamilton 1980, Clancyspecialists and Price (a–c, 1987b–d and).e–g The,f–h) TTI Hypothesis makes the Enemy-free space Diet breadth, natural Specialist are better than generalists at using The benefits of specialization for predator avoidance enemies shared plants for predator avoidance (a–e,b–f are greater on low-quality plants (c–g%d–h) than and c–g,d–h) high-quality plants (a–e,b–f) Slow-growth/high-mortality Plant quality, natural Low plant quality increases the effects of natural Low plant quality increases the effects of natural enemies enemies (a–e,c–g and b–f,d–h) enemies more for generalists (b–f%d–h) than specialists (a–e,c–g)

Parenthetical references to a–h refer to the graphical representation of these predictions shown in Fig. 1. doi:10.1371/journal.pone.0034403.t001

PLoS ONE | www.plosone.org 2 April 2012 | Volume 7 | Issue 4 | e34403 8 central prediction that host-plant quality, herbivore diet breadth, and natural enemies interactively determine herbivore performance in ways not explicitly addressed by the existing hypotheses (Mooney et al. 2012). More specifically, it predicts that specialists will escape natural enemies and outcompete generalists, and that such differences should be greater on low quality

(as compared to high quality) host plants. This recently proposed hypothesis has provided a unique framework for investigating selection by trophic levels on the maintenance and regulation of herbivore populations.

1.5. Study System

Penstemon glaber

Penstemon species are long-lived herbaceous perennials native to North America, and are chemically defended by a group of monoterpene-derived secondary metabolites known as iridoid glycosides (IGs; Stermitz et al. 1994, Nold 1999). These extremely bitter compounds are found in more than fifty plant families (Bowers 1991) and have been shown to be important mediators of tri-trophic interactions (Harvey et al 2005, Lampert and Bowers 2010). My research focused on two species of Penstemon, co-occurring in the Front Range of Colorado: Penstemon glaber var. alpinus (Pursh) and P. virgatus (A. Gray). Both species grow in mountain meadows and road cuts and appear to prefer xeric, rocky or sandy soils (Nold 1999).

Penstemon glaber var. alpinus (alpine sawsepal penstemon) is one of four known varieties of P. glaber and has broad and occasionally puberulent leaves, thick stems and inflorescences with blue or blue-purple flowers. The chemical defenses of P. glaber are less well known that those of other Penstemons (Stermitz et al., 1994). There is only one other published study on the chemistry of P. glaber (Stermitz et al., 1994), which found the IGs catalpol, lamiide,

9 globularin, and durantoside-I. However, those individuals were collected from the northern most part of the range in central Wyoming and the variety of P. glaber was not specified. Preliminary analyses from the Colorado population where I worked (the Crescent Meadows region of El

Dorado State Park, Boulder County, Colorado, USA; 39° 55' 51.60" N 105° 20' 16.80" W, elevation 2258m) suggest that P. glaber var. alpinus at this site contains large amounts of catalpol but not other IGs (M.D. Bowers, unpublished data). Documented herbivores include

Euphydryas anicia (Nymphalidae; Kelly and Bowers 2016).

Penstemon virgatus

Penstemon virgatus grows in rocky or sandy soils in mountain meadows throughout

Colorado and other southwestern states (Shonle et al. 2004). Also known as the upright blue beardtongue, P. virgatus features narrow leaves, tall stems and long inflorescences with several small to medium purple flowers that bloom from June to August (Crosswhite 1967). Previous research found two major IGs in P. virgatus: catalpol and scutellarioside-II (henceforth referred to as scutellarioside; L’Empereur and Stermitz 1990; Figure 1.2). Documented herbivores include Euphydryas anicia, Polydryas arachnae, Polydryas minuta (all family Nymphalidae) and Herbivores on chemically defended host plants 3

(a) (b) The salt marsh moth, E. acrea, is a conspicuous grazing generalist found throughout the United States, including Colorado (Bernays et al., 2004). Populations have been recorded feeding on at least 88 plant species in 33 families, and are pests on many crops (Singer et al.,2004).TheycantoleratefeedingonIG-rich plants (Bowers, 2009; Lampert & Bowers, 2010). Females lay 400–1000 eggs in one or more clusters. Larvae complete fve to seven instars and typically overwinter as mature larvae. They can complete between one and three generations year−1,depending on location. Larvae are poor sequesterers of IGs (Lampert FigureFig. 1. 1.2Chemical Chemical structures structures of: (a) of catalpol (a) catalpol and (b) and scutellarioside. (b) scutellarioside & Bowers, 2010). Larvae used in these experiments were from a laboratory colony maintained at the University of Colorado. The colony was started with eggs and larvae collected in Austin, blue or blue-purple fowers. Although the defensive chemistry of Texas. The colony was maintained on dandelion (Taraxacum spp.) leaves collected in Boulder, Colorado. P. v i rg a t u s has been previously examined (L’Empereur & Ster- mitz, 1990b; Quintero & Bowers, 2013), the chemical defences of P. g l a b e r are less well known (Stermitz et al., 1994). There Field site. The feld experiment was conducted at the Cres- is only one other published study on the chemistry of P. g l a b e r cent Meadows site. This location features a large population of (Stermitz et al., 1994), which found the IGs catalpol, lamiide, the IG specialist, Euphydryas anicia, and several of its potential globularin, and durantoside-I. However, those individuals were host plants. Multiple Penstemon species, including P. v i rg a t u s collected from the northernmost part of the range in central and P. g l a b e r , co-occur at this site and E. anicia uses both species Wyoming and the variety of P. g l a b e r was not specifed. Prelim- as host plants at this site (C.A. Kelly, pers. obs.). inary analyses with Colorado populations in recent years suggest that P. g l a b e r var. alpinus from our feld site contains large Common garden site. The common garden experiment took amounts of catalpol but not other IGs (M.D. Bowers, unpub- place in a 14 × 12 m plot with sandy soil located in Longmont, lished). Both species naturally co-occur at the feld site used in Colorado (40∘9′24.29′′N105∘12′11.87′′W, elevation 1566 m). this study. All plants used in experiments were obtained from a This site was chosen because it is outside of the range of local nursery and were maintained in 1 gallon pots outside of a herbivores naturally found on Penstemon,whichallowedus greenhouse until used in experiments. to examine the interaction of naïve herbivores with the two Penstemon species grown under identical conditions.

Insects. Euphydryas anicia occurs throughout the western half of the United States, including the mountains of Colorado Common garden experiment – chemistry of P. glaber (Cullenward et al., 1979; White, 1979; Ferris & Brown, 1981). and P. virgatus and preference of generalists Females lay large egg masses (50–200 eggs) on the underside of leaves. Larvae are gregarious in early instars and form a web To examine the IG content of both plant species and the on their host plant. They enter diapause in the fourth instar, and host preference of naïve, generalist herbivores, we planted 30 then overwinter in this stage, emerging in spring to complete P. g l a b e r and 30 P. v i rg a t u s plants in a common garden in development. At our feld site, Crescent Meadows (Eldorado Longmont, Colorado, in May 2010. The plot was rototilled Canyon State Park, Boulder County, Colorado; 39∘55′51.60′′N, prior to planting. All 60 plants were equally spaced, 1.5 m 105∘20′16.80′′W, elevation 2258 m), adults are typically found apart, in a randomly assigned order and then mulched. The in late June to early or mid-July. Euphydryas anicia is known garden was weeded and watered regularly. Plants were allowed to sequester the IGs catalpol, macfadienoside, and aucubin to grow uninhibited and free of pesticide until late August 2010, (Gardner & Stermitz, 1988; L’Empereur, 1989; L’Empereur & approximately 15 weeks. Herbivores observed on the plants Stermitz, 1990a), but there is currently no evidence that they were noted throughout the season and collected when possible. sequester scutellarioside. Analyses to be presented elsewhere There are no known IG specialists that occur at the garden site showed that larvae only sequester catalpol when reared on either and none were observed in the garden. Therefore, generalist P. g l a b e r or P. v i rg a t u s (C.A. Kelly, unpublished). Depend- herbivores were presumed to be the source of all damage. At ing on the location and host plant, E. anicia can sequester the end of the growing period, the proportion of leaf material missing due to herbivore consumption was calculated according 7.1–18.0% dry weight IGs as larvae and 0.5–10.4% dry weight to methods in Stamp and Bowers (1996). Leaves were then IGs as adults (Gardner & Stermitz, 1988; L’Empereur, 1989; harvested and dried for chemical analysis. L’Empereur & Stermitz, 1990a). Sequestered IGs are retained through metamorphosis and both larvae and adults are aposematically coloured (Stermitz et al.,1986;Gardner&Stermitz, Field experiment – oviposition choice by the specialist, 1988). Larvae used in these experiments were reared from eggs E. anicia collected at Crescent Meadows in Eldorado Canyon State Park and were maintained as separate family groups in a growth Twenty pairs of potted P. g l a b e r and P. v i rg a t u s plants were chamber at the University of Colorado. placed in sites at Crescent Meadows where E. anicia had been

Meris alticola (Geometridae; Stermitz et al. 1988, L’Empereur and Stermitz 1990b, Robinson et al. 2002).

Euphydryas anicia

Euphydryas anicia (Doubleday, Lepidoptera: Nymphalidae), the anicia checkerspot butterfly, occurs throughout the western half of the United States, including the mountains of

Colorado (White, 1979; Cullenward et al., 1979; Ferris and Brown, 1981). Females lay large egg masses (50-200 eggs) on the underside of leaves. Larvae are gregarious in early instars and form a web on their host plant. They enter diapause in the fourth instar, and then overwinter in this stage, emerging in spring to complete development. At my field site, Crescent Meadows (see information above) adults are typically found in late June through early or mid July. Euphydryas anicia is known to sequester the IGs catalpol, macfadienoside, and aucubin (Gardner and

Stermitz, 1988; L’Empereur, 1989; L’Empereur and Stermitz, 1990a), but there is currently no evidence that they sequester scutellarioside. Analyses presented in this dissertation showed that larvae only sequester catalpol when reared on either P. glaber or P. virgatus (Kelly, Chapter 4).

Depending on the location and host plant, previous research showed that E. anicia can sequester

7.1-18.0% dry weight IGs as larvae and 0.5-10.4% dry weight IGs as adults (Gardner and

Stermitz, 1988; L’Empereur, 1989; L’Empereur and Stermitz, 1990a). Sequestered IGs are retained through metamorphosis and both larvae and adults are aposematically colored (Stermitz et al., 1986; Gardner and Stermitz, 1988).

1.6 Research Overview

My dissertation research tests several components of the TTI hypothesis to determine

11 how plant secondary metabolites mediate herbivore performance on different host plants and how those compounds further mediate herbivore interactions with higher trophic levels. In

Chapter 2, I present a study investigating how differences in plant secondary metabolites mediate the host plant choice and performance of herbivores of varying diet breadth. This project tested components of the physiological efficiency (PE) and preference-performance (P-P) hypotheses using the specialist checkerspot, Euphydryas anicia, the generalist salt marsh caterpillar,

Estigmene acrea (Erebidae). Results showed that components of the physiological efficiency hypothesis were supported in this system, as the specialist outperformed the generalist on the more iridoid glycoside rich host plant species. However, there was no support for the preference- performance hypothesis, as there was no clear relationship between female preference in the field and offspring performance in the laboratory. In Chapter 3, I assess variation in the IG content of

P. virgatus across multiple growing seasons, six natural populations and three tissue types

(leaves, stems and flowers). I also present data on the amount of herbivore damage at these six locations. I found significant variation in iridoid glycoside concentrations between sampling years and among plant populations and tissue type. Results also indicated that P. virgatus plants contain unusually high concentrations of iridoid glycosides and there was very little herbivore damage at each population. Chapter 4 examines how plant secondary metabolites may influence herbivore sequestration and herbivore defenses against predators and parasitoids and how

Penstemon IGs mediate interactions between E. anicia and the third trophic level. The results of these experiments indicate that host plant diet affects larval sequestration of IGs, although predators found all concentrations of IGs unpalatable. However, I did detect a difference in the phenoloxidase activity of larvae reared on different host plants, which suggests that certain host plant diets may reduce the immune response of E. anicia larvae. Lastly, in Chapter 5, I

12 summarize the main conclusions of my dissertation research.

The recently proposed TTI hypothesis has provided a unique framework for investigating the effects of multiple trophic levels on the maintenance of herbivore populations. Given the potential implications of this hypothesis for future research, it is important to test its predictions across a diverse set of taxa. The only published test of this hypothesis used piercing/sucking herbivores and a woody shrub (Mooney et al. 2012). My research presented here provides a novel test of this hypothesis and the first to use chewing herbivores and an herbaceous perennial.

Furthermore, Mooney et al. (2012) suggested that future studies focus on the potentially different roles played by predators and parasitoids, as they may respond differently to variation in herbivore reared on different quality host plants, and my research has provided such a comparison.

CHAPTER 2

PREFERENCE AND PERFORMANCE OF GENERALIST AND SPECIALIST HERBIVORES ON CHEMICALLY DEFENDED HOST PLANTS1

ABSTRACT

Both the physiological efficiency (PE) hypothesis and preference-performance (P-P) hypothesis address the complex interactions between herbivores and host plants, albeit from different perspectives. The PE hypothesis contends that specialists are better physiologically adapted to their host plants than generalists. The P-P hypothesis predicts that larvae perform best on the host plant preferred by ovipositing females. This study tests components of both hypotheses using the specialist checkerspot, Euphydryas anicia, the generalist salt marsh caterpillar, Estigmene acrea, and host plants in the genus Penstemon, which are defended by iridoid glycosides. In laboratory experiments, the generalist preferred and performed significantly better on the less defended host plant species. This is consistent with results from a common garden experiment where the less defended Penstemon species received more damage from the local community of generalists. Larvae of the specialist checkerspot preferred the more chemically defended species in the laboratory, but performed equally well on both hosts.

However, field experiments demonstrated that adult checkerspot females preferred to oviposit on the less defended host plant. Components of the physiological efficiency hypothesis were supported in this system, as the specialist outperformed the generalist on the more iridoid glycoside rich host plant species. There was no support for the preference-performance

1 Published version: Kelly, C.A. and M.D. Bowers. (2016). Preference and performance of generalist and specialist herbivores on chemically defended host plants. Ecological Entomology. 41(3): 308-316. doi: 10.1111/een.12305

14 hypothesis, however, as there was no clear relationship between female preference in the field and offspring performance in the laboratory.

INTRODUCTION

Individual plant species may be exposed to multiple herbivores that vary widely in their degree of dietary specialization. The performance of specialist herbivores (monophagous or oligophagous species with adaptions to specific host plant defenses) on particular plant species may be greater, less than or the same as that of generalist herbivores (grazing species which feed on multiple plant species over the course of their lifetime). Several hypotheses have been proposed to explain these performance differences while providing insight into the evolution of host specificity (Krieger et al., 1971; Whittaker and Feeny, 1971; reviewed in Cornell and

Hawkins, 2003 and Ali and Agrawal, 2012). One of the most influential is the physiological efficiency (PE) hypothesis (also known as the feeding specialization hypothesis or the trade-off hypothesis; Rausher, 1984, 1988; Noriyuki and Osawa, 2012; Friberg et al., 2015), which posits that dietary specialists show superior performance (e.g. faster growth rates) compared to generalists on a shared host plant due to physiological adaptations for utilizing the host plant as food (Dethier, 1954; Scriber, 1983; Singer, 2001; Scriber, 2005). Additionally, the PE hypothesis predicts that variation in the chemical defenses of host plant taxa used by generalists and specialists will result in larger differences in generalist herbivore performance than specialist performance (Cornell and Hawkins, 2003). However, this hypothesis has a history of mixed support from empirical studies, especially those studying “composite generalists”, which are relatively specialized at the population or individual level (Fox and Morrow, 1981), as opposed to a “true” food-mixing, grazing generalist (Singer, 2001). A second hypothesis, the preference-

15 performance (P-P) hypothesis, also addresses host plant suitability, but in the context of maternal choice (reviewed by e.g. Thompson, 1988; Gripenberg et al., 2010). Specifically, the P-P hypothesis predicts that the oviposition preference of adult females corresponds to the performance of their offspring (Jaenike, 1978; Thompson, 1988). Thus, phytophagous females are expected to maximize their fitness by ovipositing on plants that will provide a better diet for their offspring (Dethier, 1959a, 1959b; Singer, 1972, Jaenike, 1978). Yet, empirical tests of the

P-P hypothesis have produced contradictory results, prompting the proposal of several theories addressing why female choice may not match offspring performance (e.g. Thompson, 1988;

Courtney and Kibota, 1990; Thompson and Pellmyr, 1991; Mayhew, 1997; Craig and Itami,

2008, Gripenberg et al., 2010). Together with physiological adaptations, female host plant choice is a necessary consideration when evaluating herbivore host plant suitability.

These two hypotheses are closely linked as they both consider the importance of maternal choice on offspring performance for specialists and generalists. The physiological suitability of a plant species is likely a major component of a mother’s host plant choice, particularly for specialists whose offspring may demonstrate a preference for the mother’s preferred host species

(e.g. Singer et al., 1988). Although females are presented with a suite of external stimuli (e.g. multiple visual and olfactory cues) to assist with host plant choice, plant secondary metabolites, as detected by a female upon landing, often determine whether a female will oviposit (e.g.

Bernays and Chapman, 1994; Macel and Vrieling, 2003; Nieminen et al., 2003). Similarly, larvae also make host plant choices based on secondary metabolite content (Da Costa and Jones,

1971; Raybould and Moyes, 2001; Bowers, 2003), which may (e.g. Singer et al., 1988; Gonzáles and Gianoli, 2003) or may not (e.g. Clark et al., 2011) reflect their mother’s preferences.

Though the presence of particular secondary metabolites in host plants may harm or deter

16 generalist herbivores, these same compounds may provide oviposition cues to specialist females with larvae capable of coping with these compounds (e.g. Bernays et al., 2003). Therefore, one may predict that the preference-performance relationships are stronger for specialist herbivores, compared to generalists feeding on the same host, as specialists may be more ‘finely tuned’ to certain characteristics (e.g. secondary metabolites) that may indicate host plant suitability

(Gripenberg et al., 2010).

Combining the generalist/specialist performance component of the PE hypothesis and the importance of maternal host plant choice from the P-P hypothesis, we can hypothesize that specialist herbivores will have a strong preference-performance relationship and will also outperform generalists on host plant(s) preferred by the specialists. To test our hypothesis, we utilized a system involving a specialist herbivore, Euphydryas anicia Doubleday (Lepidoptera:

Nymphalidae; the anicia checkerspot), two of its host plant species, and generalist herbivores, including Estigmene acrea Drury (Lepidoptera: Erebidae). Euphydryas anicia specializes on plants containing iridoid glycosides in several plant families (e.g. Plantaginaceae,

Scrophulariaceae, Orobanchaceae, and Caprifoliaceae; Cullenward et al., 1979; White, 1979) and can sequester these compounds (Stermitz et al., 1986; Gardner and Stermitz, 1988;

L’Empereur and Stermitz, 1990a). Here, we study a population of Eu. anicia that primarily uses two species, Penstemon glaber var. alpinus (Torr.) A. Gray (Plantaginaceae) and Penstemon virgatus A. Gray. Penstemon species contain iridoid glycosides (IGs), a group of monoterpene- derived compounds found in over 50 plant families (El-Naggar and Beal, 1980; Boros and

Stermitz, 1990; Bowers, 1991). These bitter compounds are important mediators of multi- trophic interactions (e.g., Harvey et al., 2005; Lampert and Bowers, 2010). For example, IGs

17 affect both the predators and parasitoids of sequestering caterpillar species (Smilanich et al.

2009).

Here we investigate how differences in plant chemical defenses mediate the host plant choice and performance of Eu. anicia and the generalist Es. acrea. For this study, the palatability of host plants is considered from the perspective of a generalist insect herbivore (see

Ali and Agrawal, 2012; Mooney et al., 2012): “more defended” plants contain higher concentrations and higher diversity of secondary metabolites, whereas “less defended” plants contain lower concentrations and less diversity of secondary metabolites. A more diverse suite of secondary metabolites is considered “more defended” due to the potential for synergistic interactions, increasing the toxicity or unpalatability of the plant (McKey, 1979; Berenbaum,

1985; Nelson and Kursar, 1999; Richards et al., 2012). First, we determined the relative levels of chemical defense, defined by IG content and composition, of P. glaber and P. virgatus when grown in a common environment. We then used a common garden to examine which of these two Penstemon species naïve generalists preferred to consume. We examined components of the

P-P hypothesis using the specialist checkerspot. We tested the prediction that female preference corresponds to larval performance by observing Eu. anicia oviposition choice in the field and measuring larval preference and performance when reared on either P. glaber or P. virgatus. To investigate components of the PE hypothesis, we compared these performance measures to those of the generalist Es. acrea, testing the prediction that differences in host plant defenses will result in greater variation in generalist preference and performance as compared to the specialist.

MATERIALS AND METHODS

Study System and Field Sites

Plants: Penstemon glaber var. alpinus and P. virgatus are herbaceous, long-lived perennials native to Colorado and the southwestern region of the United States of America

(Shonle et al., 2004). Both species grow in mountain meadows and road cuts and appear to prefer xeric, rocky or sandy soils (Quintero and Bowers, 2013). Penstemon virgatus (upright blue beardtongue) features narrow leaves, tall stems and inflorescences with several small to medium purple flowers that bloom from June to August (Crosswhite, 1967; Quintero and Bowers,

2013). Previous research found two major IGs in P. virgatus: catalpol and scutellarioside-II

(L’Empereur and Stermitz, 1990). Penstemon glaber var. alpinus (alpine sawsepal penstemon) is one of four known varieties of P. glaber and has broad and occasionally puberulent leaves, thick stems and inflorescences with blue or blue-purple flowers. Although the defensive chemistry of P. virgatus has been previously examined (L’Empereur and Stermitz, 1990;

Quintero and Bowers, 2013), the chemical defences of P. glaber are less well known (Stermitz et al., 1994). There is only one other published study on the chemistry of P. glaber (Stermitz et al.,

1994), which found the IGs catalpol, lamiide, globularin, and durantoside-I. However, those individuals were collected from the northern most part of the range in central Wyoming and the variety of P. glaber was not specified. Preliminary analyses with Colorado populations in recent years suggest that P. glaber var. alpinus from our field site contains large amounts of catalpol but not other IGs (M.D. Bowers, unpublished data). Both species naturally co-occur at the field site used in this study. All plants used in experiments were obtained from a local nursery and were maintained in one-gallon pots outside of a greenhouse until used in experiments.

Insects: Euphydryas anicia occurs throughout the western half of the United States, including the mountains of Colorado (White, 1979; Cullenward et al., 1979; Ferris and Brown,

1981). Females lay large egg masses (50-200 eggs) on the underside of leaves. Larvae are gregarious in early instars and form a web on their host plant. They enter diapause in the fourth instar, and then overwinter in this stage, emerging in spring to complete development. At our field site, Crescent Meadows (El Dorado State Park, Boulder County, Colorado, USA; 39° 55'

51.60" N 105° 20' 16.80" W, elevation 2258m) adults are typically found in late June through early or mid July. Euphydryas anicia is known to sequester the IGs catalpol, macfadienoside, and aucubin (Gardner and Stermitz, 1988; L’Empereur, 1989; L’Empereur and Stermitz, 1990a), but there is currently no evidence that they sequester scutellarioside. Analyses to be presented elsewhere showed that larvae only sequester catalpol when reared on either P. glaber or P. virgatus (Kelly, unpublished data). Depending on the location and host plant, Eu. anicia can sequester 7.1-18.0% dry weight IGs as larvae and 0.5-10.4% dry weight IGs as adults (Gardner and Stermitz, 1988; L’Empereur, 1989; L’Empereur and Stermitz, 1990a). Sequestered IGs are retained through metamorphosis and both larvae and adults are aposematically coloured

(Stermitz et al., 1986; Gardner and Stermitz, 1988). Larvae used in these experiments were reared from eggs collected at Crescent Meadows in Eldorado Canyon State Park and were maintained as separate family groups in a growth chamber at the University of Colorado.

The salt marsh moth, Estigmene acrea, is a conspicuous grazing generalist found throughout the United States, including Colorado (Bernays et al., 2004). Populations have been recorded feeding on at least 88 plant species in 33 families, and are pests on many crops (Singer et al., 2004). They can tolerate feeding on IG-rich plants (Bowers, 2009; Lampert and Bowers,

2010). Females lay 400-1000 eggs in one or more clusters. Larvae complete 5-7 instars and

20 typically overwinter as mature larvae. They can complete between one and three generations per year, depending on location. Larvae are poor sequesterers of IGs (Lampert and Bowers, 2010).

Larvae used in these experiments were from a laboratory colony maintained at the University of

Colorado. The colony was started with eggs and larvae collected in Austin, Texas, USA. The colony was maintained on dandelion (Taraxacum spp.) leaves collected in Boulder, Colorado.

Field Site: The field experiment was conducted at the Crescent Meadows site. This location features a large population of the IG specialist, Euphydryas anicia, and several of its potential host plants. Multiple Penstemon species, including P. virgatus and P. glaber, co-occur at this site and Eu. anicia uses both species as host plants at this site (C.A. Kelly, personal observation).

Common Garden Site: The common garden experiment took place in a 14 x 12m plot with sandy soil located in Longmont, Colorado (40° 9´ 24.29" N 105° 12´ 11.87" W, elevation

1566m). This site was chosen because it is outside of the range of herbivores naturally found on

Penstemon, which allowed us to examine the interaction of naïve herbivores with the two

Penstemon species grown under identical conditions.

Common Garden Experiment--Chemistry of P. glaber and P. virgatus and preference of generalists

To examine the IG content of both plant species and the host preference of naïve, generalist herbivores, we planted 30 Penstemon glaber and 30 P. virgatus, in a common garden in Longmont, Colorado in May 2010. The plot was rototilled prior to planting. All 60 plants were equally spaced 1.5 m apart in a randomly assigned order and then mulched. The garden was weeded and watered regularly. Plants were allowed to grow uninhibited and free of

21 pesticide until late August 2010, approximately 15 weeks. Herbivores observed on the plants were noted throughout the season and collected when possible. There are no known IG specialists that occur at the garden site and none were observed in the garden. Therefore, generalist herbivores were presumed to be the source of all damage. At the end of the growing period, the proportion of leaf material missing due to herbivore consumption was calculated according to methods in Stamp and Bowers (1996). Leaves were then harvested and dried for chemical analysis.

Field Experiment--Oviposition choice by the specialist, Eu. anicia

Twenty pairs of potted P. glaber and P. virgatus plants were placed in sites at Crescent

Meadows where Eu. anicia had been observed flying. Each P. glaber was paired with a similarly sized P. virgatus. Plants were checked for egg masses every other day. If a plant contained one or more egg masses, the leaves containing eggs were removed and the same number of leaves were also removed from its partner plant, so as not to bias female choice. Each egg mass was considered a single choice by a female. Plants remained at the field site for about

3.5 weeks, until adult females could no longer be found, after which the total number of choices was tallied for each Penstemon species. Plants were watered by hand every other day. Leaves containing eggs were kept in a growth chamber (Percival model LLVL, 25° C day: 20° C night,

14 hour day length) until the eggs hatched. The oviposition preference of Estigmene acrea was not tested since females are known to lay eggs on non-plant surfaces (Castrejon et al., 2012).

Laboratory Experiments--Host plant choice and suitability for a generalist and a specialist

Food plant preference was determined by presenting groups of 10 newly hatched

Estigmene acrea (n = 16) and 10 newly hatched Euphydryas anicia (n = 26) with two leaf discs

5mm in diameter, one P. virgatus and one P. glaber, for 24 hours. Groups of larvae were used because these species are gregarious in early instars (White 1979; Singer et al. 2004). At the beginning of the experiment caterpillars had not previously been exposed to any food. Each trial was conducted inside a rectangular plastic 5x3 cm container. For half of the trials, P. virgatus discs were placed on the left and P. glaber were placed on the right; switching the locations for the other half of the trials. Groups of caterpillars were placed on one end of the container approximately two centimeters and equidistant from the leaf discs placed at the opposite end.

After the 24-hour feeding period, the leaf discs were photocopied, magnified 400x, and the amount of each leaf sample missing was quantified by hand using a sampling grid.

To compare the growth of the generalist and specialist caterpillars on the two Penstemon species, groups of 10 newly hatched first instar larvae were fed either P. glaber or P. virgatus ad libitum. There were 11 groups of Eu. anicia larvae and 20 groups of Es. acrea larvae reared on each of the two Penstemon species. After six days, groups of larvae were weighed to determine the relative growth rate (mg/mg/day) and the number of living caterpillars in each group was counted to determine survivorship. Since Eu. anicia larvae enter diapause upon reaching the fourth instar, the experiment could not continue beyond the initial six days as some larvae began to diapause on day seven. The intrinsic growth rates of Es. acrea and Eu. anicia are quite different, thus we did not statistically compare them.

Chemical Analyses

Plant iridoid glycoside content was analysed by gas chromatography (GC; Gardner and

Stermitz, 1988; Bowers and Collinge, 1992; Bowers and Stamp, 1992; Fajer et al., 1992). Leaf samples were oven dried at 50°C and ground to a fine powder in a mortar. To prepare material for analysis, each sample was extracted overnight in methanol. The solid material was filtered out, and the extract evaporated to dryness. An internal standard, phenyl β-D-glucopyranoside

(PBG) (0.500mg), was added to each sample. The sample was partitioned between water and ether. The ether fraction, which contains lipophilic substances, was discarded, and the water fraction, containing primarily the iridoid glycosides and sugars, was evaporated to dryness. An aliquot of this was derivatized using Tri-Sil Z (Thermo-Fisher Chemical Company), prior to injection onto an HP 7890A gas chromatograph (Agilent Technology) using an Agilent DB-1 column (30 m, 0.320 mm, 0.25-mm particle size; Gardner and Stermitz, 1988; Bowers and

Collinge, 1992; Fajer et al., 1992). The GC was calibrated using standards of purified catalpol and scutellarioside II (hereafter scutellarioside). Amounts of catalpol and scutellarioside were quantified using ChemStation B-03-01 software and data were analyzed as percent of dry mass

(concentration).

Statistical Analyses

All statistical analyses and figures were done using R software (R-Development-Core-

Team, 2011). Welch’s Two Sample t-tests were used to compare common garden herbivory and chemistry data on the two plant species (the main effect was plant species and the dependent variable was either leaf damage or IG content). Model I regressions were used to determine the relationship between the IG content of both Penstemon species and the amount of plant damage.

Oviposition data from the field experiment were analyzed with a chi-square test comparing the number of egg masses on each host plant species. Laboratory food choice data were analyzed

24 with paired t-tests comparing the area of the leaf disc eaten for each plant species. The growth rate and survivorship data were analyzed with Welch’s Two Sample t-tests comparing either the relative growth rate or proportion of survivors (out of the 10 larvae initially put on each plant species) on the two host plant species.

RESULTS

Common Garden Experiment--Chemistry of P. glaber and P. virgatus and preference of generalists

Chemical analyses of common garden plants showed that P. glaber indeed contained catalpol (verified by retention times and co-injection with a standard of catalpol) and no other

IGs were detected. Penstemon glaber leaves contained significantly more catalpol than P. virgatus leaves (t = 18.64, df = 20.23, P < 0.001; Fig 2.1), with an average of 18.12% dry weight.

This was within the range seen previously in plants from Crescent Meadows (17 - 23% dry weight catalpol in 2007; M.D. Bowers unpublished data), but was more concentrated than past studies have found in other populations of P. glaber (Stermitz et al., 1994). Penstemon virgatus leaves from these common garden plants contained two different IGs, catalpol and scutellarioside, as well as a significantly higher overall IG concentration (average of 30.56% dry weight; t = -9.55, df = 38.86, P < 0.001; Fig 2.1), largely due to the high scutellarioside content

(average of 29.10% dry weight), which is not found in P. glaber. The IG content of P. virgatus fell within the range of variation recently seen in natural populations, including Crescent

Meadows (14-32% dry weight scutellarioside, 1-5% dry weight catalpol; C. A. Kelly. unpublished data). However, the P. virgatus in this study contained noticeably higher amounts of IGs than previous work in other populations (L’Empereur and Stermitz, 1990b) or with plants

25 in earlier developmental stages (Quintero and Bowers, 2013). Thus, from the perspective of a generalist, P. virgatus is considered the more defended (less palatable) plant because it contains both larger amounts of IGs and a higher diversity of IGs.

In the common garden experiment, Penstemon glaber leaves received significantly more damage by generalist herbivores than P. virgatus (t = 10.88, df = 29.09, P < 0.001; Fig 2.2).

Unidentified species of Phyllotreta (Coleoptera: Chrysomelidae) were the most commonly observed herbivore in the garden. This result provided ecologically relevant support for our designation of the relative palatability of these two host plants, as the less chemically defended plant species was the preferred choice of generalists. The catalpol concentration of P. glaber showed a marginally significant negative linear association with the amount of damage (F1,18 =

4.082, r2 = 0.1396, P = 0.0585). However, P. virgatus showed no significant linear associations

2 between IG content and amount of damage (total IGs: F1,18 = 0.010, r = 0.055, P = 0.922;

2 2 catalpol only: F1,18 = 1.808, r = 0.041, P = 0.196; scutellarioside only: F1,18 = 0.103, r = 0.050, P

= 0.753).

Field Experiment--Oviposition choice by the specialist, Eu. anicia

Females of the specialist, Euphydryas anicia, laid significantly more egg masses on leaves of P. glaber (41 masses) than on P. virgatus (two masses) (χ2 = 35.37, df = 1, P < 0.001).

Although in July 2010, the existing population of P. virgatus at Crescent Meadows appeared to be larger than that of P. glaber, there were very few naturally occurring P. virgatus plants found with egg masses (C. A. Kelly, personal observation). Conversely, nearly every P. glaber observed at Crescent Meadows had at least one egg mass.

Laboratory Experiments--Host plant choice and suitability for a generalist and a specialist

Larvae of the specialist and generalist behaved differently in the food choice experiments.

Larvae of the generalist, Estigmene acrea, preferred to eat P. glaber, consuming a significantly larger proportion of the leaf disk (mean ± standard error: P. glaber = 0.117 ± 0.015, P. virgatus =

0.0 ± 0.0; paired t = -7.45, df = 15, P < 0.01). In contrast, larvae of the specialist Euphydryas anicia preferred to consume P. virgatus (mean ± standard error: P. glaber = 0.043 ± 0.011, P. virgatus = 0.101 ± 0.015; paired t = -2.35, df = 25, P = 0.027).

As expected given their food preference, Es. acrea larvae had significantly higher growth rates when fed P. glaber compared to P. virgatus in the no-choice feeding experiment (t = -8.03, df = 23.68, P < 0.001; Fig. 2.3a). In contrast, although Euphydryas anicia preferred P. virgatus, there was no difference in growth rate when reared on either plant species (t = -0.68, df = 10.47,

P = 0.51; Fig. 2.3b). Consistent with the growth rate data, survivorship by day six for Es. acrea was significantly higher on P. glaber than on P. virgatus (t = -2.64, df = 30.29, P = 0.013; Fig.

2.3c), whereas survivorship for Eu. anicia was similar for both host plant species (t = -0.80, df =

16.65, P = 0.434; Fig. 2.3d).

35 Scutellarioside Catalpol 30

Percent Dry Weight Percent 10

0 P. glaber P. virgatus

Figure 2.1 The iridoid glycoside (IG) composition of Penstemon glaber and P. virgatus leaves from plants reared in the common garden. Means ± s.e.

0.35

0.30

0.25

0.20

0.15

0.10

consumed of tissue leaf Proportion 0.05

0.00 P. glaber P. virgatus

Estigmene acrea Euphydryas anicia 2.5 a 0.5 b ) 1 − 2.0 0.4 day 1 − mg

mg mg 1.5 0.3 (

1.0 0.2

0.5 0.1 Relative Growth Rate Growth Rate Relative

0.0 0.0

1.0 c d

0.8

0.6

0.4

0.2 Proportion of Survivors at 6 of Day Survivors Proportion

0.0 P. glaber P. virgatus P. glaber P. virgatus

DISCUSSION

This study provides mixed support for our initial hypothesis that specialist herbivores will have a strong preference-performance relationship and will also outperform generalists on host plant species preferred by the specialists. Penstemon glaber had a lower concentration of total

IGs and fewer IG types than P. virgatus. Our data suggest that these chemical differences may play an important role in driving host plant preferences in the herbivores Eu. anicia and Es. acrea. Both the common garden and laboratory experiments indicate that generalist herbivores prefer P. glaber to P. virgatus. The physiological efficiency hypothesis predicts that generalist herbivores will be more sensitive to variation in plant defenses than specialist herbivores. Indeed, the performance of the generalist, Es. acrea was notably decreased when reared on P. virgatus

(the more defended host), whereas specialist performance did not vary with the host plant species.

In contrast, support for components of the preference-performance hypothesis was lacking.

Ovipositing females of the specialist Eu. anicia demonstrated a preference for P. glaber; however, larvae performed equally well on the two host plants. Furthermore, specialist offspring in the laboratory preferred to feed on a different host species than the mothers preferred for oviposition in the field. Thus, we found no clear relationship between specialist female preference and offspring performance.

That host plant preference differed between parent and offspring of the IG specialist, Eu. anicia, was unexpected, given that female herbivores, particularly those with more specialized diets, are predicted to preferentially oviposit on host plants that enable their offspring to attain the highest performance (reviewed in Gripenberg et al., 2010). Although the preference- performance relationship of Eu. anicia was not negative, neither do the results suggest a strong positive relationship for this specialist. These results contrast with Jaenike’s (1978) prediction

31 that in populations where females prefer to oviposit on one host species despite the presence of other suitable hosts, the preferred host species will be most suitable for larval development.

Since there was no difference in Eu. anicia performance on the two plants in the laboratory, it is possible that maternal host plant preferences may be influenced by other factors, such as host plant nutritional quality. Non-chemical differences between these Penstemon species, including leaf structure, plant size, and visual detectability (Singer; 2004, Reudler Talsma et al., 2008), may also drive host plant choice in the field. Furthermore, the preference-performance (P-P) hypothesis assumes that larvae are unable to disperse to other plants (Clark et al., 2011), but many caterpillar taxa, including checkerspots, are able to relocate to other host plants (e.g. Cain et al., 1985). Host plant availability and enemy free space are also important components of host plant choice (Fraenkel, 1953; Jeffries and Lawton 1984; Thompson, 1988; Denno et al., 1990;

Osier and Lindroth, 2001; Holton et al., 2003; Murphy, 2004; Wiklund and Friberg, 2008) and may be factors in our system as well. Larvae preferred P. virgatus in the laboratory, an environment devoid of natural enemies. Adult females showed a strong preference to oviposit on P. glaber in their natural habitat, which likely includes predators and parasitoids. Thus, larval fitness may be higher on P. glaber when the threat of natural enemies is present.

Consistent with the predictions of the PE hypothesis (Cornell and Hawkins, 2003;

Mooney et al., 2012), Estigmene acrea demonstrated inferior performance on the more defended host (defined by iridoid glycoside amount and diversity), whereas Euphydryas anicia performed similarly on the two Penstemon species. The PE hypothesis suggests that the benefits of dietary specialization, such as faster larval development and increased survival rates, should be more pronounced on more highly defended host plants. Differences in host plant defenses were not reflected in specialist performance on the two plant species, yet, as predicted, noticeably affected

32 the generalist’s survivorship and performance. What we did not predict, however, was that the specialist would prefer P. virgatus, yet grow equally well on both plant species. Though performance differences may only be detectable in post-diapause larvae or later life stages, the focus of these experiments was female oviposition and early instar larval performance.

Additionally, caterpillars in these experiments were coping with extremely high levels of iridoid glycosides in the plants on which they were feeding. Data from a different experiment showed that Eu. anicia larvae contained only catalpol when reared on P. virgatus, and in relatively large amounts (mean = 8.83 ± 0.64% dry weight; Kelly unpublished data), while P. virgatus contains only a mean of 1.45% catalpol. This suggests that the larvae may be metabolically converting scutellarioside into catalpol by hydrolyzing the side chain. A previous study proposed that Eu. anicia is capable of metabolically converting the catalpol ester, 6-isovanillylcatalpol, into catalpol via hydrolysis, and then sequestering the catalpol (Gardner and Stermitz, 1988). A similar conversion of scutellarioside may be occurring here. Such a conversion could incur a metabolic cost; thus it was unexpected that larvae performed similarly on the two Penstemon species. Perhaps metabolizing or eliminating scutellarioside incurs a fitness cost to Eu. anicia that is not apparent with our chosen metrics, which could partially explain the female preference for P. glaber. Likewise, compensatory feeding could have allowed Eu. anicia to overcome any potential host plant deficiencies, resulting in no measurable difference in performance on the two hosts.

In concordance with our results, previous studies have also produced mixed support for the PE and P-P hypotheses (e.g. Friberg et al., 2015). Although several studies have failed to find a strong relationship between female preference and larval survival and performance

(reviewed in Futuyma, 2008; Friberg and Wiklund, 2009), field experiments (in the presence of a

33 third trophic level, natural enemies) often show tighter preference-performance relationships

(Damman and Feeny 1988; Doak et al. 2006; Wiklund and Friberg 2008). Yet, a meta-analysis found support for the P-P hypothesis in a strictly bi-trophic setting (Gripenberg et al., 2010).

Our study features a combination of field and laboratory studies, with female preference tested in presence of natural enemies and larval preference and performance tested in the absence of natural enemies. Our results, together with those of past studies, seem to emphasize the need for different sets of predictions within the P-P hypothesis: one for tri-trophic environments and another for bi-trophic. Perhaps the traditional specialist-generalist paradigm confounds the predictions of the P-P and PE hypotheses, and thus highlights the need for reconsideration of the classifications of herbivore diet breadth.

The goal of understanding the ecology and evolution of diet breadth in herbivorous insects has resulted in the development of several hypotheses that attempt to explain the observed patterns and processes. Here we combined components of both the PE and P-P hypotheses to test the hypothesis that specialist herbivores will show a strong preference- performance relationship and will also outperform generalists on host plant(s) preferred by the specialists. The results of our experiments only partially agree with these predictions. Given the increasing amount of mixed support for these seminal hypotheses, considered both separately and in combination, we suggest that consideration of the third trophic level may be important in refining them (e.g. Mooney et al., 2012). Preliminary tests of such a unified hypothesis, the Tri-

Trophic Interactions (TTI) hypothesis, showed that the interaction between herbivore diet breadth and plant quality may be altered by the presence of natural enemies (Mooney et al.,

2012). Thus, the future development of theories that better address the complexities of trophic interactions may be in the synthesis of these foundational hypotheses.

CHAPTER 3

THE UNPREDICTABLE PENSTEMON: VARIATION IN DEFENSIVE CHEMISTRY ACROSS YEARS, POPULATIONS, AND TISSUES

ABSTRACT

Plants produce a variety of secondary metabolites that function as a defense against their natural enemies. Production of these secondary metabolites is genetically controlled, but is also phenotypically plastic and varies in response to both biotic and abiotic factors. Therefore, plant species may vary widely in their chemical defenses and such variation can be evident at temporal, spatial and tissue levels. Focusing on the chemical defenses of a native Colorado wildflower,

Penstemon virgatus, we assess variation in iridoid glycoside content across two growing seasons, six natural populations and three tissue types: leaves, stems and flowers. Our results indicate that

P. virgatus plants contain high concentrations of iridoid glycosides (mean = 23.36% dry weight of leaves). Leaves contained the highest concentration of iridoid glycosides and varied between sampling years, among plant populations, and plant parts. Results also indicated that IGs were differentially allocated among tissue types. We also quantified leaf herbivore damage at all six populations but we found very little herbivore damage. Our study indicates the iridoid glycoside concentrations of P. virgatus plants are both spatially and temporally variable and that their high concentrations of secondary metabolites is an effective defense against generalist herbivores.

INTRODUCTION

Plants produce a vast array of secondary metabolites that can improve plant fitness by deterring plant enemies, including both herbivores and pathogens (Fraenkel 1953, Odum and

Pinkerton 1955, Whittaker and Feeny 1971, Karban and Baldwin 1997, Wink 2010, Iason et al.

2012). The production of these compounds is likely the outcome of coevolutionary interactions with herbivores (Ehrlich and Raven 1964). For herbivores, determination of host plant suitability is often due (at least in part) to the amounts and kinds of secondary plant metabolites (Fraenkel

1959, Scriber 1984, Zangerl and Berenbaum 1993, Denno 2012). As a result of both evolutionary interactions with other organisms and plastic responses to the biotic and abiotic environment, plant populations exhibit extensive variation in secondary metabolite content. These chemical phenotypes are important in structuring communities of herbivores, and such variation is integral to understanding the landscape in which host plants and herbivores interact.

Variation in plant secondary metabolites depends on several factors (reviewed in Moore et al. 2014), including but not limited to: ontogeny (Quintero and Bowers 2011), genotype

(Gouyon et al., 1983; Berenbaum et al., 1986; Lincoln et al., 1986; Mihaliak et al., 1989), the presence of herbivores (Rios et al. 2008) and various components of the abiotic environment, such as water and nutrient availability (Waterman and Mole 1989). Given that intraspecific host plant variation has significant ecological implications for multiple trophic levels, it is particularly important to understand the relative contributions of these multiple factors in natural populations.

This study focuses on three potential sources of variation in secondary metabolites in a long- lived herbaceous perennial: inter-annual (across years), location (among geographically distinct populations) and plant tissue type.

By definition, herbaceous perennial plants have a life cycle lasting longer than one year, with many perennials living for several years (Harper 1977). Although perennials are ubiquitous, there are surprisingly little data on the temporal variation in plant quality (i.e. variation across years) of natural populations of long-lived herbaceous plants. Many longer-term studies of plant quality examined woody plants (e.g. Nicholsorians 1991, Scogings et al. 2004, Laitinen et al.

2005), or conducted greenhouse experiments (e.g. Gols et al. 2007). Abiotic factors, such as light and temperature, and biotic factors, such as the presence of pathogens and herbivores, may change dramatically from year to year. Similarly, the secondary metabolite content of perennials may also vary considerably during their multi-year life span. As a result, different generations of herbivores may also be affected as the host plant landscape changes from one year to the next. In addition, many herbivores have life cycles that span more than one year; for example, many checkerspot butterflies (Melitaeini) have oviposition and early larval development in one year, an over-winter diapause, and complete their development the following year (Murphy et al.

2004). Thus, ovipositing females and early instar larvae may interact with host plants that are chemically different than the host plants on which post-diapause larvae depend. An herbivore may find a host plant palatable one year, but unpalatable the next. For herbivore species that have the ability to sequester plant secondary metabolites, an individual’s palatability may vary along with chemical content of the host plant. Thus, secondary metabolite variation has implications for interactions with higher trophic levels (e.g. Gols et al. 2008).

Chemical differences between plant populations are caused, at least in part, by variation in the biotic and abiotic features of the local environment. Additionally, population differences in plant secondary metabolites may be partially controlled by genetic (Berenbaum and Zangerl

1992), and developmental (Bowers and Stamp 1993, Quintero and Bowers 2012) sources of variation. This variation can be expressed at quite different spatial scales. Populations separated by only a few kilometers can show wide variation in the make-up of their secondary metabolite phenotype (e.g. Moyes et al. 2000, Newton et al. 2009a, b), whereas populations separated by over 1000km may not show any significant variation (Gols et al. 2009). Such variation, and the factors underlying its control, in plant secondary chemistry can be particularly important for

37 highly mobile herbivores that may interact with several populations of a given plant species.

Much of this spatial variation can be attributed to local adaptation or phenotypic plasticity of genotypes in response to changes in resources availability (reviewed in Moore et al. 2014). Thus, there is ample support that secondary metabolites are influenced by their environment, but not always in predictable ways (Endara and Coley 2011).

Within a single population, significant variation can be observed among as well as within individual plants. Secondary metabolites are not evenly distributed across plant tissues or organs.

The optimal defense theory (ODT; also known as the tissue value hypothesis) was developed to explain the distribution of defensive chemicals within a single plant (McKey 1974, Feeny 1991), and is particularly useful for predicting intra-plant distributions of secondary compounds (e.g.

Zangerl and Rutledge 1996). One of the main components of the ODT is that the allocation of defensive compounds within a plant is a function of tissue or organ value in terms of fitness, because secondary metabolites are costly to produce (McKey 1974, Koricheva 2002, Stamp 2003,

McCall and Fordyce 2010). Under the ODT, flowers are sometimes predicted be more valuable than leaves because they are more directly related to sexual reproduction (Strauss, Irwin and

Lambrix 2004). Similarly, new leaves are expected to be higher in levels of secondary compounds than older leaves (McKey 1979, Krischik and Denno 1983). Such intra-individual plant variation can significantly affect the fitness of its associated herbivores (Mody et al. 2007).

Herbivores that feed on multiple plant parts (e.g. Calophasia lunula, which feeds on the leaves and flowers of Dalmatian toadflax, Linaria dalmatica (Plantaginaceae) and Utetheisa ornatrix, which feeds on the leaves and seeds of Crotolaria spectabilis (Fabaceae)) may experience substantial differences in both the quantity and composition of defensive chemicals in these two different tissues (Wilson et al. 2005, Ferro et al. 2006, Martins et al. 2015, Sourakov 2015).

Our goal is to investigate potential sources of chemical variation in a long-lived perennial plant, so as to better understand the chemical landscape in which herbivores interact with this host plant. We examined intraspecific variation in the secondary metabolite content of the perennial Penstemon virgatus A. Gray (Plantaginaceae) as a function of year, population and tissue type. Penstemon virgatus is chemically defended by iridoid glycosides (IGs), a group of cyclopentanoid monoterpene-derived compounds found in over 50 plant families (Bobbitt and

Segebarth 1969, Jensen et al. 1975, El-Naggar and Beal 1980, Boros and Stermitz 1990). We determined how the concentrations of the major IGs found in P. virgatus, catalpol and scutellarioside, differ among and within a) two growing seasons, b) six Colorado populations, and c) three plant tissues, namely leaves, flowers and stems. We also assessed possible relationships between IG content and palatability to herbivores by quantifying herbivore damage in the field.

METHODS AND MATERIALS

Study Species Plants of the genus Penstemon are long-lived herbaceous perennials native to North

America, including higher elevation locations throughout Colorado. Penstemon virgatus grows in rocky or sandy soils in mountain meadows throughout Colorado and other southwestern states

(Shonle et al. 2004, Quintero and Bowers 2013). Also known as the upright blue beardtongue, P. virgatus features narrow leaves, tall stems and long inflorescences with several small to medium purple flowers that bloom from June to August (Crosswhite 1967; Quintero and Bowers 2013).

Previous research found two major IGs in P. virgatus: catalpol and scutellarioside-II (henceforth referred to as scutellarioside; L’Empereur and Stermitz 1990). Documented herbivores include

Euphydryas anicia, Polydryas arachnae, Polydryas minuta (all family Nymphalidae) and Meris

39 alticola (Geometridae) (Stermitz et al. 1988, L’Empereur and Stermitz 1990b, Robinson et al.

2002).

Collection Procedure We visited six known locations of P. virgatus throughout the Colorado Front Range and eastern Rocky Mountains (Table 1) in July 2011 and July 2013. In both years, at each location,

15 – 20 individual flowering plants were collected. Though P. virgatus is a long-lived perennial, individual plants may not return the following year. As such, we sampled different individual plants each sampling year. Plants were haphazardly chosen at each site, but were only considered if they were in bloom. If an individual plant had multiple flowering stalks, only one stalk was sampled and removed. In 2013, herbivory was quantified for those plants that were collected (see below). After collection, the above-ground tissue was separated into leaves, flowers, and stems and these different tissues analyzed for iridoid glycosides (see below).

Permits were obtained to collect plants within Rocky Mountain National Park (ROMO-2011-

SCI-0055, ROMO-2013-SCI-0046) and Eldorado Canyon State Park.

Herbivory Herbivory data were only collected in the 2013 season. Prior to harvesting individual plants for chemical analysis, we quantified herbivore damage in the field using a method similar to that of Stamp and Bowers (1996). The length of each leaf of every plant sampled was measured to the nearest millimeter and then visually examined for evidence of herbivore damage.

The amount of damage on each leaf was categorized as: 0, 1-5, 6-10, 11- 25, 26-50% of leaf tissue missing. As there was never more than 50% of leaf tissue missing due to herbivory, we did not need to include higher damage categories. The midpoint of each damage category (3%,

8%, etc.) was multiplied by leaf length and the values for all leaves on a plant were added to

40 provide an index of herbivory, or the total amount eaten for each plant. This index was then divided by the sum of the leaf lengths to calculate the proportion of leaf tissue eaten per plant.

Chemical Analysis We quantified the concentration of the IGs catalpol and scutellarioside for leaves, flowers and stems via gas chromatography (Gardner and Stermitz 1988, Bowers and Collinge 1992,

Bowers and Stamp 1992, Kelly and Bowers 2016). Plants were oven dried at 50°C and then divided into three tissue types (leaves, flowers and stems), which were extracted and analyzed separately. Samples were ground to a fine powder in a mortar. To prepare material for analysis, each sample was extracted overnight in methanol. The solid material was filtered out, and the extract evaporated to dryness. An internal standard, phenyl β-D-glucopyranoside (0.500mg), was added to each sample. The sample was partitioned between water and ether. The ether fraction, which contains lipophilic substances, was discarded, and the water fraction containing primarily the iridoid glycosides and sugars was evaporated to dryness. An aliquot of this was derivatized using Tri-Sil Z (Pierce Chemical Company), prior to injection onto the gas chromatograph

(Gardner and Stermitz, 1988; Bowers and Collinge, 1992; Fajer et al., 1992).

Statistical Analysis Iridoid glycoside (IG) content was analyzed as a concentration, specifically percent of dry sample weight. All proportion data were linearized with a logit transformation to meet assumptions of normality. Given that catalpol and scutellarioside were correlated (Pearson’s correlation coefficient = 0.28; P < 0.0001), we used a three-way multivariate analysis of variance

(MANOVA) to assess variation in IG concentration due to plant part (leaves, flowers or stems), population and year. When the MANOVA detected significant effects, we followed up with univariate, ANOVAs for each dependent variable (catalpol and scutellarioside). We used

41 repeated-measures ANOVAs for analyses that included tissue type (the repeated measure) as a main effect. For all univariate ANOVAs, Tukey’s post hoc multiple comparisons tests were used to examine pairwise differences among groups when significant differences were detected. We used a Kruskal-Wallis test to compare herbivore damage across all six populations and also to compare herbivory at the Crescent Meadows population against the other five populations, as

Crescent Meadows was the only population where we directly observed herbivory. We assessed the relationship between herbivory and leaf chemistry for the 2013 season with Pearson product- moment correlations for all populations combined and for Crescent Meadows alone. All statistical analyses were performed in R version 3.1.2.

RESULTS

The total iridoid glycoside concentration (catalpol and scutellarioside) of leaves was very high with an average of 22.77% dry weight (± 0.53 SE). Leaves had over 4.5 times more scutellarioside than catalpol. The total IG concentration of flowers ranged from 48.07% dry weight to 1.21% dry weight (mean = 16.08 ± 0.43 SE), and contained approximately equal concentrations of catalpol and scutellarioside. Leaves contained more than twice as much scutellarioside than flowers (scutellarioside: mean leaves = 18.76 ± 0.49 SE; mean flowers =

8.38 ± 0.39 SE); yet, flowers had nearly twice as much catalpol than leaves (catalpol: mean flowers = 7.68 ± 0.18 SE; mean leaves = 4.01 ± 0.14 SE). Stems contained comparatively low concentrations of total IGs (mean = 5.39 ± 0.35 SE).

Results of the MANOVA indicated a significant three-way interaction between plant part, population and year (Wilks’ λ = 0.878; F20, 1382 = 4.65, P < 0.001). Furthermore, all main effects and two-way interactions were significant (P < 0.01 for all). Based on the results of this

MANOVA, we proceeded with separate univariate ANOVAs for catalpol and scutellarioside.

Three-way ANOVAs, with tissue type as the repeated measure, revealed significant three-way interactions between tissue type, population and year for both scutellarioside (F10, 689 = 6.54, P <

0.001) and catalpol (F10, 689 = 3.05, P < 0.001; Table 3.2). Univariate ANOVAs for each tissue type indicated that catalpol and scutellarioside varied in their response to the effects of population, year, and the interaction of these two factors (Table 3.3). For all three tissues, there was a significant population by year interaction for scutellarioside. Flowers were the only tissue that showed a significant population by year interaction for catalpol. There was a significant effect of population for catalpol in leaves and stems.

There was a marginally significant difference in the amount of herbivore damage among the six populations (Kruskal–Wallis χ2 = 10.03, df = 5, P = 0.0743). Overall, there was very little herbivore damage on P. virgatus (mean = 2.61% of leaf damaged ± 0.25 SE). Crescent Meadows was the only population where we directly observed the presence of herbivores. The IG specialist checkerspot, Euphydryas anicia, has been documented using P. virgatus as a host plant at this location, although it is not the preferred host plant species (Kelly and Bowers 2016). Although there was no overall effect of population on herbivore damage, a comparison of damage at the

Crescent Meadows population with that of the other populations showed that it had significantly higher levels (Kruskal–Wallis χ2 = 6.27, df = 1, P = 0.0123). There was no significant correlation between herbivore damage and leaf chemistry for all populations combined (Total IGs: Pearson’s correlation coefficient = -0.0685, P = 0.467; Scutellarioside only: Pearson’s correlation coefficient = -0.0416, P = 0.659; Catalpol only: Pearson’s correlation coefficient = -0.110, P =

0.240), nor for the Crescent Meadows population alone (Total IGs: Pearson’s correlation coefficient = 0.277, P = 0.238; Scutellarioside only: Pearson’s correlation coefficient = 0.233, P

= 0.324; Catalpol only: Pearson’s correlation coefficient = 0.320, P = 0.169).

Table 3.1 The six Penstemon virgatus sampling populations in Colorado, USA.

Population Abbr. Location Coordinates Elevation (m)

Boulder, CO - Eldorado 39.931° N Crescent Meadows CM 2258 Canyon State Park 105.338° W

40.150° N Calwood Education Center CAL Jamestown, CO 2371 105.389° W

Estes Park, CO - Rocky 40.401° N Lumpy Ridge Trailhead LRTH 2415 Mountain National Park 105.520° W

Estes Park, CO - Rocky 40.357° N Cub Lake CUB 2481 Mountain National Park 105.619° W

38.713° N Chalk Lake Campground CLC Nathrop, CO 2670 106.232° W

39.363° N Michigan Hill MIH Jefferson, CO 2931 105.837° W

Table 3.2 Summary of three-way, repeated measures ANOVAs that tested the effects of tissue type, population and year and their interactions on the iridoid glycoside content of Penstemon virgatus. The IGs catalpol and scutellarioside were measured as a concentration (% dry weight).

Catalpol Scutellarioside df F P F P Tissue 2 83.75 <0.001 87.41 <0.001

Population 5 19.05 <0.001 15.57 <0.001

Year 1 0.18 0.67 0.62 0.43

Tissue x Population 10 3.36 <0.001 3.96 <0.001

Tissue x Year 2 2.76 0.10 2.29 0.064

Year x Population 5 1.26 <0.001 5.48 0.28

Year x Population x Tissue 10 3.05 <0.001 6.54 <0.001

Error 689

P <0.001 0.99 <0.001

Scutell.

F 16.68 0.00 7.24

P Stems <0.001 0.51 0.079

Catalpol F 15.06 0.44 2.005

1 5 5 df

229

P <0.001 <0.001 <0.001

Scutell.

F 5.37 6.73 14.34

P 0.57 Flowers 0.0039 <0.001

Catalpol F 3.57 0.33 5.48

ANOVAs comparing content type ANOVAs IGfor tissue each comparing

1 5 5 df 230 way way

P 0.50 0.074 0.043

Scutell. F 2.04 0.45 2.33

P 0.95 0.56 Summary two Summary of

Leaves <0.001

Catalpol F 4.70 0.79 0.003 Table Table 3.3 IGs sixtwo Colorado The catalpol and across sampling populations years. and (%weight). drywere a as scutellariosidemeasuredconcentration

1 5 5 df 233

Population Year x Population Year Error

CM CWD LRTH CL CHLK MIH A Population CM CWD LRTH CL CHLK MIH CM CWD LRTH CL CHLK MIH 30

30 Leaves

25 Flowers

25 Stems 20

20 15 %DW %DW scut 15 %DW %DW scut Scutellarioside Percent Dry Weight Dry Percent 10 10 5 5 0

0 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 Collection Year

B CM CWD LRTH Population CL CHLK MIH

CM CWD LRTH CL CHLK MIH 12 CM CWD LRTH CL CHLK MIH 12 Leaves

Flowers

10 Stems 8 8

%DW %DW cat 6 %DW %DW cat Catalpol 4 Percent Dry Weight 4 2 2 0 0 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 2011 2013 Collection Year

0.06

0.05

0.04

0.03

0.02 Proportio of Leaf Tissue Consumed of Tissue Leaf Proportio 0.01

0.00 CM CWD LRTH CL CHLK MIH

Population

DISCUSSION

Although previous studies have examined variation in plant secondary metabolites across different populations and over development (e.g. Brenes-Arguedas and Coley 2005, Jamieson and Bowers 2010, Quintero and Bowers 2012, Cirak et al. 2012, Fortuna et al. 2014), few, if any, have looked at long-lived, herbaceous perennials in the same populations in different years. Here, we show that catalpol and scutellarioside concentrations in P. virgatus plants vary due to tissue type, population and year. Consistent with other studies, we found significant variation across different populations (Darrow and Bowers 1997, Jamieson and Bowers 2010, Cirak et al. 2012,

Alba et al. 2013). Our results indicate that P. virgatus contains extremely high levels of iridoid glycosides and that different plant tissues vary in the relative proportion of catalpol to scutellarioside. Leaves were especially high in scutellarioside, although flowers had the highest concentrations of catalpol. We only observed herbivores using P. virgatus as a host plant at

Crescent Meadows and plants at that location had the most herbivore damage, although the amounts were still quite low. We did not directly observe herbivores on P. virgatus at the other five locations and there was little evidence of herbivore use of those plants.

Our results indicate that P. virgatus plants are both spatially and temporally variable in their IG concentration. Plant demographic studies have monitored long term variation in plant population dynamics (Harper and White 1974, Roach and Gampe 2004), but they have not investigated how chemical defenses have changed over time. Other plant species have shown IG variation across different populations throughout a single growing season (e.g. Darrow and

Bowers 1997, Jamieson and Bowers 2010, Alba et al. 2013), but IG variation had not been measured across multiple years. Longer-term studies, over multiple growing seasons, on the chemical defenses of herbaceous perennial plants are lacking. Penstemon virgatus has a multi-

49 year life cycle; thus year-to-year variation in IG content in P. virgatus could be important for the local herbivore community. In general, research investigating variation in defensive chemistry in herbaceous plants has focused on short-term temporal variation, typically examining changes within a single season (Brenes-Arguedas and Coley 2005, Jamieson and Bowers 2010, Quintero and Bowers 2012, Cirak et al. 2012). Future research should also consider monitoring variation in the defensive chemistry of individual perennial plants over multiple years, as such variation would be ecologically relevant to herbivores with multi-year life spans.

Leaves contained the highest total concentration of IGs and were particularly rich in scutellarioside. These extremely high concentrations are unprecedented for Penstemon virgatus

(L’Empereur and Stermitz 1989, L’Empereur and Stermitz 1990b, Quintero and Bowers 2013).

Flowers, however, were comparatively high in catalpol. Often, the concentration of catalpol in flowers far exceeded that in leaves. Meanwhile, stems were generally IG poor. According to the

Optimal Defense Theory (McKey 1974, Stamp 2003, McCall and Fordyce 2010), reproductive tissues are likely to be higher in constitutive defenses, as they are an essential element of plant fitness. Although it is unknown whether our chemical data represent constitutive defenses, inducible defenses or a mixture of both, the low amounts of visible damage to these plants suggests that we measured constitutive defenses. Leaves were clearly the tissue type with the highest IG concentration, and though they are not reproductive tissues, they are an important component of plant fitness as they are the main photosynthetic organs. Flowers lower IG concentrations than leaves overall, but perhaps higher proportions of catalpol carry an ecological significance. We do not yet understand the relative effectiveness of catalpol and scutellarioside as deterrents or toxins to generalist herbivores; thus it is possible one compound is more effective than the other. Some studies have found distributions of defensive compounds like

50 those found in the leaves and flowers of P. virgatus (e.g. Cirak et al. 2012), while others have found that the secondary metabolite content of leaves and flowers can be quite similar (Jamieson and Bowers 2010).

The high concentrations of IGs in P. virgatus suggest that it is well defended against generalist herbivores; a notion supported by our observation that herbivory was very low across the populations we examined. Earlier research found that P. virgatus is unpalatable to generalist insect herbivores and can reduce generalist larval performance (Kelly and Bowers 2016).

However, given the observed variation among populations and across years, it is possible that in some years, levels of IGs are low enough that generalists can use P. virgatus as a host plant. We did not observe any specialist herbivores, with the exception of the checkerspot, Euphydryas anicia, at the Crescent Meadows population. Previous accounts found checkerspots at the

Michigan Hill population in 2007 (pers. comm. with C. Quintero), but we found no evidence of checkerspots at that location in either year. Euphydryas anicia was observed feeding and ovipositing on P. virgatus at Crescent Meadows in several years, yet has been documented to prefer P. glaber at that location (Kelly and Bowers 2016).

We found no correlation between the IG concentration of leaves and the amount of herbivore damage for all populations. This was unexpected given the high concentrations of IG founds in leaves and the significant variation of IGs across the six populations. Crescent

Meadows is the only population where herbivores were directly observed using P. virgatus as a host plant, and yet there was no correlation between IG concentrations and herbivory when that population was analyzed separately. The exceedingly low amounts of herbivory we observed may be the consequence of other environmental factors and not IG concentration. Alternatively, there may be a baseline IG concentration beyond which herbivores, and generalists in particular,

51 will be completely deterred from consuming the plant. The significant variation in IG concentration may be the result of a bet hedging strategy for this perennial plant. Herbivory can be subject to temporal variation as well, with some generalists demonstrating eruptive life cycle patterns (population densities are low for extended periods of time and then intermittently rise to high densities; e.g. Wallin and Raffa 2004). In low-herbivory years, the baseline IG concentration may be sufficient to protect plants from herbivore damage. However, in high- herbivory years only plants higher concentrations of IGs may survive. It is possible that we sampled low-herbivory years and thus the selective pressures for such variable and high concentrations of IGs were not evident.

Secondary metabolites are necessary for protecting plants from attack by generalist herbivores, yet can also mediate interactions with specialist herbivores. Our study demonstrates that secondary metabolites can vary significantly over years, across populations and within individual plants. Such local adaptation has important implications for communities of herbivores, and, in particular, the specialist herbivores that use them as host plants. This large and often unpredictable variation exerts selective pressure on herbivores to successfully navigate a heterogeneous landscape. Consequently, herbivore performance will vary and, in turn, affect higher trophic levels. Finally, long-term research examining variation in secondary metabolites of herbaceous plants is lacking and future studies should focus on temporal sources of variation in these defensive chemicals.

CHAPTER 4

HOST PLANT IRIDOID GLYCOSIDES MEDIATE HERBIVORE INTERACTIONS WITH NATURAL ENEMIES

ABSTRACT

Many insect herbivores are dietary specialists and some are capable of sequestering the secondary metabolites produced by their host plants. These defensive compounds have important, but complex implications for tri-trophic interactions between plants, herbivores and natural enemies. The sequestration of host plant secondary metabolites defends herbivores from attack by generalist predators, but may also compromise the insect immune response, making insect herbivores more vulnerable to parasitism. Here we investigate the role of plant secondary metabolites in mediating interactions between specialist herbivores and their natural enemies.

We focus on two Penstemon species, P.glaber and P. virgatus, which are chemically defended by iridoid glycosides (IGs). First, we examine how Penstemon iridoid glycoside content influences the sequestration of IGs by a specialist herbivore, Euphydryas anicia. Then we performed ant bioassays and phenoloxidase assays with larvae reared on each host plant diet to assess larval vulnerability to predators and parasitoids. We found that host plant diet affects larval sequestration of IGs as E. anicia reared on P. glaber sequestered more of the IG catalpol.

Yet, ant predators found larvae unpalatable regardless of host plant diet and were also deterred by sugar solutions containing isolated IGs. However, E. anicia larvae reared on P. glaber showed more phenoloxidase activity than larvae reared on P. virgatus. Our results suggest that the sequestration of some secondary metabolites can effectively protect herbivores from predation, yet may also increase vulnerability to parasitism via a decreased immune response.

INTRODUCTION

In tritrophic systems, the bottom-up effect of host plant variation can be important in determining the performance of an herbivore as well as its associated natural enemies (e.g.

Awmack and Leather 2002; Teder and Tammaru 2002, Vogelweith et al. 2011). For instance, highly nutritional plants could improve herbivore performance and consequently increase the prey population available to their natural enemies (Awmack and Leather 2002; Coley, Bateman and Kursar 2006). In contrast, higher host plant quality could strengthen herbivore defenses, and thus negatively affect the performance of natural enemies (e.g. Barbosa et al. 1986, Sime 2002).

Although plant quality can refer to a multitude of factors, plant secondary metabolites are particularly important components of plant quality and play an important role in the evolution of plant-herbivore-natural enemy interactions (Turlings et al. 1990; Dyer 1995; Smilanich et al.

2009a; Smilanich et al 2011b). Plant secondary metabolites improve plant fitness by acting as a defense against herbivores and pathogens, but they are also important for the performance of organisms at higher trophic levels (Hare 2002). Additionally, these compounds have been shown to attract the natural enemies of insect herbivores (Turlings et al. 1990, Dicke and van Loon

2000).

Plant secondary metabolites are not produced in isolation; virtually all plants produce suites of these compounds and some produce different classes of compounds. For example,

Linaria dalmatica, Dalmatian toadflax, produces both terpenoids and alkaloids (Von Gröger and

Johne 1965). These suites of compounds provide the potential for synergistic or antagonistic interactions among individual compounds (Richards et al. 2010, 2012). In this context, synergy is defined as combined antiherbivore effects of different compounds that are greater than expected based on projected additive effects of each individual compound (Berenbaum et al.

1991, Nelson and Kursar 1999). These synergistic effects can include enhanced deterrence or toxicity, and improved or decreased herbivore performance (Dyer et al. 2003, Richards et al.

2010, 2012).

Most insect herbivores are specialists, with a narrow diet breadth that is often limited to host plants containing specific groups of plant secondary metabolites (Bernays and Chapman

1994, Bernays 2001). These compounds may serve as feeding and oviposition cues for specialists (Da Costa and Jones 1971, Raybould and Moyes 2001, Macel and Vrieling 2003,

Nieminen et al. 2003), and some specialist species have the ability to sequester these compounds for their own defense (Dyer and Bowers 1996, Camara 1997a, Theodoratus and Bowers 1999).

Bernays and Graham (1988) hypothesized that the prevalence of herbivore specialization is the result of predators preferring generalist prey, as they are less likely to be chemically defended.

The sequestration of plant secondary metabolites provides an effective defense against many different generalist predators (e.g. Rayor and Munson 2002); however, feeding on highly defended plants can render some insect herbivores more susceptible to parasitoid attack (Dyer et al. 2004, Smilanich et al. 2009a). Parasitoid larvae may be better protected from their natural enemies if they develop within a chemically defended host; those sequestered compounds allow the insect host to provide enemy-free space for the parasitoid (Dyer and Gentry 1999, Gentry and

Dyer 2002, Sznajder and Harvey 2003, Zvereva and Rank 2003, Murphy et al. 2014). Therefore, insect herbivores may experience an evolutionary trade-off between sequestering secondary metabolites for protection against predators, while becoming more vulnerable to parasitism.

Plant secondary metabolites can have bottom-up impacts on the third trophic level. In turn, the third trophic level may impact plant fitness by removing herbivores or changing herbivore behavior (e.g. Turlings et al. 1990, Dicke and van Loon 2000). More specifically, plant

55 secondary metabolites could slow the rate of herbivore development, thereby increasing their susceptibility to natural enemies (Clancy and Price 1987, Turlings and Benrey 1998, Lill and

Marquis 2001). Parasitoids are often the highest source of mortality for lepidopteran larvae

(Hawkins et al. 1997), and the insect immune response is one of the most effective defenses against these enemies (Godfray 1994). However, mounting an immune response can be costly in terms of both growth and reproduction (Ahmed et al. 2002, Freitak et al. 2003, McKean et al.

2008, Honkavaara et al. 2009, Ardia et al. 2012). Additionally, the vulnerability of specialist insects to parasitoid attack implies that the immune response may be weakened by the sequestration of secondary metabolites (Smilanich et al. 2009a). Thus, the sequestration of secondary metabolites by herbivores may result in a trade-off between defense from predators and protection from parasitoids.

This study examines how plant secondary metabolites mediate herbivore interactions of a specialist herbivore with the third trophic level, including components of defense against both predators and parasitoids. We focus on the herbivore, Euphydryas anicia Doubleday

(Lepidoptera: Nymphalidae; the anicia checkerspot), that specializes on plants containing iridoid glycosides (IGs; e.g., Plantaginaceae, Scrophulariaceae, Orbanchaceae, Caprifoliaceae,

Cullenward et al., 1979; White, 1979), and can sequester these compounds (Stermitz et al., 1986;

Gardner and Stermitz, 1988; L’Empereur and Stermitz, 1990a). Iridoid glycosides are a group of monoterpene-derived compounds found in over 50 plant families (Bobbitt and Segebarth 1969,

Jensen et al. 1975, El-Naggar and Beal, 1980; Boros and Stermitz, 1990; Bowers, 1991) and these bitter compounds are known mediators of multi-trophic interactions (e.g., Harvey et al.,

2005; Lampert and Bowers, 2010).

At our study site in Colorado, Euphydryas anicia primarily uses two host plant species,

Penstemon glaber var. alpinus (Torr.) A. Gray (Plantaginaceae) and Penstemon virgatus A. Gray.

To address how plant secondary metabolites may influence both herbivores and herbivore defenses against predators and parasitoids, we ask three questions: 1) How much variation in

IGs do E anicia herbivores encounter? 2) How does variation in Penstemon IGs influence E. anicia sequestration of IGs, and 3) how does the IG content of E. anicia larvae affect defenses against the third trophic level (predators and parasitoids)? To address these questions we: a) compare the IG content of these two Penstemon species; b) quantify sequestration of IGs by E. anicia larvae reared on either P. glaber or P. virgatus, c) compare the ability of larvae reared on these two Penstemon species to deter predatory ants, and d) examine how herbivore diet affects one component of larval defense against parasitoids, the immune response.

MATERIALS AND METHODS

Study system

Plants: Penstemon glaber var. alpinus and P. virgatus are herbaceous, long-lived perennials native to Colorado and the southwestern region of the United States of America

(Shonle et al., 2004). Both species grow in montane meadows and road cuts and feature inflorescences with several small to medium flowers that bloom from June to August (Quintero and Bowers, 2013). Penstemon virgatus (upright blue beardtongue) features narrow leaves, tall stems and purple flowers (Crosswhite, 1967). Penstemon glaber var. alpinus (alpine sawsepal penstemon) has broad and occasionally puberulent leaves, thick stems and blue or blue-purple flowers. Previous research found two major IGs in P. virgatus, catalpol and scutellarioside-II and only one major IG in P. glaber, catalpol (L’Empereur and Stermitz, 1990, Kelly and Bowers

2016). Both species naturally co-occur at the site where caterpillars were collected. We used greenhouse grown plants to control for environmental differences. All plants used in experiments were obtained from a local nursery and were maintained in one-gallon pots outside of a greenhouse.

Insects: Euphydryas anicia occurs throughout the western half of the United States, including the mountains of Colorado (White, 1979; Cullenward et al., 1979; Ferris and Brown,

51.60" N 105° 20' 16.80" W, elevation 2258m) adults are typically found in late June through mid July. Euphydryas anicia is known to sequester the IGs catalpol, macfadienoside, and aucubin (Gardner and Stermitz, 1988; L’Empereur, 1989; L’Empereur and Stermitz, 1990a), but there is currently no evidence that they sequester scutellarioside. Sequestered IGs are retained through metamorphosis and both larvae and adults are aposematically colored (Stermitz et al.,

1986; Gardner and Stermitz, 1988). Kelly and Bowers (2016) found that E. anicia larvae performed equally well when reared on P. glaber and P. virgatus in the laboratory. Larvae used in the experiments present here were reared from eggs collected on Penstemon at Crescent

Meadows and maintained as separate family groups in a growth chamber (Percival model LLVL,

25° C day: 20° C night, 14 hour day length) at the University of Colorado.

Iridoid Glycoside Chemistry

Plants Leaf samples were oven-dried at 50°C and ground to a fine powder in a mortar.

Each sample was then extracted overnight in 95% methanol. Sample purification and quantification of IGs via gas chromatography followed previously described methods (see Kelly and Bowers 2016). An internal standard, phenyl β-D-glucopyranoside (0.500 mg), was added to each sample. The sample was partitioned between water and ether. The ether fraction, which contains lipophilic substances, was discarded, and the water fraction, containing primarily the

IGs and sugars, was evaporated to dryness. An aliquot of this was derivatized using Tri-Sil Z

(Thermo-Fisher Chemical Company, Waltham, Massachusetts), prior to injection onto an HP

7890A gas chromatograph (Agilent Technologies, Santa Clara, California) using an Agilent DB-

1 column (30 m, 0.320 mm, 0.25 mm particle size; Gardner and Stermitz, 1988; Bowers and

Collinge, 1992; Fajer et al., 1992). The gas chromatograph was calibrated using standards of purified catalpol and scutellarioside II (hereafter scutellarioside). Concentrations of catalpol and scutellarioside were quantified using ChemStation B-03-01 software and data were analyzed as percentage of dry mass.

Larvae Upon reaching diapause, groups of E. anicia larvae were freeze killed and later processed for chemical analysis. Sibling groups of whole caterpillars (n = 5 individuals in each family group) were ground inside 15ml test tubes using sand and a glass rod. The mixture was extracted overnight in 95% methanol. The methanol extract was subsequently filtered to remove solid material, and the resulting residue was evaporated to dryness. Sample preparation for larvae was the same as for plant material and larval IG content was determined by gas chromatography in the same manner as the plant tissue (described above and in Kelly and Bowers 2016).

Predator bioassays

To test for deterrence of catalpol and scutellarioside and their potential function as a

59 defense against predators, we conducted bioassays with laboratory ant colonies of Formica pallidefulva Latreille. Ants were collected from six field colonies around the University of

Colorado Boulder campus in September 2014. Formica pallidefulva also occurs at our study site and has been observed on the same Penstemon plants used by E. anicia (C. Kelly, personal observation). They were maintained as distinct colonies in the laboratory in rectangular plastic bins lined with Fluon to prevent the ants from escaping. Colonies were maintained on a diet consisting of water and a 20% sucrose solution. The sugar solution was removed for 18-24 hours prior to bioassay trials. Ants were then presented with the choice of two different solutions, the contents of which varied with the trial. All experimental solutions were prepared in a solution of

20% sucrose in water. These solutions were presented to ants as 100 µl drops on top of a two small Parafilm squares (5 cm2), which were placed side by side. There was a minimum of 24 hours between trials for a given colony. Each trial was video recorded for two hours with an HD camcorder (Sony HDR-CX405). Video recordings were then observed and the number and length of individual feeding bouts by ants (i.e. total consecutive time spent at each drop by individual ants) were recorded.

To investigate the potential deterrence of E. anicia larvae, ant bioassays with the same colonies were conducted comparing solutions containing pulverized, fresh mealworm larvae (no

IGs) or E. anicia larvae reared on either P. glaber or P. virgatus. The mealworm solution

(control) was a 20% sucrose solution containing ground mealworm larvae (35 mg/ml). The experimental solutions consisted of 20% sucrose and sibling groups of freshly ground E. anicia larvae that were reared on either P. glaber (10 larvae/ml = 36mg/ml) or P. virgatus (10 larvae/ml

= 34 mg/ml). All larvae were in the 3rd instar and ground in the sucrose solution with a glass rod.

Ant colonies were given the following pairwise comparisons: A) control solution and P. glaber

60 reared caterpillar solution; B) control solution and P. virgatus reared caterpillar solution; and C)

P. glaber reared caterpillar solution and P. virgatus reared caterpillar solution.

To investigate the effects of specific IGs, assays were conducted with solutions containing isolated IGs with no insect tissue. For these trials, the control solution was a 20% sucrose solution that did not contain a protein source. The IG-containing solutions contained either catalpol (4 mg/ml) or catalpol + scutellarioside (4 mg/ml combined) in a 20% sucrose solution. Ant colonies were then presented with the following choices: D) control solution and catalpol solution, and E) catalpol solution and catalpol + scutellarioside solution.

Phenoloxidase Assays

An important component of insect immunity relies on rapidly activated cascades of enzymes, such as phenoloxidase (Cerenius and Söderhäll 2004; Siva-Jothy et al. 2005).

Phenoloxidase (PO) catalyzes the initial steps in the production of melanin, which is used to encapsulate foreign bodies in insect hemolymph (Sugumaran 2002; Siva-Jothy et al. 2005). To investigate the effects of host plant and the resulting chemical variation in sequestering caterpillars, we compared the phenoloxidase activity of caterpillars reared on P. virgatus and P. glaber. The E. anicia larvae used in this study had reached diapause and were in their fourth instar, a time when parasitoids may attack larvae. Upon hatching, groups of full siblings were divided into two groups: one group was reared exclusively on P. virgatus and the other on P. glaber. The small size of these larvae precluded extracting sufficient hemolymph from an individual larva for the immune assay; therefore, each sample consisted of four, previously frozen, full-sibling larvae. The activity of naturally activated PO enzymes was measured with a spectrophotometer using the methods described in Vogelweith et al. (2011).

For sample preparation, groups of frozen, whole caterpillars were ground on ice in microcentrifuge tubes with 20 µl of phosphate buffer solution (PBS) and then centrifuged (4000 g, 15 min, 4 °C). Ten microliters of supernatant were added to a microplate well containing 20 µl of PBS and either 140 µl of distilled water to measure PO activity only or 140 µl of chymotrypsin solution (Sigma C-7762, 0.07 mg/ml; Sigma-Aldrich, St. Louis, MO, USA) to measure total-PO activity. Last, 20 µl of L-Dopa solution (Sigma D-9628, 4 mg/ml) was added to each well. The reaction was allowed to proceed at 30 °C in a microplate reader (Synergy HTX,

BioTek, Sunnyvale, VT, USA) for 40 min. Readings were taken every 30 s at 490 nm and analyzed using the software Gen5 1.11 (BioTek Instruments Inc.). Enzyme activity was measured as the Vmax (change in absorbance unit per minute) during the linear phase of the reaction and reported per milligram of caterpillar.

Statistics

Non-parametric tests were used when the data failed to meet the assumptions of normality, as determined by Shapiro-Wilk tests. Wilcoxon-Mann-Whitney tests were used to compare plant IG concentrations, larval IG concentrations and feeding times in the ant predator bioassays. Chi-square analyses compared the number of individual ant visits for each predator bioassay trial. The PO assay data were analyzed with a Wilcoxon-Mann-Whitney test. All statistical analyses were performed in R version 3.1.2.

RESULTS

Penstemon chemistry and herbivore sequestration

Penstemon glaber leaves contained only catalpol, whereas P. virgatus leaves contained both catalpol and scutellarioside (Figure 4.1a). Penstemon glaber leaves contained significantly more catalpol than P. virgatus leaves (Wilcoxon p < 0.001), with an average of 5.33% dry weight. This is noticeably less catalpol than that measured in P. glaber in a recent study (18.12% dry weight; Kelly and Bowers 2016). Penstemon virgatus leaves had more IGs overall

(Wilcoxon p < 0.001, Figure 4.1a), due to the high concentrations of scutellarioside (average of

25.05% dry weight). The IG concentrations found in P. virgatus were similar to those found in previous research (Kelly and Bowers 2016).

Larvae reared on either P. virgatus or P. glaber only sequestered catalpol. Larvae reared on P. glaber sequestered more catalpol than those reared on P. virgatus (Wilcoxon p < 0.001,

Figure 4.1b).

30 a 14 b

Scutellarioside Catalpol 25 12

10 20

8 15

Percent Dry Weight Percent 10

Percent Dry Weight Catalpol Dry Weight Percent 4

5 2

0 0 P. glaber P. virgatus P. glaber fed larvae P. virgatus fed larvae

Predator bioassays

Ants showed a significant preference for the control solution (n = 110 visits) over the P. glaber reared caterpillar solution (n = 16 visits), as demonstrated by both mean feeding time spent at control solutions and the number of ant visits (Feeding time: Wilcoxon p = 0.0504,

Figure 4.2a; Visits: χ2 = 68.64, p < 0.0001). Similarly, ants demonstrated a clear preference for the control solution (n = 78 visits) over the P. virgatus reared caterpillar solution (n = 24 visits;

Feeding time: Wilcoxon p < 0.001, Figure 4.2b; Visits: χ2 = 27.54, p < 0.0001). Ants showed no preference when given a choice between the P. glaber reared caterpillar solution (n = 4 visits) and the P. virgatus reared caterpillar solution (n = 6 visits) and both solutions had low visitation

(Feeding time: Wilcoxon p = 0.625, Figure 4.2c; Visits: χ2 = 0.1, p = 0.75).

Assays comparing the control sugar solution to the sugar-catalpol solution found that ants were indeed deterred by presence of catalpol, as ants spent significantly more time feeding at the control solution (Wilcoxon p < 0.001, Figure 4.3a) and the control solution received more visits

(n = 67) than the catalpol solution (n = 23; χ2 = 20.54, p < 0.0001). However, ants were equally deterred by the catalpol solution (n = 7 visits) and the catalpol + scutellarioside solution (n = 5 visits; Feeding time: Wilcoxon p = 0.533, Figure 4.3b; Visits: χ2 = 0.8, p = 0.533).

Phenoloxidase assays

Penstemon glaber reared larvae had significantly higher PO activity than larvae reared on

P. virgatus (Wilcoxon p = 0.00281; Figure 4.4).

40 40 1.0 a b c

0.8 30 30

0.6

20 20

0.4 Mean Feeding Time (s) Feeding Mean 10 10 0.2

0 0 0.0 Control P. glaber fed larvae Control P. virgatus fed larvae P. glaber fed larvae P. virgatus fed larvae

60 6 a b

50 5

40 4

30 3

20 2 Mean Feeding Time (s) Feeding Mean

10 1

0 0 Control Catalpol Catalpol Catalpol and Scutellarioside

100

40 V Max (per V mg larva) Max (per

0 P. glaber fed larvae P. virgatus fed larvae

Figure 4.4 Phenoloxidase activities of larvae reared on either Penstemon glaber (n = 27) or P. virgatus (n = 27). Means ± SE.

DISCUSSION

In this study, we found that host plant diet indeed affected the IG sequestration of herbivore larvae, which, in turn, had consequences for some herbivore interactions with the third trophic level. Larvae of the IG specialist Euphydryas anicia sequestered 50% more catalpol when reared on P. glaber than larvae reared on P. virgatus, and catalpol was the only IG detected in these larvae. Yet, larvae reared on these two Penstemon species were equally unpalatable to invertebrate predators. These results suggest that there may be a threshold effect, such that levels of sequestered IGs above a certain level are equally deterrent (Del La Fuente et al. 1994/1995).

Supporting this hypothesis, ants did not distinguish between the solutions containing P. glaber reared larvae and that containing P. virgatus reared larvae. Bioassays with pure compounds indicated that ants were deterred by the presence of catalpol, but were similarly deterred by the solution containing both catalpol and scutellarioside. This suggests that no synergistic interaction exists between catalpol and scutellarioside, since the combination of these compounds did not increase their unpalatability. Parasitoids and pathogens, however, may be affected by larval diet.

Larvae reared on P. glaber had higher phenoloxidase activity than those reared on P. virgatus, which indicates that E. anicia larvae reared on P. glaber may be capable of mounting a stronger immune response.

Euphydryas anicia can sequester catalpol but not scutellarioside. Larvae reared on P. glaber, the more catalpol-rich host plant, sequestered more catalpol. However, we detected much higher concentrations of catalpol in larvae reared on either Penstemon diet than was found in either host plant (P. glaber diet: 11.84% dry caterpillar weight and 5.33% dry plant weight; P. virgatus diet: 8.37% dry caterpillar weight and 0.54% dry plant weight). Likewise, Stermitz et al.

(1986) found that E. anicia contained higher concentrations of catalpol than was in the host plant

Besseya plantaginea (James) Rydb (Plantaginaceae). Besseya plantaginea contains five other

IGs in addition to catalpol, though three of them are catalpol esters. Euphydryas anicia may be selectively sequestering and concentrating catalpol, or they may be metabolizing catalpol esters into catalpol, as suggested by the previous study of Gardner and Stermitz (1988). Gardner and

Stermitz (1988) found evidence that E. anicia is capable of such conversion, demonstrating that larvae metabolized 6-isovanillylcatalpol into catalpol and then excreted the remaining isovanillic acid. Therefore, it is possible that E. anicia larvae metabolically convert scutellarioside into catalpol by hydrolyzing the side chain.

Iridoid glycosides, including catalpol, have been shown to play an important defensive role against both vertebrate and invertebrate predators (Bowers 1991, De La Fuente et al. 1994,

1995, Dyer and Bowers 1996, Camara 1997b). Although host plant diet resulted in differences in the amount of catalpol sequestered by E. anicia, larvae reared on these two plant species appeared equally defended from invertebrate predators; the Formica pallidefulva ants used in the predator bioassays did not distinguish between sugar solutions containing P. glaber reared larvae and P. virgatus reared larvae. These data suggest that small amounts of catalpol may be sufficient to successfully deter predation by invertebrates. The results from the bioassays with isolated IGs confirmed that ant predators found catalpol unpalatable. The bioassay trials comparing a catalpol solution to a combined catalpol and scutellarioside solution showed no indication of a synergistic interaction of these two IGs. Ants were deterred by both solutions, as demonstrated by the low number of visits to each solution and the short duration of each feeding bout. Although there was a downward trend in foraging bouts to the solution that contained both catalpol and scutellarioside, there was not a statistically significant effect. It is possible that

70 catalpol and scutellarioside may interact synergistically, but this bioassay was unable to confirm such an interaction.

The results of the PO assays suggest that E. anicia larvae reared on P. virgatus may be less well defended against parasitoid attack. One possible mechanism underlying this difference may be the cost of metabolizing scutellarioside. This metabolic process may weaken the immune response of P. virgatus reared caterpillars. Similarly, a previous study with Junonia coenia (Nymphalidae), an IG specialist caterpillar (Bowers and Stamp 1997), found a significantly lower melanization response when larvae sequestered high concentrations of IGs

(Smilanich et al., 2009a). Thus, P. glaber reared larvae may be less immunocompromised, despite sequestering higher concentrations of catalpol overall. Though mounting an immune response can sometimes decrease herbivore fitness in other ways (e.g. growth rates), Kelly and

Bowers (2016) found that E. anicia larvae performed equally well when reared on P. glaber and

P. virgatus in a laboratory. So any metabolic costs incurred by P. virgatus reared larvae only seem to be detected in the context of immune defense.

We found a significant effect of host plant diet on PO activity, which may indicate that larvae reared on P. glaber are capable of mounting a stronger immune response or that the larvae used in the assay were already infected. We decreased the likelihood of prior infection affecting out results by using sibling groups reared on either host plant diet, to control for genotypic differences, and greenhouse grown plants, to control for environmental differences between the two host plants. However, the use of PO assays as the sole metric for immune response can be limiting. The innate immunity of insects can also be quantified with the counting of hemocytes, which serves a proxy for measuring cellular responses such as phagocytosis and encapsulation

(Strand and Pech 1995, Schmidt et al. 2001, Lavine and Strand 2002). Studies often use both

71 hemocyte counts and PO activity as they represent different components of arthropod immune defense (e.g. Satterfield et al. 2013, McKay et al. 2016a, 2016b). Yet, several studies have found no correlation between different measures of arthropod immune defense, including hemocyte count and PO activity (Adamo et al. 2001, Rolff 2001, Satterfield et al. 2013). Hemocyte counts can vary even when PO activity does not (McKay et al. 2016a, 2016b) and are subject to variation based on sex and infection status (Satterfield et al. 2013) as well as food restriction

(Karl et al. 2011). Nonetheless, a recent review provides evidence that individuals in better condition produce higher levels of PO (González-Santoyo and Córdoba-Aguilar 2012). The relative efficacy of hemocyte counts and PO activity for measuring an arthropod’s immune response remains resolved and warrants further research.

Plant secondary metabolites can have important bottom-up effects on insect herbivores and the third trophic level alike. Our results indicate that certain host plant diets may result in a trade-off between defense against predators and protection from parasitoids. Recent work has demonstrated that E. anicia females preferred to oviposit on P. glaber in the field, but larvae showed no preference for either P. glaber or P. virgatus in the laboratory (Kelly and Bowers

2016). The results of this study suggest that, when defense against natural enemies is considered, females are indeed choosing the host plant that is better for their offspring. While larvae reared on either host plant appear to be unpalatable to invertebrate predators, larvae reared on P. glaber were able to mount a stronger immune response. Gentry and Dyer (2002) suggest that larval sequestration is driven more by predators than parasites, since well-defended hosts provide enemy free space for parasitoids. However, there is contrary evidence demonstrating that sequestering defensive compounds protect insects from parasitoids (e.g. Sime 2002) and also helps fight microbial infections. We found that the immunocompentence of specialist larvae

72 varies with host plant species and that feeding on plants containing compounds that larvae cannot directly sequester may result in a weakened immune response. This suggests that, in the

Penstemon-Euphydryas system, the sequestration of IGs may be driven more by predators. Our study highlights the importance of plant secondary metabolites in mediating herbivore vulnerability to natural enemies and the role of sequestration in shaping the evolution of complex multi-trophic interactions.

CHAPTER 5

CONCLUSIONS

In this chapter I first give an overview of my hypotheses and findings. I then discuss the development of the tri-trophic interaction model (TTI, Mooney et al. 2012) as a conceptual framework for understanding plant-herbivore coevolution. In the next section I explore the published tests of the TTI, as these set the background for my work. I then give a brief synthesis of how my findings relate to the TTI model and other author’s studies that have tested components of the model. Finally, I give some insights into the types of future studies that will help to further improve our understanding of plant-animal interactions under the TTI model.

5.1 Principal Findings

In my dissertation, I tested several of the predictions of the tri-trophic interactions (TTI) hypothesis to determine how plant secondary metabolites mediate herbivore performance on different host plants and how those compounds further mediate herbivore interactions with higher trophic levels (Table 5.1). Thus, I use the TTI model as a context for understanding the results of my empirical studies.

Chapter 2 presents a test of the prediction that variation in the chemical defenses of host plant taxa used by generalists and specialists will result in larger differences in generalist herbivore performance than specialist performance (Mooney et al. 2012). This prediction stems from the physiological efficiency (PE) hypothesis, which is one of the three seminal hypotheses that contributed to the TTI hypothesis (illustrated in Figure 5.1). My results supported this prediction, as generalist herbivores demonstrated more variation in performance than the specialist on the two host plant species. I also tested the preference-performance (P-P) hypothesis,

CHAPTER HYPOTHESIS or GOAL PREDICTION CONCLUSION

Chapter 2 Specialist herbivores will have A) Specialist female preference My data partially a strong preference- corresponds to larval performance support this performance relationship and B) Differences in host plant hypothesis. will also outperform generalists defensive chemistry will result in on host plant(s) preferred by the greater variation in performance specialists for a generalist than a specialist herbivore Chapter 3 To assess variation in the IG P. virgatus IGs will vary across I found significant content of Penstemon virgatus growing seasons, populations and variation in iridoid across multiple growing tissue types. glycoside seasons, six natural populations concentrations and three tissue types across all three factors. Chapter 4 Plant secondary metabolites The defensive chemistry I found support for mediate interactions between a sequestration of specialist larvae this hypothesis. specialist herbivore and its influences its defense against natural enemies predators and parasitoids Table 5.1. A brief summary of the main hypotheses or goals addressed in each research chapter and the corresponding conclusions. which is not explicitly discussed by Mooney et al (2012) in the context of the TTI hypothesis but which has considerable relevance to tests of the TTI. The preference-performance hypothesis predicts that adult females preferentially oviposit on host plants that maximize offspring performance Unexpectedly, I found no relationship between female host plant preference and

75 offspring performance.

One component of the TTI Hypothesis is the prediction that intra- and inter-specific variation in host plant quality determines herbivore performance, and consequently, natural enemy performance (Figure 5.2). In Chapter 3 I examined the variation of IGs to which herbivores are exposed in several populations of Penstemon virgatus. The significant spatial and temporal variation I observed in Penstemon IGs suggests that the outcomes of tests of the TTI hypothesis may be highly specific to year and location.

Figure 5.2 Adapted from Mooney et al. 2012. Chapter 3 focused variation in host plant quality (highlighted in green), which is one of three major components of the TTI hypothesis.

Finally, research presented in Chapter 4 tested the prediction that specialist herbivore performance suffers on low quality plants and makes them more susceptible to attack by natural enemies (Figure 5.3). The specialist herbivore showed a weakened immune response when reared on the lower quality host plant. There was no clear relationship between host plant species and specialist herbivore protection from predators. Although this study focused on host plant quality and natural enemies, it did not explicitly test the slow-growth/high-mortality (SGHM)

76 hypothesis, as Figure 5.3 may suggest. A proper test of the SGHM hypothesis would include measurements of growth rate on each host and exposure to natural enemies in the field.

Figure 5.3 Adapted from Mooney et al. 2012. Chapter 4 examined the role of host plant quality in mediating specialist herbivore interactions with natural enemies (highlighted in purple).

In order to place these conclusions in the context of current ecological theory, I will now briefly review the theoretical developments that led up to my studies.

5.2 Hypotheses in Multi-Trophic Interactions

Interactions between plants, herbivores and natural enemies are as complex and variable as the environments in which they occur. Hairston et al. (1960) proposed that top-down processes control herbivores, based on the assertion that predators limit herbivores (this is also known as the “green world hypothesis”). In contrast, the “green desert hypothesis” argued that herbivore populations were more limited by bottom-up controls, such as plant nutritional quality and chemical and physical defenses (Murdoch 1966, Menge 1992). Assessing the relative strengths of top-down and bottom-up forces remains a contentious issue (Halaj and Wise 2001,

Walker and Jones 2001). Many studies acknowledge that the relative importance of top-down or bottom-up mechanisms is largely context dependent. For example, top-down controls may have a

77 stronger influence on aquatic food webs relative to terrestrial ones (Strong 1992; Polis & Strong

1996; Polis 1999; Halaj & Wise 2001, Shurin et al. 2002). Similarly, evidence suggests that top- down trophic cascades are more prevalent in agricultural systems than unmanaged systems

(Hawkins et al. 1999, Halaj and Wise 2001). In 1980, Price et al. combined theories regarding the bottom-up and top-down controls of herbivore populations and promoted the idea of tri- trophic interactions among plants, herbivores and herbivore natural enemies (also reviewed in

Ode 2006). Ecologists now recognize that both top-down and bottom-up forces maintain herbivore populations (e.g. Price et al. 1980, Bernays & Graham 1988, Hunter 2003, Singer &

Stireman 2003, Letourneau et al. 2004, Ode et al. 2004). The mechanisms that allow these forces to interact are crucial for understanding the evolution and ecology of multi-trophic interactions.

Mooney et al.’s (2012) recently proposed tritrophic interactions (TTI) hypothesis provides a new framework for investigating the effects of multiple trophic levels on the maintenance of herbivore populations. There is currently no other hypothesis that combines host plant quality, herbivore diet breadth and the presence of natural enemies together into a single theory. The TTI hypothesis consolidates the predictions of three seminal hypotheses, specifically physiological efficiency (PE), enemy-free space (EFS) and slow-growth/high-mortality (SGHM), into a single framework. This integration is novel and differs substantially from other models in its consolidation of already well-established hypotheses. The TTI hypothesis is perhaps easily accessible to a broad, ecological audience given its foundation of long-standing theories.

Furthermore, the TTI hypothesis escapes the linearity suggested in the original models of trophic interactions (top-down and bottom-up) and allows for multi-directional interactions that better parallel actual trophic webs (see figure 5.4).

The Tri-Trophic Interactions Hypothesis

78 A B Natural enemies (predators and parasitoids)

Top-down controls

Herbivores

Bottom-up controls

Host plant quality

Figure 5.4. A) A model demonstrating the linear nature of traditional trophic Figure 1. Predictions of the tri-trophic interactions (TTI) hypothesis for the interactive effects of natural enemies, host-plant quality models; B) The tri-trophic model from Mooneyand et diet al. breadth(2012). on Note herbivores. its useThree of well-studied hypotheses – the physiological efficiency (PE), enemy free space (EFS) hypotheses, and slow- overlapping circles instead of straight lines. growth/high-mortality (SGHM) – each address unique, pairwise combinations of these factors. The physiological efficiency (PE) hypothesis predicts specialists should outperform generalists on shared host plants (e.g. a.b), and that generalists should be more sensitive to variation in host-plant quality than specialists (e.g. a–c,b–d). The Enemy Free Space (EFS) hypothesis predicts natural enemies should have a stronger effect on dietary Given the integrative nature of studies of multispecialists-trophic than interactions, generalists (e.g. a–eit is, b–f).increasingly The Slow-Growth/High-Mortality (SGHM) hypothesis predicts low host-plant quality enhances the effects of natural enemies (e.g. b–f,d–h). The TTI hypothesis offers novel predictions for the three-way interaction among these factors: Dietary specialists (as compared to generalists) are predicted to escape natural enemies and be competitively dominant due to faster growth rates, and such differences important that highly influential but independently expressedshould be greater hypotheses, on low quality such (as as compared the PE, to EFS high quality) host plants. Such non-additive dynamics imply that predictions for the PE, EFS, and SGHM hypotheses are contingent upon the third, discounted factor. Natural enemies should mediate the predictions of the PE hypothesis, such that the differential effects of host-plant quality on specialists and generalists is greater in the presence of natural enemies (e–g%f–h) than in the absence and SGHM, be unified into a single, inclusive theory.of naturalThe central, enemies (a–covera,b–d).rching Host-plant prediction quality shouldof the mediate the predictions of the EFS hypothesis, such that the differential effects of natural enemies on specialist and generalist herbivores is greater on low-quality host plants (c–g%d–h) than on high-quality host plants (a–e,b–f). Herbivore diet breadth should mediate the predictions of the SGHM hypothesis, such that SGHM dynamics are stronger for dietary generalist (b– d%f–h) than specialist herbivores (a–c,e–g). TTI hypothesis is that host-plant quality, herbivore doi:10.1371/journal.pone.0034403.g001diet breadth and natural enemies interactively determine herbivore performance in waysCurrent not explicitly hypotheses addressed by the PE, EFS The PE hypothesis also offers three predictions for the The physiological efficiency (PE) hypothesis states that dietary interactive effects of host-plant quality and herbivore diet breadth. and SGHM hypotheses individually. The specific subspecialists-predictions are better under adapted the TTI than hypothesis generalists at are physiologically First, variation in host-plant quality should have stronger effects on utilizing their host plants as food [1]. As a result, specialists should dietary generalists than on better-adapted dietary specialists [28]. have superior physiological performance (e.g., more efficient As depicted in Fig. 1, the PE hypothesis predicts a–c,b–d (PE as follows: resource assimilation and faster growth rates) than generalists on effects without natural enemies) and e–g,f–h (PE effects with their shared host plants [24,25]. As depicted in Fig. 1, the PE natural enemies). However, the PE hypothesis does not offer • Specialist herbivore escape natural enemieshypothesis better predicts than generalists a.b, c.d, due e.f, to and the g .h. This central predictions for the relative magnitude of PE effects between the prediction of the PE hypothesis has found support in some (e.g. presence and absence of natural enemies. Past studies support this [26]) but not all (e.g. [24]) studies (reviewed by [27]). first prediction, showing that toxic forms of plant secondary increased mortality of physiologically inefficient and slow-growing generalists. compounds have larger effects on the performance of generalist

• Specialists competitively exclude generalistsTable on 1. Descriptionsa shared host of three plant long-standing as they are hypotheses for plant-herbivore and herbivore-predator interactions and their relation to the tri-trophic interactions hypothesis. physiologically better able to utilize the host plant.

Predictions under tri-trophic interactions Original hypothesis • These differences between generalists and specialists should be greater on lower hypothesis Name Factors considered Predictions quality host plants. The performance gapPhysiological between efficiency specialists Dietand breadth, generali plant sts shouldSpecialists are better adapted than generalists at The benefits of specialization for performance are quality using shared plants as food (a.b, c.d, e.f, greater in the presence of natural enemies (e–g%f–h) and g.h) and variation in host-plant quality than absence of natural enemies (a–c,b–d) be smaller on higher quality host plants. should have stronger effects on generalists than specialists (a–c,b–d and e–g,f–h) Enemy-free space Diet breadth, natural Specialist are better than generalists at using The benefits of specialization for predator avoidance The TTI hypothesis will provide invaluable contributions to the understandingenemies of theshared plants for predator avoidance (a–e,b–f are greater on low-quality plants (c–g%d–h) than and c–g,d–h) high-quality plants (a–e,b–f) ecology of multi-trophic interactions, but only if it is rigorouslySlow-growth/high-mortality tested in a Plant variety quality, of natural plant- Low plant quality increases the effects of natural Low plant quality increases the effects of natural enemies enemies (a–e,c–g and b–f,d–h) enemies more for generalists (b–f%d–h) than specialists (a–e,c–g)

Parenthetical references to a–h refer to the graphical representation of these predictions shown in Fig. 1. doi:10.1371/journal.pone.0034403.t001

PLoS ONE | www.plosone.org 2 April 2012 | Volume 7 | Issue 4 | e34403 79 herbivore systems. Mooney et al. (2012) acknowledged the need for a “synthesis of results across many systems.” Specifically, they suggest factorial studies comparing multiple specialist/generalist herbivore pairs, community-level studies, and a meta-analysis that combines the results from previous experiments. Once large quantities of data are accumulated from different systems, the universality of the TTI hypothesis can be assessed. The current lack of such data suggests that it be challenging to find a system capable of testing all components of this hypothesis. In the next section (5.3), I expand the discussion on the limitations of empirically testing the TTI hypothesis.

5.3 Tests of the TTI Hypothesis

Surprisingly, Mooney et al.’s paper (2012) is the only published test of the TTI hypothesis to date. Using an aphid-Baccharis study system, they found support for the overarching prediction that a three-way interaction exists among host-plant quality, herbivore diet breadth and natural enemies. The generalist aphid varied more in performance between the different host plants than the specialist aphid. Host plant quality largely determined vulnerability to predation for the generalist, but not for the specialist. Plant quality can be measured in a wide variety of ways (e.g. nitrogen content, water content, defensive chemistry), but for the purpose of their empirical studies, Mooney et al. (2012) defined plant quality as defensive chemistry content.

Not all of their predictions were supported by their findings (Mooney et al. 2012).

Generalists were expected to perform worse than specialists on the lower quality host plant when in the presence of natural enemies. Aphid performance was calculated as per capita daily population growth. Yet, the generalist aphid species performed better than expected on lower quality host plants when natural enemies were present. The unexpectedly high performance of

80 the generalist aphid was likely the result of a mutualistic interaction with ants (Mooney et al.

2012). Ant tending can improve aphid performance by providing protection from natural enemies (Stadler and Dixon 2008) and removal of competing herbivores (Smith et al. 2008). This symbiotic ant-generalist aphid relationship may be what allows for the coexistence of these two aphid species on shared host plants.

Finding other study systems that allow sufficient tests of all of the predictions of the TTI hypothesis is potentially problematic. The prediction that specialist herbivores are competitively dominant over generalists dictates the use of generalist and specialist species that coexist on plants with observable variation in quality. The empirical test in Mooney et al.’s (2012) paper utilized two aphid species (one a specialist and the other a generalist) that co-exist within the same host plant stand, and occasionally on the same individual plant. They speculated that the increased performance of the generalist aphid in the presence of ant mutualists might cause these two herbivore species to successfully coexist. An ideal study system would feature a generalist and a specialist herbivore that coexist without the confounding influence of symbiotic relationships. Trophic relationships that meet these requirements may be common in nature, but identifying them and developing them as study systems has proven challenging. This greatly limits the number of study systems available to test the TTI hypothesis and, in turn, makes future research on this hypothesis even more challenging.

Competitive asymmetries between sympatric species may lead to complete competitive exclusion on individual host plants or in whole populations. This may partially explain why systems with coexisting generalist and specialist species are seemingly rare. This may also be why generalists are less common overall, as generalist herbivores likely have a competitive disadvantage when co-occupying a plant with a specialist herbivore (Egas et al. 2004). The TTI

81 hypothesis predicts that specialists will outperform generalists in all situations, yet also acknowledges that specialists and generalists do co-exist (e.g. Strong et al. 1984, Novotny et al.

2006, Dyer et al. 2007, Mooney et al. 2012). Mooney et al. (2012) theorize that generalists may outcompete specialists under certain ecological conditions, particularly fluctuating environments

(Wilson and Yoshimura 1994, Abrams 2006).

Mooney et al. (2012) also argued that competition plays a particularly large role in structuring communities of sap-feeding herbivores. Yet sap-feeding herbivores, such as aphids, are only a small subset of generalist and specialist herbivores. Many of the tests of the hypotheses featured in the TTI hypothesis were done with leaf-feeding herbivores (e.g. Atsatt

1981, Gross and Price 1988, Denno et al. 1990, Benrey and Denno 1997, Lill and Marquis 2001,

Singer 2001, Cornell and Hawkins 2003, Fordyce and Shapiro 2003, Murphy 2004, Singer et al.

2004, Cornelissen and Stiling 2006, Friberg and Wiklund 2009). The success of any major hypothesis partially lies in its relevance to multiple study systems. It is currently difficult to extend the TTI hypothesis to other guilds of herbivores since larger herbivores (such as leaf- feeders) are less likely to coexist on the same individual plant, or even in the same plant stand. A woody plant-based study system could offer a possible solution to the problems of differences in competitive ability, as multiple communities may be more likely to coexist on larger, longer- lived plants than herbaceous plants. Woody plants may also allow for more differentiation among herbivores that specialize on particular parts of the plant.

Since the publication of the Mooney et al. paper in 2012, 36 other scientific papers have cited it (confirmed via Google Scholar on November 6, 2016; Figure 5.5). Only four of these publications specifically address components of the TTI hypothesis in a significant way (Mason et al. 2014; Abdala-Roberts et al. 2014; Muller et al. 2015; Katsanis et al. 2016), but none of

14 12 10 8 6 4

Number of Citations 2 0 2012 2013 2014 2015 2016 Publicaon Year

Figure 5.5 The number of primary publications that cite Mooney et al. 2012 by publication year. None of these publications explicitly test all components of the TTI hypothesis. them tested all of the components. Katsanis et al. (2016) tested the prediction that dietary specialization mediates the effects of plant defenses on tri-trophic interactions. This is not an explicit test of the TTI hypothesis, but it is the most similar to Mooney et al. (2012).

Katsanis et al. (2016) found that generalist and specialist aphid performances were more similar than expected and were unaffected by variation in host plant quality. Their results generally supported the broader prediction that host plant quality, herbivore diet breadth and natural enemies together mediate multi-trophic interactions in ways not addressed by previous hypotheses. In sum, Mooney at al. (2012) has been cited many times, but the TTI hypothesis is not as prevalent as those citations would suggest, possibly due to challenges with finding an appropriate study system.

5.4 Synthesis

To summarize, my dissertation research advances knowledge in the field of ecology and evolutionary biology by combining chemical and behavioral data to test components of the TTI hypothesis in a novel system. Chapters 2 and 3 establish necessary background for my study

83 system (Chapter 3) and vital insight into plant-herbivore relations in a tri-trophic context

(Chapter 2). The research in Chapter 2 also examined the effects of variation in host plant quality on oviposition preference, but the TTI hypothesis does not address the role of host plant preference in tritrophic interactions. However, enemy free space may play a significant role in both oviposition choice and larval performance. Accordingly, I contend that host plant preference has large implications for multi-trophic interactions. Chapter 4 relates more directly to the TTI hypothesis, but I was limited by the lack of a biologically relevant generalist herbivore.

Since I was unable to explicitly test the main prediction that host-plant quality, herbivore diet breadth and natural enemies interactively determine herbivore performance, it is difficult to directly compare the results of my dissertation to the results from the only published empirical test (Mooney et al. 2012). However, the components I did test were largely in agreement with the predictions under the TTI hypothesis. My research also emphasizes some of the difficulties in finding a study system that can appropriately test this hypothesis. The results of this work provide support for several predictions of the TTI hypothesis and also highlight the importance of unified theories when investigating multi-trophic interactions.

5.5 Future Directions and Closing Remarks

The predictions of the TTI Hypothesis focus on the bottom up effects of plant quality on herbivore growth and development, which then affects the natural enemies of these herbivores.

However, there is evidence that higher trophic levels are able to affect plants on a chemical and molecular level (Felton and Tumlinson 2008, Howe and Jander 2008). For example, plants may respond to chemical cues, in the form of larval salivary compounds or other fluids, or physical cues, in the form of mechanical damage from chewing insects or the scratching of leaves by

84 ovipositing females (reviewed in Karban and Myers 1989, Hilker and Meiners 2002). There are an increasing number of studies examining induction of plant secondary compounds by herbivory but fewer that examine induction via oviposition. It is even more rare that both of these factors are combined in a single study. As plants typically host both egg masses and subsequently larvae, it is crucial to examine these influences together and not in isolation. As the

TTI Hypothesis does not form predictions regarding the top-down effects of herbivore on plant quality, future research is needed fill in this gap.

The TTI hypothesis is a relatively recent addition to the field of community ecology and more tests of its predictions are necessary. In particular, the prediction tested in Chapter 4 should be tested with generalist herbivores. The TTI hypothesis posits that low plant quality increases the effects of natural enemies more for generalists than specialists. We can more explicitly test that prediction by examining how host plant secondary metabolites affect generalist herbivore vulnerability to predators and parasitoids. Grazing generalist herbivores that are tolerant of chemically defended host plants, such as Grammia incorrupta (Arctiidae), would be an ideal study system for such a study.

Future studies should also acknowledge the bias towards examining the separate effects of each secondary compound on higher trophic levels (Dyer 2011). Despite increasing evidence that plant secondary compounds can be synergistic (Hay et al. 1994, Stamp and Osier 1998,

Richards et al. 2010, 2012), most research on plant chemical defenses focuses on individual compounds. Thus it is becoming increasingly important to examine phytochemical synergy.

Additionally, there is growing evidence that synergistic defenses also affect specialists and generalists differently (Hagele and Rowell-Rahier 2000, Dyer et al. 2003). Specialists tend to be better adapted to the variety of secondary metabolites within their host plants, whereas

85 generalists are more likely to be negatively affected when feeding on the same host. Therefore, it is increasingly important that tritrophic research with multiple plant defensive compounds test for potential synergies.

In conclusion, many hypotheses have been proposed to explain controls over the outcome of multi-trophic interactions and the TTI hypothesis is one of the most recent attempts at such a synthesis. While there may be some novel predictive ability with the TTI hypothesis, it is clear from my dissertation research and attempts by other investigators that problems with explicitly testing this hypothesis currently limit its application to other studies. This hypothesis offers a novel, more integrative approach to understanding tritrophic interactions, but its fate as a seminal hypothesis in ecology remains unclear. Future research that focuses on simplifying and expanding tests of this hypothesis will provide key new insights into the complex interactions of organisms across multiple trophic levels.

LITERATURE CITED

Ali, J.G. and Agrawal, A.A. (2012) Specialist versus generalist insect herbivores and plant defence. Trends in Plant Science, 17, 239-242. Agrawal, A.A. (2011) Current trends in the evolutionary ecology of plant defence. Functional Ecology, 25, 420-432. Ahmed, A. M., Baggott, S.L., Maingon, R. and Hurd, H. (2002) The costs of mounting an immune response are reflected in the reproductive fitness of the mosquito Anopheles gambiae. Oikos, 97, 371-377. Alba, C., Prioreschi, R. and Quintero, C. (2013) Population and leaf-level variation of iridoid glycosides in the invasive weed Verbascum thapsus L.(common mullein): implications for herbivory by generalist insects. Chemoecology, 23(2), 83-92. Ardia, D. R., Gantz, J.E., Schneider, B.C. and Strebel, S. (2012) Costs of immunity in insects: an induced immune response increases metabolic rate and decreases antimicrobial activity. Functional Ecology, 26, 732-739. Awmack, C.S. and S.R. Leather. 2002. Host plant quality and fecundity in herbivorous insects. Annual Review in Entomology 47:817-844. Barbosa, P., Saunders, J.A., Kemper, J., Trumbule R., Olechno, J. and Martinat, P. (1986) Plant allelochemicals and insect parasitoids effects of nicotine on Cotesia-congregata (say) (Hymenoptera, Braconidae) and Hyposoter annulipes (Cresson) (Hymenoptera, Ichneumonidae). Journal of Chemical Ecology, 12, 1319-1328. Berenbaum, M.R., Zangerl, A.R. and Nitao, J.K. (1986) Constraints on chemical coevolution: wild parsnips and the parsnip webworm. Evolution, 1215-1228. Berenbaum, M.R., Nitao, J.K. and Zangerl, A.R. (1991) Adaptive significance of furanocoumarin diversity inPastinaca sativa (Apiaceae). Journal of chemical ecology, 17(1), 207-215. Berenbaum, M.R. and Zangerl, A.R. (1992) Genetics of secondary metabolism and herbivore resistance in plants. Herbivores: their interactions with secondary plant metabolites, 2, 415-438. Berenbaum, M.R. and Zangerl, A.R. (1998) Chemical phenotype matching between a plant and its insect herbivore. Proceedings of the National Academy of Sciences of the United States of America, 23, 13743-13748. Bernays, E.A. (1998) Evolution of feeding behavior in insect herbivores. Bioscience, 48(1), 35- 44. Bernays, E.A. and Chapman, R.F. (1994) Host-plant selection by phytophagous insects. Chapman and Hall, New York, New York. Bernays, E.A. Chapman, R.F., Lamunyon, C.W. and Hartmann, T. (2003) Taste receptors for pyrrolizidine alkaloids in a monophagous caterpillar. Journal of Chemical Ecology, 29, 1709-1722. Bernays, E. and Graham, M.. (1988) On the evolution of host specificity in phytophagous arthropods. Ecology, 69, 886-892.

Bernays, E.A. Singer, M.S., Rodrigues, D. (2004) Foraging in nature: foraging efficiency and attentiveness in caterpillars with different diet breadths. Ecological Entomology, 29, 389- 397. Bobbitt, J.M. and Segebarth, K.P. (1969) Iridoid glycosides and similar substances. in: Taylor, W.I., and A.R. Battersby (eds.). Cyclopentanoid Terpene Derivatives. Marcel Dekker, New York. pp. 1-145. Boros, C.A. and Stermitz, F.R. (1990) Iridoids. An updated review, Part I. Journal of Natural Products, 53, 1055–1147. Bowers, M.D. (1991) Iridoid glycosides. Herbivores: Their Interactions with Secondary Plant Metabolites (ed. by G. A. Rosenthal and M. R. Berenbaum), pp. 297 – 326. Academic Press, New York, New York. Bowers, M.D. (1992) The Evolution of unpalatability and the cost of chemical defense in insects. In: B. D. Roitberg and M. B. Isman (eds) Insect Chemical Ecology: An Evolutionary Approach. Chapman and Hall: New York, pp 216-243. Bowers, M.D. (2003) Hostplant suitability and defensive chemistry of the Catalpa sphinx, Ceratomia catalpae. Journal of Chemical Ecology, 29, 2359-2367. Bowers, M.D. (2009) Chemical defences in woolly bears: sequestration and efficacy against predators and parasitoids. Tiger Moths and Woolly Bears: Behavior, Ecology, and Evolution of the Arctiidae (ed. by W.E. Conner), pp. 83–102. Oxford University Press, Oxford, U.K. Bowers, M.D. and Collinge, S.K. (1992) Fate of iridoid glycosides in different life stages of the buckeye, Junonia coenia (Lepidoptera, Nymphalidae). Journal of Chemical Ecology, 18, 817-831. Bowers, M.D. and Stamp, N.E. (1992) Chemical variation with and between individuals of Plantago laceolata. Journal of Chemical Ecology, 18, 989-995. Bowers M.D. and Stamp, N.E. (1993) Effects of plant age, genotype, and herbivory on Plantago performance and chemistry. Ecology, 74, 1778-1791. Brenes-Arguedas, T. and Coley, P. D. (2005) Phenotypic variation and spatial structure of secondary chemistry in a natural population of a tropical tree species. Oikos 108, 410 - 420. Cain, M.L., Eccleston, J. and Kareiva, P.M. (1985) The influence of food plant dispersion on capterpillar searching success. Ecological Entomology, 10, 1-7. Camara, M.D. (1997a) Physiological mechanisms underlying the costs of chemical defence in Junonia coenia Hubner (Nymphalidae): a gravimetric and quantitative genetic analysis. Evolutionary Ecology, 11, 451-469. Camara, M. D. (1997b) Predator responses to sequestered plant toxins in buckeye caterpillars: Are tritrophic interactions locally variable? Journal of Chemical Ecology, 23, 2093-2106. Castrejon, F. Virgen, A. and Rojas, J.C. (2012) Oviposition of the saltmarsh caterpillar moth (Lepidoptera: Arctiidae) is influenced by the presence of host plant and time of day. Southwestern Entomologist, 37, 103-113.

Cerenius, L. and Söderhäll, K. (2004) The prophenoloxidase‐activating system in invertebrates. Immunological reviews, 198(1), 116-126. Cirak, C., Radusiene, J., Stanius, Z., Camas, N., Caliskan, O. and Serhat Odobas, M. (2012) Secondary metabolites of Hypericum orientale L. growing in Turkey: variation among populations and plant parts. Acta Physiologiae Plantarum, 34, 1313-1320. Clancy, K.M. and Price, P.W. (1987) Rapid herbivore growth enhances enemy attack: Sublethal plant defenses remain a paradox. Ecology 68: 733–737. Clark, K.E., Hartley, S.E. and Johnson, S.N. (2011) Does mother know best? The preference– performance hypothesis and parent–offspring conflict in aboveground–belowground herbivore life cycles. Ecological Entomology, 36, 117–124. D Coley, P., Bateman, M. L. and Kursar, T.A. (2006) The effects of plant quality on caterpillar growth and defense against natural enemies. Oikos, 115(2), 219-228. Cory, J. S. and Hoover, K. (2006) Plant-mediated effects in insect–pathogen interactions. Trends in ecology & evolution, 21(5), 278-286. Cornell, H.V. and Hawkins, B.A. (2003) Patterns of herbivorous insect response to plant secondary compounds: a test of phytochemical coevolution theory. American Naturalist, 161, 507-522. Courtney, S.P. and Kibota, T.T. (1990) Mother doesn’t know best: selection of hosts by ovipositing insects. Insect–Plant Interactions (ed. By E.A. Bernays). pp. 161–188. CRC Press, Boca Raton, FL, USA. Craig, T.P. and Itami, J.K. (2008) Evolution of preference and performance relationships. Specialisation, Speciation, and Radiation: The Evolutionary Biology of Herbivorous Insects (ed. by K.J. Tilmon), pp. 20–28. University of California Press, Berkeley, USA. Crosswhite, F.S. (1967) Revision of Penstemon Section Habroanthus (Scrophulariaceae). III: Series Virgati. American Midland Naturalist, 77, 28-41. Cullenward, M.J., Ehrlich, P.R., White, R.R. and Holdren, C.E. (1979) The ecology and population genetics of an alpine checkerspot butterfly, Euphydryas anicia. Oecologia, 38, 1-12. Da Costa, C.P. and Jones, C.M. (1971) Cucumber beetle resistance and mite susceptibility controlled by the bitter gene in Cucumis sativus L. Science, 172, 1145-1146. Damman, H. and Feeny, P. (1988) Mechanisms and consequences of selective oviposition by the zebra swallowtail butterfly. Animal Behavior, 36, 563–573. Darrow, K., and Bowers, M.D. (1997) Phenological and population variation in iridoid glycosides of Plantago lanceolata (Plantaginaceae). Biochemical Systematics and Ecology 25, 1-11. De la Fuente, M.A., Dyer, L.A., and Bowers, M.D. (1994) The iridoid glycoside, catalpol, as a deterrent to the predator Camponotus floridanus (Formicidae). Chemoecology, 5, 13-18. De la Fuente, M.A., Dyer, L.A., and Bowers, M.D. (1995) The iridoid glycoside, catalpol, as a deterrent to the predator Camponotus floridanus (Formicidae). Chemoecology, 5, 13-18.

Denno, R. (Ed.). (2012) Variable plants and herbivores in natural and managed systems. Elsevier. Denno, R.F., Larsson, S. and Olmstead, K.L. (1990) Role of enemy-free space and plant quality in host-plant selection by willow beetles. Ecology, 71, 124–137. Dethier, V.G. (1954) Evolution of feeding preferences in phytophagous insects. Evolution, 8, 33– 54. Dethier, V.G. (1959a) Egg-laying habits of Lepidoptera in relation to available food. Canadian entomologist, 91(9), 554-561. Dethier, V.G. (1959b) Food-plant distribution and density and larval dispersal as factors affecting insect populations. Canadian Entomologist, 91(9), 581-596. Dicke, M. and van Loon, J.J.A. (2000) Multitrophic effects of herbivore-induced plant volatiles in an evolutionary context. Entomologia Experimentalis et Applicata, 97, 237-249. Doak, P., Kareiva, P. and Kingsolver, J. (2006) Fitness consequences of choosy oviposition for a time-limited butterfly. Ecology, 87, 395– 408. Dobler, S., Daloze, D. and Pasteels, J. (1998) Sequestration of plant compounds in a leaf beetle's defensive secretion: cardenolides in Chrysochus. Chemoecology, 8(3), 111-118. Dyer, L.A. (2011) New synthesis—Back to the future: New approaches and directions in chemical studies of coevolution. Journal of chemical ecology, 37(7), 669-669. Dyer, L.A. (1995) Tasty generalists and nasty specialists? A comparative study of antipredator mechanisms in tropical lepidopteran larvae. Ecology , 76, 1483-1496. Dyer, L.A., Dodson, C.D, Stireman, J.O, Tobler, M.A, Smilanich, A.M., Fincher, R.M. and Letourneau, D.K. (2003) Synergistic effects of three Piper amides on generalist and specialist herbivores. Journal of Chemical Ecology, 29, 2499-2514. Dyer, L. A., Letourneau, D.K, Dodson, C.D, Tobler, M.A, Stireman, J.O. and Hsu, A. (2004) Ecological causes and consequences of variation in defensive chemistry of a neotropical shrub. Ecology 85, 2795-2803. Dyer, L.A. and Bowers, M.D. (1996) The importance of sequestered iridoid glycosides as a defense against an ant predator. Journal of Chemical Ecology, 22, 1527-1539. Dyer, L.A. and Gentry, G. (1999) Predicting natural-enemy responses to herbivores in natural and managed systems. Ecological Applications, 9, 402-408. Ehrlich, P. R. and Raven, P.H. (1964) Butterflies and plants: a study in coevolution. Evolution, 18, 568-608. El-Naggar, L.J. and Beal, J.L. (1980) Iridoids: a review. Journal of Natural Products, 43, 649- 707. Endara, M.J. and Coley, P.D. (2011) The resource availability hypothesis revisited: a meta- analysis. Functional Ecology, 25, 389–398. Faria, M.L. and Fernandes, G.W. (2001) Vigour of a dioecious shrub and attack by a galling herbivore. Ecological Entomology 26, 37–45.

Fajer, E.D., Bowers, M.D. and Bazzaz, F.A. (1992) The effect of nutrients and enriched CO2 environments on production of carbon-based allelochemicals in Plantago – a test of the carbon nutrient balance hypothesis. American Naturalist, 140, 707-723. Farrell, B.D. (1998) " Inordinate fondness" explained: Why are there so many beetles?.Science, 281(5376), 555-559. Feeny, P. (1991) Chemical constraints on the evolution of swallowtail butterflies. Plant-animal interactions: Evolutionary ecology in tropical and temperate regions (ed. by P.W. Price et al.), pp 315-340. John Wiley and Sons Inc., New York, USA. Felton, G.W. and Tumlinson, J.H. (2008) Plant–insect dialogs: complex interactions at the plant– insect interface. Current opinion in plant biology, 11(4), 457-463. Ferris, C.D. and Brown, F.M. (1981) Butterflies of the Rocky Mountain States. University of Oklahoma Press, Norman, Oklahoma. Ferro, V. G., Guimarães, P. R., & Trigo, J. R. (2006) Why do larvae of Utetheisa ornatrix penetrate and feed in pods of Crotalaria species? Larval performance vs. chemical and physical constraints. Entomologia experimentalis et applicata, 121(1), 23-29. Fraenkel, G. (1953) The nutritional value of green plants for insects. Proceedings of the 9th International Congress of Entomology, 2, 90-100. Fraenkel, G. (1959) The raison d'etre of secondary plant substances. Science, 121, 1466-1470. Friberg, M. and Wiklund, C. (2009) Host plant preference and performance of the sibling species of butterflies Leptidea sinapis and Leptidea reali: a test of the trade-off hypothesis for food specialization. Oecologia, 159(1), 127-137. Friberg, M., Posledovich, D., and Wiklund, C. (2015) Decoupling of female host plant preference and offspring performance in relative specialist and generalist butterflies. Oecologia, 178, 1181-1192. Freitak, D., Ots, I., Vanatoa, A. and Horak, P. (2003) Immune response is energetically costly in white cabbage butterfly pupae. Proceedings of the Royal Society B-Biological Sciences, 270, S220-S222. Fritz, R.S., Crabb, B.A. and Hochwender, C.G. (2000) Preference and performance of a gall- inducing sawfly: a test of the plant vigor hypothesis. Oikos, 89, 555–563. Futuyma, D.J. (2008) Sympatric speciation: norm or exception? Specialisation, Speciation, and Radiation: The Evolutionary Biology of Herbivorous Insects (ed. by K.J. Tilmon), pp. 136-202. University of California Press, Berkeley, USA. Futuyma, D. J. and Moreno, B. (1988) The evolution of ecological specialization. Annual Review of Ecology and Systematics, 19, 207-233. Gardner, D.R. and Stermitz, F.R. (1988) Host plant utilization and iridoid glycoside sequestration by Euphydryas anicia (Lepidoptera, Nymphalidae). Journal of Chemical Ecology, 14, 2147-2168. Gentry, G. and Dyer, L.A. (2002) On the conditional nature of neotropical caterpillar defenses against their natural enemies. Ecology, 83, 3108-3119. Gillespie, J. P., Kanost, M.R. and Trenczek, T. (1997) Biological mediators of insect immunity. Annual Revieiw of Entomology, 42, 611-643.

Godfray, H.C.J. (1994) Parasitoids: behavioral and evolutionary ecology. Princeton University Press, Princeton, New Jersey. Gols, R., Raaijmakers, C.E., van Dam, N.M, Dicke M., Bukovinszky, T. and Harvey, J.A. (2007) Temporal changes affect plant chemistry and tritrophic interactions. Basic and Applied Ecology. 8, 421-433. Gols, R., Wagenaar, R., Bukovinszky, T., van Dam, N.M, Dicke, M., Bullock, J.M. and Harvey, J.A. (2008) Genetic variation in defense chemistry in wild cabbages affects herbivores and their endoparasitoids. Ecology, 89, 1616-1626. Gols R., van Dam N. M., Raaijmakers C. E, Dicke M. and Harvey, J.A. (2009) Are population differences in plant quality reflected in the preference and performance of two endoparasitoid wasps? Oikos, 118, 733–743. Gonzáles, W.L. and Gianoli, E. (2003) Evaluation of induced responses, insect population growth, and host-plant fitness may change the outcome of tests of the preference- performance hypothesis: a case study. Entomologia Experimentalis et Applicata, 109, 211–216. González‐Santoyo, I. and Córdoba‐Aguilar, A. (2012) Phenoloxidase: a key component of the insect immune system. Entomologia Experimentalis et Applicata, 142(1), 1-16. Gouyon P .H., Fort, P .H., and Caraux, G. (1983) Selection of seedlings of Thymus vulgaris by grazing slugs, Journal of Ecology, 73, 299-306. Gripenberg, S., Mayhew, P.J., Parnell, M. and Roslin, T. (2010) A meta-analysis of preference– performance relationships in phytophagous insects. Ecology Letters, 13, 383-393. Hagele, B. F. and Rowell-Rahier, M. (2000) Choice, performance and heritability of performance of specialist and generalist insect herbivores towards cacalol and seneciphylline, two allelochemicals of Adenostyles alpina (Asteraceae). Journal of Evolutionary Biology, 13(1), 131-142. Hairston, N.G., Smith, F.E. and Slobodkin, L.B. (1960) Community structure, population control, and competition. American Naturalist, 94, 421-424. Halaj, J. and Wise, D.H. (2001) Terrestrial trophic cascades: How much do they trickle? American Naturalist, 157, 262–281 . Hare, J.D. (2002) Plant genetic variation in tritrophic interactions. Multitrophic level interactions, 8-43. Harper, J.L. (1977). Population biology of plants. Academic Press, London, UK. Harper, J.L. and White, J. (1974) The demography of plants. Annual review of ecology and systematics, 419-463. Harvey, J.A., Van Nouhys, S. and Biere, A. (2005) Effects of quantitative variation in allelochemicals in Plantago lanceolata on development of a generalist and a specialist herbivore and their endoparasitoids. Journal of Chemical Ecology, 31, 287-302. Hawkins, B.A., Cornell, H.V. and Hochberg, M.E. (1997) Predators, parasitoids, and pathogens as mortality agents in phytophagous insect populations. Ecology, 78, 2145-2152.

Hay, M.E., Dufy, J.E., Paul, V.J., Renaud, P.E. and Fenical, W. (1990) Specialist herbivores reduce their susceptibility to predation by feeding on the chemically defended seaweed Avrainvillea longicaulis. Limnology and Oceanography, 35(8), 1734-1743. Hay, M.E., Kappel, Q.E. and Fenical, W. (1994) Synergisms in plant defenses against herbivores: interactions of chemistry, calcification, and plant quality. Ecology, 75(6), 1714-1726. Hilker, M. and Meiners, T. (2002) Induction of plant responses to oviposition and feeding by herbivorous arthropods: a comparison. Entomologia Experimentalis et Applicata, 104(1), 181-192. Holton, M.K., Lindroth, R.L. and Nordheim, E.V. (2003) Foliar quality influences tree-

herbivore-parasitoid interactions: effects of elevated CO2, O3, and plant genotype. Oecologia, 137, 233-244. Honkavaara, J., Rantala, M.J. and Suhonen, J. (2009) Mating status, immune defence, and multi-parasite burden in the damselfly Coenagrion armatum. Entomologia Experimentalis Et Applicata, 132, 165-171. Howe, G.A. and Jander, G. (2008) Plant immunity to insect herbivores. Annual Review of Plant Biology, 59, 41-66. Iason, G.R., Dicke, M. and Hartley, S.E. (Eds.). (2012) The ecology of plant secondary metabolites: from genes to global processes. Cambridge University Press, Cambridge, UK. Jaenike, J. (1978) On optimal oviposition behavior in phytophagous insects. Theoretical Population Biology, 14, 350–356. Jamieson, M.A. and Bowers, M.D. (2010) Iridoid glycoside variation in the invasive plant Dalmatian toadflax, Linaria dalmatica (Plantaginaceae), and sequestration by the biological control agent, Calophasia lunula. Journal of chemical ecology, 36(1), 70-79. Jeffries, M.J. and Lawton, J.H. (1984) Enemy free space and the structure of ecological communities. Biological Journal of Linnean Society, 23, 269-286. Jensen, S.R., Franzyk, H. and Wallander, E. (2002) Chemotaxonomy of the Oleaceae: iridoids as taxonomic markers. Phytochemistry, 60(3), 213-231. Karl I., Stoks, R., De Block, M., Janowitz, S.A. and Fischer, K. (2011) Temperature extremes and butterfly fitness: Conflicting evidence from life history and immune function. Global Change Biology, 17, 676–687. Karban, R. and Baldwin, I.T. (1997) Induced responses to herbivory. University of Chicago Press. Karban, R., and Myers, J.H. (1989) Induced plant responses to herbivory. Annual Review of Ecology and Systematics, 331-348. Kelly, C.A. and Bowers, M.D. (2016) Preference and performance of generalist and specialist herbivores on chemically defended host plants. Ecological Entomology, 41(3), 308-316.

Koricheva, J., Nykanen,, H. and Gianoli, E. (2004)vMeta-analysis of trade-offs among plant antiherbivore defenses: Are plants jacks-of-all-trades, masters of all? American Naturalist, 163, E64-E75. Krieger, R.I., Feeny, P.P. and Wilkinson, C.F. (1971) Detoxication enzymes in the guts of caterpillars: an evolutionary answer to plant defences? Science, 172, 579–580. Krischik, V.A. and Denno, R.F. (1983) Individual, population, and geographic patterns in plant defense (Vol. 463). Academic Press, New York. Laitinen, M. L., Julkunen-Tiitto, R., Tahvanainen, J., Heinonen, J. and Rousi, M. (2005) Variation in birch (Betula pendula) shoot secondary chemistry due to genotype, environment, and ontogeny. Journal of Chemical Ecology, 31, 697-717. Lampert, E.C. and Bowers, M.D. (2010) Host plant influences on iridoid glycoside sequestration of generalist and specialist caterpillars. Journal of Chemical Ecology, 36, 1101-1104. Lavine, M.D. and Strand, M.R. (2002) Insect hemocytes and their role in immunity. Insect Biochemistry and Molecular Biology, 32, 1295–1309. L’Empereur, K.M. (1989) Iridoid glycoside and pyrrolizidine alkaloid sequestration by specialist Lepidoptera. PhD dissertation, Department of Chemistry, Colorado State University, Fort Collins, CO, USA. L’Empereur, K.M. and Stermitz, F.R. (1990a) Iridoid glycoside content of Euphydryas anicia (Lepidoptera: Nymphalidae) and its major host plant, Besseya plantaginea (Scrophulariaceae), at a high plains Colorado site. Journal of Chemical Ecology, 16, 187- 197. L’Empereur, K.M. and Stermitz, F.R. (1990b) Iridoid glycoside metabolism and sequestration by Poladryas minuta (Lepidoptera: Nymphalidae) feeding on Penstemon virgatus (Scrophulariaceae). Journal of Chemical Ecology, 16, 1495-1506. Lill, J.T. and Marquis, R.J. (2001) The effects of leaf quality on herbivore performance and attack from natural enemies. Oecologia, 126(3), 418-428. Lincoln, D. E., Murray, M. J., and Lawrence, B.M. (1986) Chemical composition and genetic basis for the isopinocamphone chemotype of Mentha citrata hybrids. Phytochemistry, 25,1857- 1863. Macel, M. and Vrieling, K. (2003) Pyrrolizidine alkaloids as oviposition stimulants for the cinnabar moth, Tyria jacobaeae. Journal of Chemical Ecology, 29, 1435-1446. Malcolm, S.B. and Brower, L.P. (1989) Evolutionary and ecological implications of cardenolide sequestration in the monarch butterfly. Experientia, 45(3), 284-295. Martins, C.H.Z., Cunha, B.P., Solferini, V.N. and Trigo, J.R. (2015) Feeding on Host Plants with Different Concentrations and Structures of Pyrrolizidine Alkaloids Impacts the Chemical- Defense Effectiveness of a Specialist Herbivore. PLoS ONE 10(10): e0141480. Mayhew, P.J. (1997) Adaptive patterns of host plant selection by phytophagous insects. Oikos, 79, 417–428. McCall, A.C. and Fordyce, J.A. (2010) Can optimal defence theory be used to predict the distribution of plant chemical defences?. Journal of Ecology, 98(5), 985-992.

McKay, A.F., Ezenwa, V.O. and Altizer, S. (2016a) Unraveling the costs of flight for immune defenses in the migratory monarch butterfly. Integrative and comparative biology, icw056. McKay, A.F., Ezenwa, V.O. and Altizer, S. (2016b). Consequences of food restriction for immune defense, parasite infection, and fitness in monarch butterflies. Physiological and Biochemical Zoology, 89(5), 389-401. McKean, K. A., Yourth, CP.P, Lazzaro, B.P and Clark, A.G. (2008) The evolutionary costs of immunological maintenance and deployment. Bmc Evolutionary Biology 8, 19. McKey, D. (1974) Adaptive patterns in alkaloid physiology. American Naturalist, 108, 305-320. McKey, D. (1979) The distribution of secondary compounds within plants. In: G. A. Rosenthal and D. H. Janzen (eds) Herbivores: Their interactions with secondary plant metabolites. Academic Press: New York, pp 55-133. Menge, B.A. (1992) Community regulation: under what conditions are bottom-up factors important on rocky shores? Ecology, 73, 755-765. Mihaliak, C. A., Couvet, C., and Lincoln, D.E. (1989) Genetic and environmental contributions to variation in leaf mono- and sesquiterpenes of Heterotheca subaxillaris. Biochemical Systematics and Ecology, 17, 529-533. Mody, K., Unsicker, S.B. and Linsenmair, K.E. (2007) Fitness related diet-mixing by intrapsecific host‐plant‐switching of specialist insect herbivores. Ecology, 88(4), 1012- 1020. Mooney, K.A., Pratt, R.T. and Singer, M.S. (2012) The tri-trophic interactions hypothesis: Interactive effects of host plant quality, diet breadth and natural enemies on herbivores. PloS One, 7, e34403. Moran, N. (1986) Benefits of host plant specificity in Uroleucon (Homoptera, Aphididae). Ecology, 67, 108–115. Moran, N. and Hamilton, W.D. (1980) Low nutritive quality as defense against herbivores. Journal of Theoretical Biology 86: 247–254. Moore, B.D., Andrew, R.L., Külheim, C. and Foley, W.J. (2014) Explaining intraspecific diversity in plant secondary metabolites in an ecological context. New Phytologist, 201(3), 733-750. Moyes, C.L., Collin, H.A., Britton, G. and Raybould, A.F. (2000) Glucosinolates and differential herbivory in wild populations of Brassica oleracea. Journal of Chemical Ecology, 26(11), 2625-2641. Murdoch, W.W. (1966) Community structure, population control, and competition. American Naturalist 100, 219-226. Murphy, S. M. (2004) Enemy-free space maintains swallowtail butterfly host shift. Proceedings of the National Academy of Sciences, 101(52), 18048-18052. Murphy, D.D., Wahlberg, N. and Hanski, I. (2004) Introducing Checkerspots. In Ehrlich, P. R. and Hanski, I. (Eds.) On the wings of checkerspots: a model system for population biology. Oxford University Press. pp 17.

Murphy, S.M., Lill, J.T., Bowers, M.D. and Singer, M.S. (2014) Enemy-free space for parasitoids. Environmental Entomology, 43(6), 1465-1474. Nelson, A. C. and Kursar, T.A. (1999) Interactions among plant defense compounds: a method for analysis. Chemoecology, 9, 81-92. Newton E., Bullock J.M. and Hodgson, D.J. (2009a) Bottom-up effects of glucosinolate variation on aphid colony dynamics in wild cabbage populations. Ecological Entomology, 34, 614– 623. Newton E. L., Bullock J. M. and Hodgson, D.J. (2009b) Glucosinolate polymorphism in wild cabbage (Brassica oleracea) influences the structure of herbivore communities. Oecologia, 160, 63–76. Nicholsorians, C.M. (1991) Environmentally induced differences in plant traits: consequences for susceptibility to a leaf-cutter ant. Ecology, 72, 1609-1623. Nieminen, M., Suomi, J., Van Nouhuys, S., Sauri, P. and Riekkola, M.L. (2003) Effect of iridoid glycoside content on oviposition host plant choice and parasitism in a specialist herbivore. Journal of Chemical Ecology, 29, 823-844. Nishida, R. (2002) Sequestration of defensive substances from plants by Lepidoptera. Annual Review Of Entomology. 47, 57-92. Nold, R. (1999). Penstemons. Timber Press. OR, USA. Noriyuki, S. and Osawa, N. (2012) Intrinsic prey suitability in specialist and generalist Harmonia ladybirds: a test of the trade-off hypothesis for food specialization. Entomologia Experimentalis et Applicata, 144, 279–285. Odum, H. T. and Pinkerton, R.C. (1955) Times speed regulator-the optimum efficiency for maximum power output in physical and biological systems. American Scientist, 43, 331- 343. Opitz, S.E.W. and Muller, C. (2009) Plant chemistry and insect sequestration. Chemoecology, 19, 117-154. Osier, T.L. and Lindroth, R.L. (2001) Effects of genotype, nutrient availability and defoliation on aspen phytochemistry and insect performance. Journal of Chemical Ecology, 27, 1289- 1313. Papaj, D. R. and Prokopy, R.J. (1989) Ecological and evolutionary aspects of learning in phytophagous insects. Annual Review of Entomology, 34, 315-350Pereyra, P.C. and Bowers, M.D. (1988) Iridoid glycosides as oviposition stimulants for the buckeye butterfly, Junonia coenia (Nymphalidae). Journal of Chemical Ecology, 14, 917- 928. Petschenka, G. and Agrawal, A. A. (2016). How herbivores coopt plant defenses: natural selection, specialization, and sequestration. Current Opinion in Insect Science, 14, 17-24. Price P. W., Bouton, C.E., Gross, P., McPheron, B.A, Thompson, J.N. and Weis, A.E. (1980) Interactions among three trophic levels: Influence of plants on interactions between insect herbivores and natural enemies. Annual Review of Ecology and Systematics, 11, 41–65.

Quintero, C. and Bowers, M.D. (2013) Effects of insect herbivory on induced chemical defences and compensation during early plant development in Penstemon virgatus. Annals of Botany, 112, 661-669. Quintero, C. and Bowers, M.D. (2012) Changes in plant chemical defences and nutritional quality as a function of ontogeny in Plantago lanceolata (Plantaginaceae). Oecologia, 168, 471-481. Quintero, C. and Bowers, M. D. (2011) Plant induced defenses depend more on plant age than previous history of damage: implications for plant-herbivore interactions. Journal of chemical ecology, 37(9), 992-1001. R Development Core Team. (2011) R: a language and environment for statistical computing. R Foundation for Statistical Computing http://www.R-project.org. R Foundation for Statistical Computing, Vienna. Rausher, M.D. (1979) Larval habitat suitability and oviposition preference in 3 related butterflies. Ecology, 60, 503–511. Rausher, M.D. (1984) Tradeoffs in performance on different hosts: evidence from within and between-site variation in the beetle Deloyala guttata. Evolution, 38, 582–595. Rausher, M.D. (1988) Is coevolution dead? Ecology, 69, 898–901. Raybould, A.F. and Moyes, C.L. (2001) The ecological genetics of aliphatic glucosinolates. Heredity, 87, 383-391. Rayor, L.S. and S. Munson. (2002). Larval feeding experience influences adult generalist predators to accept chemically defended prey. Entomologia Experimentalis et Applicata 104, 193-201. Renwick, J. A. A. and Chew, F. S. (1994). Oviposition behavior in Lepidoptera. Annual review of entomology, 39(1), 377-400. Renwick, J.A.A. and Radke, C.D.D. (1988) Sensory cues in host selection for oviposition by the cabbage butterfly, Pieris rapae. Journal of Insect Physiology, 34, 251-257. Reudler Talsma, J.H., Biere A., Harvey, J.A. and van Nouhuys, S. (2008) Oviposition cues for specialist butterfly-plant chemistry and size. Journal of Chemical Ecology, 34,1202–1212. Richards, L.A., Dyer, L.A., Smilanich, A.M. and Dodson, C.D. (2010) Synergistic effects of amides from two piper species on generalist and specialist herbivores. Jounral of Chemical Ecology, 36, 1105–1113. Richards, L.A., Lampert, E.C., Bowers, M.D., Dodson, C.D., Smilanich, A.M. and Dyer, L.A. (2012) Synergistic Effects of Iridoid Glycosides on the Survival, Development and Immune Response of a Specialist Caterpillar, Junonia coenia (Nymphalidae). Journal of Chemical Ecology, 38, 1276–1284. Rios, R. S., Marquis, R. J. and Flunker, J.C. (2008) Population variation in plant traits associated with ant attraction and herbivory in Chamaecrista fasciculata (Fabaceae). Oecologia, 156, 577-588. Roach, D.A. and Gampe, J. (2004) Age‐specific demography in Plantago: uncovering age‐ dependent mortality in a natural population. The American Naturalist, 164(1), 60-69.

Robinson, G.S., Ackery, P.R., Kitching, I.J., Beccaloni, G.W. and Hernández, L.M. (2010). Hosts ― A Database of the World’s Lepidopteran Hostplants. Natural History Museum, London. Satterfield, D.A., Wright, A.E., & Altizer, S. (2013) Lipid reserves and immune defense in healthy and diseased migrating monarchs Danaus plexippus. Current Zoology, 59(3), 393-402. Schmidt, O., Theopold, U. and Strand, M.R. (2001) Innate immunity and evasion by insect parasitoids. BioEssays, 23, 344–351. Schmid-Hempel, P. and Ebert, D. (2003) On the evolutionary ecology of specific immune defense. Trends in Ecology & Evolution, 18, 27-33. Scogings, P.F., Dziba, L.E. and Gordon, I.J. (2004) Leaf chemistry of woody plants in relation to season, canopy retention and goat browsing in a semiarid subtropical savanna. Austral Ecology, 29, 278-286. Scriber, J.M. (1983) The evolution of feeding specialization, physiological efficiency, and host races. Variable Plants and Herbivores in Natural and Managed Systems (ed. By R.F. Denno and M.S. McClure), pp. 373–412. Academic Press, New York, New York. Scriber, J. M. (1984) Host-plant suitability. In Chemical ecology of insects (pp. 159-202). Springer US. Scriber, J.M. (2005) A mini-review of the “feeding specialization/physiological efficiency” hypothesis: 50 years of difficulties, and strong support from the North American Lauraceae-specialist, Papilio troilus (Papilionidae: Lepidoptera). Trends in Entomology, 4, 1-42. Scriber, J. M. and Feeny, P. (1979) Growth of herbivorous caterpillars in relation to feeding specialization and to the growth form of their food plants. Ecology, 60(4), 829-850. Shonle, I., Klett, J.E. and Vickerman, L.G. (2004) Native Herbaceous Perennials for Colorado Landscapes. http://www.ext.colostate.edu/pubs/garden/07242.html. Colorado State University Cooperative Extension, Fort Collins, CO, USA. Siemens, D. H. and Mitchell-Olds, T. (1996) Glucosinolates and herbivory by specialists (Coleoptera: Chrysomelidae, Lepidoptera: Plutellidae): Consequences of concentration and induced resistance. Environmental Entomology, 25, 1344-1353. Sime, K. (2002) Chemical defence of Battus philenor larvae against attack by the parasitoid Trogus pennator. Ecological Entomology, 27, 337-345. Singer, M.C. (1972) Complex components of habitat suitability within a butterfly colony. Science, 176, 75-77. Singer, M.S. (2001) Determinant of polyphagy by a woolly bear caterpillar: a test of the physiological efficiency hypothesis. Oikos, 93, 194-204. Singer, M.C., Ng, D. and Thomas, C.D. (1988) Heritability of oviposition preference and its relationship to offspring performance within a single insect population. Evolution, 42, 977-985.

Singer, M.S., Rodrigues, D., Stireman III, J.O. and Carriere, Y. (2004) Roles of food quality and enemy-free space in host use by a generalist insect herbivore. Ecology, 85, 2747-2753. Siva-Jothy, M.T., Moret, Y. and Rolff, J. (2005) Insect immunity: an evolutionary ecology perspective. Advances in insect physiology, 32, 1-48. Smilanich, A. M., Dyer, L. A., Chambers, J. Q. and Bowers, M. D. (2009a) Immunological cost of chemical defence and the evolution of herbivore diet breadth. Ecology Letters, 12, 612–621. Smilanich, A.M., Dyer, L.A. and Gentry, G.L. (2009b) The insect immune response and other putative defenses as effective predictors of parasitism. Ecology, 90(6), 1434-1440. Smilanich, A.M., Vargas, J., Dyer, L.A. and Bowers, M.D. (2011) Effects of Ingested Secondary Metabolites on the Immune Response of a Polyphagous Caterpillar Grammia incorrupta. Journal of Chemical Ecology, 37, 239–245. Sourakov, A. (2015) You are what you eat: native versus exotic Crotalaria species (Fabaceae) as host plants of the Ornate Bella Moth, Utetheisa ornatrix (Lepidoptera: Erebidae: Arctiinae), Journal of Natural History, 49:39-40, 2397-2415. Stamp, N. (2003) Theory of plant defensive level: example of process and pitfalls in development of ecological theory. Oikos, 102, 672-678. Stamp, N.E. and Bowers, M.D. (1996) Consequences for plantain chemistry and growth when herbivores are attacked by predators. Ecology, 77, 535-549. Stamp, N.E. and Osier, T.L. (1998) Response of five insect herbivores to multiple allelochemicals under fluctuating temperatures. Entomologia experimentalis et applicata, 88(1), 81-96. Stermitz, F.R., Gardner, D.R., Odendaal, F.J. and Ehrlich, P.R. (1986) Euphydryas anicia (Lepidoptera: Nymphalidae) utilization of iridoid glycosides from Castilleja and Besseya (Scrophulariaceae) host plants. Journal of Chemical Ecology, 12, 1459-1468. Stermitz, F.R., Gardner, D.R. and McFarland, N. (1988) Iridoid glycoside sequestration by two aposematic Penstemon-feeding geometrid larvae. Journal of chemical ecology, 14(2), 435-441. Stermitz, F.R., Blokhin, A., Poley, C.S., Krull, R.E. (1994) Iridoid glycosides of additional Penstemon species. Phytochemistry, 37, 1283-1286. Strand, M.R. (2008) The insect cellular immune response. Insect Science, 15, 1–14. Strand, M.R. and Pech, L.L. (1995) Immunological basis for compatibility in parasitoid–host relationships. Annual Review of Entomology. 40, 31–56. Strauss, S. Y., Irwin, R.E. and Lambrix,V.M. (2004) Optimal defence theory and flower petal colour predict variation in the secondary chemistry of wild radish. Journal of Ecology, 92, 132-141. Sugumaran, M. (2002) Comparative biochemistry of eumelanogenesis and the protective roles of phenoloxidase and melanin in insects. Pigment Cell Research, 15(1), 2-9.

Sznajder, B. and Harvey, J.A.. (2003) Second and third trophic level effects of differences in plant species reflect dietary specialisation of herbivores and their endoparasitoids. Entomologia Experimentalis et Applicata, 109, 73-82. Teder, T. and Tammaru, T. (2002) Cascading effects of variation in plant vigour on the relative performance of insect herbivores and their parasitoids. Ecological Entomology, 27(1), 94- 104. Theodoratus, D. H. and Bowers, M.D. (1999) Effects of sequestered iridoid glycosides on prey choice of the prairie wolf spider, Lycosa carolinensis. Journal of Chemical Ecology, 25, 283-295. Thompson, J.N. (1988) Evolutionary ecology of the relationship between oviposition preference and performance of offspring in phytophagous insects. Entomologia Experimentalis et Applicata, 47, 3-14. Thompson, J.N. and Pellmyr, O. (1991) Evolution of oviposition behavior and host preference in Lepidoptera. The Annual Review of Entomology, 36, 65–89. Turlings, T.C.J., Tumlinson, J.H. and Lewis, W.J. (1990) Exploitation of herbivore-induced plant odors by host-seeking parasitic wasps. Science, 250, 1251-1253. Turlings, T.C.J. and Benrey, B. (1998) Effects of plant metabolites on the behavior and development of parasitic wasps. Ecoscience, 5, 321-333. Valladares, G. and Lawton, J.H. (1991) Host plant selection in the holly leaf-miner: does mother know best? Journal of Animal Ecology, 60, 227– 240. Vogelweith, F., Thiery, D., Quaglietti, B., Moret, Y. and Moreau, J. (2011) Host plant variation plastically impacts different traits of the immune system of a phytophagous insect. Functional Ecology, 25(6), 1241-1247. Von Gröger, D. and Johne, S. (1965) Uber das Vorkommen von peganin in Linaria-Arten. Planta Medica 13, 182-188. Underwood, D.L.A. (1994) Intraspecific variability in host plant quality and ovipositional preferences in Eucheira socialis (Lepidoptera: Pieridae). Ecological Entomology, 19, 245–256. Wallin, K.F. and Raffa, K.F. (2004). Feedback between individual host selection behavior and population dynamics in an eruptive herbivore. Ecological Monographs, 74(1), 101-116. Walker, M. and Jones, T.H. (2001) Relative roles of top-down and bottom-up forces in terrestrial tritrophic plant–insect herbivore– natural enemy systems. Oikos, 93, 177–87. Waterman, P.G. and Moles, S. (1989) Extrinsic factors influencing production of secondary metabolites in plants, In: E. A. Bernays (ed.). Insect-Plant Interactions. CRC Press, Boca Raton, FL, pp. 107-134. White, R.R. (1979) Foodplants of alpine Euphydryas anicia (Nymphalidae). Journal of the Lepidopterist Society, 33, 170-173. Whittaker, R.H. and Feeny, P.P. (1971) Allelochemics: chemical interactions between species. Science, 171, 757–770.

100

Wiklund, C. (1974) Oviposition preferences in Papilio machaon in relation to host plants of larvae. Entomolgia Experimentalis Et Applicata, 17, 189-198. Wiklund, C. (1981) Generalist vs. specialist oviposition behaviour in Papilio machaon (Lepidoptera) and functional aspects on the hierarchy of oviposition preferences. Oikos, 163-170. Wiklund, C. and Friberg, M. (2008) Enemy-free space and habitat-specific host specialisation in a butterfly. Oecologia, 157, 287–294. Wilson, L.M., Sing, S.E., Piper, G.L., Hansen, R.W., De Clerck-Floate, R., MacKinnon, D.K., and C. Randall. (2005) Biology and Biological Control of Dalmatian and Yellow Toadflax. USDA Forest Service, FHTET-05-13. Wink, M. (2010) Introduction. Annual Plant Reviews Volume 39: Functions and Biotechnology of Plant Secondary Metabolites, Second edition, 1-20. Zangerl, A.R. and Rutledge, C.E. (1996) The probability of attack and patterns of constitutive and induced defense: a test of optimal defense theory. American Naturalist, 599-608. Zvereva, E. L. and Rank, N.E. (2003) Host plant effects on parasitoid attack on the leaf beetle Chrysomela lapponica. Oecologia, 135, 258-267.