Eco-Physiological Causes and Consequences of Sexually Selected Color Variation in Dragonflies

ECO-PHYSIOLOGICAL CAUSES AND CONSEQUENCES OF SEXUALLY SELECTED COLOR VARIATION IN DRAGONFLIES

MICHAEL P. MOORE

Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy

Advisor: Ryan A. Martin

Department of Biology

CASE WESTERN RESERVE UNIVERSITY

August, 2019

CASE WESTERN RESERVE UNIVERSITY

SCHOOL OF GRADUATE STUDIES

We hereby approve the dissertation of

Michael P. Moore

Candidate for the degree of Doctor of Philosophy*

Committee Chair

Dr. Ryan A. Martin

Committee Member

Dr. Sarah E. Diamond

Committee Member

Dr. Michael F. Benard

Committee Member

Dr. Patrick D. Lorch

Date of Defense

May 13, 2019

* We also certify that written approval has been obtained for any proprietary material contained therein 1

TABLE OF CONTENTS

LIST OF TABLES ...... 8

LIST OF FIGURES ...... 10

ACKNOWLEDGEMENTS ...... 12

ABSTRACT ...... 14

CHAPTER 1: AN INTRODUCTION TO THE ECO-PHYSIOLOGICAL

CAUSES AND CONSEQUENCES OF SEXUALLY SELECTED COLOR

VARIATION IN DRAGONFLIES ...... 16

BACKGROUND ...... 16

Ecological causes of variation in sexually selected characters ...... 18

Consequences of sexual selection for variation in ecological adaptation ...... 21

ECO-PHYSIOLOGICAL CAUSES AND CONSEQUENCES OF

DRAGONFLY WING COLOR VARIATION ...... 23

A primer on wing coloration in dragonflies and damselflies ...... 24

Pachydiplax longipennis: a future model system? ...... 25

Research Overview ...... 27

Sexual selection on wing coloration ...... 27

Potential eco-physiological causes of sexually selected color variation ...... 28

Potential eco-physiological consequences of sexual selection on wing coloration .....30 2

CHAPTER 2: INTRASEXUAL SELECTION FAVORS AN

IMMUNE-CORRELATED ORNAMENT IN A DRAGONFLY...... 33

ABSTRACT ...... 33

INTRODUCTION ...... 34

METHODS ...... 37

Study System ...... 37

Study Site, Population Monitoring, and Phenotypic Measurements ...... 38

Do territory holding success and longevity vary with wing melanization? ...... 39

Do males with greater wing melanization receive less aggression from territorial rivals? ...... 41

Is wing melanization associated with fighting ability? ...... 43

Does wing melanization correlate with melanin-based immune responses? ...... 44

Statistical Analyses ...... 45

RESULTS ...... 46

Phenotypic variation ...... 46

Do territory holding success and longevity vary with wing melanization? ...... 46

Do males with greater wing melanization receive less aggression from territorial rivals? ...... 47

Is wing melanization associated with fighting ability? ...... 47 3

Does wing melanization correlate with melanin-based immune responses? ...... 48

DISCUSSION ...... 48

CHAPTER 3: TEMPERATURE SHAPES THE COSTS, BENEFITS, AND

GEOGRAPHIC DIVERSIFICATION OF SEXUAL COLORATION IN A

DRAGONFLY ...... 59

ABSTRACT ...... 59

INTRODUCTION ...... 60

METHODS ...... 63

Wing coloration in Pachydiplax longipennis ...... 63

Prediction 1: Wing coloration increases male body temperatures ...... 63

Prediction 2: Thermal consequences of wing coloration translate into performance variation ...... 66

Prediction 3: Wing coloration provides its greatest territorial advantages on the coldest and/or most thermally variable days ...... 67

Prediction 4: Reduced wing coloration in hottest parts of range ...... 69

RESULTS ...... 72

Prediction 1: Wing coloration increases male body temperatures ...... 72

Prediction 2: Thermal consequences of wing coloration translate into performance variation ...... 72

Prediction 3: Wing coloration provides its greatest territorial advantages 4

on the coldest and/or most thermally variable days ...... 73

Prediction 4: Reduced wing coloration in hottest parts of range ...... 73

DISCUSSION ...... 74

CHAPTER 4: TRADE-OFFS BETWEEN LARVAL SURVIVAL AND

ADULT ORNAMENT DEVELOPMENT DEPEND ON PREDATOR REGIME

IN A TERRITORIAL DRAGONFLY ...... 84

ABSTRACT ...... 84

INTRODUCTION ...... 85

METHODS ...... 88

Focal species and study overview ...... 88

Aim 1: ornament development and viability selection ...... 89

Ornament development ...... 91

Larval Survival ...... 92

Aim 2: performance, morphological, and physiological correlates of body condition ...... 93

RESULTS ...... 95

Aim 1: ornament development and viability selection ...... 95

Ornament development ...... 95

Aim 2: Performance and morphological correlates of body condition ...... 96 5

DISCUSSION ...... 97

CHAPTER 5: IMMUNE DEPLOYMENT INCREASES LARVAL

VULNERABILITY TO PREDATORS AND INHIBITS

ADULT LIFE-HISTORY TRAITS IN A DRAGONFLY ...... 106

ABSTRACT ...... 106

INTRODUCTION ...... 107

METHODS ...... 110

Study Species and Study Overview ...... 110

Collection and Manipulating Immune Deployment ...... 112

Melanin deposition and immune challenge duration ...... 113

Laboratory Experiment: Does larval immune deployment affect predator avoidance traits? ...... 114

Mesocosm Experiment: Is larval immune deployment costly for other larval life-history traits? ...... 115

Mesocosm Experiment: Does predation risk intensify any trade-offs between larval immune deployment and adult life-history traits? ...... 117

Statistical Analyses ...... 120

RESULTS ...... 121

Melanin deposition and immune challenge duration ...... 121

Laboratory Experiment: Does larval immune deployment affect 6 predator avoidance traits? ...... 121

Mesocosm Experiment: Is larval immune deployment costly for other larval life-history traits? ...... 121

Mesocosm Experiment: Does predation risk intensify trade-offs between larval immune deployment and adult life-history traits? ...... 122

DISCUSSION ...... 123

CHAPTER 6: EVOLUTIONARY TRADE-OFFS BETWEEN

LARVAL IMMUNE DEFENSE AND ADULT WING COLORATION

ACROSS DRAGONFLIES ...... 131

ABSTRACT ...... 131

INTRODUCTION ...... 132

METHODS ...... 135

Species, Collection, Maintenance ...... 135

Immune Assay, Wing Coloration, and Analysis ...... 136

RESULTS ...... 138

DISCUSSION ...... 138

CHAPTER 7: CONCLUSIONS AND FUTURE DIRECTIONS ...... 146

What is the adaptive significance of wing coloration in

Pachydiplax longipennis? ...... 146 7

Eco-physiological causes of variation in sexually selected traits ...... 150

Eco-physiological consequences of sexual selection ...... 154

Coda ...... 156

APPENDIX FOR CHAPTER 2 ...... 158

APPENDIX FOR CHAPTER 3 ...... 161

Measuring wing coloration and its association with other male traits ...... 161

Scoring wing coloration ...... 161

Relationship with thorax darkness ...... 163

Relationship with body mass ...... 164

Testing for the effects of body size on maximum lifting force ...... 164

Assessing potential thermal costs of wing coloration in the American Southwest ...165

APPENDIX FOR CHAPTER 4 ...... 174

Validation of OLS residuals as an estimate of body condition ...... 174

Comparing within-pool selection gradients ...... 181

APPENDIX FOR CHAPTER 5 ...... 190

APPENDIX FOR CHAPTER 6 ...... 203

Testing within-year immune differences among species ...... 203

Testing for the effects of body condition ...... 204

BIBLIOGRAPHY ...... 208

LIST OF TABLES

Chapter 2

Table 2.1…………………………………………………………………….54

Table S2.1………………………………………………………………….158

Chapter 3

Table 3.1…………………………………………………………………….79

Table S3.1………………………………………………………………….167

Table S3.2………………………………………………………………….168

Table S3.3………………………………………………………………….169

Table S3.4………………………………………………………………….170

Table S3.5………………………………………………………………….171

Table S3.6………………………………………………………………….172

Table S3.7………………………………………………………………….173

Chapter 4

Table 4.1…………………………………………………………………...102

Table S4.1………………………………………………………………….184

Table S4.2………………………………………………………………….185

Chapter 5

Table S5.1………………………………………………………………....192 9

Table S5.2………………………………………………………………….193

Table S5.3………………………………………………………………….194

Table S5.4………………………………………………………………….195

Table S5.5………………………………………………………………….196

Table S5.6………………………………………………………………….197

Table S5.7………………………………………………………………….198

Table S5.8………………………………………………………………….199

Table S5.9………………………………………………………………….200

Chapter 6

Table S6.1…………………………………………………………………..206

LIST OF FIGURES

Chapter 2

Fig. 2.1………………………………………………………………………55

Fig. 2.2………………………………………………………………………56

Fig. 2.3………………………………………………………………………57

Fig. 2.4………………………………………………………………………58

Fig. S2.1……………………………………………………………………159

Fig. S2.2……………………………………………………………………160

Chapter 3

Fig. 3.1………………………………………………………………………80

Fig. 3.2………………………………………………………………………81

Fig. 3.3………………………………………………………………………82

Fig. 3.4………………………………………………………………………83

Chapter 4

Fig. 4.1……………………………………………………………………...103

Fig. 4.2……………………………………………………………………...104

Fig. 4.3……………………………………………………………………...105

Fig. S4.1…………………………………………………………………….186

Fig. S4.2…………………………………………………………………….187

Fig. S4.3…………………………………………………………………….188 11

Fig. S4.4…………………………………………………………………….189

Chapter 5

Fig. 5.1……………………………………………………………………...128

Fig. 5.2……………………………………………………………………...129

Fig. 5.3……………………………………………………………………...130

Fig. S5.1…………………………………………………………………….201

Fig. S5.2…………………………………………………………………….202

Chapter 6

Fig. 6.1……………………………………………………………………...144

Fig. 6.2……………………………………………………………………...145

Fig. S6.1…………………………………………………………………….207

ACKNOWLEDGEMENTS

The work presented in this dissertation would not have been possible without the guidance and support of many people.

Case Western Reserve University supported much of this research financially, including with Oglebay Grants from the Department of Biology. The Theodore Roosevelt

Award from the American Museum of Natural History also provided key funding.

R. Martin invested in my ideas and always served as a valuable sounding board on projects and manuscripts. By empowering me to blaze my own path as I developed these research themes, he enabled me to exceed even my loftiest professional goals.

Despite my penchant for long, information-dense committee meetings, S. Diamond, M.

Benard, and P. Lorch were always extremely generous with their time and thoughts.

Their insight and ideas enhanced the research in this dissertation immeasurably.

I was fortunate to get scientific and professional guidance from three post-doctoral researchers during my time at CWRU: M. Dugas, K. Krynak, and L. Chick. Their insightful comments, kind encouragement, and, in one case, persistent nagging made me a better thinker and writer even if I didn’t realize or appreciate it at the time.

Some of the research presented in this dissertation was enriched by collaborations with C. Lis and I. Gherghel. Without their efforts, I’m not sure how parts of this dissertation would have come together.

I am deeply grateful for the huge amount of logistical and technical support that I received while conducting this work. A. Locci, J. Koonce, S. Brown, J. Miller, C. Brown and all of the staff at Squire Valleevue farm could not have been more accommodating of my field research. C. Miller and the Hathaway Brown Science Research & Engineering 13

Program enabled C. Lis to collaborate with me on portions of this work. A. Gilmore, whose own research gave me the idea of working on dragonflies, generously provided crucial data on fat stores during one particularly harrowing peer-review process. M.

Willis and P. Kalyanasundaram lent equipment and expertise that helped me acquire spectrophotometric readings. J. Larson, L. Robinson, and A. Wiecek all expertly assisted with animal care and data collection.

Finally, I have benefited from a tremendous network of people who support me. M. &

B. Moore provided unwavering encouragement during this boondoggle. H. & L. Rollins are the best companions that I could have ever hoped for in spite of, or perhaps because of, their hijinks. N. Kathman drank beers on the porch with me, even if he showed up later than I usually expected. B. Tumolo commiserated with me about peer reviewers and things not being so simple. Lastly, I remain indebted to H. Whiteman at Murray State

University for giving me the opportunity to start graduate school when no one else did.

Eco-physiological Causes and Consequences of Sexually Selected Color Variation

in Dragonflies

Abstract

MICHAEL P. MOORE

Many animals use elaborate adult traits to attract mates and intimidate rivals. However, the development of these sexually selected traits, and the reproductive interactions that confer their benefits, occur against a complex backdrop of environmental factors. When such features of the habitat modify the costs and benefits of displaying and developing these traits, environmental variation across space and time can shape their diversification.

Likewise, sexual selection on these characters may have consequences for how organisms interact with and adapt to different ecological contexts. Here, I explore these themes by investigating the interplay between sexually selected coloration, the external environment, and physiology in dragonflies. I begin by examining the adaptive function of wing pigmentation in a dragonfly (Pachydiplax longipennis), finding that this trait is intrasexually selected. Using this dragonfly system, I then document how interactions between the environment and an organism’s physiological state can drive the divergence in its sexually selected wing coloration. I first show that thermal physiology causes the performance benefits of wing coloration to depend on ambient temperature, and, as a result, males in the warmest parts of North America nearly lack this trait all together. I 15 next illustrate how improving an individual’s physiological condition to develop better sexually selected coloration can harm its juvenile survival in the presence of predators. I then consider how sexual selection could feed back to influence ecological adaptation by examining links between wing coloration and immune defense. I find that, in addition to several important ecological costs of deploying immune defenses during the larval stage

(e.g. predation vulnerability, delayed emergence), producing a strong immune response directly inhibits wing color development. Moreover, when comparing across species, I show that those species with more wing coloration tend to have weaker immune responses. Thus, due to proximate trade-offs with immune defense, sexual selection on wing coloration can slow ecological adaptation, or even potentially drive maladaptation, to parasites and pathogens. Overall, this work demonstrates how eco-physiology may be an important nexus for the ecological causes and consequences of sexually selected trait variation. 16

CHAPTER 1: AN INTRODUCTION TO THE ECO-PHYSIOLOGICAL CAUSES

AND CONSEQUENCES OF SEXUALLY SELECTED COLOR VARIATION IN

DRAGONFLIES

BACKGROUND

Professional and amateur naturalists have long been inspired by the elaborate, conspicuous traits that many adult animals display. Darwin (1871) proposed that the evolutionary origins of these characters are rooted in “sexual selection”, differential mating success that is non-random with respect to phenotype. Myriad studies indeed show that an individual’s ability to ward off rivals and attract mates increases with the size or exaggeration of these characters (Andersson 1994). Synthetic research further indicates that such advantages are consistently large: sexual selection tends to be stronger than components of natural selection that act through variation in survival and fecundity

(Kingsolver et al. 2001; Kingsolver et al. 2012; but see Crone 2001). This strong source of selection should promote the evolution of larger and/or more conspicuous sexually selected traits as a result (Lande 1981; Andersson 1994). However, both the development of these traits, and the reproductive interactions that confer their benefits, occur against a complex backdrop of environmental conditions to which animals must respond and adapt

(Endler 1992; Zuk & Kolluru 1998; Miller & Svensson 2014). Such factors can modify the relative survival costs and reproductive benefits of developing and possessing sexually selected traits. Consequently, spatial variation in the environment has the opportunity to potentiate the diversification of sexually selected characters (Maan &

Seehausen 2011). Likewise, as an individual’s sexually selected traits can influence its 17 survival in a habitat, the evolutionary effects of sexual selection may also feed back to shape how organisms adapt to ecological selective pressures (Lande 1980; Bonduriansky

2011).

In the ensuing chapters of this dissertation, I explore several ecological causes and consequences of sexually selected trait variation. I specifically focus on how such patterns can be mediated through “eco-physiology”—an organism’s adjustment or adaptation of its physiological state to differences in the external environment (Tracy et al. 1982). This work documents multiple ways in which the interplay between the environment and an organism’s physiology can cause divergence in sexually selected traits. It also shows that sexual selection can have consequences for interactions between an organism’s physiological state its environment. I begin this introductory chapter by summarizing major themes of research into the ecological causes of variation in sexually selected traits, as well as the consequences of sexual selection on these characters for ecological adaptation. I conclude each of these subsections by outlining the potential role of eco-physiological processes. Next, I move to a synopsis of my research into the eco- physiological causes and consequences of sexually selected trait variation in dragonflies.

I first describe the dragonfly research system that I use to address these topics. I then conclude with a narrative overview of the chapters of my dissertation—emphasizing how each chapter connects to the broader theme of eco-physiological causes and consequences of variation in sexually selected traits.

Ecological causes of variation in sexually selected characters

Sexually selected characters often exhibit remarkable variation among populations and closely related species (e.g. Craig & Foote 2001; Arnegard et al. 2010; Weber et al.

2016). Environmental variation is cited as one important cause of this diversification because features of the habitat can determine the balance between the survival costs and reproductive benefits of developing and displaying an elaborate sexual character

(Cornwallis & Uller 2010; Miller & Svensson 2014). For instance, in habitats where parasites or predators are abundant, increased conspicuousness to these enemies can make it more costly to possess an exaggerated sexually selected trait (Endler 1980; Zuk

& Kolluru 1998). Geographic variation in the prevalence or efficacy of these enemies then promotes the character’s divergence across the landscape (Lande 1981). Such a scenario has been observed in crickets that invaded islands with an acoustically orienting parasite and subsequently lost the ability to call (Zuk et al. 2006; Pascoal et al. 2014). In other environments, the benefits of displaying a very exaggerated character may be increased when features of the habitat reduce the effectiveness of less conspicuous traits

(Maan & Seehausen 2011). Environmental characteristics that disrupt the transmission of signals illustrate how this mechanism might shape variation in sexually selected characters (e.g. Morton 1975; Endler 1992). For example, bird vocalizations are diverging along urbanization gradients in response to the pressures associated with having to be heard in the noisier novel habitats (Patricelli & Blickley 2006; Halfwerk et al. 2011). While these case studies have greatly illuminated our understanding of how sexually selected traits diversify in response to concurrent reproductive and ecological selective pressures, recent reviews have highlighted that our understanding of the 19 environmental factors underlying among-population variation is mostly limited to these examples of species’ enemies and the signaling environment (Cornwallis & Uller 2010;

Maan & Seehausen 2011; Miller & Svensson 2014).

Interactions between the environment and an individual’s physiological state may be one important cause of divergence in sexually selected characters, though attention has largely been limited to a single process: condition-dependent phenotype expression

(Cornwallis & Uller 2010; Hill 2011; Morehouse 2014). For instance, differences in physiological condition among individuals commonly generate variation in sexually selected characters within a population (Morehouse 2014; but see Cotton et al. 2004).

This occurs because individuals in better condition have more energetic resources to allocate towards developing elaborate sexually selected characters (Hill 2011; Morehouse

2014). As physiological condition will almost always vary among individuals due to different experiences of the external environment (Cotton et al. 2004) and to mutational differences across the genome (Dugand et al. 2019), links to condition provide a nearly perpetual source of sexually selected trait variation within populations (Rowe & Houle

1996). Such processes can also translate into variation among populations (Maan &

Seehausen 2011). When the available dietary or nutritional resources differs among habitats, all else being equal, the ensuing variation in individuals’ mean condition can generate differences in sexually selected characters among populations as well (e.g. Craig

& Foote 2001; Allen & Miller 2014). Collectively, variation in physiological condition is one well-documented way in which the interplay between the environment and physiological state can drive variation in sexually selected traits. 20

Despite this focus on variation in physiological condition, interactions between the external environment and an individual’s physiological state are far more diverse than just variation in absolute levels of energetic reserves (Tracy et al. 1982; Hill 2011). This broader suite of eco-physiological processes could play an important role in the generation and maintenance of sexually selected trait variation (Hill 2011; Morehouse

2014; Vitousek et al. 2014). For example, circulating hormone levels used to develop exaggerated morphologies (e.g. testosterone in vertebrates, juvenile hormone in arthropods) can depend on the frequency of antagonistic interactions with rivals and enemies within a habitat (Vitousek et al. 2013; Tibbetts et al. 2016). Differences in the abundance of rivals may then cause variation in sexually selected traits by generating differences in the hormonal processes underlying their development. Similarly, in environments that increase the oxidative stress of their inhabitants, natural selection may favor preferential allocation of key molecules, like carotenoids, to combatting free radicals rather than producing vibrant coloration (von Schantz et al. 1999; Alonso-

Alvarez et al. 2004). Divergent adaptation of a shared metabolic pathway among habitats may then indirectly drive variation in sexually selected traits among populations.

Ultimately, many physiological processes associated with the expression of sexually selected traits are themselves quite responsive to the prevailing environmental conditions

(Hill 2011; Vitousek et al. 2014), and such interactions could shape the diversification of sexually selected traits. However, research considering how such mechanisms generate variation among populations remains relatively rare.

Consequences of sexual selection for variation in ecological adaptation

In addition to ecological factors causing variation in sexually selected traits, sexual selection may influence an organism’s ability to adapt to its environment (Darwin 1871;

Lorch et al. 2003; Bonduriansky 2011; Giery & Layman 2019). Sexual selection is thought to primarily hinder ecological adaptation because the average phenotype of individuals who are more successful at mating can diverge from the optimal phenotype for survival and fecundity (Lande 1980). Empirical evidence suggests this pattern may be common and can arise in at least a few different ways. Differences in sexual selection among populations encountering even the same ecological conditions, perhaps due to divergent female preferences (Rodríguez et al. 2013), can lead to different levels of ecological “maladaptation” in those habitats (Lande 1980). For example, if females strongly prefer traits that impose direct survival costs to those males that possess them, like brightly colored fish that are more conspicuous to predators (e.g. Endler 1980), then the average phenotype that evolves will be offset from the phenotype that confers the highest population-mean survival (Lande 1980; Bonduriansky 2011). Another ecological consequence of sexual selection arises when intense competition among males to achieve copulations leads to the harassment and harm of females (Chapman 2006). The ensuing, direct fitness costs incurred by females can be strong enough to reduce population-mean fitness (e.g. Rundle et al. 2006; Takahashi et al. 2014). Through these mechanisms, and others, sexual selection inhibits ecological adaptation.

In contrast to these scenarios, there are also several conditions under which sexual selection aids ecological adaptation (Lorch et al. 2003; Bonduriansky 2011; Servedio & 22

Boughmann 2017). For instance, when sexually selected traits are condition dependent, sexual selection that directly favors more elaborate phenotypes will also indirectly favor higher condition (van Nordwijk & de Jong 1986; Lorch et al. 2003). As an individual’s condition should reveal the effects of many loci on its performance in its habitat, sexual selection favoring higher condition can purge deleterious mutations and accelerate ecological adaptation (Lorch et al. 2003; Servedio & Boughmann 2017). Patterns of experimental evolution in systems with low sexual conflict support this hypothesis

(Jacomb et al. 2016; Yun et al. 2018; Dugand et al. 2019). Additionally, even if sexual selection at first hinders ecological adaptation, it can also enable populations to explore the adaptive landscape and evolve towards previously inaccessible ecological niches

(Bonduriansky 2011). When the reproductive benefits gained by maximizing sexually selected traits balance the viability and fecundity costs, the population will be displaced from its local optimum on the adaptive landscape and can evolve through phenotypic regions of low viability and fecundity. In some of these cases, the population will drift towards new “adaptive zones” where the sexually selected trait is co-opted for ecological function that it did not have in the population’s previous ecological niche (Giery &

Layman 2019). In this way, the multiple sources of natural selection on the character become aligned, and the evolution of higher population-mean fitness is possible

(Bonduriansky 2011). While this is an intriguing hypothesis, it has yet to be directly tested across multiple taxa (but see Svensson & Waller 2013). Nonetheless, although sexual selection has considerable opportunity to obstruct ecological adaptation, it also can hasten it under some conditions. 23

Because of the intimate link between an individual’s sexually selected traits and various components of its physiological state (Hill 2011; Kasumovic 2013), any changes in these characters may feed back to affect how an organism’s physiological state adjusts or adapts to prevailing environmental conditions (Rubenstein & Hauber 2008; Vitousek et al. 2014). For instance, experimentally augmenting the plumage of barn swallows causally decreases their circulating testosterone levels and their oxidative stress (Safran et al. 2008; Vitousek et al. 2013). Similarly, inducing male damselflies to produce greater wing ornamentation reduces the fat reserves that they have available for other ecological demands (Contreras-Garduño et al. 2011). Changes in an individual’s sexually selected traits therefore affect diverse components of its underlying physiological state. All else being equal, the evolution of sexually selected traits could then affect the evolution of those physiological parameters, potentially influencing how an organism interacts with other aspects of its environment (Giery & Layman 2019). In such scenarios, eco- physiological variation would not only be an ecological cause of variation in sexually selected traits, but it would be a conduit for the ecological consequences of sexual selection as well (Vitousek et al. 2014).

ECO-PHYSIOLOGICAL CAUSES AND CONSEQUENCES OF DRAGONFLY

WING COLOR VARIATION

To explore reciprocal interactions between eco-physiology and sexual selection, I developed a research program using wing coloration in dragonflies. To provide some background on what is known about the interplay between wing coloration and eco- physiology in this lineage, I first summarize some of the previous research into odonate 24 wing coloration and into the primary focal species of my work (Pachydiplax longipennis

Burmeister). I then outline the empirical research of my dissertation into these patterns.

A primer on wing coloration in dragonflies and damselflies

Many species of dragonflies and damselflies display pigmentation on the surface of their wings (Corbet 1999). Across lineages, these pigmented regions come in diverse colors and sizes, and can be expressed by one or both sexes (Svensson & Waller 2013).

Research dating back several decades shows odonate wing pigmentation functions in interactions with territorial rivals and potential mates (Waage 1979b; Grether 1996b;

Córdoba-Aguilar 2002; Guillermo-Ferreira et al. 2015). As these patches are energetically costly to produce and maintain, individuals with greater energetic stores are able to produce larger regions of pigmentation (Hooper et al. 1999; Contreras-Garduño et al. 2008). Thus, the trait has the capacity to convey information about the bearer’s physiological state to hetero- and conspecifics (Contreras-Garduño et al. 2006). During contests over reproductive territories, males of some species will use these color patters to signal their own fighting ability, as well as to assess the fighting ability of rivals

(Grether 1996b; Guillermo-Ferreira et al. 2015). Additionally, in some species, females appear to directly prefer to mate with males with more coloration (e.g. Moore 1990).

Melanin wing pigmentation is especially common among odonates (Hooper et al.

1999; Svensson & Waller 2013; Futahashi 2016). This pigment is particularly suitable for signaling to rivals and potential mates because: 1) it is energetically costly to synthesize melanin from dietary and metabolic precursors (reviewed in González-Santoyo &

Córdoba-Aguilar 2012); and 2) arthropods directly use the melanin-synthesis pathway for 25 many other functions, including defense against parasites and pathogens (Rolff & Siva-

Jothy 2003). Because so many vital physiological functions rely on this single, costly metabolic pathway (González-Santoyo & Córdoba-Aguilar 2012), it may be unsustainable for individuals to allocate disproportionally more of the shared dietary and metabolic precursors to wing coloration than to other traits (Rantala et al. 2000; Siva-

Jothy 2000). Melanin wing coloration may therefore readily evolve a communication function because it will convey accurate information about the bearer’s physiological condition (Lawniczak et al. 2007). However, beyond sexual selection (Waage 1979b;

Moore 1990; Cordoba-Aguilar 2002), many ecological selective pressures are thought to shape the evolution of melanin wing patterns, including predation (Svensson & Friberg

2007; Kuchta & Svensson 2014) and climate (Outomuro & Ocharan 2011; Svensson &

Waller 2013). Given these potential interactions between sexual selection, ecology, and physiology, melanin wing coloration in odonates could be well suited for examining eco- physiological causes and consequences of variation in sexually selected trait variation.

Pachydiplax longipennis: a future model system?

For the research of this dissertation, I primarily focused on the blue dasher dragonfly

(Pachydiplax longipennis Burmeister). This medium-sized libellulid dragonfly is abundant in northeast Ohio and has a broad distribution across central North America— encompassing at least southern Canada, most of the United States, and northern Mexico

(Paulson 2009; Paulson 2012; Moore et al. 2019). P. longipennis is univoltine throughout much of its range, including northeast Ohio, and is relatively unusual among odonates in that it can successfully inhabit ponds both with and without fish (McCauley et al. 2008). 26

Previous research using this species has considered its population ecology (e.g. Van

Buskirk 1987), energy budgets (e.g. Fried & May 1983), behavioral thermoregulation

(e.g. May 1976), and intraguild interactions (e.g. Crumrine 2005).

As adults, this species engages in resource defense polygyny (Emlen & Oring 1977), with males battling to found and defend reproductive territories on emergent vegetation.

Once a male selects a pond to attempt to reproduce in, it rarely disperses to others

(McCauley 2010). As with most territorial dragonfly species, females spend most of their time feeding in adjacent field and/or forest habitats (Johnson 1962; Sherman 1983;

Corbet 1999). When a female is ready to mate, she flies to the pond, rapidly finds a territorial male, directly oviposits in his territory under his guard, and then leaves

(Sherman 1983). Territorial males consequently gain the vast majority of the matings

(Sherman 1983). Moreover, because male odonates have structures to dislodge sperm from previous matings (Waage 1979a), the last male to mate with a female typically sires nearly all of her offspring (McVey & Smittle 1984; Hooper & Siva-Jothy 1996).

This highly aggressive, territorial species is sexually dichromatic. Males exhibit bright blue bodies and dark brown wing pigmentation, while females possess brown and black bodies and clear wings (Rosche et al. 2008; Paulson 2012). Such dichromaticism suggests a role for sexual selection (Andersson 1994). However, prior to the work described in this dissertation, very little was known what, if any, role these distinct sex- specific morphological features play in mating interactions (but see Robey 1975). In particular, nothing was known about the species’ wing coloration except that males mostly lack wing pigmentation in the western half of the United States (Paulson 2009).

Research Overview

Sexual selection on wing coloration

I began by describing the wing pigmentation of P. longipennis and testing one potential reproductive function: signaling of fighting ability during male-male competition

(Chapter 2; Moore & Martin 2016). This adaptive benefit of wing pigmentation had been documented in other territorial odonates (Grether 1996b; Córdoba-Aguilar 2002), and it therefore seemed like a plausible candidate function for P. longipennis. My study shows that males with more wing coloration are able to defend reproductive territories for longer throughout each day. An experimental manipulation revealed that males with more wing coloration receive less aggression from territorial rivals, demonstrating that this trait has some signaling function that intimidates other males. Males with more wing coloration also were shown to win more aerial battles with rivals, suggesting that wing coloration is associated with a male’s fighting ability. Finally, I found that males with more wing coloration are able to produce a larger melanin immune response. Given that the melanin is costly to produce in insects (González-Santoyo & Córdoba-Aguilar 2012), this suggests that males with more wing coloration are in better energetic states. Thus, one adaptive function of wing coloration is through signaling of condition and fighting ability to reproductive rivals.

Potential eco-physiological causes of sexually selected color variation

After documenting that wing pigmentation in male P. longipennis is an intrasexually selected character, the next two chapters of my dissertation examine mechanisms by which eco-physiological interactions can cause variation in wing coloration.

In chapter 3 (“Temperature shapes the costs, benefits, and geographic diversification of sexual coloration in a dragonfly; Moore et al. 2019), I explored the role that thermal physiology may play in shaping variation in wing coloration across North America.

Because darker coloration usually heats individuals by absorbing more incident light

(Clusella Trullas et al. 2007; Stuart-Fox et al. 2017), I specifically tested whether or not male wing coloration has heating properties that could influence their territorial interactions and drive diversification across the species’ range. Comparing males with natural differences and experimentally manipulated differences in wing coloration revealed that wing coloration heats males—the magnitude of which improves male flight performance under cool conditions but hinders it under warm conditions. Behavioral observations of a wild population indicated that these thermal effects translate into an additional territorial advantage for males with more wing coloration under cooler conditions. Finally, photographs taken by citizen scientists showed that males almost always produce extensive wing coloration in the coolest parts of the range, but display drastically reduced coloration in the hottest parts of the species’ range. Collectively, these results demonstrate that thermal physiology has the potential to underlie the diversification of a secondary sexual trait.

In chapter 4 (“Trade-offs between larval survival and adult ornament development depend on predator regime in a territorial dragonfly”, Moore & Martin 2018), I evaluated 29 evidence for another potential eco-physiological cause of variation in sexually selected traits: juvenile predation costs of the physiological condition used for developing wing color. While having a higher physiological condition tends to improve the development of an individual’s sexual phenotype (Hill 2011; Kasumovic 2013; Morehouse 2014), carrying around large amounts of accumulated resources can have detrimental consequences for an individual’s morphology and performance in some habitats (Gosler et al. 1995; Zamora-Camacho et al. 2014a). Trade-offs between survival and reproduction could then manifest over how an individual’s physiological condition directly interacts with features of the environment. To evaluate the potential for such environment-specific trade-offs, I used a wading pool experiment and several small laboratory studies. Comparing among those males that successfully emerged from this study indicated that those that began the study as larvae with better body condition tended to have better initial wing color development. However, in wading pools with large predatory heterospecifics (Anax junius), viability selection opposed high body condition.

Thus, there are habitat-specific costs to accumulating and storing energetic resources.

Finally, while laboratory studies demonstrated that high-condition individuals do not have inhibited escape performance, these larvae did have much larger abdomens, perhaps making them easier targets for capture by A. junius. Because differences in the trade-offs between juvenile survival and adult phenotype development can shape phenotypic variation among habitats, predation costs of high condition in the juvenile stage could be an important eco-physiological cause of variation in sexually selected traits.

Potential eco-physiological consequences of sexual selection on wing coloration

The final two empirical chapters of my dissertation provide insight into some of the eco-physiological consequences of sexual selection on wing coloration. Specifically, these projects by examine interactions between sexual selection and immune defense against parasites and pathogens.

Chapter 5 (“Immune deployment increases larval vulnerability to predators and inhibits adult life-history traits in a dragonfly”; Moore et al. 2018a) reports on a broader study of ecological and sexually selected costs of using melanin for immune defenses in the larval stage. To test one ecological cost of melanin immune deployment, I first examined how predation risk can limit the benefits of mounting a large immune response in the larval stage. I showed that mounting a large immune response reduces larval escape performance and delays emergence time. I also find that, in wading pools with predatory heterospecifics, those individuals that I induced to mount a larger immune response had lower survival. These results emphasize the potential for ecological context to limit the benefits of immune defense. From this same project, I also tested for sexually selected costs of mounting a stronger immune response. Such a cost could enable sexual selection to limit adaptation of immune defenses—a potentially important eco- physiological consequence of sexual selection. Indeed, individuals that I induced to mount a larger immune response during the larval stage subsequently had poorer wing color development as adults. Importantly, inducing larger immune responses did not reduce individuals’ conditions at emergence, on average, which suggests that these trade- offs between immune defense and wing coloration are not solely the result of the immune deployment depleting and individual’s total energetic reserves. Trade-offs between 31 immune defense and sexually selected wing coloration thus appear to directly arise over melanin synthesis and use. Overall, in addition to documenting ecological costs of melanin immune deployment, this study reveals that sexual selection can limit the benefits of mounting a larger melanin immune response and may slow the evolution of stronger immune defenses as a result.

In light of these apparent proximate trade-offs between sexually selected wing coloration and larval immune defense, Chapter 6 (“Evolutionary trade-offs between larval immune defense and adult wing coloration across dragonflies”) explores the co- evolution of adult wing coloration and larval immune defense. Evolutionary theory provides three hypotheses for how these traits may co-evolve. First, the adaptive decoupling hypothesis predicts that metamorphosis should break apart developmental genetic associations across life stages and enable these traits to evolve independently

(Moran 1994). In this scenario, there should be no relationship between these two traits across evolutionary timescales. Second, if these traits cannot be decoupled, allocation trade-offs may enable sexual selection to drive the evolution of greater allocation of melanin precursors to wing coloration at the expense of larval immune defense (van

Noordwijk & de Jong 1986; Stearns 1992). Here, we should observe a negative relationship between wing coloration and immune defense through evolutionary time.

Finally, if these traits cannot be decoupled, allocation trade-offs may force sexual selection to increase the acquisition of resources or the efficiency with which melanin is produced (van Noordwijk & de Jong 1986; Baydaev 2004). In this way, there will be a positive relationship between wing coloration and immune defense across evolutionary time scales (Reznick et al. 2000). To test these competing hypotheses, I compared the 32 larval immune defenses of six closely related dragonfly species that differed widely in their extents of male wing coloration. I found that species with more melanin wing coloration as adults had weaker melanin immune responses during the larval stage. This indicates that proximate trade-offs over melanin resources are accommodated into evolutionary trade-offs across species. Thus, sexual selection on wing coloration has the capacity to inhibit adaptation along an important eco-physiological dimension.

CHAPTER 2: INTRASEXUAL SELECTION FAVORS AN IMMUNE-

CORRELATED ORNAMENT IN A DRAGONFLY

Published in Journal of Evolutionary Biology Vol. 29, No. 11, pp. 2256-2265, 2016.

DOI: 10.111/jeb.12953

Authors: Michael P. Moore & Ryan A. Martin

Department of Biology, Case Western Reserve University, Cleveland, OH, USA

(Submitted: 6 May 2016; Returned for Revision: 3 July 2016; Accepted 19 July 2016)

ABSTRACT

Sexual signaling is predicted to shape the evolution of sex-specific ornamentation, and establishing the costs and benefits of ornamentation and the information that ornamentation provides to receivers is necessary to evaluating this adaptive function.

Here, we assessed the adaptive function of a common color ornament in insects, melanin wing ornamentation, using the dragonfly Pachydiplax longipennis. We hypothesized that greater ornamentation would improve territory holding success by decreasing aggression that males receive from territorial rivals, but that more ornamented males may have shorter lifespans. Using mark-recapture field observations, we found that more ornamented males had greater territory holding success, and that viability selection did not act on wing melanization. We then compared the aggression of territorial rivals to decoy males before and after experimentally augmenting wing melanization, finding that 34 males significantly reduced aggression following the manipulation. We next hypothesized that wing melanization would signal fighting ability to territorial rivals by reflecting condition via investment in the costly melanin-synthesis pathway. We observed a positive relationship between ornamentation and the likelihood of winning territorial disputes, suggesting that wing melanization provides information about fighting ability to rivals. We also found a positive relationship between melanin-based immune defense and ornamentation, supporting a link between the signal and condition. We conclude that wing melanization is a condition-related signal of fighting ability, and suggest that this may be a common mechanism promoting the evolution of melanin ornamentation.

INTRODUCTION

Sex-specific ornamentation is predicted to affect reproductive success through sexual signaling (Darwin 1871; Andersson 1994; Maynard Smith & Harper 2003). Specifically, sexual signals in animals functionally shape mating success by providing information about the individual to rivals or potential mates, thereby directly moderating the outcomes of behavioral interactions over territories or copulation opportunities (Maynard

Smith & Harper 2003; Searcy & Nowicki 2005; Lailvaux & Irshick 2006). For example, sex-specific ornamentation may influence competition over territories by signaling the bearer’s fighting ability to rivals, and facilitating the resolution of disputes quickly and without injury (Maynard Smith & Harper 2003; Arnott & Elwood 2009). As sexual selection should therefore strongly favor individuals with large, conspicuous ornaments, the adaptive value and evolutionary maintenance of signal function crucially depends on its ability to convey reliable information (Maynard Smith & Harper 2003). Understanding 35 the adaptive function of sex-specific ornaments thus necessitates directly evaluating the relationship between ornamentation and components of reproductive success, as well as characterizing the information being signaled to rivals and potential mates. However, investigations of the adaptive function of sex-specific ornamentation rarely integrate both approaches (Lailvaux & Irschick 2006)

Melanin is a phylogenetically widespread pigment underlying sex-specific ornaments in many animals, yet its role in sexual signaling has been relatively controversial (Stoehr

2006; Roulin 2016). While any positive covariance between ornamentation and aspects of condition may promote sexual signaling functions, insects use the melanin-synthesis enzymatic pathway not only for coloration but also for directly mounting immune responses and repairing wounds (Schmid-Hempel 2005; Siva-Jothy et al. 2005), providing a strong proximate link by which ornamentation may intrinsically reflect aspects of condition (Hill 2011; Roulin 2016). For instance, allocation trade-offs of melanin precursors (e.g. tyrosine) may promote signal reliability such that only males in the best condition can invest precursors in both immunocompetence and ornamentation

(i.e. Y-shaped acquisition-allocation model; Stoehr 2006; Hill 2011). Additionally, the high energetic costs of melanin synthesis may maintain signal reliability (e.g. Moret &

Schmid-Hempel 2000; Fedorka et al. 2004; Schwarzenbach & Ward 2006), whereby only males with the greatest energetic reserves can activate and maintain high overall levels of melanin synthesis (González-Santoyo & Córdoba-Aguilar 2012). As a consequence of this proximate link, melanin ornamentation may often reliably signal information about an individual’s condition, directly moderating the outcomes of interactions with rivals or potential mates, and causally shaping reproductive outcomes (Hooper et al. 1999; Siva- 36

Jothy 2000; Wittkopp & Beldade 2009). However, the relatively few tests of this sexually selected function of melanin coloration in insects have been equivocal (Lawniczak et al.

2007; Punzalan et al. 2008b; Izzo & Tibbets 2012), and its explicit function in intrasexual selection has received little attention in particular. Given this, and the growing recognition of similar pleiotropic effects in invertebrates and vertebrates (Ducrest et al.

2008; Roulin 2016), studies that directly estimate phenotypic selection on, and the information content of, melanin ornaments remain critical to understanding the adaptive function of this widespread pigment.

Using the sexually dimorphic dragonfly, Pachydiplax longipennis (Burmeister), we tested the adaptive function of a common melanin ornament in arthropods, wing melanization (Fig. 2.1). Similar to damselflies that display red wing ornaments (e.g.

Hetaerina americana, Grether 1996b), we hypothesized that wing melanization improves territory holding success, a vital component of mating success in odonates (reviewed in

Koenig 2008; Suhonen et al. 2008), by decreasing energetically costly and potentially injurious aggressive interactions with rivals. We also predicted that males with greater wing melanization would be more susceptible to predators and have higher mortality (e.g.

Grether 1997; Kuchta & Svensson 2014), and therefore have shorter reproductive lifespans. We further hypothesized that wing melanization signals fighting ability to rivals by proximately reflecting energetic reserves via the shared melanin-synthesis pathway. We thus predicted that ornamentation would be positively associated with the likelihood of winning territorial contests and with melanin-based immune defense.

METHODS

Study System

Pachydiplax longipennis (Burmeister) is a medium-sized dragonfly distributed broadly across North America. Females are brown with longitudinal yellow stripes, while males have blue abdomens and, in the eastern extent of their range where our study was conducted, express melanin coloration on the distal portion of their wings (Fig. 2.1;

Paulson 2012). Detailed descriptions of the mating system can be found elsewhere

(Johnson 1962; Robey 1975; Fried & May 1983; Sherman 1983). Briefly, on sunny days between mid-June and early August, males arrive at the pond between 0800 and 0900

EST, and defend territories on emergent vegetation around the perimeter of the pond until approximately 1530 EST. Males spend time in their territories perching on emergent vegetation and patrolling. Males frequently encroach upon rivals’ territories, engaging in aggressive disputes where the winner retains the territory and the loser either searches for a new territory or leaves the pond entirely. Once males have established at a pond for breeding, they rarely disperse (McCauley 2010). Females arrive at the pond throughout the day, mate with a male, and then oviposit in his territory while he hovers above her

(Sherman 1983). There are seldom more than one or two females on a pond at a given time, and copulations are very short, rarely lasting more than two minutes (Sherman

1983; Paulson 2012).

Study Site, Population Monitoring, and Phenotypic Measurements

We conducted this study at a small (perimeter = 140.2 m) research pond at Case

Western Reserve University’s Squire Valleevue Farm (Hunting Valley, Ohio, USA).

Males defend territories along two regions of emergent vegetation on opposite sides of the pond (primarily Typha sp.; 13.8 and 35.7 m, respectively). We captured males and uniquely marked their abdomens with four dots of acrylic paint (randomly chosen from among five different colors; Anderson et al. 2011). We kept males on ice in plastic bags before processing to slow their movement and facilitate safe handling (McCauley 2010).

To assess variation in wing melanization, body size, and wing size (two traits associated with odonate reproductive success; Koenig 2008), we took digital photographs (Canon

G15; Canon, USA, Inc., Lake Success, NY) of each male’s wings and body. We standardized the lighting conditions by taking pictures of males against a standard white background (DGK Color Tools®; Fig. 2.1) in a dark box that excluded ambient light. We attempted to include only males with little wing wear and fully developed abdomen coloration to minimize potential variation due to age differences (Grether 1996a;

Contreras-Garduño et al. 2008; McCauley 2010). In two cases, we recaptured and reprocessed males that had not developed their full abdomen coloration. Following processing, we released males from a common location approximately 10 m from the pond. While several ponds were within the average dispersal distance of breeding males

(430 m, McCauley 2010), no marked males were ever observed at any of these other ponds.

We quantified all traits from digitized photographs in ImageJ (Rasband 2012). Body size was calculated as the distance (mm) from the front of the head to the tips of the cerci. 39

Wing area was scored as the total area (mm2) of all four wings. Wing area and body size are highly correlated (r = 0.834, P < 0.001), and thus we calculated relative wing size by taking the standardized residuals from the linear regression of body size on total wing area. To quantify the extent of wing melanization, we identified the highest mean grey value (0-255, 0 = most opaque; 255 = most transparent) in ImageJ of the pigmented portion of each wing (i.e. least darkly pigmented), converted the photograph to binary black and white with this value as the threshold for black, and calculated the size (mm2) of the digitized black area. Wing melanization was then estimated as the proportion of the total wing area that was pigmented. To ensure consistency of measurements, we calculated repeatability by randomly resampling a subset of 16 individuals, and blindly rescoring each trait. The repeatability was high for all traits (all R > 0.99, all F15,16 >

259.2, all P < 0.001; Lessels & Boag 1987). We evaluated the phenotypic variation in the population by considering how wing melanization varied with body size and relative wing size using Pearson’s product-moment correlation coefficients, with individuals included as the unit of replication.

Do territory holding success and longevity vary with wing melanization?

We examined the fitness costs and benefits of our focal traits using behavioral observations of marked individuals. One observer (MPM) continuously circled the pond during the peak activity hours (0900-1530 EST) on all sunny days between 23-June and

24-July (n = 18), recording each marked male’s location, territorial behavior, and the time.

On most days, approximately 50% of the territorial males at the pond were marked.

Territorial males exhibit characteristic, unambiguous behaviors, such as perching in an 40 obelisk position or chasing other males (see Johnson 1962; Robey 1975). We quantified a male’s within-day territorial tenure as the amount of time (minutes) that it was territorial during that day. Males that were sighted only once were assigned a territorial tenure of zero minutes. We could not directly evaluate mate choice as we observed only four marked males copulating with females. However, as females choose among territorial males, a male’s territorial tenure is strongly correlated with mating success (Sherman

1983).

We assessed how daily territorial tenure varied with wing melanization, body size, and relative wing size using a generalized linear mixed-effects model with a negative binomial error distribution to account for overdispersion. We did not test interactive effects among traits because models including these terms did not converge, and preliminary visual assessment of the interactions suggested there were no strong effects between any combinations of the traits. To account for multiple territorial tenures of the same male among days, and the non-independence among males within a given day, we included random intercepts for individual identity and observation date, respectively. As territorial tenures of zero minutes were potentially misidentified males, we excluded these observations from analyses. However, our results are qualitatively robust to their inclusion. Using individuals’ daily territorial tenures, as opposed to total time spent defending a territory over the flight season, reduces bias resulting from variation in longevity (Hamon 2005).

We evaluated how wing melanization, body size, and relative wing size varied with minimum-estimated longevity using a generalized linear model with a quasipoisson error distribution to account for overdispersion. Each male’s minimum-estimated longevity 41

(hereafter: “longevity”) was calculated as the number of days between marking and the last day it was observed. As we conducted field observations regularly through the end of the flight period, this sample of days reflects a realistic period over which viability selection may act through differences in longevity. This metric is commonly used in odonate studies (e.g. Grether 1996a; Córdoba-Aguilar 2002), and is relatively robust to low resighting probabilities (Waller & Svensson 2015). We only included males marked prior to 5-July to control for differences in the available number of days between marking date and the end of the flight season. We only considered males that were observed at least twice after marking to ensure that we exclusively included resident males.

To estimate the strengths of intrasexual and viability selection on our focal traits, we used standard regression techniques to calculate selection gradients (Lande & Arnold

1983), dividing each individual’s fitness component (territorial tenure or longevity) by the population mean (i.e. relative fitness), and converting phenotypic values to mean of 0 and unit variance. To compare overall fitness variation acting through intrasexual selection versus viability selection, we also calculated the opportunities for sexual and viability selection by dividing the variance in each fitness component by its squared mean

(Arnold & Wade 1984).

Do males with greater wing melanization receive less aggression from territorial rivals?

To assess whether wing melanization has direct effects on the aggression received from rival males, we presented decoys (previously frozen males) to territorial males, experimentally augmented the ornamentation of the decoys, presented them to new males, 42 and compared the aggression received before and after the experimental manipulation

(see also Anderson & Grether 2010; Guillermo-Ferreira et al. 2015). We tethered decoys to a 2 m aluminum pole using clear nylon line (diameter = 0.18 mm), and presented them to territorial males. Typical of natural encounters between rivals (Suhonen et al. 2008;

McCauley 2010), males engaged the decoys by aggressively making direct, physical contact multiple times (mean number of strikes ± SD = 4.4 ± 5.1) over short durations

(mean number of seconds ± SD = 6.2 ± 4.6). For each presentation, we evaluated: whether the territorial male engaged the decoy (“engagement”, y/n); the time between the territorial male engaging the decoy and returning to a perch in its territory (“engagement duration”, seconds); how many times it struck the decoy (“strikes”, n); and the number of strikes per second (“aggression rate”, strikes per second). After presenting the decoy to several different territorial males (median: 4, range: 2 - 5), we augmented the ornamentation by homogenously coloring the wings distally from the nodus (Fig. S2.2) with a felt-tip marker chosen to approximate the natural color (Crayola® Cuppa’

Cappucino; sensu Anderson & Grether 2010; Guillermo-Ferreira et al. 2015). This degree of ornamentation is within the natural phenotypic range. We then presented the decoys to a different set of territorial males and again evaluated the interactions. While the same territorial males were often presented multiple decoys, many of these males were presented post-manipulation decoys before pre-manipulation decoys, and thus any observed effects of the manipulation on territorial male aggression are not confounded with territorial males becoming acclimated to the decoys.

We compared interactions before and after the manipulation using mixed-effects models with decoy identity as a random effect to account for multiple presentations of the 43 same decoy to the different males. For error distributions, we specified the binomial for engagement, negative binomial for number of strikes, and gaussian for engagement duration and aggression rate. Engagement duration was natural log transformed to improve normality of the residuals. Significance was assessed with likelihood ratio tests of models with and without the fixed effect for generalized linear mixed-effects models, and F-tests with the Kenward-Roger degrees of freedom approximation for linear mixed- effects models (Kenward & Roger 1997).

Is wing melanization associated with fighting ability?

To assess the potential for wing melanization to signal fighting ability to territorial rivals, we observed naturally occurring, aggressive interactions (e.g. chasing, striking) between males during the field observations described above. We defined the winner of a territorial contest as the male that succeeded in forcing the other male out of the territory.

Many contests included marked males, and we attempted to capture all unmarked males from contests. However, as contest losers frequently fly to another location on the pond or off the pond altogether, we rarely knew the phenotypic values of both males in a contest.

We analyzed the likelihood of winning a territorial contest using a generalized linear mixed-effects model with a binomial error distribution and contest outcome (win = 1, lose = 0) fitted as the response variable. We included wing melanization, body size, and relative wing size as fixed effects. To account for repeated measures of some individuals across disputes, and the non-independence of two individuals in the same contest when 44 both were known, we included random intercepts for both individual and contest identity, respectively.

Does wing melanization correlate with melanin-based immune responses?

To evaluate if wing melanization proximately signals an individual’s condition via investment in the shared melanin-synthesis pathway (sensu Siva-Jothy 2000), we considered the relationship between a male’s wing melanization and its ability to mount a melanin-based immune responses. We captured and assayed the melanin-based immune responses of 33 territorial males over two days (28-July and 3-August). To assay immunocompetence, we inserted a piece of sterilized nylon monofilament (mean length ±

SD = 2.70 ± 0.38 mm, diameter = 0.18 mm) into the body cavity dorsally between the fifth and sixth abdominal segments, and allowed the males’ immune systems to react to it for 24 hours. A pilot study indicated that 24 hours provided the strongest and most variable immune responses in this species (now reported in Chapter 5). During this period, males were kept in plastic bags in a dark climate chamber set to 8.2 °C. While this treatment may have slowed the rate of melanin immune responses, it was necessary to prevent adults from injuring themselves during or after the surgical implantation, and we do not expect that any temperature effect will have differentially affected males with different phenotypes. We then dissected out the implants and stored them in a freezer set to -22.8 °C. In four cases, the implant fell out or was accidentally inserted into the midgut, and these males were not considered further.

We quantified the melanin-based immune responses from digitized photographs of the implants. Using a dissecting microscope with a brightfield background, we took one 45 photograph of the implant, rotated the implant 90°, and then took another photograph. In every photograph, we also had a nylon monofilament that was not inserted into any males as a negative control. We used ImageJ to assess the mean gray values (0 = opaque, 255 = transparent) of each implant. To calculate an immune response score for each male, we subtracted the mean gray values of the implant from the control nylon for each picture, and averaged the two pictures. Higher scores indicate a darker implant and therefore a stronger melanin-based immune response, and previous work indicates this standard technique reflects resistance to naturally occurring pathogens (Rantala & Roff 2007).

Using the procedure described above (see also Lessells & Boag 1987), the repeatability of this metric was high (R = 0.995, F15,16 = 404.00, P < 0.001).

We used a linear model to consider how immune response scores varied with wing melanization, body size, and their interaction. To account for potential variation between the two dates when males were captured, we also initially included date and its interactions in the model. As these interactions were not significant (all F < 1.75, P >

0.201), we removed them from the model and re-tested effects. The immune score from one very large male was an outlier, and was not included in the analysis to improve the normality of model residuals. However, our results are qualitatively robust to its inclusion.

Statistical Analyses

All statistical analyses were conducted using R v. 3.1.2 (R Core Team, 2014). Mixed- effects models were fit using the ‘lme4’ package (Bates et al. 2014). To account for large scaling differences among three focal phenotypes, wing melanization, body size and 46 relative wing size were z-transformed for all analyses (Schielzeth 2010). All model parameter estimates and selection gradients are reported as estimate ± SE.

RESULTS

Phenotypic variation

We first assessed the phenotypic variation of males in our population. The mean body size ± SD of males was 38.06 ± 2.18 mm, and the mean wing melanization ± SD was

0.41 ± 0.09 (n = 115). There was no relationship between body size and wing melanization (r = -0.090, t114 = -0.96, P = 0.339; Fig S2.1a), or between relative wing size and wing melanization (r = -0.143, t114 = -1.54, P = 0.126; Fig S2.1b).

Do territory holding success and longevity vary with wing melanization?

In total, we observed 126 territorial tenures across 45 males (mean ± SD per individual: 2.8 ± 2.2 tenures), with a mean tenure ± SD of 104.9 ± 84.2 minutes. Males with greater wing melanization had longer territorial tenures (0.348 ± 0.090; χ2 = 13.99, df = 1, P = 0.002; Fig. 2.2). In contrast, territorial tenure did not vary with body size (-

0.023 ± 0.077; χ2 = 0.09, df = 1, P = 0.767) or relative wing size (-0.057 ± 0.088; χ2 =

0.43, df = 1, P = 0.515). Overall, the opportunity for intrasexual selection was 0.707, and the estimated strength of intrasexual selection on wing melanization was relatively strong

(β = 0.326 ± 0.104; cf. Kingsolver et al. 2012). We report non-significant intrasexual selection gradients in Table S2.1. 47

Of the 24 males included in the longevity analyses, the mean longevity ± SD was 17.4

! ± 5.4 d. Longevity was not associated with wing melanization (0.078 ± 0.066, �! = 1.43,

! P = 0.232), body size (-0.096 ± 0.079, �! =1.49, P = 0.222), or relative wing size (-0.022

! ± 0.093, �! = 0.05, P = 0.817). Overall, the opportunity for viability selection was 0.097, and we report the non-significant viability selection gradients in Table S2.1.

Do males with greater wing melanization receive less aggression from territorial rivals?

Regardless of experimental treatment, decoys were equally likely to be engaged by

! territorial males (Pre vs Post: 0.719 ± 0.479, �! = 2.31, P = 0.129). However, engagement durations were shorter following the manipulation (-0.429 ± 0.201, F1,46.9 =

! 4.46, P = 0.040). Similarly, decoys received fewer strikes (-0.951 ± 0.367, �! = 6.12, P =

0.0134; Fig. 2.3a) and lower aggression rates (-0.441 ± 0.121, F1,43.7 = 12.98, P < 0.001;

Fig. 2.3b) after augmenting the decoy’s ornamentation.

Is wing melanization associated with fighting ability?

We observed 155 territorial contests (94 winners, 73 losers; mean contests ± SD per

! individual: 2.8 ± 3.1). Males with greater wing melanization (0.484 ± 0.181; �! = 6.50, P

! = 0.011) and larger body sizes (0.433 ± 0.176; �! = 5.39, P = 0.020) were more likely to win contests (Fig. 2.4). Relative wing size was not associated with the probability of

! winning contests (0.213 ± 0.166; �! = 1.72, P = 0.189).

Does wing melanization correlate with melanin-based immune responses?

Immune response scores increased with wing melanization, and marginally decreased with body size (Table 2.1). Immune response scores were also marginally different between the two capture dates (Table 2.1).

DISCUSSION

We investigated the adaptive function of melanin ornamentation in the sexually dimorphic dragonfly, Pachydiplax longipennis. Following patterns in damselflies with red wing ornaments (Grether 1996b), we hypothesized that greater wing melanization would improve territory holding success by decreasing aggression that males receive from territorial rivals, but that benefit may come at the cost of a shorter reproductive lifespan (Grether 1997). We also hypothesized that wing melanization would signal fighting ability to territorial rivals by proximately reflecting condition via investment in the costly melanin-synthesis pathway (Hooper et al. 1999; Siva-Jothy 2000; Rantala et al.

2000). Overall, our results provide broad support for the hypothesized adaptive function of male wing melanization as a sexual signal of fighting ability to territorial rivals.

Male fitness is determined by the combined effects of intrasexual, intersexual, and viability selection (Arnold & Wade 1984; Hamon 2005). Intrasexual selection via territorial occupancy is crucial to male reproductive success in territorial odonates

(Moore 1990; Grether 1996b), and especially in P. longipennis (Sherman 1983). Among the traits considered, only wing melanization improved territory holding success (Fig.

2.2), and our selection analysis indicated strong intrasexual selection on this ornament (cf.

Kingsolver et al. 2012). Although we are unable to assess the strength of intersexual 49 selection in this study, other selection analyses in odonates have found positive (e.g.

Córdoba-Aguilar 2002) or no (e.g. Grether 1996b) intersexual selection on wing ornamentation. Contrary to many studies of other odonates (reviewed in Koenig 2008), and despite a positive relationship between body size and the likelihood of winning territorial disputes (Fig. 2.4b), we did not observe selection on body size. As small males are likely to lose fights and never acquire a territory, it is possible that, by including primarily males that had already acquired territories, our estimates of selection on body size (and all other traits) may be conservative. While we also predicted that increased melanization may come at the cost of reproductive lifespan (e.g. Grether 1997; Kuchta &

Svensson 2014), we found no evidence for viability selection against wing melanization.

However, the opportunity for viability selection (0.097) was considerably lower than the opportunity for intrasexual selection (0.707), and therefore intrasexual selection may be relatively more important for generating variance in male fitness in this population, further suggesting strong overall fitness benefits of ornamentation.

Our results indicate that the functional benefit of wing melanization in P. longipennis is through decreased aggression received from rival males. Territorial disputes in odonates are extremely energetically costly, and consume most of an individual’s daily energy budget (Fried & May 1983; Koskimäki et al. 2004). As males do not feed while they are defending territories (Fried & May 1983), any decrease in the energy expended on battling intruders may increase the time over which a male holds its territory, and therefore also increase the likelihood of mating (Suhonen et al. 2008). Moreover, in contrast with many other odonates, P. longipennis males make physical contact during territorial disputes, greatly increasing the chances of severe injury or death (Sherman 50

1983; McCauley 2010; Paulson 2012). Our results suggest that rivals challenge males with greater ornamentation less frequently and with lower intensity (Fig. 2.3), potentially improving a male’s ability to remain in a territory and/or avoid injury. Similar to damselflies expressing red wing ornaments (Grether 1996b; Guillermo-Ferreira et al.

2015), it seems likely that a primary functional advantage of greater melanin ornamentation is through decreased aggression from territorial rivals.

The information provided to rivals and potential mates is crucial to the evolution of sexual signals, and those involved in male-male competition are predicted to evolve to ensure that disputes are resolved as cheaply as possible for both parties (Maynard Smith

& Harper 2003; Searcy & Nowicki 2005). In particular, territorial males should assess the fighting ability of their rival and avoid engaging them if they are unlikely to win the dispute or if the cost of winning the dispute is too high (Maynard Smith & Harper 2003;

Arnott & Elwood 2009). As wing melanization is associated with a male’s ability to win contests (Fig. 2.4), rivals would benefit from avoiding costly and/or potentially injurious disputes with highly ornamented males (Junior & Peixoto 2013; Guillermo-Ferreira et al.

2015). Indeed, males with experimentally increased ornamentation received significantly lower aggression from rivals (Fig. 2.3). As the fitness benefits of displaying a large ornament are great, the signal must be reliable, on average, to be evolutionarily maintained, otherwise males would cease to respond (Maynard Smith & Harper 2003). In many cases, melanin ornaments in arthropods may signal condition reliably because of the biochemical link between immune defense and coloration (Table 2.1). While a male’s immunocompetence may not be specifically informative to rivals, the high costs of the melanin-synthesis pathway (González-Santoyo & Córdoba-Aguilar 2012) will ensure that 51 only males in the best condition have the capacity to allocate sufficient resources (e.g. energy, precursors, enzymes) to support robust immune function and large wing ornaments (e.g. Hooper et al. 1999; Rantala et al. 2000; Siva-Jothy 2000). The maintenance of melanin wing ornaments in insects also requires some degree of constant pigment deposition (Hooper et al. 1999; True et al. 1999), further enabling the ornament to reflect the individual’s present physiological state. As there is a strong relationship between condition and fighting ability in insects (reviewed in Vieira & Peixoto 2013), and investment in many other condition-related traits ends at metamorphosis or maturity

(e.g. body size), the melanin-synthesis pathway may be a common mechanism by which arthropods reliably signal proximate information about their condition and fighting ability to rivals.

While the shared biochemical pathway linking immune defense and coloration promotes signal reliability, intrasexual selection will also favor males that maximize signal efficiency (Badyaev 2004; Stoehr 2006), which will have consequences across the melanin-synthesis pathway. Indeed, artificial selection experiments (Armitage & Siva-

Jothy 2005) and studies of natural populations experiencing divergent selection pressures

(Fedorka et al. 2013) have found that selection on melanin coloration often promotes the correlated evolution of melanin immune defense. As with most sexual signals (reviewed in Hill 2011; Morehouse 2014; but see Craig & Foote 2001), the specific targets of selection for improving the efficiency of melanogenesis are unknown. However, two evolutionary outcomes for the melanin-synthesis pathway seem most likely: 1) energetic resources may be more efficiently allocated to the production of melanin-synthesis enzymes (e.g. phenoloxidase), and 2) amino acid precursors may be more readily 52 available for conversion to melanin. It remains to be seen which of these two outcomes is more likely or whether either is general across species, since some studies of the melanin- synthesis pathway indicate energetic limitations (e.g. Cotter et al. 2010), while others report precursor limitations (e.g. Srygley et al. 2009). Importantly, as both immune defense and coloration depend on many of the same precursors and enzymes, selection for increased signaling efficiency may be unable to completely erode the association between coloration and condition, and the signal will remain reliable, on average.

Nevertheless, if intrasexual selection favors greater efficiency of signal production, then populations exhibiting stronger intrasexual selection on wing melanization should express more efficient melanogenesis.

The evolution of sex-specific ornamentation depends in part on the functional mechanisms by which the ornament affects reproductive success (Andersson 1994;

Lailvaux & Irschick 2006). We found strong support for a sexual signaling function shaping the evolution of sex-specific wing melanization in a territorial dragonfly.

Melanin coloration exhibits a diverse suite of adaptive functions in arthropods (e.g.

Punzalan et al. 2008c; Fedorka et al. 2013; Debecker et al. 2015), but given its frequently observed condition dependence (e.g. Talloen et al. 2004; Punzalan et al. 2008a), the ability to act as a sexual signal of condition may commonly shape its evolution. Although any direct or indirect positive covariance between ornamentation and aspects of condition may ultimately facilitate condition-related sexual signaling of an ornament, linkages underlain by shared condition-dependent developmental pathways are predicted to most readily evolve via sexual signaling functions (Hill 2011; Roulin 2016). As the melanin- synthesis pathway links ornamentation and immunocompetence in arthropods, sexual 53 signaling, like that observed here, may indeed prove to be a common adaptive function shaping the evolution of sex-specific melanin coloration.

Table 2.1. Variation in immune response scores of 28 territorial males as a function of

2 wing melanization, body size, and date (multiple R = 0.306, F3,24 = 3.53, P = 0.030).

Model estimates (± SE) were obtained from after removing the non-significant wing melanization x size interaction. Wing melanization and body size were z-transformed prior to analysis, and date represents the difference between the two days over which males were captured. All partial F-tests were on 1 and 24 degrees of freedom.

Effect Estimate F P Wing Melanization 7.381 ± 3.110 5.69 0.026 Body Size -5.418 ± 2.808 3.76 0.065 Date 9.466 ± 5.414 2.94 0.100 Wing Melanization x Body Size 1.26 0.274

55 a) b)

Figure 2.1. Wings of two mature males with relatively low (a) and high (b) wing melanization. Note that the pigmentation at the base of the wing is less variable among males, and is also expressed in females. Pigmentation in the distal portion of the wing is sexually dimorphic and highly variable among males.

300

200

100 Territorial tenure (mins) tenure Territorial

0 −2 −1 0 1

Wing melanization [z−transformed]

Figure 2.2. Territorial tenure increased with wing melanization. Each point represents the duration of time within a day that an individual held a territory, and points are jittered horizontally by 0.05 to improve visual clarity. The regression line is fitted from the mixed-effects model reported in the Results, and wing melanization was z-transformed to improve scaling among explanatory variables.

57 a) 15

10 Strikes (n) Strikes 5

0 b)

2.0

1.5

1.0

0.5

Aggression rate (strikes/second) Aggression rate 0.0 Pre Post

Manipulation

Figure 2.3. Territorial males exhibited reduced aggression (a: total number of strikes; b: number of strikes per second) to decoy males following experimental increase of wing melanization. Each circle represents a presentation of a decoy to a territorial male, and points in panel a are jittered vertically by 0.1 to improve visual clarity. Squares represent the model-estimated means ± SE.

a) 1.00

0.75

0.50

0.25

Probability of winning contest 0.00

−2 −1 0 1 2

Wing melanization [z−transformed] b) 1.00

0.75

0.50

0.25

Probability of winning contest 0.00

−3 −2 −1 0 1 2

Body size [z−transformed (mm)]

Figure 2.4. The likelihood of winning a territorial contest increased with wing melanization (a) and body size (b). Each point represents the outcome of a contest for an individual (0 = contest loss, 1 = contest won), and points were jittered vertically by 0.1 to improve visual clarity. Regression lines are fitted from the mixed-effects model reported in the results, and both explanatory variables were z-transformed.

CHAPTER 3: TEMPERATURE SHAPES THE COSTS, BENEFITS, AND

GEOGRAPHIC DIVERSIFICATION OF SEXUAL COLORATION IN A

DRAGONFLY

Published in Ecology Letters Vol. 22, No. 3, pp. 437-446, 2019. DOI: 10.1111/ele.13200

Authors: Michael P. Moore1, Cassandra Lis2, Iulian Gherghel1, and Ryan A. Martin1

1. Department of Biology, Case Western Reserve University, Cleveland, OH USA

2. Hathaway Brown School, Shaker Heights, OH USA

(Submitted: 3 July 2018; Returned for Revision: 19 October 2018; Accepted: 10

November 2018)

ABSTRACT

The environment shapes the evolution of secondary sexual traits by determining how their costs and benefits vary across the landscape. Given the thermal properties of dark coloration generally, temperature should crucially influence the costs, benefits, and geographic diversification of many secondary sexual color patterns. We tested this hypothesis using sexually selected wing coloration in a dragonfly. We find that greater wing coloration heats males—the magnitude of which improves flight performance under cool conditions but dramatically reduces it under warm conditions. In a colder region of the species’ range, behavioral observations of a wild population show that these thermal effects translate into greater territorial acquisition on thermally variable days. Finally, 60 geo-referenced photographs taken by citizen scientists reveal that this sexually selected wing coloration is dramatically reduced in the hottest portions of the species’ range.

Collectively, our results underscore temperature’s capacity to promote and constrain the evolution of sexual coloration.

INTRODUCTION

Many animals produce elaborate, sex-specific traits to attract mates and intimidate rivals

(Darwin 1871). While more exaggerated characters typically improve mate acquisition

(Andersson 1994), reproductive interactions occur against a complex backdrop of environmental factors that can modify the relative costs and benefits of these traits

(Cornwallis & Uller 2010; Maan & Seehausen 2011; Miller & Svensson 2014). Some environments magnify the reproductive benefits of elaboration, as when only the loudest vocalizations can be transmitted through the noisiest habitats and detected by females

(Patricelli & Blickley 2006; Halfwerk et al. 2011). Other environments increase the costs of elaboration, for example when more intense sexual signaling increases detection by parasites and predators (Endler 1980; Zuk et al. 2006). In the many cases where the prevailing conditions of the habitat govern the magnitude of the costs and benefits of a secondary sexual trait, environmental variation across space and time will promote the phenotype’s diversification (Andersson 1994; Wiens 2001). However, despite the wide array of environmental factors that could plausibly regulate the fitness effects of these traits (Cornwallis & Uller 2010), our understanding of the environmental causes of their diversification remains largely limited to variation in the signaling environment and/or species’ enemies (Maan & Seehausen 2011; Miller & Svensson 2014). 61

For the sexually selected color patterns displayed by many animals, temperature could be one widely overlooked driver of their costs, benefits, and ultimately diversification. As darker colors absorb more emitted light and lead to greater heating for a given level of solar radiation (Watt 1968; Clusella Trullas et al. 2007; Stuart-Fox et al. 2017), coloration for signaling rivals and potential mates could also raise body temperatures above ambient environmental temperatures. Such effects of sexual coloration may confer pronounced reproductive benefits at low environmental temperatures by raising an individual’s body temperature closer to its optimum for fighting rivals and courting potential mates (e.g. Huey & Kingsolver 1989; Punzalan et al. 2008c). In contrast, at high environmental temperatures, thermal effects of sexual coloration could increase an individual’s body temperature beyond the optimum for mate acquisition and/or for other vital functions like feeding or predator avoidance (West & Packer 2002). Given the considerable differences in environmental temperature between many populations and between many closely related species (Sunday et al. 2012, 2014; Kingsolver et al. 2013), such effects could underlie geographic patterns of sexually selected coloration in a diverse suite of animals. Furthermore, these potential thermal effects may have outsized evolutionary consequences as global temperatures continue to rise (IPCC 2014). Yet, while temperature is often linked to the evolution of color patterns that are produced by both sexes (Watt 1968; Kingsolver 1995; Clusella Trullas et al. 2007; Stuart-Fox et al.

2017), its role in the geographic diversification of sexually selected coloration has received far less attention (West & Packer 2002; Punzalan et al. 2008c; Svensson &

Waller 2013). 62

In this study, we explored temperature’s potential to influence the costs, benefits, and geographic diversification of sexual coloration in a North American dragonfly

(Pachydiplax longipennis Burmeister). Males of this medium-sized species produce dark, condition-dependent coloration on the their wings (Moore & Martin 2018), which intimidates rivals and improves territorial success (Moore & Martin 2016). As territorial success determines a substantial amount of mating success in this (Sherman 1983) and similar species (reviewed in Suhonen et al. 2008), such advantages strongly increase male fitness. Males exhibit continuous variation in wing coloration within populations across the northern and eastern portions of the species’ range (e.g. Moore & Martin 2016).

However, males possess drastically reduced coloration in western regions (e.g. Paulson

2009). This geographic variation makes P. longipennis well suited for testing temperature’s ability to modify the costs and benefits of sexually selected coloration.

Previous work in odonates also suggests that the evolution of sexually selected wing coloration can be sensitive to thermal conditions (Outomuro & Ocharan 2011; Svensson

& Waller 2013), making them good candidates for such tests generally. If temperature indeed has the capacity to determine the relative costs and benefits of wing coloration in

P. longipennis, and ultimately shapes its diversification across the species’ range, then at least four predictions should be supported. First, greater wing coloration should increase male body temperatures. Second, any color-induced heating should meaningfully affect male flight performance. Third, in a cold portion of the species’ range, the territorial advantages of greater wing coloration should be largest on the overall coolest days and/or on the most thermally variable days (because of frequent drops below requisite 63 temperatures for territorial acquisition and defense). Fourth, male wing coloration should be substantially reduced in warm regions.

METHODS

Wing coloration in Pachydiplax longipennis

Males can display dark brown coloration across their entire wing surface, with particularly dense regions of pigmentation between the nodus and pterostigma, as well as at the base of the wing. Females sometimes, though not always, produce a small amount of basal wing pigmentation. Unless otherwise noted, we report wing coloration as the percent of the total wing area with pigmentation, as this metric is a known target of intrasexual selection (Moore & Martin 2016). Appendix 3 provides detailed methods for these measuring this trait from digitized photographs in ImageJ (Rasband 2012). While distal wing coloration is much more variable than basal wing coloration (coefficients of variation: distal = 0.324, basal = 0.191), they are fairly strongly correlated (r = 0.668, P <

0.001), suggesting they are unlikely to have completely independent evolutionary consequences. The total proportion of wing coloration is not correlated with total wing area (Moore & Martin 2016), total body mass (r = -0.092, P = 0.669), or dorsal thorax darkness (r = -0.269, P = 0.175).

Prediction 1: Wing coloration increases male body temperatures

We first compared thoracic heating between males with naturally high and low levels of wing coloration. Over the course of five days between 10:00 and 11:30 EDT, we captured 64

22 mature males from a population in northeastern Ohio USA (Hunting Valley, OH,

USA). To best approximate the large differences in wing coloration observed across the species’ range, we compared males with as divergent extents of coloration as possible

(mean % coloration ± SD: high = 57.9 ± 7.6%, n = 11; low = 25.5 ± 7.9%, n = 11). We placed captured males in plastic bags on ice to ease handling and brought them into a field room at the Squire Valleevue Farm (Hunting Valley, OH USA). Once a male had cooled enough to safely handle without injuring it, we tied a loop of nylon line around its legs and placed it back in its bag on ice until its heating trial began. After each male’s trial, we removed one of its middle legs to prevent recapturing, and released it.

We also conducted an experimental manipulation of wing coloration and compared male heating. Over two days, we captured 20 males with low coloration. We then visually paired each male with another captured on the same day that was a similar size and had similar natural extent of wing coloration. On one male within the pair, we experimentally augmented its extent of coloration to that of a high-coloration male by coloring its wings with a brown felt tip marker (~60%, Crayola® Cuppa’ Cappucino). On the other male within each pair, we colored in the same wing area with a colorless blending marker (Copic® 0-S), which has the chemical properties of marker ink but lacks dye. This controlled for any effect of the marker on the wing. Males were otherwise captured and handled as described above. Males in each treatment did not differ in mass

(F1,19 = 0.46, P = 0.507) or basal wing coloration (F1,19 = 0.18, P = 0.677).

For each heating trial, we carefully removed the male from its bag, touching only the nylon line, and tethered it to a metal eyelet in a white Stryofoam box (L x W x D: 17 cm x 17 cm x 20 cm), which had the top and one side removed. Using a 60 W lamp 65 positioned 20 cm above the eyelet (Punzalan et al. 2008; Hegna et al. 2013), we heated each male and recorded the surface temperature of the lateral side of its thorax every 30 seconds for five minutes with a thermal imaging camera (FLIR® C2). Trials were always performed between 12:00 and 15:00 EDT, when room temperatures vary little (mean ± sd: 23.10 ± 0.40 °C). While surface temperature recordings in odonates follow internal temperatures taken with thermal probes very precisely (R2 = 0.988, Samejima & Tsubaki

2010), they somewhat exaggerate the magnitude of warming (slope of internal temperature regressed on surface temperature = 0.836, Samejima & Tsubaki 2010).

Surface temperature differences therefore may overstate the true magnitude, but not the presence or direction, of internal temperature differences.

We then assessed how natural and experimental differences in wing coloration affected males’ maximum temperatures and heating rates. We first fit an asymptotic non- linear model to each male’s heating curve, including each temperature recording as the response (‘nlme’, Pinheiro et al. 2015, Table S3.1). Using the parameter estimates from each male’s fitted heating curve, we then compared the asymptotes (maximum temperature) and rate constants (heating rate) between the groups using likelihood ratio tests of models with and without each term. We fit models using ‘metafor’ (Viechtbauer

2010) to directly incorporate the standard error of the parameter estimates into the analyses. We included date as random effect in these models to account for any non- independence among trials performed on the same day (e.g. acclimation to ambient temperature). We also included pair as a random effect in the analyses of the experimental males to explicitly test for differences between each manipulated male and its control counterpart. Because comparisons between males with naturally high and low 66 levels of coloration are inherently correlative, we included mass as a covariate in this analysis. Conversely, we did not include mass in analyses of experimental males because statistical comparisons are between individuals that were already paired by size.

Significant differences between groups in the asymptote or rate constant indicate differences in maximum temperatures or heating rate, respectively. Initial temperature was not strongly associated with asymptotic temperatures in either comparison (both t <

2.1, P > 0.058).

Prediction 2: Thermal consequences of wing coloration translate into performance variation

We next used a modification of the protocol described by Samejima & Tsubaki (2010) to measure how temperature influenced lifting force, a known performance target of intrasexual selection in dragonflies (Marden & Cobb 2004). Over the course of 10 days, we captured 70 males and acclimated them to one of seven temperatures (25, 29, 33, 37,

41, 45, 49 °C, n = 10 males each) for 90 minutes in lighted incubators (DigiTherm® DT2-

MP-47L). To each male, we then attached a nylon string with evenly spaced bundles of small sequins (~0.030 g per bundle). We perched the male on a plastic cylinder that was positioned vertically, induced it to take off, and recorded how many bundles it lifted.

Because males must take off from perches frequently over short time spans to engage territorial intruders, we used the average mass lifted across three trials as proxy of a male’s potential flight performance at a temperature at an ecologically relevant task (see also Marden 1995). All males acclimated to 49 °C surpassed their critical thermal maxima before performance trials, and we scored their performance as 0 (Kingsolver et 67 al. 2013). Results are similar if these males are excluded (cf. Tables S3.2, S3.3). We fit an exponentially modified Gaussian model to the relationship between acclimation temperature and flight performance (‘nlme’; Table S3.2 compares potential thermal performance curves). From this thermal performance curve, we estimated the maximum average performance across all temperatures (PMAX), the temperature of PMAX (thermal optimum), and the range of temperatures at which males perform at least 80% PMAX

(performance breadth, sensu Huey & Kingsolver 1989). To explore how a plausible magnitude of color-induced heating may influence performance, we quantified how a 1-

2 °C increase (see Results) alters performance relative to PMAX at: 1) 25 °C; 2) the thermal optimum; and 3) the upper and lower bounds of the performance breadth. While non-linear models of flight performance that directly included body size would not converge, several supplemental analyses demonstrated that any effects of body size did not confound our results (see Appendix 3).

Prediction 3: Wing coloration provides its greatest territorial advantages on the coldest and/or most thermally variable days

To assess if temperature modifies the relationship between wing coloration and territorial success, we re-analyzed a dataset described in Moore & Martin (2016). During the 2015 reproductive season, we captured, marked, phenotyped (Appendix 3), and observed the daily territorial interactions of mature males at a pond in northeastern Ohio (perimeter =

140.2 m, area = 0.248 ha). Prior to release, we marked each male’s abdomen with a unique combination of acrylic paint colors. Then, every sunny day throughout the species’ 68 flight period that year (n = 18 days), one observer (MPM) slowly circled the pond during peak activity hours (1000 - 1630 EDT) and recorded each sighted male’s location, territorial behavior, and the time of the observation.

We assessed two components of territorial success from these field observations: territorial acquisition and the duration of territorial defense. For every day between a male’s capture and its last-known day alive (its last resighting), we assessed whether or not it acquired a breeding territory (n = 45 males, 309 potential male-by-day combinations). A male was designated as having acquired a territory for a day if we observed it perching or patrolling within one of two regions of emergent vegetation (13.8 and 35.7 m, respectively) where males always set up territories. As we captured and marked exclusively territorial males, each male’s first recorded territorial acquisition was always the day on which it was captured. Male P. longipennis do not come to the pond except for reproduction (Fried & May 1983), and “sneaker” or “satellite” tactics were never observed. For males that were subsequently resighted in an acquired territory, we also scored the duration (mins) over which they were able to defend the territory for that day (n = 126 territorial tenures). This duration was estimated as the time between the first and last observation of the male in the territory. Then, for the activity period of each day that males were observed defending territories, we obtained the means and SDs of temperature (°C) and solar radiation (W * m-2) from a weather station (HOBO® U30

Station) that was located ~1.5 km away in a similarly open habitat. Mature males rarely migrate and defend territories at new ponds within a single reproductive season

(McCauley 2010), making territoriality elsewhere after the last resighting unlikely. 69

We used mixed-effects models (‘lme4’, Bates et al. 2015) to test if the daily means and SDs of temperature and/or solar radiation altered the relationships between wing coloration and our measures of territorial success. In each model, we included individual as a random effect to account for the repeated measures of individuals across days. As fixed effects in both models, we included a male’s total wing coloration and body length

(each scaled to mean of 0 and SD of 1; Lande & Arnold 1983), mean temperature of the day, SD of temperature of the day, mean solar radiation of the day, SD of solar radiation of the day, and all trait by environment two-way interactions. We modeled the probability of acquiring a territory for a given day as a binary response (1 = did acquire territory, 0 = did not acquire territory) and tested the significance of terms with likelihood ratio tests of models with versus without the effect. We loge transformed the duration of territorial defense and tested the significance of terms using F-tests with the Kenward-Rogers degrees of freedom approximation (Kenward & Roger 1997). In these analyses, a significant trait by environment interaction indicates that the environmental factor is modifying the relationship between wing coloration and the aspect of territorial success

(MacColl 2011). Mean daily temperature was the only environmental variable to show a temporal trend across the season (Table S3.4), and previous analysis of this population showed no relationship between longevity and either male trait (Moore & Martin 2016).

Prediction 4: Reduced wing coloration in hottest parts of range

We used photographs taken by citizen scientists to characterize environmental correlates of the reported variation in male wing coloration. We examined the first 480 geo- referenced photographs of male P. longipennis that had been uploaded to iNaturalist.org 70 and scored if: 1) the male was mature (indicated by a waxy blue prunescence over its abdomen) and 2) it had produced distal wing coloration (1 = yes, 0 = no). While both distal and basal wing coloration can vary, we used this qualitative measure of wing color variation because it can be reliably assessed from the unstandardized photographs of wild males. Importantly, because the regions of pigmentation are correlated, and distal coloration is much more variable, this metric should enable us to generally assess if wing coloration is indeed reduced in warm regions. As males can also produce wing coloration in this region of the wing prior to maturation (Moore & Martin 2018; Moore et al. 2018a), we positively scored males that were not mature but exhibited such coloration (~2% of all observations). Immature males without any wing coloration were excluded. We also haphazardly chose and scored 55 additional photographs in regions that were underrepresented in our initial sample. Similar procedures have been validated for examining clines in other color polymorphisms (Leighton et al. 2016). No duplicated photographs were observed. We also extracted the annual mean temperature and precipitation of the driest quarter (aridity) for the location of each observation from

WorldClim1.4 (Hijmans et al. 2005). We computed annual mean solar radiation with

ArcGIS using standard techniques (Böhner & Antonić 2009). Results are similar for environmental data from WorldClim2 (Fick & Hijmans, 2017; Tables S3.5, S3.6). We used annual means of temperature and solar radiation because differences in the duration of the flight season across the species’ range may expose adults of some populations and not others to subsets of year (e.g. hottest or coldest quarters). As P. longipennis requires at least some permanent standing water to complete its life cycle, we included precipitation of the driest quarter as our measure of aridity because it should best reflect 7 1 the input into ponds necessary for maintaining water levels during larval development than annual means.

We used a mixed-effects logistic regression (‘lme4’, Bates et al. 2015) to explore how the production of wing coloration varies with annual mean air temperature, aridity, annual mean solar radiation, and their interactions. We z-transformed each variable prior to analyses to make parameter estimates more comparable among these environmental factors (Schielzeth 2010). Correlations among environmental variables were modest (all r

< 0.242), indicating that multicollinearity should not be a large concern (Graham 2003).

Models including the interaction between solar radiation and aridity did not converge and were not tested. We modeled each geo-referenced observation as a response. Because population boundaries are not discernable from these geo-referenced observations, we included a random term for municipality (county or its equivalent) nested within state

(USA, Mexico) or province (Canada). This nesting structure accounts for non- independence of observations of males from within the same populations, and also for observations within a single state being more correlated to each other than observations from different states (Outomuro & Ocharan 2011). We tested the significance of fixed effects by comparing models with and without the effect using likelihood ratio tests. We also conducted a spatially explicit logistic regression using the ‘ggwr’ function from the

‘spgwr’ package in R (Bivand & Yu 2017). As test statistics have not yet been implemented for this package, we compared the parameter estimates of this model to those generated by the nested mixed-effects model. In all cases, the parameter estimates between to the two approaches were similar (Tables S3.5, S3.6). We therefore report on only the nested mixed-effects model in the main text. 72

RESULTS

Prediction 1: Wing coloration increases male body temperatures

Males with naturally greater wing coloration reached higher asymptotic temperatures than males with less wing coloration, but heating rates did not differ between them (Fig.

3.1a, Table 3.1). Heavier males also reached marginally higher asymptotic temperatures

! (β = 9.646 ± 5.147, �! = 3.7, P = 0.054) and heated marginally slower (β = -5.876 ±

! 3.325, �! = 3.1, P = 0.077). Males with experimentally augmented coloration both reached higher asymptotic temperatures and heated faster than their size-matched, control counterparts (Fig. 3.1b, Table 3.1).

Prediction 2: Thermal consequences of wing coloration translate into performance variation

Lifting force increased with body temperature, reaching 0.252 g (PMAX) at 39.4 °C, before precipitously declining (Fig. 3.2). Flight performance was 80% of PMAX between 33.4 and

42.5 °C (performance breadth, PBR). At 25 and 33.4° C, color-induced heating of 1-2 °C

(Table 3.1) improves relative flight performance by 2.6 - 5.4% and 3.9 - 7.9%, respectively. In contrast, at 39.4 and 42.5 °C, that same range of heating reduces relative flight performance by 1.9 - 8.1% and 15.3 - 32.3%, respectively.

Prediction 3: Wing coloration provides its greatest territorial advantages on the coldest and/or most thermally variable days

The probability of a male acquiring a territory on a day increased strongly with wing coloration on the most thermally variable days and slightly decreased on the least

! variable days (�! = 7.6, P = 0.006; Fig. 3.3). This relationship was unaffected by body

! size, daily mean temperature, or solar radiation (all �! < 2.6, P > 0.108, Table S3.7). Of the males that acquired territories on a given day, those with more wing coloration were able to defend them for longer (F1,26.4 = 10.4, P = 0.003, see also Chapter 2), but this relationship was not modified by body size, temperature, or solar radiation (all F < 2.6, P

> 0.107, Table S3.7).

Prediction 4: Reduced wing coloration in hottest parts of range

Consistent with reports in field guides (e.g. Paulson 2009), males typically produce at least some coloration across their wing area in northern and eastern portions of the species’ range, but coloration is usually reduced in western regions (Fig. 3.4a). These observations also revealed that mature males also frequently exhibit reduced coloration in southern regions. After controlling for the spatial non-independence of observations, males typically had reduced wing coloration in regions with the greatest annual mean air

! temperatures (β = -0.502 ± 0.103, �! = 21.4, P < 0.001, Fig. 3.4c). Reduced wing

! coloration was also common in the driest areas (β = 0.273 ± 0.101, �! = 6.5, P = 0.011).

! There were no effects of annual mean solar radiation (β = -0.089 ± 0.069, �! = 1.8, P =

! 0.184) or any interaction (all �! < 1.0, P > 0.328).

DISCUSSION

Given its role in the evolution of coloration shared by both sexes (Watt 1968; Clussella

Trullas et al. 2007; Stuart-Fox et al. 2017), temperature should also have important effects on the costs, benefits, and ultimately diversification of coloration produced by one sex to attract mates and intimidate rivals (West & Packer 2002; Ellers & Boggs 2003;

Svensson & Waller 2013). Here, we found that: 1) the sexually selected wing coloration of P. longipennis increases male body temperatures; 2) heating, like that associated with greater wing coloration, modestly enhances flight performance under cool conditions and dramatically reduces it under warm conditions; 3) wing coloration most greatly improves territory acquisition on thermally variable days in a cold region of the species’ range; and

4) male wing coloration is greatly reduced in the hottest parts of the species’ range. These results collectively support temperature’s capacity to modify the strength and direction of selection on wing coloration and, therefore, to shape this trait’s evolution across the landscape.

When sexual coloration raises male body temperatures, any fitness effects will depend on environmental temperatures, especially in ectotherms (Clussella Trullas et al. 2007;

Stuart-Fox et al. 2017). Both our correlative and manipulative experiments indicate that the presence of dark coloration across the entire wing surface can elevate male temperatures, consistent with direct and indirect evidence from other Holarctic odonates

(e.g. Outomuro & Ocharan 2011; Svensson & Waller 2013) and some lepidopterans (e.g.

Hanley et al. 2013; Hegna et al. 2013; Brashears et al. 2016). Beyond governing the physiological consequences of their thermal effects, environmental temperature may also affect how secondary sexual color patterns function during behavioral interactions. For 75 example, in the many other territorial animals that display coloration to signal fighting ability during male-male competition (Cuthill et al. 2017), body temperature could be one component of the information that is signaled to rivals. If this is the case, the way that males respond to a rival’s coloration in cool regions, where color-induced heating improves male fighting ability, should differ from warm regions, where such heating diminishes it. Consequently, sexual coloration may signal strength and deter territorial intruders in one part of the range, but signal weakness and invite them in others. While the possible signaling of male body temperatures remains to be explored, our results nevertheless underscore the potential thermal consequences of sexually selected coloration (see also West & Packer 2002; Punzalan et al. 2008c).

For temperature to determine the reproductive benefits and performance costs of sexually selected traits, their thermal effects must translate into variation in performance.

We found that heating, like that induced by sexually selected wing coloration, modestly improves male flight performance under cool conditions and dramatically reduces it under hot conditions. As locomotor performance boosts mating success in dragonflies

(Marden & Cobb 2004) and many other animals (Lailvaux & Irschick 2006), even small thermal effects on territorial defense or mate attraction should have substantial fitness consequences (Punzalan et al. 2008c). Indeed, in one of the coolest portions of the species’ range, we found that wing coloration most improved territorial acquisition on the most thermally variable days. Thus, at least one reproductive benefit of wing coloration depends on the thermal conditions, perhaps because color-induced heating buffers against fluctuations below lower thermal thresholds for territorial activities. However, because males also behaviorally thermoregulate via adjustments of position and posture (May 76

1976), exploring how behavior and coloration act together to affect male body temperatures in the field remains necessary to elucidating the precise mechanism underlying this pattern. Intriguingly, once a male successfully acquired a territory for a day, the benefits of wing coloration to the duration of territorial defense did not depend on thermal conditions, indicating that not all of this trait’s territorial benefits are temperature dependent. Nonetheless, these observations collectively reveal that the thermal effects of wing coloration can translate into variation in intrasexually selected performance in the wild.

In contrast to the territorial benefits conferred under the coolest conditions, wing coloration likely imposes intense thermal costs to both reproduction and survival in the warmest regions. For example, in parts of the American Southwest where wing coloration is reduced (e.g. Sonoran Desert, USA, 32.8 N, 112.8 W), one to two months every year have maximum daily temperatures that average over 41.5 °C (Appendix 3). Thus, on most days for long stretches of the summer, color-induced heating of 1-2 °C would cause males to reach or exceed the upper limit of their performance breadth. This in turn would force males to abandon territories and seek shade during the peak activity hours (i.e. reproductive cost; Huey et al. 2009; Sunday et al. 2014) and/or induce unsustainably high metabolic rates (i.e. survival cost; Fried & May 1983; Gillooly et al. 2001; Dillon et al.

2010). As animal performance commonly plummets at high temperatures (Huey &

Kingsolver 1989), similar thermal costs may often cause both sexual selection and viability selection to disfavor secondary sexual coloration in warm regions (see also West

& Packer 2002; Svensson & Waller 2013). These thermal costs should also intensify with 77 the rising global temperatures, potentially making them an important force on the future evolutionary trajectories of sexual coloration in many animals as well.

Environmental variation in the costs and benefits of a sexually selected trait will promote its diversification across the species’ range (Wiens 2001; Maan & Seehausen

2011). Consistent with the thermal modifications to the relative costs and benefits of wing coloration that we observed, we also found substantial reductions in this sexually selected trait in the hottest portions of the species’ North American range. While the relative importance of evolved, constitutive differences versus temperature-induced plasticity to this geographic pattern remains unknown (Watt 1969), any such plasticity would still likely be a signature of temperature-mediated differences in selection

(Kingsolver 1995). Additionally, recent work in butterflies shows that genetic variation and plasticity combine to facilitate adaptation to temperature (Kingsolver & Buckley

2017), demonstrating that their evolutionary effects are not mutually exclusive. Our analysis also cannot exclude other factors that are themselves correlated with temperature, but several alternatives seem unlikely. For instance, defending against infection inhibits wing color production in odonates (Siva-Jothy 2000; Moore et al. 2018a). The observed geographic pattern could then plausibly reflect variation in parasite load. However, as odonates’ primary parasites have additional aquatic hosts and/or free-swimming life stages (Corbet 1999; Forbes & Robb 2008), this explanation seems at odds with the pronounced wing color reduction occurring in the most arid regions. Interspecific character displacement of wing coloration could also potentially explain these patterns

(Tynkkynen et al. 2004; Anderson & Grether 2010; Hassall 2014), yet no similarly colorful dragonfly inhabits only the regions where P. lonigpennis typically lack distal 78 wing coloration (Paulson 2009). Thus, although other explanations cannot yet be definitively rejected, the observed geographic variation is broadly consistent with temperature-mediated variation in selection on wing coloration translating into diversification across the landscape.

A complex interaction of environmental factors ultimately controls the magnitude of the reproductive benefits conferred by secondary sexual characters, as well as the intensity of any performance costs that they inflict (Miller & Svensson 2014). The optimal extent of sexually selected traits is thus expected to track spatial or temporal variation in any of these abiotic or biotic features (Maan & Seehausen 2011). Our results broadly support temperature’s capacity to shape the performance costs, reproductive benefits, and geographic diversification of sexually selected color patterns. Given the pervasiveness of sexual coloration (Cuthill et al. 2017), and the well-known thermal effects of dark coloration (Clusella Trullas et al. 2007), similar effects could readily drive geographic patterns of sexual selection and diversification across the animal kingdom.

Additionally, in light of the central role that sexually selected color patterns play in assortative mating and other mechanisms of pre-zygotic reproductive isolation (Servedio

& Boughmann 2017), variation in temperature-mediated selection could have underappreciated effects on the promotion and/or maintenance of current reproductive boundaries between species. Moreover, if rising global temperatures further modify the costs and benefits of sexually selected traits, subsequent evolutionary changes in these traits may also promote novel pre-zygotic barriers between populations that currently can interbreed (Boughman 2001) or weaken barriers between those that presently cannot

(Seehausen et al. 1997). 79

Table 3.1. Least-squares mean asymptotes and rate constants (± SE) from asymptotic models fit to each male’s heating response. Comparisons between high vs low natural coloration (a) and experimentally augmented and control coloration (b) were made using the R package ‘metafor’. A higher asymptote indicates greater maximum thoracic temperature. A larger rate constant indicates faster heating. We also included each male’s body mass as a covariate in analyses of natural variation, and we report these effects in the main text.

High/ Low/ Comparison Parameter χ2 P Augmented Control

Asymptote (°C) 33.16 ± 0.87 31.38 ± 0.87 51.6 < 0.001 a) Natural Variation Log - Rate Constant e -3.74 ± 0.55 -3.50 ± 0.56 2.1 0.143 (°C * s-1)

Asymptote (°C) 31.93 ± 0.20 30.83 ± 0.23 18.8 < 0.001 b) Experimental Manipulation Log - Rate Constant e -4.40 ± 0.12 -4.78 ± 0.14 4.2 0.040 (°C * s-1)

A) B) C) 29.1 36

30 27

Thoracic Surface Temperature (C) Temperature Surface Thoracic High Augmented

27 Low Control 24 23.0 0 100 200 300 0 100 200 300 Time (s) Time (s)

Figure 3.1. Wing coloration increases male body temperatures. a) Males with naturally greater coloration reached higher thoracic surface temperatures (n = 11) than those with less coloration (n = 11) when warmed under a lamp (Table 3.1). b) Males with experimentally augmented coloration (n = 10) also reached higher thoracic surface temperatures (n = 10) than control males when warmed (Table 3.1). In both panels a and b, dotted lines represent individual males, and solid lines illustrate heating curves with the mean asymptotes and heating constants that were estimated from mixed-effects models. c) Thermal image of male P. longipennis with naturally high coloration being warmed under a lamp. Temperature gradient is in °C.

0.45

0.3 TOpt

PBr Mass lifted (g) MassLifted (g) 0.15

25 33 41 49 Acclimation temperature (C) Acclimation Temperature (°C)

Figure 3.2. Thermal flight performance of male P. longipennis. Circles are the mean mass lifted by individual males that were acclimated to one of seven temperatures (n = 10 males per temperature). The blue line is the best fitted non-linear relationship. TOPT is the temperature at which performance is maximized, and PBR is the range of temperatures at which performance is at least 80% of the maximum.

A) B) O – Acquired a territory x – Did not acquire a territory

0.45 0.6 0.75 ● ● ● ● ●

0.8 ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● 1 0.7 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● 0.6 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 0.5 0 ● ● ● ● ● ● ● ● ●

● ● ●● ● 0.4 ●● ● ● ● 0.45 0.3 ● ● ● ● ● ● −1 ● ● Wing Coloration WingColoration 0.2 ● ● ● ● ● ● ● ●

Probability of acquiring a territory Probability of territory acquisition territory of Probability ● 0.3

0.6● ● ● ● ● ● ● ● ● ● ● ● ● 1 −2 0 1 Wing Coloration ●

−1 0 ● ●

−2 −1 −1 0 1 2 Daily SD Temp DailyDaily SD SDTemperature Temp

Figure 3.3. Wing coloration conferred the greatest improvements to territorial acquisition on the most thermally variable days. Probabilities in both (a) surface and (b) contour plots are those estimated from the generalized linear-mixed effects model. Points (b) represent whether a male of known wing coloration did (circles) or did not (x’s) acquire a territory on a given day (n = 309 male by day combinations).

-125° -110° -95° -80°

(!(!(! (!(!(! (!(! (! (!(!(! A) (! (! (! (! (! (!(!(! (! (! (! (! (! (! (!(! (! (! (!(!(!(! (! (! (!(! !(! (!(! (!(! (! (!(!(! (!(! (!((! (! (!( (!(! (! (!(!(! (!(!(!(! (! ° (! (! (!(! (!(!(! (! (!(!(!(! ° 0 (!(! (! (! (!(!(! (!(! (!(!(! 0 4 (! ! (! (! 4 (! ( (! (!(!(! (!(!(!(!(!(!(!(!(! (!(!!(!(! (! (!(! ! (! (! (! (!(!(!(! (!(!(! (! (! (! (!(! (! ((!(!(! (! (!(! (!(! (! (! (! (!(! (!(! (!(! (!(! (!(!(! (! (! (!(! (! (! (!(! (!(!(!(!(! (! (! (! (!(! (! (!(!(!(!(!(! (! (! (! (!(! (! (! (! (!(!(!(!(!(!(! (!(!(! (!(!(!(!(!(!(!(!(!(!(!(! (! (! (! (!(! (!! (! (! (!(! (!(! ( (! (! (! (! (! (!! (! (!(!(!(! (!(!(! (! (! (!(! (!( (! (!(! (! (! (!(! (!(! (! (! (! (! (! (! (!!(!(! (!! (!((!(!(! ° ( (!(! ° 5 5 2 2 ± (! 0 300 600 (! km -125° -110° -95° -80° B) C) 1.0 || || | | ||||||| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||| ||||||||| |||||||| || | ||||||||||||||||||||||||||||| || | | || ||| |

0.5

Probabilitycolorof across thewing 0.0 ||| | ||||||||| |||| || |||| ||||||||||||||||| |||||||||||||||||||||||| ||||||||||||| ||||||||||||||||| ||| | 5 10 15 20 25 Annual Mean Temperature (°C)

Figure 3.4. Geographic variation of wing coloration. a) Males typically produce at least some coloration across their entire wing area (black circles) in the northern and eastern portion of the species range, but this coloration is often greatly reduced (grey circles) in the southern and western portions of the range (n = 535 observations). Annual mean temperature is shown as a color gradient with the coldest regions being shown in blue and the hottest being shown in red. b) Mature males with high wing coloration (top) and reduced coloration (bottom) captured in northeast Ohio. c) After accounting for spatial non-independence, the probability of a mature male producing at least some distal wing coloration declines as annual mean temperature increases. Each tick mark corresponds to a geo-referenced photograph of a male from the citizen science website iNaturalist.org.

The blue line shows the fitted line from the mixed-effects logistic regression. 84

CHAPTER 4: TRADE-OFFS BETWEEN LARVAL SURVIVAL AND ADULT

ORNAMENT DEVELOPMENT DEPEND ON PREDATOR REGIME IN A

TERRITORIAL DRAGONFLY

Published in Oecologia Vol. 188, No. 1, pp. 97-106, 2018. DOI: 10.1007/s00442-018-

4171-x.

Authors: Michael P. Moore & Ryan A. Martin

Department of Biology, Case Western Reserve University, Cleveland, OH USA

(Submitted: 21 October 2017; Returned for Revision: 23 February 2018; Accepted: 21

May 2018)

ABSTRACT

Trade-offs between juvenile survival and the development of sexually selected traits can cause ontogenetic conflict between life stages that constrains adaptive evolution.

However, the potential for ecological interactions to alter the presence or strength of these trade-offs remains largely unexplored. Antagonistic selection over the accumulation and storage of resources could be one common cause of environment-specific trade-offs between life stages: higher condition often simultaneously enhances adult ornament development and increases juvenile vulnerability to predators. We tested this hypothesis in an ornamented dragonfly (Pachydiplax longipennis): higher larval body condition should improve the development of its intrasexually selected wing coloration, but also 85 lower survival in the presence of large aeshnid dragonfly predators. Higher body condition indeed enhanced the initial development of wing coloration, but was opposed by viability selection in the presence of large aeshnid predators. In contrast, viability selection did not oppose body condition in pools when aeshnids were absent, and was not affected when we manipulated cannibalism risk. Trade-offs between larval survival and ornament development, mediated through the conflicting effects of body condition, therefore occurred only under high predation risk. We additionally characterized how body condition influences several traits associated with predator avoidance. Although body condition did not affect burst distance, it did increase larval abdomen size, potentially making larvae easier targets for aeshnid predators. As high body condition similarly increases vulnerability to predators in many other animals, predator-mediated costs of juvenile resource accumulation could be a common, environment-specific limitation on the elaboration of sexually selected traits.

INTRODUCTION

Despite sexual selection favoring continuous exaggeration of secondary sexual traits

(Andersson 1994), these traits often exhibit tremendous diversity in size, shape, and intensity across natural populations (e.g. Moczek & Nijhout 2003; Svensson et al. 2004;

Martin et al. 2014). Evaluating the environmental and developmental causes of this variation can highlight important factors that promote and constrain phenotypic diversification. For example, spatiotemporal variation in the strength of mate competition or the efficiency of signal transmission can drive diversification by altering the strength 86 or form of sexual selection (Endler 1992; Miller & Svensson 2014). Spatiotemporal differences in natural selection opposing secondary sexual traits can also shape the extent of elaboration by causing trade-offs between survival and mate acquisition (Endler 1980;

Andersson 1994; Zuk et al. 2006; Heinen-Kay et al. 2015). Moreover, as sexually selected traits commonly require resources that are accumulated throughout the life cycle

(Kasumovic 2013; Morehouse 2014), viability selection may further oppose features of their development in non-reproductive seasons or life stages (Jennions et al. 2001;

Kotiaho 2001). In theses cases, the trait values that improve reproductive success in one life stage ultimately reduce survival in another, and there is “ontogenetic conflict” over the development of the sexually selected trait (sensu Sinervo & Calsbeek 2003; Calsbeek

& Goedert 2017; see also Chippindale et al. 2001). However, the potential for ecological and/or developmental factors to moderate these fitness trade-offs between life stages or seasons has received only limited attention.

Diversification of sexual phenotypes may be especially influenced by variation in the strength of trade-offs between the development of sexually selected traits and juvenile survival (Jennions et al. 2001; Cornwallis & Uller 2010; Mojica & Kelly 2010). For instance, differences in predators, pathogens, or other sources of mortality among juvenile habitats could alter survival costs of sexual phenotype development. Then, for a genotype that produces an advantageous sexual phenotype, the likelihood of ever reaching maturity and even being exposed to sexual selection would differ widely among juvenile environments (Hadfield 2008; Mojica & Kelly 2010; Johnson & Hixon 2011).

Indeed, such trade-offs between life stages are known to maintain genetic variation in secondary sexual traits within natural populations (Brooks 2000; Robinson et al. 2008) 87 and inhibit phenotypic exaggeration under even strong artificial selection (Hine et al.

2011). Examining how the juvenile environment modifies the strength, or even presence, of trade-offs between juvenile survival and sexual phenotype development could therefore illuminate important, but perhaps overlooked, factors underlying the diversity of secondary sexual traits. Yet, the potential for these environment-dependent trade-offs between life stages remains largely unexplored.

As predators preferentially consume prey that are easier to capture and handle (Gosler et al. 1995; Van Buskirk et al. 1997; Mikolajewski et al. 2006), environment-specific trade-offs between juvenile survival and sexual phenotype development could commonly arise over the resource accumulation used in the production of sexually selected traits.

For example, higher juvenile body condition often enhances the development of sexually selected traits (Kasumovic 2013; Morehouse 2014), but also increases vulnerability to predators by making individuals slower and/or easier targets (Gosler et al. 1995; Zamora-

Camacho et al. 2014a). The strength of the resulting ontogenetic conflict over the development of secondary sexual traits, mediated by this potential functional constraint, would then depend on the predators in the juvenile environment. Here, we experimentally test for predator-mediated variation in trade-offs between larval survival and adult ornament development using the territorial dragonfly, Pachydiplax longipennis

(Burmeister). This broadly distributed, North American dragonfly inhabits ponds with a wide range of aquatic predators as a larva (McCauley et al. 2008), and produces intrasexually selected melanin wing ornaments that vary substantially in size and intensity across its range (Paulson 2012; Moore & Martin 2016; Chapter 3). Wing pigmentation in odonates is well suited for examining such patterns because these 88 ornaments have long been a model system for understanding trade-offs between survival and reproduction (e.g. Grether 1996a; Svensson et al. 2004), and aspects of melanin- synthesis pathway are strongly affected by the larval environment (Stoks & Córdoba-

Aguilar 2012; Debecker et al. 2015). In P. longipennis, this wing coloration is essential during male-male competition over the reproductive territories in which males copulate with females and into which females oviposit (Moore & Martin 2016). If condition- mediated trade-offs between larval survival and adult ornament development depend on predation risk, then high-condition larvae will have improved adult ornament development if they emerge, but also lower survival in the presence of larval predators.

METHODS

Focal species and study overview

Pachydiplax longipennis (Burmeister) is a medium-sized dragonfly with a wide distribution across North America (Paulson 2012). In the northern extent of the range, where this study was conducted, P. longipennis is univoltine (Wissinger 1988b; MPM, personal observation). After overwintering, the larvae of P. longipennis and most other pond-dwelling dragonflies move towards the shore and rapidly complete the final three to four development stages before emergence (Wissinger 1988b). Larval dragonfly assemblages subsequently exhibit tremendous hetero- and conspecific size asymmetries during this period, leading to extremely high rates of intraguild predation and cannibalism (Wissinger 1992; Hopper et al. 1996; Crumrine et al. 2008).

In this study, we first examined how larval body condition influences ornament development and how environmental context modifies selection on larval body condition 89 by rearing uniquely marked individuals in outdoor wading pools that differed in predator regime. We then assessed several performance and morphological correlates of larval condition that can affect vulnerability to predators (Mikolajewski et al. 2006; Strobbe et al. 2009): escape distance, abdomen size, and cuticle darkness.

Aim 1: ornament development and viability selection

We collected P. longipennis larvae from a pond at Case Western Reserve University’s

Squire Valleevue Farm (Hunting Valley, Ohio, USA), at three different times (16 - 18

May, 28 - 29 May, and 31 May 2016). We maintained larvae in 473 mL opaque plastic cups filled with dechlorinated water at a 15:9 L:D photoperiod in a laboratory at Case

Western Reserve University (Cleveland, Ohio, USA). We uniquely marked each larva by injecting a coded wire tag (Northwest Marine Technology Inc., Shaw Island, Washington,

USA) ventrally into its abdomen (Catania & McCauley 2015). We measured head width, a proxy for body size (Corbet 1999) using an ocular micrometer. We blot dried larvae, gently induced them to expel water from their rectal branchial chambers, and measured mass to the nearest 0.0001 g using an electronic balance. To assess the repeatability of this method, we massed 86 larvae twice with a week between measurements.

Repeatability was high (R = 0.921, F84,85 = 24.27, P < 0.001; Lessells & Boag 1987). We estimated larval body condition for each individual using the residuals from a loge – loge regression of body mass on head size (Jakob et al. 1996). We provide a detailed rationale for the validity of this condition estimate in Appendix 4. Briefly, in P. longipennis, this

-11 metric 1) is independent of head size (r = -8.125 x 10 , t476 = 0, P > 0.999); 2) increases with food quantity received (0.064 ± 0.025, F1,42 = 6.72, P = 0.013, n = 44); 3) is 90 positively associated with producing an energetically costly melanin immune response

(75.529 ± 30.755, F1,40 = 5.90, P = 0.020, n = 44), and is positively associated with time until emergence overall and enables developmental acceleration under high predation risk

(Moore et al. 2018b). Additionally, this metric is also positively correlated with fat stores in both highly similar and very disparate odonates (see Appendix 4).

To then investigate ornament development and viability selection, we raised these marked and measured larvae in outdoor wading pools where we manipulated predation risk by altering the presence/absence of A. junius and the size variation of our focal P. longipennis larvae (n = 4 pools per A. junius x conspecific size variation combination).

For pools with A. junius (Anax-present, n = 8 total pools), we introduced two ultimate instar A. junius larvae one hour after the release of our focal larvae, and replaced them when they were emerged (n = 6) or were found dead (n = 3). To manipulate conspecific size variation, we divided focal larvae into four groups within each collection period: relatively large and small ultimate instars, and relatively large and small penultimate instars. We randomly assigned larvae within each group to either high- or low-size variation treatments, where high-size variation pools had a 2:1 ratio of large to small ultimate instars and a 1:2 ratio of large to small penultimate instars, and low-size variation pools had the opposite (Fig. S4.4). Because the available larvae differed among the collection periods, the ratio of instars differed slightly among blocks (Table S4.1), but was the same between treatments within a block. Between treatments, mean head sizes were equal (mean ± SD: low = 4.961 ± 0.107; high = 4.972 ± 0.151; F1,13 = 0.04, P =

0.840), and head size variation differed by ~24% (mean coefficient of variation ± SD: low = 0.116 ± 0.015, high = 0.144 ± 0.023; F1,13 = 7.62, P = 0.0162). 91

Three days after marking, we introduced 30 P. longipennis larvae by their assigned treatment each to 100 L plastic wading pools that we filled with aged water, ~0.003 m3 of aquatic vegetation, ~0.011 m3 of leaf litter, and ~1400 cm3 of benthic substrate. To provide a self-sustaining prey base for the larvae, we introduced 100 mL of concentrated

Daphnia magna 10 days before releasing P. longipennis larvae. We also permitted the natural colonization of prey species (e.g. Culicidae, Chironomidae, and Hylidae spp.) up until the day prior to dragonfly introduction, after which we covered pools with window screen. We checked for adults daily and sacrificed them in a freezer set to -22.8 °C immediately upon recovery. As coded wire tags could be found in adults or their exoskeletons, we collected both when possible. After 20 days, we recovered surviving larvae, and stored them in 95% ethanol in the same freezer. Two larvae accidentally were not measured prior to release into the wading pools at the beginning of the experiment and were excluded from all analyses. Additionally, five Anax larvae had to be replaced from one Anax-present pool, and we have therefore excluded it. However, all results are robust to its inclusion.

Ornament development

To characterize how larval body condition influences the initial stages of ornament development, we gently removed the wings of the sacrificed males that successfully emerged from our experimental wading pools (n = 26 males), and took pictures of them in a dark box against a standard white background. We assessed if larval body condition influenced the initiation of ornament development on the day of emergence (1 = initiated development, 0 = had not initiated development) using a generalized linear-mixed effects 92 model with a binomial error distribution and pool as a random effect. Given that males without coloration are largely unable to defend breeding territories, males that initiate ornament development earlier will be able to occupy and hold territories earlier in their adult lifespans (Sherman 1983; Moore & Martin 2016). As fixed effects, we included larval body condition, head size, Anax presence, conspecific size variation, and the interactions between the traits and the treatments. Models including interactions between

Anax presence and head size would not converge, likely due to relatively few males successfully emerging from the Anax-present treatment (n = 6 males), and we thus only included the main effect of Anax presence and its interaction with larval body condition.

We report overall condition-dependent patterns of emergence elsewhere (Moore et al.

2018b).

Larval Survival

We evaluated how predator treatment modified viability selection on larval body condition using a generalized linear mixed-effects model with each individual’s survival as the response (Lande & Arnold 1983; Chenoweth et al. 2012). We designated all individuals whose tags were not recovered as dead. In some cases, adults emerged successfully from their exoskeletons but died after falling into the water, and we scored these individuals as “survivors” because they likely would have survived if not for the window screen covering the tanks. We included body condition, head size, the quadratic effects of body condition and head size, Anax presence, conspecific size variation, and all interactions as fixed effects, and pool nested within block as a random effect. We scaled individual’s phenotypes to a mean of zero and standard deviation of one within each pool 93

(Lande & Arnold 1983). When a significant trait by treatment interaction was observed, we compared the strength and direction of the linear and non-linear selection coefficients within each of the treatments that differed (Chenoweth et al. 2012). We also computed the opportunity for selection in such treatments (variance in survival dived by its squared mean), a standardized measure of fitness variation that quantifies the upper potential limit to the strength of selection (Arnold & Wade 1984). We calculated selection coefficients and their standard errors with the logistic regression approach (Janzen & Stern 1998;

Mitchell et al. 2013) and used projection pursuit regression to visualize the fitness surface

(Schluter & Nychka 1994). Mixed-effects models were fit with ‘lme4’ (Bates et al. 2014) and projection pursuit regression with ‘stats’ (R Core Team 2014). Selection coefficients and standard errors that were alternatively calculated from generalized additive mixed- effects models (Morrissey & Sakrejda 2013; ‘gsg’ package’, Morrissey & Sakrejda 2014) were highly similar (Table S4.2).

Aim 2: performance, morphological, and physiological correlates of body condition

We also quantified the effects of larval body condition on several traits known to affect vulnerability to predators. We first evaluated how body condition (calculated using residuals from loge-loge regression as described above) affected escape performance using 60 ultimate instar larvae that we collected in July 2016 from the same source pond as for the wading pool experiment. We placed each larva next to a ruler in the center of a clear plastic box (34.6 cm L x 21 cm W x 12.4 cm H) filled with 1 L of aged water, gently prodded its abdomen, and measured the distance (mm) of its initial escape burst three times. This performance metric should reflect an individual’s ability to move out of 94 the attack radius of other dragonfly larvae (Corbett 1999). We used a general linear model to consider how mean burst distance (loge transformed) varied with body condition and head size.

We also characterized how body condition influences morphological traits that affect predation risk. We collected larvae in Oct 2016 from the same source pond as the wading pool study, and reared 44 of them in the laboratory in 473 mL plastic cups under a 13:11

L:D photoperiod (to match conditions at time of this experiment). To generate variation in condition, we provided larvae with a 1.5 mL aliquot of concentrated D. magna (mean

± SD: 15.9 ± 3.8) either 6 (high food) or 2 (low food) times per week (see Appendix 4).

After 21 days, we took digital photographs of each larva against a brightfield background with a dissecting microscope. We calculated relative abdomen size using the residuals from a loge-loge regression of dorsal area (measured from digital photographs in Image J;

Rasband 2012) on head size. From the photographs, we also scored cuticle darkness, as

255 (maximum white value) - the mean grey value of the abdomen (e.g. Fedorka et al.

2013). Darker cuticles are typically thicker and thus more difficult to puncture (Corbet

1999). We used separate regressions to assess the relationships between larval body condition and relative abdomen size and cuticle darkness. We also included food treatment in models to account for any effects that acted in addition to its direct influence on body condition.

RESULTS

Aim 1: ornament development and viability selection

Ornament development

The probability of a male initiating ornament development on the day that it emerged increased with larval body condition (parameter estimate ± SE, hereafter: 1.939 ± 0.974,

! �! = 6.46, P = 0.011, Fig. 4.1, n = 26), but was not influenced by larval head size, Anax

! treatment, conspecific size variation treatment, or any interactions (all �! < 2.01, P >

0.156).

Larval Survival

When considering how predator regime influences viability selection, we detected a

! significant interaction among head size, body condition, and Anax presence (�! = 4.75, P

= 0.029; Fig. 4.2, n = 448), indicating that Anax presence altered selection on head size and body condition. Conspecific size variation did not affect viability selection (all χ2 <

1.11, P > 0.292), and there was no evidence of stabilizing or disruptive selection on either trait (both χ2 < 0.29, P > 0.591). We therefore compared selection on larval condition and head size between pools with and without Anax predators. Anax presence decreased overall survival (present: 46.2%; absent: 60.8%), resulting in ~80% greater among- individual fitness variation (present = 1.6876; absent = 0.9319). When Anax were present, larval survival increased strongly with head size and decreased strongly with body condition (Fig 4.2a,b, Table 4.1, n = 208). In contrast, in pools without A. junius, the fitness landscape showed marginal evidence of correlational selection that favored large 96 larvae with intermediate body condition and intermediately sized larvae with very high body condition (Fig. 4.2c,d, Table 4.1, n = 240). However, in these pools without A. junius, larval survival did not decrease with larval body condition overall, and also the positive relationship between larval survival and head size was >37% weaker (Table 4.1).

Qualitative comparisons of just high-condition larvae between these two predator treatments revealed similar patterns of survival. Larvae with a body condition 1 SD above average had 42% (15/36) and 26% (8/31) survival in the Anax-absent pools and Anax- present pools, respectively. Moreover, 71% (5/7) and 0% (0/5) of larvae with a body condition 2 SD above average survived in the Anax-absent pools and Anax-present pools, respectively. The average selection gradient calculated within each pool also showed that directional selection opposed body condition in pools with, but not without, A. junius

(Appendix 4). Together, this broadly supports differences in directional selection on body condition between predator treatments.

Aim 2: Performance and morphological correlates of body condition

Mean escape distance was not associated with larval body condition (F1,57 = 1.40, P =

0.241, Fig. 4.3a, n = 60), but tended to increase with head size (1.800 ± 0.969, F1,57 =

3.45, P = 0.068). Relative abdomen size increased with body condition (0.571 ± 0.097,

F1,40 = 34.87, P < 0.001, Fig. 4.3b, n = 44), but was not additionally influenced by food treatment or the interaction with condition (both F1,40 < 0.90, P > 0.349). Cuticle darkness marginally increased with body condition (24.166 ± 12.744, F1,40 = 3.66, P = 0.063, n =

44), but was not additionally affected by food treatment or the interaction with condition

(both F1,40 < 1.72, P > 0.197). 97

DISCUSSION

While the evolution of sexually selected traits may be substantially influenced by differences in the strength of trade-offs between their development and juvenile survival

(Jennions et al. 2001; Robinson et al. 2008; Mojica & Kelly 2010; Hine et al. 2011), the potential for, and ecological causes of, variation in the resulting ontogenetic conflict between the juvenile and adult life stages remains largely unexplored. Using an ornamented dragonfly, we examined how the larval environment could influence trade- offs between juvenile survival and the development of sexually selected male wing coloration, which is vital to an important adult fitness component (territorial success).

Specifically, we considered: 1) how larval body condition influenced the development of intrasexually selected wing ornaments, and 2) how viability selection on larval body condition varied with larval predation risk. Higher larval body condition improved the early development of the intrasexually selected wing ornaments (Fig. 4.1), and was directly opposed by viability selection only in pools with predatory A. junius (Fig. 4.2a,b).

The accumulation of resources used for producing intrasexually selected adult wing coloration is therefore associated with lower larval survival when larval predation risk is high. Conversely, when the large predatory A. junius was not present, viability selection did not directly oppose larval body condition overall (Fig. 4.2c,d). Thus, trade-offs between larval survival and the development of intrasexually selected wing coloration can depend on the larval predation regime. This highlights the potential for context dependence in trade-offs between adult and juvenile fitness components, and further indicates the importance of considering antagonistic selection over features of ornament development prior to maturity (Jennions et al. 2001; Cornwallis & Uller 2010). 98

The extent to which ontogenetic conflict between the juvenile and adult life stages will limit the adaptive evolution of sexually selected traits depends largely on the relative magnitudes of the fitness costs incurred through lower juvenile survival versus the fitness benefits gained through improvements to adult fitness components (Mojica & Kelly

2010; Johnson & Hixon 2011). Although we did not directly estimate the relationship between early ornament development and territorial success in this study, producing coloration sooner after emergence will enable a male to begin holding breeding territories earlier in the adult stage (Moore & Martin 2016), which in turn will greatly enhance its opportunity to acquire mates over its adult lifespan (Sherman 1983; see also Koenig &

Albano 1987; Moore 1990; Suhonen et al. 2008). In pools with A. junius, males with a larval body condition that was 1 SD above average had 38% lower larval survival than those males that were 1 SD below average (Table 4.1, Fig. 4.2a,b). Conversely, the probability of beginning ornament development on the day of emergence was 63.8% greater for those high-condition males than the low-condition males (Fig. 4.1). All else being equal (van Tienderen 2000), the benefits of this earlier ornament development for those high-condition males would then have to improve territorial success by 59.6% to offset the observed survival costs—equivalent to an additional 1.7 days on which a male holds a territory in our population (mean number of days holding a territory ± SD = 2.8 ±

2.2; Moore & Martin 2016). While increases to territorial success of this magnitude are certainly plausible (Sherman 1983; Grether 1996b), and other adult fitness benefits of high larval body condition may further offset these costs (e.g. female fecundity, Stoks &

Córdoba-Aguilar 2012), our results indicate that predator-mediated viability selection 99 against larval body condition can substantially limit the net benefits of the early development of an intrasexually selected trait.

Any costs of developing or possessing sexually selected traits that cannot be mitigated by increasing body condition impose strong constraints on the elaboration of these traits

(Jennions et al. 2001; Kotiaho 2001; Hine et al. 2011; Morehouse 2014). In pools with A. junius, high-condition larvae had lower survival (Fig. 4.2a,b), despite the fact that high- condition larvae tended to have darker, thicker cuticles (Fig. 4.3c) and are able to accelerate development in the presence of A. junius (Moore et al. 2018b). Previous work has shown that aeshnid dragonfly larvae preferentially capture and consume prey with bodies that are easier to grab and hold, such as other odonates with relatively large abdominal spines (Mikolajewski et al. 2006) and tadpoles with relatively wide and deep bodies (Van Buskirk et al. 1997). Consequently, it may have been the relatively large abdomens of high-condition larvae (Fig. 4.3b) that increased vulnerability to predation by providing an easier target for the extendable, grasping mouthparts of A. junius. Although other aspects of being large overall could improve performance and decrease predation risk, larger abdomens are unlikely to enhance other features of performance to offset any predator-mediated costs of being an easier target. Given the role of aeshnids as predators to other aquatic invertebrates (Strobbe et al. 2009), amphibians (Anderson & Semlitsch

2016), and even small fish (Marchinko 2009), this potential functional constraint of high body condition could similarly constrain the benefits of sexual phenotypes in many animals. Furthermore, as the accumulation of large energetic stores often increases vulnerability to predators by making animals slower (Zamora-Camacho et al. 2014) and/or easier to catch and handle (Gosler et al. 1995), environment-specific trade-offs 100 between juvenile survival and the development of sexually selected traits, mediated through the effects of body condition, could be common.

While our results show that differences in juvenile predation regime can modify the survival costs of developing sexually selected traits, they also indicate that differences in predators’ foraging ability may have important consequences for variation in ontogenetic conflict. For instance, despite cannibalism rates typically increasing with size asymmetry in dragonflies (Crumrine et al. 2008) and other animals with considerable among- individual size differences (e.g. fish: Persson et al. 2003; salamanders: Wissinger et al.

2010), a 24% increase in conspecific size variation did not change the strength or form of viability selection on larval body condition or head size—perhaps because of the gape limitations of P. longipennis (Wissinger 1992; Corbet 1999). Consequently, in contrast with the presence or absence of A. junius, which can consume any naturally occurring larval P. longipennis, differences in conspecific size variation are unlikely to modify the juvenile survival costs of the resource accumulation used in ornament production. This suggests that functional differences among putative predators, such as differences in gape-limitation or foraging strategy (see also Miehls et al. 2014), could crucially affect spatiotemporal variation in trade-offs. More generally, as our understanding of the effects of ecological agents of selection continues to expand (Moore et al. 2016; Siepielski et al.

2017; Caruso et al. 2017), considering variation within simplified habitat categories (e.g. predation or competition) could further illuminate the shared and unique components of the ecological causes of diversification (Langerhans & DeWitt 2004; Oke et al. 2017).

Trade-offs between life stages are likely to be an important constraint on adaptive evolution generally (Schluter et al. 1991; Marshall & Morgan 2011), and the elaboration 101 of secondary sexual traits specifically (Jennions et al. 2001). However, their underlying ecological causes have received only limited attention (see also Crean et al. 2011; Monro

& Marshall 2014). Overall, our results highlight 1) how viability selection against body condition can lead to trade-offs between larval survival and the development of sexually selected traits, and 2) how variation in an important ecological factor, predation, could alter the strength of these trade-offs. Although many animals accumulate resources for the production of the adult phenotype primarily during the juvenile stage (Morehouse

2014), the potential for ecological interactions during this stage to generate fitness variation that constrains the benefits of sexually selected traits has been largely ignored

(Jennions et al. 2001; Cornwallis & Uller 2010; Kasumovic 2013). Further examination of how ecological factors across the life cycle influence trade-offs between juvenile survival and sexual phenotype development therefore could ultimately yield exciting, novel insights into the evolution of secondary sexual traits, and the optimization of trade- offs between life stages more broadly.

102

Table 4.1. Viability selection gradients on head size (loge transformed) and body condition in pools with and without Anax junius. There was a significant interaction

! among head size, body condition, and Anax presence in the full model (�! = 4.75, P =

0.029), indicating differences in selection between these treatments (Chenoweth et al.

2012). Model estimates, their standard errors, and the significance tests are those from the logistic regression model. Selection coefficients and their standard errors are calculated from the logistic regression model estimates using the techniques advocated by Janzen &

Stern (1998) for binomially distributed fitness metrics. Directional selection coefficients were estimated after removing the cross-product (correlational) terms.

Model Selection Treatment Trait χ2 P Estimate ± SE Gradient ± SE Head Size 0.608 ± 0.172 0.344 ± 0.097 14.39 < 0.001 Anax Body Condition -0.338 ± 0.157 -0.191 ± 0.089 4.87 0.027 Present Head Size x Body Condition 0.259 ± 0.204 0.146 ± 0.115 1.64 0.200 Head Size 0.481 ± 0.136 0.215 ± 0.061 14.26 < 0.001 Anax Body Condition -0.213 ± 0.138 -0.096 ± 0.062 2.51 0.113 Absent Head Size x Body Condition -0.309 ± 0.167 -0.137 ± 0.073 3.67 0.062

103

● ●● ● ● ● 1.0 ● ● ● ● ● ● ● ● ● ●

0.5

● ● 0.0 ● Probability of initiating ornament production ● ● ● ●● ●

−1 0 1 Body condition

Figure 4.1. Males with higher larval body condition were more likely to initiate ornament development on the day of emergence. Each point represents whether or not a male began developing ornamentation (1 = initiated ornament production, 0 = did not initiate ornament production), and points are jittered vertically by 0.1 to improve visual clarity.

104

a) b) 1.0 1.2 1.4 1.2 1.0 0.8 0.6 ● ● ● ●● ● ● ● ● ● 1 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●●●● ● ●●● ● ● ● ● ● ● ● 2.0 ● ● ●● 0.4 ● ● ●● ● 1.6 1.4 0.2 0 1.5 ● ● ● ● ● ● 1.0 −1 ● ● ● ● ● ● ●

ANAX Survival (w) Survival ● 0.5 Head size

● −2 0.0 ● 1 0 3 2 −3 Head size−1 1 0 −2 −1 −2 0 2 −2 −3 −3 Body condition Body condition c) d) 1.2 1.4 0.8 1.0 ● ● 1.0 ● ●● ●●● ● ● ●● ● ● ● ● ●● ● ● ● 1 ● ● ● ● 1.2 ● ● ● ● ● ● ● ● ● ●●● ● ●●● ● ● ● ● ●●●● ● ●● ● ● ● ●● ● ● ● ● ● ● 2.0 ● ● ●● ● ● ● ● ● ● 1.4 ● ● ● ● ● ●● ● ● 1.6 0 1.5 ● ● ● 0.8 ● ● 0.6 ● ● ● ● ● 1.0 0.4 ●● ● ●● ● −1 ● ● 0.2 ● ● ● ● ● ● ●

Survival (w) Survival ● ● 0.5 Head size ● ● ● NO ANAX NO ● ● 0.2 −2 0.0 1 3 ● 0 2 −3 Head size 1 −1 0 −2 0 2 −2 −1 −2 −3 Body condition Body condition

Figure 4.2. Fitness surfaces for larvae in pools with (a,b) or without (c,d) Anax junius.

Surfaces were calculated with projection pursuit regression using cubic splines estimated by generalized cross validation. In contour plots showing the fitness surfaces in 2 dimensions (b,d), points represent individuals that either survived (circles) or died (x’s), and lines illustrate regions of relative fitness.

105

c) a) ● b) ● ● 200 ● ● ● 4.1 ● ● ● ● 0.1 ● ●● ●

[mm]) ● ● ● e ●● ● ● ● ●● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 190 ● ● ● ● ● ● ● ● ● ● ●● ● ● ● ● ● ● ● ●● ●● ● ●● ● ● ● ● ●● 3.6 ● ● ● ●●● ●● ● ●● ● ● ●● ● 0.0 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ●● ●● ● ● ● ● ● ● ● ● ● ● 180 ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● ● 3.1 ● ● ● −0.1 ● MeanEscape Distance (log ● AbdomenSize (residuals) Relative ● ● ● CuticleDarkness (255-mean gray value) 170 ● ● ● ●

−0.2 0.0 0.2 −0.2 −0.1 0.0 0.1 −0.2 −0.1 0.0 0.1 Body Condition BodyBody Conditioncondition Body Condition

Figure 4.3. The effects of body condition on a larva’s mean escape distance (a) and relative abdomen size (b). Each point represents an individual, and individuals are not the same between two panels. Mean escape distances were loge transformed to improve normality.

106

CHAPTER 5: IMMUNE DEPLOYMENT INCREASES LARVAL

VULNERABILITY TO PREDATORS AND INHIBITS ADULT LIFE-HISTORY

TRAITS IN A DRAGONFLY

Published in Journal of Evolutionary Biology Vol. 31, No. 9, pp. 1365-1376, 2018. DOI:

10.1111/jeb.13337

Authors: Michael P. Moore1, Cassandra Lis2, and Ryan A. Martin1

1. Department of Biology, Case Western Reserve University, Cleveland, OH USA

2. Hathaway Brown School, Shaker Heights, OH USA

(Submitted: 11 April 2018; Returned for Revision: 5 June 2018; Accepted: 19 June 2018)

ABSTRACT

While deploying immune defenses early in ontogeny can trade-off with the production and maintenance of other important traits across the entire life cycle, it remains largely unexplored how features of the environment shape the magnitude or presence of these lifetime costs. Greater predation risk during the juvenile stage may particularly influence such costs by 1) magnifying the survival costs that arise from any handicap of juvenile avoidance traits and/or 2) intensifying allocation trade-offs with important adult traits.

Here, we tested for predator-dependent costs of immune deployment within and across life stages using the dragonfly, Pachydiplax longipennis. We first examined how larval immune deployment affected two traits associated with larval vulnerability to predators: 107 escape distance and foraging under predation risk. Larvae that were induced to mount an immune response had shorter escape distances but lower foraging activity in the presence of predator cues. We also induced immune responses in larvae and reared them through emergence in mesocosms that differed in the presence of large predatory dragonfly larvae

(Aeshnidae spp.). Immune-challenged larvae had later emergence overall and lower survival in pools with predators. Immune-challenged males were also smaller at emergence and developed less sexually selected melanin wing coloration, but these effects were independent of predator treatment. Overall, these results highlight how mounting an immune defense early in ontogeny can have substantial ecological and physiological costs that manifest both within and across life stages.

INTRODUCTION

The persistent threat of parasites and pathogens should favor increasingly robust immunity, yet immune function is often imperfect and highly variable among individuals within and across populations (Sheldon & Verhulst 1996; Rolff & Siva-Jothy 2003). One important constraint on the evolutionary maximization of immune function may be the inherent trade-offs between mounting immune responses and other important traits that arise due to resource limitations (Zuk & Stoehr 2002; Schmid-Hempel 2005; McKean et al. 2008) and/or incidental tissue damage (“self-harm” or “autoreactivity”, Sadd & Siva-

Jothy 2006; Khan et al. 2017). Indeed, the deployment of a robust immune response often comes at a substantial cost to traits such as locomotor performance (Adamo et al. 2008;

Zamora-Camacho et al. 2014), investment in offspring size (Uller et al. 2006), sexual 108 displays (Jacot et al. 2004, 2005), and longevity (Armitage et al. 2003, Krams et al.

2013; Khan et al. 2017). Characterizing trade-offs between immune deployment and other functionally important traits can thus illuminate the forces maintaining variation in immunity.

The costs of deploying immune defenses during the juvenile stage may have particularly vital, but often overlooked, effects on the evolution of immunity (Jacot et al.

2005; McNamara et al. 2013; Kelly et al. 2014). Because the resources accumulated early in ontogeny must be partitioned among traits influencing juvenile survival, the transition to adulthood, and the development of adult phenotypes (Kasumovic 2013), immune deployment during the juvenile stage could trade-off with fitness components both within and across life stages (McNamara et al. 2013). Moreover, vulnerability to predators is often highest during this stage because individuals are smallest (e.g. Werner & Gilliam

1984; Wellborn 1994; Sogard 1997), and predation risk may further modify the strength, or even presence, of such costs across the rest of the life cycle (Zuk & Stoehr 2002). For instance, greater predation risk may increase the survival costs of trade-offs between mounting an immune defense and important avoidance or defensive traits (Adamo et al.

2008; Otti et al. 2012; Janssens & Stoks 2014). Additionally, predator-induced behavioral or developmental modifications (Benard 2004; Relyea 2007) may also deplete resources that would otherwise be available for the trade-offs between immune deployment and important adult traits. Yet, despite the widespread importance of the lethal and non-lethal effects of predation risk on juvenile life stages (Lima 1998; Abrams 2000; Moore et al.

2016), few studies have experimentally assessed how predation shapes the costs of immune deployment generally (Rantala et al. 2010; Otti et al. 2012; Janssens & Stoks 109

2014), and no studies have considered how it affects the costs of immune deployment both within and across life stages under ecologically relevant conditions.

An important component of the arthropod immune system, the melanotic encapsulation response, is well suited for examining how environmental conditions, such as predation risk, influence the costs of immune deployment across the life cycle more generally (Rolff & Siva-Jothy 2003; Schmid-Hempel 2005). When challenged by a parasite or pathogen, arthropods encapsulate the intruder in an aggregation of melanin- producing haemocytes, which ultimately asphyxiates and kills the foreign invader

(Schmid-Hempel 2005). This melanin-based immune response is energetically costly

(Freitak et al. 2003; Ardia et al. 2012; González-Santoyo & Córdoba-Aguilar 2012) and highly sensitive to environmental conditions (including predation risk: Stoks et al. 2006;

Duong & McCauley 2016), making it likely to be involved in many allocation trade-offs within life stages. Because of the cytotoxic properties of the melanin-synthesis pathway, this non-specific immune response can also result in considerable incidental tissue damage (Nappi & Vass 1993; Sadd & Siva-Jothy 2006; Khan et al. 2017). Moreover, the functioning of the melanin-synthesis pathway directly carries over from the larval to the adult life stage in arthropods (Debecker et al. 2015) and is used in the production of sexually selected coloration (Hooper et al. 1999; Lawniczak et al. 2007; Wittkopp &

Beldade 2009), making the overall encapsulation response likely to affect fitness components across life stages as well.

Here, we experimentally tested how larval predation risk influences the costs of immune deployment within and across life stages using a dragonfly, Pachydiplax longipennis, that produces dark intrasexually selected wing coloration (Moore & Martin 110

2016). We specifically addressed three questions. First, does larval immune deployment affect predator avoidance traits? We predicted that larvae that were experimentally induced to mount an immune defense would have traits conferring greater vulnerability to predators, including reduced larval escape performance and greater foraging activity under predation risk. Second, is larval immune deployment costly for other larval life- history traits? We predicted that immune deployment would reduce larval survival when we reared larvae in experimental wading pools with predators and cause later emergence times. Finally, does predation risk intensify any trade-offs between larval immune deployment and adult life-history traits? We predicted that immune deployment would inhibit adult life-history traits, such as adult size and initial ornament development, and that these costs would be greatest for individuals from wading pools with predators.

METHODS

Study Species and Study Overview

Pachydiplax longipennis (Burmeister) is a medium-sized dragonfly with a wide distribution across North America (Paulson 2012). In the northern extent of the range, where this study was conducted, P. longipennis is univoltine, commonly emerging from ponds with and without fish between late-May and mid-July (McCauley et al. 2008).

Prior to emergence, the developing larvae primarily inhabit aquatic vegetation for most of the larval stage (Paulson 2012). Following emergence, males in this part of the range produce dark, intrasexually selected wing coloration. A male producing more wing coloration intimidates its rivals, reducing the aggression it receives, and enabling it to 111 hold a territory for longer throughout each day (Moore & Martin 2016). Males without wing coloration are largely unable to hold breeding territories. All else being equal, producing as much wing coloration as soon after emergence as possible then substantially improves male reproductive success by 1) expanding the number of days in its adult lifespan that it can defend the breeding territories that are necessary for successful mating and 2) increasing the duration within each day that it can hold a territory (Sherman 1983;

Moore & Martin 2016; see also Moore 1990; Grether 1996b; Córdoba-Aguilar 2002).

We used a combination of laboratory and mesocosm approaches to address our three questions. To answer if larval immune deployment affects predator avoidance traits, we used a laboratory approach where we induced an immune response and then compared two performance traits known to proximately affect vulnerability to predation: 1) escape performance (e.g. Strobbe et al. 2009) and 2) foraging in the presence of predator cues

(e.g. Start 2018). Next, to examine if larval immune deployment is costly for other larval life-history traits, we reared immune-challenged larvae through emergence in 12 mesocosms (~100 L, 100 cm D x 17 cm H) that differed in the presence of aeshnid dragonfly larvae (n = 6 pools per predator treatment), important intraguild predators of P. longipennis (Wissinger 1992; Moore & Martin 2018), and compared larval survival and emergence times. Finally, using those individuals that survived and emerged from mesocosms, we tested if predation risk intensifies any trade-offs between larval immune deployment and adult size, adult condition, and male wing color development.

112

Collection and Manipulating Immune Deployment

We collected all P. lonigpennis larvae for this study from a single large pond at Case

Western Reserve University’s Squire Valleevue Farm (Hunting Valley, Ohio, USA;

41.4929 N, 81.4234 W; pond area: ~2074 m2). While both aeshnid species that we used

(Anax junius and Aeshna umbrosa) inhabit this pond despite the presence of fish, we collected these predators from a nearby pond where they are more numerous. After collection, we kept all larvae under a 15:9 L:D photoperiod in opaque plastic cups filled with 473 mL of dechlorinated water. We randomly assigned these P. longipennis larvae to either an immune-challenge or sham treatment.

We experimentally induced melanotic encapsulation responses with sterilized, roughened, knotted nylon monofilaments (1.84 mm length, 0.18 mm diameter). We inserted each implant up to the knot between the fifth and sixth tergites of each larva’s abdomen and then removed the implant after 24 hours (Moore & Martin 2016). This

“24HR-challenge” treatment closely mimics a natural immune insult by a common and important source of parasitism in dragonflies, parasitic mites (reviewed in Forbes & Robb

2008). The amount of melanin deposited on the nylon implant is associated with the strength of the immune response (Rantala & Roff 2007). While the energetic costs of the subsequent encapsulation response are known to be substantial (e.g. Freitak et al. 2003;

Ardia et al. 2012), incidental tissue damage caused by auto-immunity may also contribute to the total costs observed here (Sadd & Siva-Jothy 2006; Khan et al. 2017). For a control treatment, we inserted an implant but then removed it immediately. This “sham-challenge” treatment controls for any potential wounding effects of the implant insertion on muscles and/or fat bodies. Because we were interested in estimating just those costs of mounting 113 an immune response, we did not include an unmanipulated control, which often differs from sham treatments in many ways beyond closing the wound caused by the sham treatment (e.g. stress response, Sadd et al. 2006; Rantala et al. 2010). Although we cannot rule out the possibility of causing an incidental infection during the sham and challenge procedures, we punctured the cuticle of all individuals in the study, and any effects of the immune-challenge treatments therefore act in addition to any such incidental infection. Our design here thus allows us to isolate just the potential costs of the immune deployment (see Ardia et al. 2012 for a similar design and rationale).

Melanin deposition and immune challenge duration

To ensure that the production and allocation of melanin for immune defense increases with the duration of the challenge, we also challenged 49 final instar P. longipennis larvae for 5, 8, 12, or 24 hours using a knotted nylon monofilament. We kept larvae in pairs in 500 mL plastic containers (one larva died during immune challenge). After the implant was removed, we removed and digitally photographed it at three angles against a standard brightfield lighting condition with a dissecting microscope (Moore & Martin

2016). For each picture, we used ImageJ (Rasband 2012) to calculate the difference in implant darkness (mean grey value) between the area that was versus was not inserted into the larva. Higher scores indicate greater melanin deposition (Rantala & Roff 2007).

We assessed the relationship between melanin deposition and challenge duration using a linear mixed-effects model with the mean melanin deposition across the three photographs as the response variable, challenge duration as a fixed effect, and container as a random effect. 114

Laboratory Experiment: Does larval immune deployment affect predator avoidance traits?

We first examined how the costs of immune deployment manifest on two traits important to larval survival under elevated predation risk: escape performance and foraging behavior. We collected ultimate larval instar individuals (also called F-0 larvae; e.g.

Corbet 1999) on 10 July 2016, kept them separately in the laboratory as described above for two days, and then subjected them to their randomly assigned immune-challenge treatment. Twenty-four hours after the implants were removed from the 24HR-challenged larvae, we compared the escape performances of the 24HR- and sham-challenged larvae.

We placed each larva next to a ruler in the center of one of two clear plastic boxes (34.6 cm L x 21 cm W x 12.4 cm H) each filled with 1 L of dechlorinated tap water. After approximately one minute, we gently prodded each larva’s abdomen with forceps and measured the distance (mm) of its initial escape burst by eye along the length of the ruler three times in rapid succession. This performance metric is significantly repeatable (R =

0.414, F64,129 = 3.12, P < 0.001; Lessells & Boag 1987) and should reflect an individual’s ability to move out of the attack radius of other dragonfly larvae (Corbet 1999). We used a general linear model to consider how mean burst distance (loge transformed) differed between immune treatments.

We then investigated whether immune deployment altered foraging behavior in the presence of predator cues. Using the same individuals right after the escape performance trials, 24HR- and sham-challenged larvae were placed individually in clear plastic boxes

(dimensions as above) filled with 1 L of dechlorinated tap water, and provided 15 adult 115

Daphnia magna (length ± SD, n: 3.24 ± 0.35 mm, 15) and 3 mL of highly concentrated predator cues. To generate predator cues, we took water from 500 mL impermeable plastic containers that each held one aeshnid larva that had consumed at least one P. longipennis larva the previous night (Brodin et al. 2006, Duong & McCauley 2016).

Because we were specifically interested in testing if larval immune deployment promotes riskier behavior (Start 2018), we only conducted these trials in the presence of predator cues. We allowed each larva to forage freely in arena for 90 minutes, after which we removed it and counted the remaining D. magna. All trials occurred simultaneously with larvae in separate boxes. We compared foraging between 24HR- and sham-challenged and larvae using a generalized linear model with the number of D. magna eaten as the response variable and treatment as the explanatory variable. We used a quasipoisson error distribution to account for overdispersion.

Mesocosm Experiment: Is larval immune deployment costly for other larval life- history traits?

We next experimentally tested how predators influence the costs of immune deployment on larval survival and emergence. We collected ultimate larval instar and penultimate larval instar (next-to-last instar before emergence, also called F-1; Corbet 1999) individuals on 16 - 17 June 2017, kept them separately in plastic cups for four days, and then subjected them to their randomly assigned immune-challenge treatments. To account for any potential differences in the date of collection, larvae collected on each day were kept in separate temporal blocks. To manipulate allocation to immune deployment, we randomly assigned larvae within each assigned pool to one of three immune treatments: 116

1) a sham challenge, 2) a 12-hour immune challenge (“12HR-challenge”), or 3) a 24HR- challenge. We included the additional 12HR-challenge treatment to further consider how intermediate allocation towards an immune defense can influence larval and adult life- history traits (see also Krams et al. 2013). One day after removing implants, we marked P. longipennis larvae within pools by their immune treatment by removing one leg using microdissecting scissors. To avoid confounding immune treatment with leg removal, we randomized which leg was removed for each immune treatment for each pool. Survival in the presence of predatory aeshnids does not depend on which leg is removed (Appendix

5), and we did not observe additional mortality due to leg removal prior to releasing larvae.

Two days after implants were removed, we released nine P. longipennis larvae of each immune treatment into each plastic wading pool (n = 27 larvae per pool), which were filled with dechlorinated water, ~0.003 m3 of aquatic vegetation, ~0.011 m3 of leaf litter, and ~1400 cm3 of benthic substrate. The ratio of ultimate larval instar to penultimate larval instar individuals released into the wading pools did not differ among immune

! treatments (�! = 1.04, P = 0.593). To provide a self-sustaining prey base for the larvae, we introduced 100 mL of concentrated D. magna 10 days before releasing P. longipennis larvae (~100 daphnids). We also permitted the natural colonization of prey species (e.g. culicidae, chironomidae, and hylidae spp.) up until the day prior to dragonfly introduction, after which we covered pools with window screen. In the aeshnid-present pools, we introduced two aeshnid larvae (one Anax junius and one Aeshna umbrosa) ~36 hours after the P. longipennis larvae and subsequently replaced them when they emerged or were found dead. We checked pools for adults at both dawn and dusk every day and 117 measured the mass of each collected adult to the nearest 0.0001 g on the day of emergence. We maintained adults in the laboratory in plastic cups for three days (see below), after which we sacrificed them. After 27 days, we emptied all tanks and recovered survivors that had not emerged.

We assessed differences in survival, the probability of emergence, and emergence date among treatments using mixed-effects models with predator and immune treatments as fixed effects, and pool nested within block as random effects. When an individual was recovered as an emerged adult or as a larva at the end of the experiment, it was scored as a “survivor”. We used binomial error distributions for the analyses of survival and emergence probability, whereby all of the individuals released in the experiment were included in the analysis of survival (n = 324), and all of the individuals that were scored as survivors were included in the analysis of emergence probability (n = 96). We scored an individual’s emergence day as the number of days (to the nearest 0.5) after release into the wading pools that the individual emerged. As sex could be determined via eye color and cerci morphology for individuals that successfully completed emergence (nmales = 37, nfemales = 36), we also considered how the effects of the immune and predator treatments on emergence day differed between the sexes.

Mesocosm Experiment: Does predation risk intensify any trade-offs between larval immune deployment and adult life-history traits?

To evaluate how the costs of mounting an immune defense interact with the larval environment to shape adult life-history traits, we considered how immune deployment and predation risk influenced adult size and body condition for both sexes and 118 intrasexually selected wing color development in the males that successfully emerged from the wading pool study. After sacrificing adults, we gently removed and photographed their wings against a standard white background (DGK Color Tools®) in a dark box that excluded any light besides the flash. Because of the very strong correlation between wing area and body size (r = 0.834; Moore & Martin 2016), we used the average area of the rear wings as an estimate of adult size in all analyses (Corbet 1999). We analyzed differences in adult size using a mixed-effects model with immune treatment, predator treatment, and their interaction as fixed effects and tank nested within block as random effects. We also considered how effects of the immune and predator treatments differed between males and females by fitting sex and its interactions with the treatments as fixed effects. We included emergence day as covariate to test for trade-offs between age and size at emergence. We further analyzed differences in mass at emergence using a mixed-effects model with immune treatment, predator treatment, sex, and their interactions as fixed effects and tank nested within block as random effects. To isolate the potential effects of the treatments on adult mass independent of adult size (i.e. body condition), we included adult size as a covariate (García-Berthou 2001). We again included emergence day as a covariate.

To quantify the extent of male wing coloration on the third day after emergence, we used ImageJ (Rasband 2012) to identify the highest mean grey value (0-255, 0 = most opaque; 255 = most transparent) of the pigmented portion of each wing (i.e. least darkly pigmented), converted the photograph to binary black and white with this value as the threshold for black, and calculated the size (mm2) of the digitized black area. Wing coloration was then estimated as the proportion of the total wing area that was pigmented, 119 which, unlike darkness or homogeneity, is a known target of intrasexual selection (Moore

& Martin 2016). These pictures are taken against a standard white background in a box that excludes any incident light besides the flash, and this procedure has previously been shown to be highly repeatable by our lab group (Moore & Martin 2016). We characterized wing coloration at this point in its development because 1) we were interested in isolating any direct trade-off over melanin production between larval immune deployment and adult wing color development (Siva-Jothy 2000), and 2) adult survival in the lab declines markedly after the third day without additional feeding.

Nonetheless, while the extent of wing coloration on the third day after emergence was typically, but not always, less than that of mature males defending territories (mean ± SD, range; third day of emergence = 0.052 ± 0.045, 0.008-0.181; territorial males = 0.411 ±

0.089, 0.150-0.673, Moore & Martin 2016), males with greater coloration sooner after emergence will be able to occupy and hold territories earlier, increasing the number of mates they can acquire (Sherman 1983; Moore & Martin 2016). We first evaluated the probability of an individual having produced at least some wing coloration by the third day after emergence (1 = had wing coloration, 0 = did not have wing coloration) using a generalized linear mixed-effects model with immune treatment, predator treatment, and their interaction as fixed effects, tank nested within block as a random effect, and a binomial error distribution. Then, among those males that had initiated wing color development, we compared the extent of wing coloration among immune treatments, predator treatments, and their interaction using a linear mixed-effects model with tank nested within block as a random effect. Wing coloration was arcsine square-root transformed prior to analysis. 120

Statistical Analyses

We conducted all analyses in R v.3.3.3 (R Core Team, 2017), fitting mixed-effects models using the ‘lme4’ package (Bates et al. 2015) and conducting Tukey post-hoc tests with ‘lsmeans’ package (Lenth 2016). To test the significance of fixed effects in mixed- effects models, we used likelihood ratio tests and F-tests with the Kenward-Roger denominator degrees of freedom approximation for generalized linear and general linear models, respectively. In one aeshnid–absent treatment pool, five aeshnids and no P. longipennis were recovered at the end of the experiment, and we have therefore removed this pool from all analyses. Aeshnids were not found in any other aeshnid-absent treatment pool. Additionally, we excluded one sham and two 24HR-challenge individuals that emerged within 5 days of being released from all analyses. Pachydiplax longipennis requires ~3 - 4 days to undergo the metamorphic process, and responses to immune challenge of these individuals would therefore not reflect much exposure to our predator manipulation (Corbet 1999). Additionally, these outliers had very high influence on statistical results (Cook’s Distances: sham individual = 0.049, 24HR individuals = 0.193,

0.164; threshold for this analysis [4/n-k-1] = 0.052), rendering analyses that included them unreliable. Because some immune and predator treatment combinations produced fewer than 3 individuals of one sex, we never fit immune treatment by predator treatment by sex interactions to avoid overfitting. We report all parameter estimates as model estimate ± SE, and, for brevity, we report full results from post hoc comparisons in the online appendix (Tables S5.1 - S5.8).

121

RESULTS

Melanin deposition and immune challenge duration

As in other studies (Kapari et al. 2006; Contreras-Garduño et al. 2008; Krams et al.

2013), melanin deposition increased with the challenge duration (0.027 ± 0.008, F1,23.4 =

11.16, P = 0.003; Fig. S5.1).

Laboratory Experiment: Does larval immune deployment affect predator avoidance traits?

Sham-challenged larvae had 33.2% longer escape distances (0.287 ± 0.056, F1,63 = 26.10,

! P < 0.001; Fig. 5.1a) and consumed 59.9% more D. magna (0.470 ± 0.199, �! = 5.87, P

= 0.015; Fig. 5.1b) than 24HR-challenged larvae.

Mesocosm Experiment: Is larval immune deployment costly for other larval life- history traits?

! Survival costs of immune deployment depended on predation risk (interaction: �! = 6.72,

P = 0.035). Sham-challenged larvae were more than twice as likely to survive in aeshnid- present pools than 24HR-challenged larvae, but survival of 12HR-challenged larvae was not different from either of the other immune treatments (Fig. 5.2a, Table S5.1). In pools without predators, survival did not differ among immune treatments. (all P > 0.324).

Additionally, the survival of each immune treatment never differed significantly between predator treatments (all P > 0.063). 122

! The probability of emergence depended on immune treatment (�! = 10.07, P = 0.007),

! but not predator treatment (�! < 0.01, P = 0.966). Likely due to the high overall probability of emergence among survivors (83.3%), we could not test the interaction because models including this term would not converge. Sham-challenged larvae were more likely to emerge than either 12HR- (P = 0.045) or the 24HR-challenged larvae (P =

0.031), which were not different from each other (P = 0.978; Table S5.2). Emergence day also depended on immune treatment (F2,60.4 = 4.49, P = 0.015), but not predator treatment

(F1,9.2 = 0.92, P = 0.363), or the interaction between predator and immune treatment

(F2,60.1 = 0.06, P = 0.939). Sham-challenged individuals emerged earlier than 24HR- challenged individuals (P = 0.013), and 12HR-challenged individuals were not different than either treatment (both P > 0.130; Fig 5.2b; Table S5.3). There was also a non- significant trend of females emerging earlier than males (-2.180 ± 1.077, F1,61.4 = 3.73, P

= 0.058), but this effect did not depend on predator (F1,62.8 = 0.349, P = 0.557) or immune treatment (F1,60.0 = 0.22, P = 0.805).

Mesocosm Experiment: Does predation risk intensify trade-offs between larval immune deployment and adult life-history traits?

Immune treatment had sex-specific effects on adult size (immune: F2,56.4 = 2.32, P =

0.108; immune x sex: F2,55.1 = 4.91, P = 0.011), but adult sizes were not otherwise affected by the predator treatment, the interaction between immune and predator treatment or the interaction between the predator treatment and sex (all P > 0.399; Table

S5.4). While female size was unaffected by immune treatment, sham-challenged males were larger than 24HR-challenged males (P = 0.006), but 12HR-challenged males did not 12 3 differ from either (both P > 0.207, Fig. 5.3a, Table S5.5). Sham-challenged males were also larger than their female counterparts (P = 0.006), but sizes did not differ between the sexes within either the 12HR-challenge, or 24HR-challenge treatments (both P > 0.354,

Table S5). Adult size did not vary with emergence day (F1,57.0 = 0.91, P = 0.344).

Although mass at emergence increased with wing size (3.057 x 10-4 ± 1.127 x 10-4,

F1,56.9 = 6.57, P = 0.013), it was not affected by immune treatment, predator treatments, sex, emergence day, or any of the interactions (all P > 0.063; Table S5.6), indicating no effects on body condition (García-Berthou 2001).

Among the 37 males that successfully emerged, the probability of producing at least some wing coloration by the third day after emergence also was not affected by immune treatment, predator treatment, or their interaction (all P > 0.078, Table S5.7). However, among the 25 males that had initiated color development, the extent of wing coloration differed among immune treatments (F2,14.2 = 5.21, P = 0.020; Fig. 5.3b), whereby sham- challenged males had produced ~3.5 times more wing coloration than 24HR-challenged males (P = 0.008), while 12HR-challenged males did not differ from either (both P >

0.271, Table S5.8). Neither predator treatment (F1,1.8 = 4.08, P = 0.191) nor the interaction between the immune and predator treatments (F2,13.0 = 0.14, P = 0.868) influenced the extent of wing coloration.

DISCUSSION

As the costs of mounting an immune defense early in ontogeny can lead to fitness trade- offs both within and across life stages, exploring how these deployment costs manifest across the life cycle is broadly important to understanding the potential constraints on the 124 evolution of immunity. Here, P. longipennis larvae that were induced to mount the largest immune responses (24HR-challenge) exhibited lower larval foraging ability and smaller escape distances, as well as later emergence, smaller adult sizes, and weaker ornament development in males. Survival of these larvae also strongly depended on predation risk, likely due to trade-offs between immune deployment and larval performance traits. Thus, trade-offs between immune deployment and these performance traits resulted in larval survival costs only in the presence of predators. In contrast, while immune deployment strongly inhibited the development of key adult life-history traits, we did not observe any further synergistic effects with predation risk on these traits, suggesting that predators do not magnify any costs that carry over to adulthood. Ultimately, evaluating the costs of immune deployment remains central to understanding the evolutionary forces shaping the optimization of immunity (Schmid-Hempel 2005; Urban et al. 2013), and these results indicate that future estimation of the ecological and physiological costs of immune deployment should consider trade-offs both within and across life stages.

The costs of mounting an immune response early in ontogeny manifested immediately through trade-offs with performance, lowering larval survival when predation risk was high. As aeshnid larvae are sit-and-wait ambush predators that capture prey using rapid extension of their grasping mouthparts (Corbet 1999), any reduction in the speed or distance of initial escape responses, like those observed here, may greatly lower the chances of eluding attacks by these voracious predators (Strobbe et al. 2009).

Additionally, because 24HR-challenged individuals emerged later overall, they spent longer in the risky larval environment than their control-treatment counterparts, thereby increasing the duration of vulnerability. Another study of odonates (Janssens & Stoks 125

2014) found that immune deployment against non-pathogenic bacteria increases foraging activity, elevating the encounter rate with ambush predators and potentially resulting in higher mortality. However, 24HR-challenged larvae actually consumed fewer daphnids in the presence of aeshnid predator cues, and reduced vigilance consequently seems unlikely to increase vulnerability to aeshnids (see also König & Schmid-Hempel 1995).

Predator-mediated costs of immune deployment thus probably stemmed more from trade- offs with escape performance and emergence time. While the substantial energetic costs of immune deployment seem likely to primarily underlie these trade-offs (Freitak et al.

2003; Ardia et al. 2012), evaluating any additional contribution of autoreactive tissue damage (Sadd & Siva-Jothy 2006; Khan et al. 2017) or shared reliance on transport proteins (Adamo et al. 2008) remains necessary. Nevertheless, as trade-offs between performance and immune deployment have also been observed in other organisms

(Adamo et al. 2008; Otti et al. 2012; Zamora-Camacho et al. 2014b), predator-dependent costs of immune deployment like those observed here could commonly constrain the evolutionary maximization of immune responses.

In addition to increasing vulnerability to predators, the costs of larval immune deployment also carried over to inhibit the development of two male traits that selection in the adult stage typically favors: size and sexual coloration. Sexual selection on adult males consequently should not only promote the evolution of weaker larval immune responses in males but also drive the evolution of sub-optimal female immunity when between-sex genetic correlations are high (e.g. Rolff et al. 2005; McKean & Nunney

2008). Thus, even though there were not strong costs of immune deployment for female traits, sexually antagonistic fitness optima may constrain the evolutionary maximization 126 of juvenile immune defenses and ultimately maintain phenotypic variation (Bonduriansky

& Chenoweth 2009; Cox & Calsbeek 2009). Proximately, the depletion of specific nutrients provides one plausible explanation for the observed sex-specific adult costs, as immune deployment did not reduce adult body condition. For instance, the shared reliance of immune defenses and sexual coloration on the melanin-synthesis pathway may cause limitations of amino acids such as tyrosine (see also Hooper et al. 1999;

Debecker et al. 2015; Evison et al. 2017). Furthermore, sex-specific allocation of particular nutrients for even the same suites of traits (e.g. Maklakov et al. 2008;

Morehouse et al. 2010; Reddiex et al. 2013) could underlie sex-specific costs of immune deployment to adult size. While any such proximate mechanisms remain to be elucidated, larval immune deployment has significant adult costs, and evolution of these early ontogenetic defenses is therefore likely to be shaped by the unique and varying selective pressures across the entire life cycle.

Contrary to our predictions, non-lethal effects of predation risk did not intensify the costs of immune deployment to adult traits. Despite strong theoretical predictions, predation risk frequently has equivocal effects on life-history traits in organisms with complex life cycles (reviewed in Benard 2004; Relyea 2007), perhaps because behavioral modification offset any effects (McPeek et al. 2001; Touchon et al. 2015; Moore et al.

2018b) or trade-offs occur with traits experiencing more diffuse selective pressures than life-history traits (e.g. oxidative stress, Slos et al. 2009; digestive physiology, Stoks et al.

2005). Nonetheless, once the juvenile environment becomes unsuitable for survival and further growth, organisms often increase energetic investment in rapid development (e.g.

Morey & Reznick 2000; Dugas et al. 2016), which could exacerbate trade-offs between 127 immune deployment and important adult traits. While predators do not appear to magnify such trade-offs across the life cycle, other features of the larval environment could (e.g. seasonal time constraints: Mikolawjewski et al. 2015) and may then mediate the lifetime costs of immune deployment.

Although consistently producing the maximum possible immune response may ensure the clearance of parasites and pathogens, subsequent trade-offs can limit the net benefits of mounting very robust immune defenses (Schmid-Hempel 2005). The optimal magnitude of each immune response therefore must balance the risk of an insufficient defense with the costs incurred to the production or maintenance of other phenotypic targets of selection (Zuk & Stoehr 2002; Urban et al. 2013). Mounting a strong immune defense early in ontogeny had ecological and physiological costs both within and between life stages—representing a potentially strong constraint on the evolutionary maximization of immune defense. However, larvae that were challenged for 12 hours had intermediate larval survival, emergence time, and male size and color development. Our results thus suggest that allocating less towards an immune deployment when possible can provide higher net fitness returns across environments and life stages (Krams et al.

2013), but direct estimation of phenotypic selection on immune deployment remains necessary in this and most other animals (Seppälä 2015). Overall, the adaptive evolution of immunity in any organism depends on a complex suite of interacting genomic and selective factors (Sheldon & Verhulst 1996; Zuk & Stoehr 2002; Urban et al. 2013), and our findings demonstrate how ecologically dependent trade-offs both within and across life stages could be one potentially important constraint.

128 a) 65

35 mean escape distance (mm)

20 b) control 24−hour challenge 15

5 number of daphnia consumed number

sham 24HR

Figure 5.1. Immune deployment reduced escape distances (a) and the number of daphnids consumed in 90 minutes (b). Circles depict the value for an individual larva and the squares represent the mean ± 95% confidence intervals. For each individual, mean escape distances (a) were the average of three trials.

129 a) 0.6

0.4 probability of survival

0.2

b) no aeshnids aeshnids 27

emergence day 15

sham 12HR 9 24HR

no aeshnids aeshnids

Figure 5.2. Relative to sham-challenged larvae, 24HR-challenged larvae had lower survival in the presence of aeshnid predators (a), and later emergence across all larval environments (b). The probabilities of survival (a) are ± 95% confidence intervals, and are those estimated from the mixed-effects model that was reported in the Results.

Individuals’ emergence days (b) are illustrated with circles, and the squares represent the treatment means ± 95% confidence intervals that were estimated from the mixed-effects model reported in the Results.

130

a) sham b) 12HR 0.18 240 24HR ) 2

m 220 0.12 m (

a e r a

g n i w 200 r

a 0.06 e r proportion of wing area with color 180

0.00 females males sham 12HR 24HR c)

5 mm 5 mm 5 mm

Figure 5.3. Males from the 24HR-challenged treatment were smaller at emergence (a) and produced less wing coloration than sham-challenged males by the third day after emergence (b). Each circle depicts the trait values for an individual, and the squares represent the treatment means ± 95% confidence intervals that were estimated from the mixed-effects model reported in the Results. Representative examples of male wing coloration (c) for: the average extent of coloration of mature males (left), an above- average extent of color development on the third day after emergence (middle), and no coloration on the third day after emergence (right). 131

CHAPTER 6: EVOLUTIONARY TRADE-OFFS BETWEEN LARVAL IMMUNE

DEFENSE AND ADULT WING COLORATION ACROSS DRAGONFLIES

Authors: Michael P. Moore & Ryan A. Martin

Department of Biology, Case Western Reserve University, Cleveland OH USA

ABSTRACT

When organisms face contrasting selective pressures across ontogeny, they are expected to evolve mechanisms that erode developmental links between life stages and enable the independent evolution of each. The dramatic metamorphosis undertaken by many animals is thought to facilitate the independent evolution of traits in different parts of the life cycle, yet empirical evidence remains mixed. Here, we test for independent evolution across life stages in dragonflies. We specifically focus on two key traits that rely on a single metabolic pathway: larval melanin immune defense and adult melanin wing coloration. If metamorphosis decouples these traits and enables their independent evolution, we should observe that they do not co-vary across species. Rather than decoupling, we find that species with more adult wing coloration tend to have weaker larval immune defenses. Thus, trade-offs over melanin synthesis span metamorphosis and guide the evolution of these traits over long timescales. This finding reveals that some genetic linkages across life stages may be robust to adaptive decoupling. 132

INTRODUCTION

Organisms often face contrasting selective pressures as they progress through their life cycles (Schluter et al. 1991; Monro & Marshall 2014; Moore & Martin 2018). When the development of phenotypes in one life stage is associated with the development of phenotypes in others, antagonistic selection between the stages can impede or even preclude adaptive evolution in one stage or both (Mojica & Kelly 2010; Marshall et al.

2016). One way that this potential constraint can be resolved is through the erosion of genetic and developmental linkages between traits in different life stages—a hypothesis known as adaptive decoupling (Moran 1994). In principle, adaptive decoupling could benefit any organism that encounters varying selective pressures across its life cycle.

Seemingly in support of this evolutionary resolution, many animals dramatically remodel their body plans during transitions between life stages and habitats (e.g. marine invertebrates, Marshall & Morgan 2011; holo- and hemi-metabolous insects, Yang 2001; amphibians, Wollenberg Valero et al. 2017). Because these whole-body transformations are thought to confer developmental independence to each life stage (Moran 1994), animals with “complex life cycles” (sensu Wilbur 1980) are well suited for testing if apparent decoupling can limit the effects of conflicting selective pressures between life stages.

Despite the morphological re-arrangement undertaken during metamorphosis in many animals, it remains unclear whether or not their life stages can actually evolve independently (Marshall & Morgan 2011; Collet & Fellous 2019). For instance, genetic correlations frequently span these life-history transitions (Fellous & Lazzaro 2011;

Aguirre et al. 2014), even for morphological characters (Phillips 1998; Watkins 2001). 133

This indicates that selection in one stage can affect the evolution of traits in others irrespective of metamorphosis. Conversely, because genetic variances and co-variances are often transient through space and time (Arnold et al. 2008; Wood & Brodie 2015), correlations across life stages may not persist over timescales that are long enough to guide macro-evolutionary patterns. Comparative studies of adaptive decoupling across closely related species have indeed revealed independent evolution in each life stage (e.g.

Smith et al. 1995; Sherratt et al. 2017; Wollenberg Valero et al. 2017; but see Kolker et al. 2019). However, most previous studies have focused on the evolution of integrated morphological characteristics, such as overall body size and shape (but see the re-analysis of Mitchell et al. 2013 in Freda et al. 2017). As the evolution of these characters may not be representative of all phenotypes (e.g. Leinonen et al. 2008; Hansen et al. 2011), comparative studies with other classes of traits are necessary to further illuminate the power and limits of adaptive decoupling.

The melanin-synthesis pathway in insects is a strong candidate for tests of adaptive decoupling. Insects produce melanin for many functions in both the juvenile and adult stages, including 1) dark coloration used in thermoregulation and sexual signaling (Watt

1968; Moore et al. 2019); 2) encapsulation of parasites and pathogens (Schmid-Hempel

2005); and 3) wound healing and the structural integrity of the cuticle (Siva-Jothy et al.

2005). These melanin-based traits share an underlying genetic basis and evolve in concert within life stages (e.g. Armitage & Siva-Jothy 2005; Cotter et al. 2008; Fedorka et al.

2013). However, because any linkages between melanin-based traits across metamorphosis may constrain these diverse traits from adapting to the highly divergent demands imposed in each life stage, decoupling should be strongly favored. Despite this 134 pressure, the shared metabolic precursors and small number of loci involved in melanogenesis may limit the capacity for decoupling (González-Santoyo & Córdoba-

Aguilar 2012; Signor et al. 2016). Indeed, proximate trade-offs over melanin resources are known to span metamorphosis in insects. For instance, experimentally inducing the allocation of melanin resources to juvenile traits, such as cuticle darkness and immune defense, often inhibits the production of melanin-based traits in the adult stage (e.g.

Debecker et al. 2015; Moore et al. 2018a). Nevertheless, the extent to which these proximate trade-offs across life stages translate into evolutionary trade-offs across populations or species remains unknown.

Here, we tested for adaptive decoupling of melanin-based traits by comparing larval melanin immune responses among dragonflies with divergent adult melanin wing coloration. Across species, differences in sexual selection should favor the diversification of wing coloration (Svensson & Waller 2013), while similar parasite-mediated selection should favor more robust immune defenses (Forbes & Robb 2008; Seppälä 2015). Given the contrasting demands between life stages, and the ancientness of the Odonata lineage

(Letsch et al. 2016; Fig. 6.1), melanin-based traits in dragonflies are good candidates for adaptive decoupling. If the production and allocation of melanin for traits in different life stages has been adaptively decoupled, there should be no relationship between larval immune defense and the extent of adult wing coloration across species. In contrast, if these traits exhibit an evolutionary trade-off, then species with greater wing coloration should tend to possess weaker larval immune defenses.

135

METHODS

Species, Collection, Maintenance

In October 2015 and 2016, we collected larvae of six dragonfly species (superfamily:

Libelluloidea) that are known to differ in adult wing coloration in our study region (Fig.

6.1). In the area of North America where this study was conducted, all six species inhabit the same microhabitats within the same ponds (Wissinger 1988b; MPM, pers. obs.). For this study, we collected larvae from across three ponds at our field site to reduce the impact on any one (Squire Valleevue Farm, Hunting Valley, OH, USA: 41.4929 N,

81.4234 W). These species are univoltine in this region, with eggs hatching in late summer. To minimize any potential variation due to differences in age, we used individuals that were in either their fourth-to-last or third-to-last larval instar (Wissinger

1992). Consequently, sample sizes depended on the availability of these developmental stages in each year.

To equilibrate any large differences in condition or nutritional resources among the field-caught individuals, we maintained the larvae in a laboratory at Case Western

Reserve University for three weeks under a 13:11 L:D photoperiod. In 2015, we kept each individual in a plastic bin (34.6 cm L x 21 cm W x 12.4 cm H) with 1 L of dechlorinated tap water. In 2016, we kept each individual in a 473 mL plastic cup to accommodate a larger number of larvae (2015: n = 43; 2016: n = 164). In both years, we fed each larva a 1.5 mL concentrated aliquot of Daphnia magna (mean ± SD: 15.9 136 daphnids ± 3.8) every other day, on average. Two species (Fig. 6.1) were used in both

2015 and 2016 to ensure that any differences between years did not have species-specific effects. While the mean immune defense differed between 2015 and 2016 for these two species (F1,101 = 469.7, P < 0.001), the magnitude of the difference between the species

(F1,101 = 8.9, P = 0.003) did not vary across years (interaction: F1,101 = 0.7, P = 0.408).

This indicates that species responded similarly to the differences in rearing or other factors, and differences between years should not bias our results.

Immune Assay, Wing Coloration, and Analysis

After three weeks of maintenance in the lab, we inserted a roughened, sterilized knotted nylon monofilament (2.70 mm length, 0.18 mm diameter) into each individual between the fifth and sixth tergites of its abdomen (Rantala & Roff 2007; Duong & McCauley

2016; Moore et al. 2018a). We allowed the insect’s immune system to react to the implant for 24 hours, and then we dissected it out. Next, we took two photographs of each implant against a brightfield background with a dissecting microscope. To assess the strength of the melanin immune response, we calculated the difference between the mean grey value of the implant and the mean grey value of an unused, control implant in

ImageJ (Rasband 2012). Larger values indicate more melanin was deposited on the implant and, therefore, the larva produced a stronger melanin immune response.

Because we expected that quantitative differences in wing pigmentation could be associated with variation in immune defense among species, rather than just the rank order of total pigmentation, we estimated the proportion of each species’ wing area that is melanized. For most species (see below), we used ImageJ to quantify this value from 137 digitized versions of plates in a field guide of odonates specific to Ohio, USA (Rosche et al. 2008). Similar methods are commonly employed in comparative studies such as ours

(see also Zeuss et al. 2014; Dale et al. 2015; Miles & Fuxjager 2018). While wing coloration values estimated from this local field guide should be viewed as approximate, we expect that they generally reflect both the rank-order of wing coloration and the magnitude of differences of these species in our study area. However, the proportion of melanized wing area of one species depicted in this local field guide diverged considerably from our own extensive records of the focal population (P. longipennis;

Moore & Martin 2016; Moore et al. 2019). Such discrepancies between field guides and local observations are common for P. longipennis because this species’ wing pigmentation varies to an unusual degree across its range (Moore et al. 2019). We therefore used the mean proportion of melanized wing area from our own records of this focal population, which were assessed in an analogous manner using digitized photographs of wild-caught males in ImageJ (see details in Moore & Martin 2016; Moore et al. 2019). Coloration for all other species in the local field guide was consistent with our personal observations of the focal assemblage.

We first compared melanin immune responses among species using a mixed-effects model with year as a random factor to account for year-to-year differences. Denominator degrees of freedom were estimated with the Kenward-Rogers approximation (Kenward &

Rogers 1997). To then assess if species with greater wing melanization tend to have weaker larval melanin immune function, we used an ordered heterogeneity test (Rice &

Gaines 1994). This post-hoc analysis integrates output from correlation coefficients and general linear models to test for directional trends in the ordering of a small number of 138 group means. For this directional post-hoc test, we computed phylogenetically independent contrasts (Felsenstein 1985) for each species’ mean immune response and wing melanization using a time-calibrated odonate phylogeny (Letsch et al. 2016). While specific pairwise comparisons among species are not a focus of this study, we report this information in Table S6.1. If larval immune responses trade off with adult wing melanization across species, then species with more wing pigmentation should tend to have lower mean immune responses.

RESULTS

We tested for adaptive decoupling between adult wing melanization and larval melanin immune responses across dragonfly species. Larval melanin immune responses differed significantly among species (F5,200.1 = 5.49, P < 0.001), and there was an overall directional trend to these interspecific differences. Larval immune responses were weakest in the species with the most adult wing coloration (OH = -0.784, P = 0.007; Fig.

S6.1). Ordered differences remain even if phylogeny is not accounted for (OH = -0.559,

P = 0.024; Fig. 6.2). This trend was also robust to supplemental analyses that explicitly compared individuals’ immune responses within each year (Appendix 6). Supplemental analyses further indicate that this pattern is not due to species-level differences in energetic reserves (Appendix 6).

DISCUSSION

The dramatic metamorphosis that many organisms undergo as they progress through ontogeny is thought to evolve, in part, as an adaptive strategy to erode trait associations 139 between life stages and enable independent evolution of each (Moran 1994). Here, we tested the potential for evolutionary independence between two traits that exhibit trade- offs over a shared metabolic pathway: larval melanin immune defense and adult melanin wing coloration (Moore et al. 2018a). Without adaptive decoupling, the contrasting demands imposed on these characters in each life stage should prevent either trait from reaching its stage-specific fitness optimum. For instance, differences in sexual selection promote diversification in wing coloration (Svensson & Waller 2013), yet similar exposure to parasites and pathogens favors more robust immune defenses (Forbes &

Robb 2008; Seppälä 2015). However, rather than adaptive decoupling of these two traits, we found evidence of an evolutionary trade-off across life stages (sensu Stearns 1992).

Species with greater adult wing coloration tended to have weaker larval immune responses. Given the timescale over which these lineages diverged (Letsch et al. 2016), our results indicate that some linkages across life stages are quite robust to erosion by natural selection.

Previous work in odonates has documented inducible trade-offs across life stages over the allocation of melanin resources (Debecker et al. 2015; Moore et al. 2018a), but the evolutionary impact of these proximate trade-offs across life stages is poorly understood. When allocation trade-offs cannot be decoupled, theory predicts two evolutionary outcomes. First, selection may promote improved resource acquisition and more efficient synthesis of metabolic products (van Nordwijk & de Jong 1986). This circumvents the negative effects of the trade-off by increasing the resources available for allocation to all traits. As a result, evolutionary increases in one trait are associated with increases in the other (Reznick et al. 2000). Second, the proximate trade-offs over how 140 the resources are allocated can be recapitulated as evolutionary trade-offs (Stearns 1992;

Roff & Fairbairn 2007). Here, the pool of resources for allocation cannot be adaptively increased, and selection for increases in one trait indirectly leads to decreases in the other

(e.g. Gomez-Mestre & Buccholz 2006). Our study is consistent with this second outcome.

As natural selection acting on the many traits that rely on the melanin-synthesis pathway should improve the efficiency of the pathway overall (see also Craig & Foote 2001), it is plausible that neither the acquisition nor synthesis of precursors can be made more efficient in this very old lineage. While the continued dissection of the molecular genetics of melanin synthesis and allocation remains necessary in odonates (Chauhan et al. 2014), our results suggest that proximate trade-offs over the dietary and metabolic precursors of melanin are accommodated into evolutionary trade-offs among species.

Although larval immune defense trades off with adult wing coloration across species, it remains to be determined the extent to which this depresses larval performance for any particular species. For instance, because these six focal species share aquatic microhabitats, weaker larval immunity should be opposed to a similar extent by parasite- mediated selection (Seppälä 2015). Given the exceptionally high mortality of dragonfly larvae (Wissinger 1988b), one possible explanation for this pattern is that the reproductive value is much greater in adults than larvae (Fisher 1930). In this way, viability selection in the juvenile stage could be relatively inconsequential, and even severe costs to larval survival may not effectively drive phenotypic evolution when compared to modest differences in sexual selection (van Tienderen 2000). Weaker larval immunity may also arise if there are broad regions of relatively high fitness on the adaptive landscape. Here, deviations from the optimum should incur only small fitness 141 costs (Hendry & Gonzalez 2008; Haller & Hendry 2014). As the disadvantage the weaker immune defenses observed here may therefore be trivial, wing coloration could evolve towards its sexually selected optimum with relatively little opposition. Measuring phenotypic selection and reproductive value across the life cycle will help distinguish between these possibilities, and may illuminate the fitness consequences of trade-offs between life stages in odonates more generally (Stoks & Córdoba-Aguilar 2012).

Beyond optimizing traits for the unique selective pressures within each life stage, maximizing fitness across the life cycle entails adapting to the many contrasting demands encountered across ontogeny (Schluter et al. 1991; Moran 1994; Marshall & Morgan

2011). Natural selection is expected to favor developmental independence of different life stages as a result (Moran 1994). Previous studies of animals with complex life cycles have supported this prediction, finding that juvenile and adult morphological features evolve independently (e.g. Smith et al. 1995; Sherratt et al. 2017; Wollenberg Valero et al. 2017; but see Kolker et al. 2019). Our findings offer two potential caveats to the generalizability of these earlier studies. First, some types of characters may have greater capacity for decoupling than others. Most previous macro-evolutionary studies have tested for adaptive decoupling in very complex, integrative phenotypes, such as body shape (e.g. Smith et al. 1995; Sherratt et al. 2017). These characters are likely controlled by numerous developmental processes and underlying loci, many of which are redundant and can be substituted for one another without dramatically altering the phenotypic outcome (Nowak et al. 1997; Kitano 2004; Barghi et al. 2019). Such redundancy could provide an avenue to decoupling between stages. In contrast, the potential for decoupling of traits linked to a single metabolic pathway, like melanin synthesis in arthropods, may 142 be limited by shared reliance on metabolic precursors and a small number of underlying loci (Armitage & Siva Jothy 2005; Cotter et al. 2008; Lindstedt et al. 2016; Signor et al.

2016). Second, the severity of the metamorphosis may be associated with the developmental independence of the life stages that it separates. For instance, hemi- metabolous insects, such as odonates, undergo an “incomplete” metamorphosis, which could permit more robust links across life stages than are found in animals with more dramatic metamorphoses, such as holo-metabolous insects and marine invertebrates

(Yang 2001; Marshall & Morgan 2011). Phenotypic links across metamorphosis are indeed common in odonates (Stoks & Córdoba-Aguilar 2012), and the adult reproductive structures of some damselflies even appear to co-evolve with larval ecotypes in certain cases (Stoks et al. 2005). If decoupling is weaker in lineages with less drastic metamorphoses, then, for example, there should be greater evolutionary independence of life stages in holo-metabolous insects and frogs than in hemi-metabolous insects and salamanders, respectively (Yang 2001; Wollenberg Valero et al. 2017).

Overall, our results demonstrate evolutionary trade-offs between larval and adult traits across dragonflies. While this rejects the adaptive decoupling hypothesis as commonly stated (Moran 1994), we also did not find a perfectly negative relationship between these traits across species. This suggests that there is some capacity for independent evolution despite the overall decreasing trend we observed. Likewise, in studies that report micro- or macro-evolutionary independence between the adult and juvenile stages, correlations are often modest but not zero (e.g. Watkins 2001; Fellous & Lazzaro 2011; Freda et al.

2017; Wollenburg Valero et al. 2017). Thus, even in putative cases of adaptive decoupling, the stages are not totally evolutionarily independent. It may therefore be 143 more fruitful to consider adaptive decoupling in terms of extent rather than presence (see also Collet & Fellous 2019). In doing so, we may better illuminate the power of developmental and genetic constraints across the life cycle as forces shaping patterns of phenotypic diversity. 144

Species Wing Melanization Year(s) N

LibellulaLibellula pulchella pulchella 28% 2016 34

ErythemisErythemis simplicicollis simplicicollis 0% 2016 48

LeucorrhiniaLeucorrhinia intacta intacta 1% 2015 10

PachydiplaxPachydiplax longipennis longipennis 49% 2015, 2016 61

TrameaTramea lacerata lacerata 16% 2015 10

EpithecaEpitheca cynosura cynosura 4% 2015, 2016 44

150 100 50 0 Millions of Years Ago

Fig. 6.1. Time-calibrated phylogeny of the six Libelluloidea dragonfly species used in this study (derived from full phylogeny reported in Letsch et al. 2016), wing coloration, and sample sizes for each year.

145

60 Larval Melanin Immune Response Larval Melanin Immune

−6 −4 −2 0 Proportion Wing Melanization (logit−transformed)

Fig. 6.2. Trade-off across dragonflies between melanin immune defense and proportion of melanized wing area. Each square represents a species, with error bars reflecting the estimated standard errors from the linear mixed-effects model reported in the main text.

Colors correspond to the tip labels and text in Fig. 6.1. The linear regression line is shown only to help visualize the overall decreasing trend.

146

CHAPTER 7: CONCLUSIONS AND FUTURE DIRECTIONS

In this final chapter, I briefly summarize some of the conclusions that can be drawn from my findings. In particular, I focus on three major themes: 1) the adaptive significance of wing coloration in Pachydiplax longipennis; 2) eco-physiological causes of sexually selected color variation; and 3) eco-physiological consequences of sexual selection on color variation.

What is the adaptive significance of wing coloration in Pachydiplax longipennis?

While wing pigmentation had previously been reported in male P. longipennis (Rosche et al. 2008; Paulson 2009, 2012), what function it had, if any, was unclear. My research indicates that male wing color is favored by intrasexual selection. Males with more wing coloration are able to acquire territories over more days (Chapter 3) and defend them for longer each day (Chapter 2). As females primarily mate with males that possess territories (Sherman 1983), this advantage should translate into a substantial fitness benefit. Wing coloration appears to provide this advantage, in part, by intimidating rivals.

Males with more wing coloration receive less aggression from other territorial males

(Chapter 2), likely enabling them to devote less energy fighting off weaker rivals.

In order for wing coloration to evolve this role in communication, it must be beneficial for a male’s territorial rivals to adjust their behavior in response to this signal (Maynard

Smith & Harper 2003). My work provides some insight into this (Chapters 2, 3). As an individual’s wing coloration is correlated with its fighting ability (Chapter 2), males that are less aggressive to more colorful rivals will: 1) expend less energy engaging in 147 unwinnable fights, and 2) incur less physical damage even in contests that they ultimately could win. Thus, natural selection should favor males who assess their rivals’ wing color patterns and avoid those rivals that they are unlikely to beat (Maynard Smith & Harper

2003). In this way, wing coloration should evolve a function in communication among territorial males.

Of course, if males have evolved to avoid challenging rivals with more wing coloration because it signals greater fighting ability, it would be beneficial for males with poor fighting ability to cheat and produce large amounts of wing pigmentation. What then maintains the honesty of this signal? Because disputes among rivals over reproductive territories are very energetically costly in this and other territorial odonates (Fried & May

1983; Corbet 1999), any display trait whose development is highly sensitive to an individual’s physiological condition could then reliably signal a male’s fighting ability to his rivals (Hill 2011). Because of its high energetic cost (González-Santoyo & Córdoba-

Aguilar 2012) and its use in a diverse suite of crucial traits (Wittkopp & Beldade 2009), the melanin-synthesis pathway could be readily co-opted for this purpose. In this way, only those males in the best physiological condition would be able to produce high levels of wing coloration and satisfy all of the other demands on the melanin-synthesis pathway that are required for basic organismal functioning (Chapters 2, 4, 5). Moreover, even if sexual selection could drive disproportionate allocation of melanin resources to wing coloration over other traits (Chapter 6; see also Cotter et al. 2008), the total energetic costs required to produce any amount of melanin should maintain some minimum level of signal honesty. Such mechanisms of signal reliability have been termed the “Shared

Pathway Mechanism” (Hill 2011). 148

While the research presented here indicates that wing coloration signals condition and fighting ability during male-male contests over breeding territories, it is not yet clear if this trait functions in communication with females. It is plausible that it does not. For instance, males compete intensely for the best territories where females oviposit and offspring develop (Sherman 1983; Suhonen et al. 2008). As a result, females likely get considerable direct and indirect fitness benefits by simply mating with one of the small number of males that possess territory, irrespective of their wing color phenotypes

(Andersson 1994; Córdoba-Aguilar 2002). Being a choosy female may also have direct fitness costs because con- and hetero-specific males are so aggressive when defending territories (Sherman 1983). For these reasons, wing coloration does not appear to confer benefits to males via female choice in some other territorial odonates (e.g. Grether

1996b). However, given the proximate link between wing coloration and immune defense

(Chapter 2, 5), females could gain also indirect fitness benefits by mating with males with more wing coloration (Rantala et al. 2000; Siva-Jothy 2000; Lawniczak et al. 2007).

While the heritability of wing coloration and immune defense in odonates is currently unknown, genetic variation in melanin-based traits is common in other insects (e.g. Ellers

& Boggs 2002; Rolff et al. 2005; Lindstedt et al. 2016), and indirect fitness benefits to females are therefore possible. Nevertheless, unless the strength and direction of any female choice on male wing coloration counteracts the benefits via male-male competition, the total force of sexual selection should still, on average, favor male greater wing coloration.

Beyond signaling, my work reveals that the thermal properties of wing coloration also provide territorial benefits to males under cool conditions (Chapter 3). For instance, I 149 found that color-induced heating of 1-2 °C enhances flight performance by ~4-8% when males are cool. This heating advantage also appears to improve territorial acquisition on the most thermally variable days in the field. Given the infrequency with which females come to the ponds to mate (Johnson 1962; Sherman 1983), simply possessing and holding a breeding territory over as many days as possible provides a large fitness advantage in the event a female does arrive (Sherman 1983). Although long-term studies that identify parentage and recruitment into the next generation remain necessary in this and most other odonates (Lowe et al. 2009; Thompson et al. 2011), color-induced heating should substantially increase male fitness in cool portions of the species’ range.

Given the thermal benefits conferred by wing coloration in the coolest regions of the species’ range, it is intriguing that females have clear wings. This indicates that even the thermal advantages of wing coloration are sex-specific. A couple of non-mutually exclusive explanations for this seem possible. First, it could be that females have very different thermal requirements than males. For instance, while males exude a highly reflective blue prunescence over their bodies (Robey 1975), both the thorax and abdomen are brown and black in females (Paulson 2009, 2012). Females probably then have an easier time heating and staying warm, as light absorbed on the body is converted into body heat more efficiently than when it is absorbed on appendages (e.g. Kingsolver

1983). Any marginal thermal benefits of wing coloration for females may therefore be outweighed by energetic costs of producing it (see also Svensson & Waller 2013).

Second, it is possible that males cannot distinguish females with some wing coloration from rival males with little coloration. Such a scenario could lead to males attacking those females as if they were weak, rival males. However, as male body coloration is 150 visually striking, it would be surprising if males solely distinguished between rivals and females based on wing coloration. Nonetheless, if males attack those females as if they were territorial rivals, there could be substantial direct fitness costs to females that produce wing pigmentation (Chapman 2006; Takahashi et al. 2014; Le Rouzic et al.

2015).

Overall, wing coloration confers advantages through both sexual signaling and thermoregulation in P. longipennis. One interesting implication of these findings is that the same sexually selected trait provides multiple advantages simultaneously (see also

Maan & Cummings 2008). Under a fairly broad range of scenarios, the co-opting of sexually selected traits for ecological and other reproductive functions may be an important way that populations evolve to higher absolute fitness (Bonduriansky 2011).

While evidence for this process is mixed for odonate wing coloration (Svensson &

Waller 2013), the multiple functions of wing pigmentation observed here make it an intriguing possibility for future research. P. longipennis may be particularly well suited for investigating such patterns because of the phenotypic variation that it exhibits across the species’ range (Chapter 3). When combined with phylogeographic information of how the species spread across North America after the last glacial maximum, this geographic variation in wing coloration could help us uncover how the unique functions of a sexually selected character are gained or lost throughout a trait’s evolutionary history.

Eco-physiological causes of variation in sexually selected traits 151

The environment in which sexually selected traits are developed and displayed encompasses a complex suite of interacting factors (Cornwallis & Uller 2010; Miller &

Svensson 2014). While many of these factors could shape variation in sexually selected characters, most previous research has focused either on the costs imposed on the already expressed phenotype by species’ enemies (e.g. parasites, predators, interspecific competitors) or on the benefits of greater exaggeration in habitats that disrupt signal transmission (Maan & Seehausen 2011). To expand our understanding of the ecological causes of variation in sexually selected traits, my dissertation explored several ways in which sexually selected trait variation could be generated and maintained by interactions between the environment and an organism’s physiological state.

One crucial aspect of an organism’s physiological ecology is its thermal physiology

(Huey & Kingsolver 1989; Angilletta 2009). As dark coloration can heat up individuals by absorbing more incident light (Clussella Trullas et al. 2007; Stuart-Fox et al. 2017), the costs, benefits, and diversification of sexually selected coloration may be highly sensitive to variation in an organism’s thermal physiology. Indeed, Chapter 3 shows that thermal physiology can generate geographic variation in this sexually selected character by determining its costs and benefits across the landscape (Moore et al. 2019). Although this represents the most comprehensive test of the effects of temperature on sexual selection to date (cf. West & Packer 2002; Punzalan et al. 2008c; Outomuro & Ocharan

2011; Svensson & Waller 2013), many questions about the evolutionary causes of this geographic variation remain to be answered. Did males gain wing coloration in the coolest regions, did they lose it in the warmest, or was it some combination of the two patterns? What factors account for the apparent mismatches between ambient air 152 temperature and wing coloration (e.g. the Pacific coast)? Given the apparent signaling benefits of wing coloration, to what extent do ambient temperature differences prevent the southward spread of color alleles? While there is considerable opportunity for further investigation in this system, Chapter 3 provides a compelling, and relatively complete, illustration of how variation in thermal physiology can underlie the diversification of a sexually selected color pattern.

An organism’s energetic state or “condition” is another important component of its overall physiological state (Hill 2011), and it has received by far the most attention with respect to the role of eco-physiological processes in sexually selected trait variation.

Differences among habitats in average individual condition have even been implicated as one way that sexually selected trait variation can be governed by interactions between the external environment and an organism’s physiological state (Maan & Seehausen 2011).

However, features of an individual’s condition can modify how an animal interacts with its environment in ways beyond the amount of resources available for allocation towards other traits (Gosler et al. 1995; Zamora-Camacho et al. 2014a). Natural selection may then directly target the resource accumulation and storage used to develop secondary sexual traits for other reasons (see also Cox & Calsebeek 2015). In such cases, spatial variation in ecological selection can shape sexually selected trait variation by altering the costs and benefits of accumulating resources to develop these characters. Chapter 4 specifically documents how an individual’s larval body condition both improves its subsequent wing color development and reduces its likelihood of surviving to adulthood in habitats with predators (Moore & Martin 2018). Sexually selected wing coloration could therefore diverge across the landscape as a consequence of how resource 153 accumulation and storage adapts to local differences in larval predation regime. Juvenile survival costs of developing adult traits, such as this, represent important constraints on the efficiency of sexual selection because they eliminate reproductively beneficial phenotypes from the population before reproduction can even begin (Hadfield 2008;

Mojica & Kelly 2010). Thus, if juvenile body condition often increases vulnerability to predators directly, as it did here, such interactions between an individual’s physiological state and the environment in which it develops could be an important cause of sexually selected trait variation.

These findings highlight just a couple of ways that eco-physiological processes can shape variation in sexually selected traits. However, there are likely to be many other pathways through which eco-physiological processes promote variation in these characters. For instance, circulating hormone levels and oxidative stress are both sensitive to an individual’s experience of its environment and important to the development of its sexually selected traits (e.g. Folstad & Karter 1992; von Schantz et al.

1999). Examining interactions between other eco-physiological processes and sexually selected traits is therefore likely to broaden our understanding of the ways in which the environment can cause variation in sexually selected traits. Moreover, given the independent roles of secondary sexual traits and physiological adaptation to reproductive isolation (Coyne & Orr 2004), exploring the broader suite of eco-physiological processes involved in generating variation in these phenotypes has the potential to blur the distinction between ecological and reproductive causes of speciation (Servedio &

Boughmann 2017). Thus, future research in this area may expand our understanding of macro-evolutionary diversification as a whole. 154

Eco-physiological consequences of sexual selection

A second important question about sexual selection concerns its consequences for ecological adaptation (Servedio & Boughmann 2017; Giery & Layman 2019; Svensson

2019). Recent research has found that changes in an individual’s sexually selected traits can feed back to affect its physiology (reviewed in Rubenstein & Hauber 2008; Vitousek et al. 2014). Thus, one way that sexual selection may affect ecological adaptation is by indirectly shaping how an organism’s physiological state evolves and subsequently interacts with the prevailing environmental conditions (Giery & Layman 2019). By focusing on the links between sexually selected wing coloration and immune defense, my research provides some insight into one way that sexual selection may have consequences for eco-physiological adaptation.

Evolving a more robust immune defense should provide ecological benefits via elevated population-mean survival in the presence of more abundant or more virulent parasites and pathogens (Sheldon & Verhulst 1996). However, if mounting stronger immune responses inhibits the development of sexually selected characters, sexual selection will limit the overall fitness benefits of a robust immune defense. The adaptive evolution of immune function will then be slower than if these trade-offs were not present.

In addition to uncovering diverse ecological costs of mounting strong immune responses,

Chapter 5 indicates that there are direct trade-offs between the production of immune defenses and wing coloration over melanin resources rather than solely over total energetic reserves (Moore et al. 2018a). Selection will therefore be unable to completely 155 alleviate the detrimental consequences of this trade-off by simply increasing overall resource acquisition (van Nordwijk & de Jong 1986; Stearns 1992; Kotiaho 2001).

Unless melanin synthesis can become more efficient or specific melanin precursors can be acquired at a higher rate, strong sexual selection for more wing coloration could impede adaptation of immune defense.

Trade-offs between sexually selected traits and physiological traits may do more than slow the rate of ecological adaptation (Lande 1980; Bonduriansky 2011). For instance, when sexual selection is much stronger than parasite-mediated selection acting through variation in survival and fecundity, proximate trade-offs between these characters could allow sexual selection to drive maladaptation of immune defense. Alternatively, when these characters trade off, sexual selection may promote increases in the accumulation of the underlying pool of resources (van Nordwijk & de Jong 1986; Reznick et al. 2000) or in the efficiency with which these resources are used (see also Craig & Foote 2001).

Rather than hindering ecological adaptation, this second pathway enhances both ecological and reproductive fitness (Lorch et al. 2003; Servedio & Boughmann 2017).

Chapter 6 indicates that proximate trade-offs over the allocation of melanin resources translate into evolutionary trade-offs across species, whereby species with greater wing coloration tend to have weaker immune responses. As a result, strong sexual selection for more wing coloration appears to have the potential to indirectly cause the evolution of weaker of immune defenses. While the performance consequences of species-level differences in immune function remain to be quantified, one eco-physiological consequence of sexual selection on wing coloration is the potential maladaptation of immune defense. 156

Overall, my research indicates that sexual selection can have consequences for eco- physiological traits that shape ecological adaptation. However, immune function is just one of many axes along which physiology adapts to different ecological contexts. Other linkages between sexually selected characters and vital components of physiology likely exist in P. longipennis, and certainly exist in other species (Rubenstein & Hauber 2008;

Vitousek et al. 2014). Exciting insight about the consequences of sexual selection for ecological adaptation may therefore be gained by examining how the evolution of sexually selected traits modifies interactions between other physiological processes and the external environment. Additionally, given the role that some of these same physiological components play in shaping population dynamics and community assembly

(Duckworth et al. 2015; Start et al. 2018), physiological variation generated by sexual selection could underlie many eco-evolutionary processes as well (Giery & Layman

2019; Svensson 2019).

Coda

When organisms adjust or adapt to the external environment, they typically change along more than one phenotypic dimension. Interactions among these phenotypic dimensions may therefore be vital to understanding adaptation and diversification. Ultimately, the research presented in this dissertation highlights how the interplay between sexually selected coloration, the external environment, and physiology can generate and maintain phenotypic diversification in dragonflies and, probably, other organisms as well.

However, interactions between sexually selected traits and other phenotypic dimensions are unlikely to be confined to physiology, and studying these relationships will deepen 157 our understanding of how sexual selection affects organisms as an integrated unit.

Moreover, through the exploration of these research topics, this work underscores how dragonflies can be used to investigate general ecological and evolutionary processes. In particular, due to its broad geographic range, remarkable phenotypic variation, and large body size relative to its life-cycle duration, P. longipennis appears to have considerable potential as a future model system for many questions in evolutionary, physiological, and behavioral ecology. Future research with dragonflies into these themes seems likely to broaden our understanding of the interactions between ecology and reproduction in the origins and maintenance of diversity.

158

APPENDIX FOR CHAPTER 2

Table S2.1. All variance-standardized selection gradients for intrasexual and viability selection (Lande & Arnold 1983).

Trait Territorial Tenure Longevity Body Size -0.003 ± 0.094 0.012 ± 0.133 Relative Wing Size -0.035 ± 0.101 0.096 ± 0.125 Wing Melanization 0.326 ± 0.104 0.133 ± 0.125

159

a) ● ●● 0.6 ● ● ● ● ● ● 0.5 ● ●● ● ●● ● ●● ●●●●●●● ●● ●● ●● ● ●●●●●●●●●●●●●●●● ● ● 0.4 ●● ●●● ● ●● ●●●● ● ●● ●● ● ●●●●● ● ● ●● ●●●●● ●● ● ●● ●● 0.3 ● ● ● ●● ● Wing melanization ●● ● ● ● ● ●●●●● ● 0.2 ● ● ● 32.5 35.0 37.5 40.0 42.5 45.0

Body size (mm) b) ● ● ● 0.6 ● ● ● ● ● ● 0.5 ● ●●● ●●● ● ● ●●● ●●●●●●●●● ●● ● ●●●●●●●●●●●●●●●● ● ●● ● 0.4 ● ●●●●●●●●●●●● ● ●●● ●● ●● ●●● ●●●● ●●●●● ● ● ●● ●● 0.3 ● ●●●● ● Wing melanization ●●● ●● ● ● ● ●● ● ● 0.2 ● ● ● −2 0 2 4

Relative Wing Size

Figure S2.1. There was no relationship between a male’s wing melanization and its body size (a) or its relative wing size (b). Each point represents an individual. 160

Figure S2.2. Rear wings of territorial P. longipennis showing examples of weak natural melanization (top), experimentally manipulated melanization (middle), and strong natural melanization (bottom).

161

APPENDIX FOR CHAPTER 3

Measuring wing coloration and its association with other male traits

Scoring wing coloration

Males can produce dark brown coloration across their entire wing area, with the most dense pigmentation occurring in a distal region centered between the nodus and the pterostigma as well as in a region at the base of the wing. To score the extent of each male’s wing coloration, we took digital photographs of its wings against a standard white background in a dark black box. From these photographs, we used ImageJ (Rasband

2012) to identify the highest mean grey value (0 - 255, 0 = most opaque; 255 = most transparent) of the pigmented portion of each wing (i.e. least darkly pigmented). To more easily and objectively distinguish the pigmented regions from non-pigmented regions, we converted the photograph to black and white using this range of values as the threshold for black, thereby making pigmented areas display as black and non-pigmented regions display as white. We then outlined the perimeter of the black regions, calculated the size of the black areas within that perimeter (mm2), and estimated wing coloration as the proportion or percent of the total wing area that is pigmented (see also Hanley et al.

2013). Earlier work by our lab group showed that this measurement is highly repeatable and that there is no correlation between total wing area and the proportion of the wing area with pigment (Moore & Martin 2016). While previous studies have considered only variation in the distal region, because it varies most among males (Moore & Martin 2016,

2018; Moore et al. 2018a), the coloration in the basal region likely also has important thermal properties and is included in all measurements in this study. Additionally, within our focal population in northeast OH, USA, the basal and distal regions of coloration are 162

substantially correlated (r = 0.668, t93 = 8.7, P < 0.001), suggesting that any selective forces on one will have evolutionary consequences for both.

To further understand the potential thermal properties of wing coloration, we estimated the difference in potential light absorption between males with naturally high and naturally low wing coloration using an ILT 950 Spectroradiometer (International

Light Technologies®). We estimated how light could be absorbed by naturally high coloration males and naturally low coloration males by calculating the product of the potentially absorptive surface area (total mm2 of pigmented wing area) and the absorptivity of the brown pigment for each group (eq. 1a).

! ! Eq. 1a. !"#$% !! !"#$%&'('")& !"# !"#! !"#"$ !"#$% ∗ !"#$%&' !"#$%&'()('* – [!"#$% !! !"#$%&'('")& !"# !"# !"#"$ !"#$% ∗ !"#$%&' !"#$%&'()('* ] [!"#$% !!! !"#$%&'('")& !"# !"# !"#"$ !"#$% ∗ !"#$%&' !"#$%&'()('*]

Spectrophotometric measurements showed that nearly all light is transmitted through the clear portions of the wings. The light absorption of the brown pigment is ~0.372 on the distal portion of the wing and ~0.471 on the basal portion of the wing. Males with naturally high levels of wing coloration have 371.7 mm2 and 62.1 mm2 of distal wing coloration and basal wing coloration, respectively. Males with naturally low levels of wing coloration have 147.3 mm2 and 41.9 mm2 of distal and basal wing coloration, respectively. Plugging these values into eq. 1a, we get (eq. 1b):

!"#.! !!! ∗ !.!"# ! !".! !!! ∗ !.!"# – [ !"#.! !!! ∗ !.!"# ! !".! !!! ∗ !.!"# ] Eq. 1b. !"#.! !!! ∗ !.!"# ! !".! !!! ∗ !.!"#

163

Thus, males with naturally high levels of coloration are able to absorb ~125% more light via their wings than males with naturally low levels of coloration.

Relationship with thorax darkness

Male P. longipennis produce a waxy blue prunescence on their bodies. Males with more of this waxy prunescence look brighter to the human eye and reflect more UV light

(Robey 1975). We captured and sacrificed 26 territorial males from our focal population in northeast Ohio, USA. We removed each male’s wings from its body, and then took separate digital photographs of the male’s dorsal side of its thorax and wings against a standard white background in a dark black box. We scored the mean grey value of the dorsal area of each male’s thorax in ImageJ (0 = black, 255 = white; Rasband 2012). We also scored the mean grey value of rectangles along the length of both sides of the male’s body to control for any small variation in the lighting condition of the standard white background. We then scored each male’s thorax darkness by subtracting the mean grey value of the male’s thorax from the mean grey value of the white background—males with darker thoraxes have higher scores. Wing coloration was scored as described above.

Because the thorax and the abdomen heat independently in dragonflies (May 1976), we did not consider abdomen darkness here. A Pearson’s product-moment showed no significant correlation between thorax darkness and wing coloration (r = -0.269, t25 = -

1.40, P = 0.175).

164

Relationship with body mass

Among those males included in the heating comparisons of natural variation in wing coloration, we measured their body masses on an electronic microbalance to the nearest

0.0001 g. We then used a Pearson’s product-moment correlation to assess the relationship between each male’s body mass and the extent of its wing coloration. These two traits were not correlated (r = -0.092, t22 = -0.43, P = 0.669). Similarly, there was no significant difference in mass between males with naturally high coloration and naturally low coloration (F1,22 = 0.579, P = 0.455).

Testing for the effects of body size on maximum lifting force

Lifting force is known to vary with body size in odonates (Samejima & Tsubaki 2010), and we therefore also tried to account for it directly in non-linear models. However, we encountered substantial convergence issues when attempting to fit non-linear models that incorporated both acclimation temperature and body size as explanatory variables of lifting performance. We therefore considered any potential effects of body size in several indirect ways. To measure male size, we measured rear wing area from digitized photographs in ImageJ for the males used in the study. We did not measure rear wing area of males that we attempted to acclimate to 49 °C, which all died and were discarded.

To first test if acclimation temperature was confounded with male size, we compared the rear wing area of males acclimated to the six temperatures using ANOVA. Male size did not differ significantly among acclimation temperatures (F5,55 = 0.44, P = 0.820), indicating that size does not exclusively explain any patterns in performance. We next explored how important male size is to lifting performance using a linear regression 165 without accounting for acclimation temperature. There was no significant relationship (R2

= 0.044, F1,58 = 2.67, P = 0.108). Finally, we tested if male size explained variation in lifting performance after accounting for acclimation temperature. In this test, we regressed the residual lifting performance (from the exponentially modified Gaussian model, i.e. after accounting for effects of acclimation temperature) on male size. There

2 was a modest relationship between residual lifting performance and size (R = 0.090, F1,58

= 5.72, P = 0.020). Collectively, these findings suggest that: 1) size is not especially important to performance across the males used in this study and 2) the effects of size are only apparent after already controlling for acclimation temperature.

Assessing potential thermal costs of wing coloration in the American Southwest

For illustrative purposes, we estimated how frequently high levels of wing coloration would cause males to suffer severe performance declines in the hottest parts of the species’ range. For a region of the American southwest where males typically lack coloration (Arizona USA, 32.838 N, 112.792 W), we used the CRU TS3.10 dataset

(Harris et al. 2013) to calculate how many months averaged maximum daily temperatures of at least 41.5 °C. We chose 41.5 °C because, at this environmental temperature, a 1 –

2 °C increase would cause males to reach or surpass the upper limit of their performance breadth based on the thermal performance curve we estimated. While thermal performance curves evolve across environmental gradients (Sinclair et al. 2012), upper limits are known to diversify less than lower limits (Hoffmann et al. 2013). Thus, for illustrative purposes like ours, the upper limit of the performance breadth estimated for a population in northeast Ohio probably provides a reasonable approximation for the upper 166 limit of the performance breadth across the species’ range. Nonetheless, until thermal performance curves are characterized in other parts of this species’ range, we urge caution when incorporating these values into climate models or projections where greater precision is necessary. Over the last 1308 months (109 years), maximum daily temperatures have averaged at least 41.5 °C in 158. Thus, a color-mediated heating of 1 –

2 °C would cause males to reach or surpass the upper limit of their performance breadth on most days for one to two months every year.

167

Table S3.1. AIC comparison of candidate negative exponential and logistic functions that were fit to the heating responses of male P. longipennis. Non-linear models were fit to the thoracic heating of the males used in the comparison between males with naturally high versus low coloration, and included random effects of individual and date to account for the repeated measures and any non-independence of heating responses measured on the same day.

Rank Function (df) AIC Δ 1 Asymptotic (10) 698.71 0 2 Gompertz (10) 699.10 0.39 3 Logistic (10) 749.86 51.15 4 Michaelis-Menten (6) 1770.22 1071.50 168

Table S3.2. AIC comparison of five non-linear models of the relationship between acclimation temperature and maximum lifting force as suggested in Angilletta (2006).

Models were fit either including males that were acclimated to 49 °C (a) or excluding those males (b).

Model Rank Model AIC Δ

1 Exponentially modified Gaussian -137.06 0

2 Weibull -133.45 3.61

a) 3 Quadratic -133.08 3.98

4 Modified Gaussian -131.88 5.18

5 Gaussian -128.37 8.69

1 Exponentially modified Gaussian -104.30 0

2 Weibull -101.99 2.31

b) 3 Quadratic -101.28 3.02

4 Gaussian -101.27 3.03

5 Modified Gaussian -99.24 5.06

169

Table S3.3. Parameters and change in relative flight performance from an exponentially modified Gaussian model of the relationship between maximum lifting force and acclimation temperature that was fit without males acclimated to 49 °C.

% Change in Relative % Change in Relative Temperature Parameter Performance from Performance from ( °C) 1 °C Increase 2 °C Increase

Lowest acclimation temperature tested 25 + 2.7% + 5.4%

Lower bound of performance breadth 33.2 + 3.1% + 7.2%

Thermal Optimum 39.2 - 1.6% - 7.2%

Upper bound of performance breadth 42.4 - 14.6% - 30.8%

170

Table S3.4. Summary statistics and temporal trends in environmental variables across the

2015 flight season of P. longipennis. F tests had 1 and 16 degrees of freedom.

Temporal Trend Environmental variable Overall Mean + SD F P (Parameter estimate ± SE) Daily Mean Temperature (°C) 23.082 ± 1.344 0.122 ± 0.055 4.875 0.042 Daily SD Temperature (°C) 1.337 ± 0.140 0.007 ± 0.006 1.162 0.297 Daily Mean Solar Radiation (W * m-2) 720.995 ± 66.032 -1.300 ± 3.045 0.182 0.675 Daily SD Solar Radiation (W * m-2) 223.282 ± 65.394 -0.419 ± 3.091 0.018 0.894

171

Table S3.5. Parameter estimates for the relationship between environmental variables and the probability of a male having produced at least some coloration across the entire wing from a geographically nested mixed-effects logistic regression (see main text) using temperature data from WorldClim1.4 and WorldClim2.

Term Parameter estimate ± SE χ2 P Temperature -0.502 ± 0.103 21.4 < 0.001 WorldClim1.4 Solar radiation -0.089 ± 0.069 1.8 0.184 PPT in Driest Qt 0.273 ± 0.101 6.5 0.011 Temperature -0.526 ± 0.104 23.6 < 0.001 WorldClim2 Solar radiation -0.086 ± 0.070 1.6 0.206 PPT in Driest Qt -0.269 ± 0.101 6.3 0.012

172

Table S3.6. Parameter estimates for the relationship between environmental variables and the probability of a mature male having produced at least some coloration across the entire wing from a spatially explicit logistic regression fit with the function ‘ggwr’ in the

‘spgwr’ package using temperature data from WorldClim1.4 and WorldClim2. Global mean is the mean parameter estimate.

1st 3rd Global Term Min Median Max Quantile Quantile Mean Temperature -0.858 -0.529 -0.413 -0.226 0.226 -0.377 WorldClim1.4 Solar radiation -2.019 -1.277 -0.149 -0.030 0.481 -0.065 PPT in Driest Qt -1.514 -0.189 0.037 0.297 1.271 0.474 Temperature -0.780 -0.525 -0.457 -0.289 0.213 -0.397 WorldClim2 Solar radiation -2.073 -1.460 -0.145 -0.037 0.438 -0.067 PPT in Driest Qt -1.722 -0.192 0.032 0.245 1.301 0.462

173

Table S3.7. The effects of the environment on the relationship between a male’s traits and its territorial acquisition and territorial defense within a day. If temperature and/or solar radiation modify the territorial benefits of male traits, we should observe a significant interaction between those traits and the environmental variable. Test statistics for models of territory acquisition are χ2 from likelihood ratio tests of models with versus without the effect. Because of the significant interaction between wing coloration and SD temperature, we obeyed the law of marginality and did not test main effects of either variable. Test statistics for models of the duration of territorial defense are F-tests using the Kenward-Roger denominator degrees of freedom approximation.

Aspect of Term Test Statistic DF P Territory Success body size 0.03 1 0.859 mean temperature < 0.01 1 0.972 mean solar radiation 2.08 1 0.149 SD solar radiation 1.13 1 0.288 wing coloration x mean temperature 0.05 1 0.831 Territory wing coloration x SD temperature 7.57 1 0.006 Acquisition wing coloration x mean solar radiation 2.58 1 0.108

wing coloration x SD solar radiation 1.89 1 0.169 body size x mean temperature 0.21 1 0.648 body size x SD temperature 0.09 1 0.759 body size x mean solar radiation 0.45 1 0.501 body size x SD solar radiation <0.01 1 0.973 wing coloration 10.4 1, 26.4 0.003 body size 0.02 1, 34.3 0.888 mean temperature < 0.01 1, 12.4 0.975 SD temperature 0.10 1, 12.3 0.759 mean solar radiation 0.02 1, 9.6 0.884 SD solar radiation 0.16 1, 14.1 0.692 Duration of wing coloration x mean temperature 0.29 1, 106.3 0.591 Territory Defense wing coloration x SD temperature 0.42 1, 109.0 0.521 wing coloration x mean solar radiation 0.53 1, 110.5 0.469 wing coloration x SD solar radiation 0.19 1, 101.5 0.666 body size x mean temperature 0.52 1, 97.5 0.474 body size x SD temperature 0.93 1, 105.5 0.338 body size x mean solar radiation 2.64 1, 106.1 0.107 body size x SD solar radiation 0.63 1, 109.6 0.430 174

APPENDIX FOR CHAPTER 4

Validation of OLS residuals as an estimate of body condition

In this study, we use the residuals from an ordinary least-squares (OLS) loge-loge regression of body mass on head size (a measure of body size in odonates; Benke &

Benke 1975; Wissinger 1988a) as our estimate of body condition (Jakob et al. 1996).

While no consensus has emerged about the best body condition metric to use in aquatic insects, body condition indices are generally intended to provide an estimate of an individual’s pool of accumulated resources that 1) is independent from its structural body size and 2) reflects variation in an individual’s physiological status (Jakob et al. 1996;

Schulte-Hostedde et al. 2005; Peig & Green 2009, 2010).

An important criterion of any body condition estimate is to be independent of structural body size, such that two individuals of different sizes could have the same body condition. The body condition estimate using OLS residuals (hereafter, “residual condition” for brevity) is both linear and independent of head size for our dataset, satisfying this first criterion. Head size and mass are linearly associated following loge

2 transformation (linear regression: 3.02 ± 0.05, F1,476 = 3295.6, P < 0.001, R = 0.8738).

Residual condition is also independent of head size (r = -8.125 x 10-11, P > 0.999). The scaled mass index has also been proposed as a suitable and perhaps preferable body condition estimate (Peig & Green 2009, 2010). However, in our dataset, this condition estimate is significantly correlated with head size (r = -0.180, P < 0.001), making it unsuitable as an estimate of potential energetic resources that is independent of variation in structural body size. Others have suggested simply including both mass and structural size in any statistical models. However, given the very strong correlation between head 175 size and body mass in our dataset (r = 0.935, P < 0.001), any statistical models that include both terms as fixed effects will provide unreliable standard errors around parameter estimates and significance tests with extremely low power due to the high multicollinearity (Mitchell-Olds & Shaw 1987; Graham 2003; Cox & Calsbeek 2015).

Consequently, with respect to its potential to reflect energetic resources independently of variation in structural body size in our dataset, the residual condition metric is preferable to the several other recently supported indices “body condition” for our purposes.

Another goal of any estimate of body condition is to reflect variation in an individual’s physiological status, typically stemming from variation in an individual’s total pool of accumulated resources that can be used for allocation towards energetically costly functions. While many previous studies have shown that residual condition reflects variation in physiological status across a wide taxonomic array of animals (e.g. Schulte-

Hostedde et al. 2005; Birkhead et al. 2006; Kaufman et al. 2007; Mills et al. 2008;

Gilbert et al. 2016), other studies find no such relationship (e.g. Kelly et al. 2014).

Consequently, we provide several lines of evidence that support residual condition’s association with physiological status in odonates.

If residual condition is associated with an individual’s total pool of accumulated resources, then individuals that are fed more should have greater residual condition. To test this first prediction, we collected Pachydiplax longipennis larvae in early Oct 2016 from the same source pond as for the experiments described in the main text. We reared larvae individually in our laboratory at Case Western Reserve University (Cleveland, OH

USA) in 473 mL plastic cups under a 13:11 LD photoperiod. We randomly assigned larvae to either a high food or low food treatment, and them fed larvae a 1.5 mL aliquot 176 of concentrated Daphnia magna (mean ± SD: 15.9 daphnids ± 3.8) either 6 times (high food) or 2 (low food) times per week. After 21 days, we massed individuals with an electronic balance (as described in methods of main text), and used ImageJ (Rasband

2012) to measure their head widths from digitized photographs taken with dissecting microscope against a brightfield background. We assessed how residual condition differed between the food treatments using ANOVA. Larvae fed more frequently had higher body condition after 21 days (0.064 ± 0.025, F1,42 = 6.72, P = 0.013, n = 44, Fig.

S4.1a). This strongly indicates that residual condition is associated with accumulated resources, but it could be that those resources were allocated toward muscle development and are not necessarily available for use in growth, development, and physiological maintenance as if they were stored as fat reserves.

If residual condition is associated with accumulated resources that are available for future allocation and is not simply a reflection of greater muscle density, then larvae with greater residual condition should also perform better at energetically costly functions. We show this to be the case in several ways.

First, melanin-based immune responses are known to be energetically costly in insects

(reviewed in González-Santoyo & Córdoba-Aguilar 2012). If residual condition reflects the accumulation of resources that are available for future allocation towards growth, development, and physiological maintenance, then larvae with higher residual condition should be able to mount larger melanin immune responses. Using the same 44 larvae as described above, we considered the relationship between larval body condition and melanin immune defense. We assayed melanin immune responses by inserting a sterilized, roughened nylon monofilament (2.70 mm length, 0.18 mm diameter) into the 177 body cavity. After 24 hours, we dissected out the monofilament, photographed it, and measured the difference in mean grey value between this monofilament and a piece of control monofilament that had not been inserted into a dragonfly larva using ImageJ.

Larger values indicate that more melanin was deposited on the monofilament, which corresponds to greater encapsulation, and thus that the individual had a stronger immune response (Rantala & Roff 2007; Moore & Martin 2016). We used multiple regression to estimate how melanin immune defenses vary with larval body condition. We also included the food treatment to account for any additional effects of this treatment that acted in addition to the food treatment’s effect on residual condition itself. We found that larvae with greater residual condition deposited more melanin on the monofilament

(75.529 ± 30.755, F1,40 = 5.90, P = 0.020, n = 44, Fig. S4.1b). We also found that, after accounting for these effects of residual condition on immune defense, larvae in the low food treatment had marginally lower immune defenses (low vs high: 9.946 ± 5.268, F1,40

= 3.49, P = 0.069), but there was no interaction between condition and food treatment

(F1,40 = 0.09, P = 0.766). While it is unclear what effects the food treatment has on melanin immune defense that are independent of the variation explained by residual condition, this finding overall strongly supports the prediction of a relationship between residual condition and the energetically costly melanin immune defense. If residual condition reflected only accumulated resources that were allocated to muscle development, it is unclear why such an association would exist in these larvae.

Second, the process of emergence in odonates is very energetically costly and requires substantial resource accumulation before it can proceed (Corbet 1999). If residual condition reflects the accumulation of resources available for future allocation, then 178 larvae with initially higher residual condition should be able to emerge sooner after overwintering than larvae with initially lower residual condition (because they would have accumulated fewer resources that were available for allocation towards growth, maintenance and reproduction). In another study of Pachydiplax longipennis, we showed this pattern: larvae with higher initial residual condition were more likely to emerge within 20 days of the experiment and emerged earlier within those 20 days than larvae with lower initial residual condition (Moore et al. 2018b). If residual condition reflected only accumulated resources that were allocated to muscle development, it is unclear why we would have observed such a pattern.

Third, the resources accumulated to be used for future allocation towards growth, development, and physiological maintenance, should primarily be stored as fat in the abdomen of larval odonates. If residual condition reflects the accumulation of resources for use in these functions, then residual condition should be associated with an individual’s fat stores. While we could not assess this relationship for the individuals in our experiment, we assessed it using two available datasets from experiments in other odonates where fat stores were directly measured.

In Gilmore (2014), the author examined how the combined effects of predation risk and the herbicide atrazine affected physiological status of the larval dragonfly Ladonna deplanata. This libellulid dragonfly has a similar body shape and larval ecology to P. longipennis, making it a very reasonable approximation. As part of this study, the author measured fat stores by extracting lipids with a 24-hour chloroform reflux using a Soxhlet apparatus. The author also measured head width and body mass, thereby allowing us to calculate residual condition and directly examine the relationship between residual 179 condition and fat weight. According to Gilmore (2014), it was discovered after the experiment had ended that an accident in the lab had caused the balance to read consistently higher values for data collected during a particular sampling point (Gilmore

2014). The balance was fixed after that sampling point, and the author was able to identify the problematic data points. To be conservative, we dealt with this data set in two ways. First, we used a linear regression to examine how the weight of an individual’s fat stores vary with residual condition and head width while excluding these problematic data points. Second, we used a regression to examine how fat weight varies with residual condition and head width across all of the individuals, while including a categorical covariate for these problematic data points. In the model with only the unproblematic data points, we found a highly significant relationship between residual condition and fat

2 weight (β ± SE = 0.0072 ± 0.0019, F1,50 = 13.79, P < 0.001, model R = 0.25, n = 53, Fig.

S4.2a). In the model where we included all of the points and accounted for them with a categorical factor, we observed that the relationship between residual condition and fat

2 stores remained significantly positive overall (F1,81 = 10.57, P = 0.002, model R = 0.56, n = 86). Additionally, this relationship was also significantly positive even within just the

2 problematic data points (β ± SE = 0.0238 ± 0.0095, F1,30 = 6.26, P = 0.018, model R =

0.24, n = 33, Fig. S4.2b). Together, this demonstrates that the residual condition metric that we used in our focal species, P. longipennis, is strongly associated with fat weight in a larval odonate that has both a very similar body shape and ecology.

In Tüzün et al. (2017), the researchers explored how habitat fragmentation altered patterns of sexual selection in the damselfly, Coenagrion puella. They extracted lipids by dissolving the fats with dichloromethane, and subtracting the individual’s mass after the 180 lipid extraction from the individual’s mass before the lipid extraction. Given the use of energetic stores in the drastic development of flight muscles following emergence in all odonates, and the general difference in body shape between damselflies and dragonflies, we expected the relationship between residual condition and fat weight in adult damselflies to be much more modest that that found in larval dragonflies. However, we felt that if this relationship were also present in an adult damselfly in spite of all of these considerable differences, it would provide stronger evidence for the validity of residual condition more generally for odonates. We calculated residual condition from the loge – loge regression of mass on structural body size (measured here with wing size, see also

Corbet 1999). In the dataset available online, the authors provide a measure that they call

“relative fat content”, whereby they divided total fat by total mass. To then calculate total fat weight, we multiplied mass by relative fat content. Here, we again used linear regression to examine how total fat weight varied with residual condition and structural body size. We found that residual condition was positively associated with fat weight

2 (2.464 ± 0418, F1,567 = 34.79, P < 0.001, model R = 0.13, n = 570; Figure S4.2c).

Taken together, these results collectively indicate that residual condition reflects an individual’s accumulated resources that can be used for future allocation towards growth, development or physiological maintenance.

181

Comparing within-pool selection gradients

While our analysis that characterizes the strength and direction of selection across all of the pools within each treatment is adequately designed for estimating both linear and non-linear selection coefficients (Brodie et al. 1995; Chenoweth et al. 2012), it could also be sensitive to a small number of highly influential points in one or a few pools. An alternative method for comparing the strength and direction among treatments is to estimate the selection coefficients within each pool, and then use those estimates of selection within each pool as the response variable and the treatments as explanatory variables (e.g. Svensson & Sinervo 2004; Calsbeek & Smith 2007; Martin & Pfennig

2012; Benkman & Mezquida 2015). This method should mitigate the collective influence of any small number of influential points because their influence would either 1) be divided across the separate estimation of selection within several independent pools or 2) all be concentrated in just one out of eight pools for a given treatment. However, an important drawback of this approach in our case is that each individual selection coefficient is calculated using a smaller sample size (30 larvae per pool) than is typically recommended for directional selection coefficients, and much smaller than required to estimate correlational selection coefficients with any precision (i.e. >100, Brodie et al.

1995; Kingsolver et al. 2001). Our ability to compare correlational selection gradients using this method is therefore substantially limited. Nevertheless, to ensure that our conclusions about directional selection arising from the selection analyses presented in the main text were not unduly influenced by any small number of influential points, we validated our across-pool selection analysis with a within-pool selection analysis. 182

We quantified directional selection on body condition for the 30 individuals within each pool using the standard Janzen & Stern (1998) technique described in the main text.

This produced a selection coefficient on body condition and associated standard error for each pool, as opposed to a single selection coefficient across all of the pools like was described in the main text of the manuscript. As with the analysis described in the main text, we did not include the tank where the Anax kept emerging and/or dying, and we therefore calculated selection coefficients on body condition for 15 pools (8 Anax-absent,

7 Anax-present).

We compared the average within-pool selection coefficient between the two treatments in two ways. First, we used a meta-analytic model in the R package (‘metafor’,

Viechtbauer 2010) with predator treatment as the explanatory variable and the within- pool selection coefficients as the response. This method also allowed us to directly incorporate the standard errors of the selection coefficients into the analysis such that estimates with smaller standard errors are more influential to the parameter estimation.

Because our within-pool selection coefficients were calculated on a smaller number of individuals than has been recommended for robust estimation of selection (~40, Brodie et al. 1995), our standard errors for these within-pool selection coefficients should be quite large, and this should be a conservative analysis for the estimation of the average strength and direction of selection for both treatments. However, due to the relatively small sample sizes within each pool, the experiment was not well designed to specifically compare within-pool selection coefficients in this way, and this analysis could be overly conservative for exploring the importance of any small number of potentially influential points. We therefore also conducted a less conservative test that did not account for the 183 standard errors of the within-pool selection coefficients using a standard ANOVA model.

This second analysis should provide the most generous test of the notion that patterns of selection on larval body condition in the Anax-absent pools are consistent with patterns in

Anax-present pools, and should therefore provide a conservative test of our overall conclusion of differences in directional selection on body condition between treatments.

In both the conservative and less conservative analyses, directional selection on larval body condition in the Anax-present pools was significantly negative (metafor: mean β ±

SE = -0.221 ± 0.085, z = -2.59, P = 0.010; lm: mean β ± SE = -0.219 ± 0.079, t = -2.76,

P = 0.016). In contrast, directional selection on larval body condition in the Anax-absent pools was more than 50% weaker than in the Anax-present pools, and was not significantly different from zero in either analysis (metafor: mean β ± SE = -0.100 ±

0.077, z = -1.30, P = 0.194; lm: mean β ± SE = -0.091 ± 0.074, t = -1.22, P = 0.243; Fig

S4.3).

If a small number of influential points were the solely responsible for the differences in directional selection observed between the treatments in the across-pool approach reported in the main text, then, when their influence is diffused across the independent estimation of selection within different pools or concentrated in one out of eight pools, we should have found broadly similar patterns of directional selection on body condition between the treatments in this within-pool approach. As we do not find this, differences in directional selection on body condition between the treatments do not appear to be solely due to a small number of highly influential points. 184

Table S4.1. Numbers of relatively large and small ultimate and penultimate larvae used for the high- and low-size variation treatments in each block. Depending on the available ultimate and penultimate larvae, relatively large and small larvae of each instar were allocated to tanks in ratios of 2:1, 1:2 for high-size variation or 1:2, 2:1 for low-size variation.

Number of Larvae

Size Variation Ultimate Penultimate Block Treatment Large Small Large Small High 14 7 3 6 Block 1 Low 7 14 6 3 High 10 5 5 10 Block 2 Low 5 10 10 5 High 12 6 4 8 Block 3 Low 6 12 8 4 High 14 7 3 6 Block 4 Low 7 14 6 3

185

Table S4.2. Viability selection gradients on head size (loge transformed) and body condition in pools with and without Anax junius calculated from generalized additive mixed-effects models using methods described in Morrissey & Sakrejda (2013; ‘gsg’ package [Morrisey & Sakrejda, 2014]). There was a significant interaction among head

! size, body condition, and Anax presence in the full model (�! = 4.90, P = 0.027), indicating differences in selection between these treatments (Chenoweth et al., 2012).

Standard errors and significance tests were calculated using 1000 parametric bootstrap replicates.

Treatment Trait Selection Gradient ± SE P Head Size 0.351 ± 0.095 < 0.001 Anax Body Condition -0.200 ± 0.086 0.018 Present Head Size x Body Condition 0.103 ± 0.102 0.262 Head Size 0.214 ± 0.057 <0.001 Anax Body Condition -0.091 ± 0.060 0.114 Absent Head Size x Body Condition -0.123 ± 0.058 0.058

186

A) B) ● 0.05 ●

60 ● ● ● ●● ● ● ● ● ● ● ● ● 0.00 ● ● 40 ● ●●●● ● ● ● ● ● ● ● ● ● ● Immune Response Immune Residual Condition ● ●● 20 ● ● ● −0.05 ● ● ● ● ●

● 0 High Food Low Food −0.2 −0.1 0.0 0.1 Residual Condition

Figure S4.1. a) Residual condition was higher for P. longipennis larvae that were fed 6 times per week (high food) than for those larvae that were fed 3 times per week (low food). b) Melanin immune response increases with larval residual condition. Circles represent individual larvae.

187

A) B) C) 0.006 0.08 0.020

0.004 0.015 0.06

0.010 0.04 Fat Stores (g) Fat Stores (g) Fat Fat Stores (g) Fat 0.002

0.005 0.02

0.000 0.000 −0.4 0.0 0.4 0.8 −0.4 −0.2 0.0 0.2 0.4 −8 −4 0 4 Residual Condition Residual Condition Residual Condition

Figure S4.2. Fat stores increase with residual condition in two odonates. Circles represent individuals. Positive relationship between fat weight and residual condition in L. deplanata larvae is seen for both the properly calibrated (a) and poorly calibrated (b) points. c) Positive relationship between fat weight and residual condition in adult C. puella.

188

● 0.2 ●

● 0.0 ● ● ●

● β ● −0.2 ● ●

● ●

●

−0.4

●

No Anax Anax

Figure S4.3. Differences in within-pool directional selection coefficients on Pachydiplax longipennis body condition between pools with (n = 7) and without (n = 8) predatory

Anax junius larvae. Circles represent selection coefficients calculated within a pool.

Diamonds show the mean ± 95% confidence intervals from the analysis accounting for each selection coefficient’s associated standard error

189

1.00 D C B A Low: 1 2 2 1 High: 2 1 1 2 0.75

0.50 Density

0.25

0.00 3.5 4.0 4.5 5.0 5.5

Head Size (mm)

Figure S4.4. Example density plot of Block 2 individuals (n = 120). The dotted line separates ultimate (right) from penultimate (left) larvae based on head size. Relatively large and small ultimate and penultimate larvae are represented with the letters, and numbers indicate the within-instar ratios of larvae assigned to each treatment. See Table

S4.1 for exact numbers in each treatment in each block.

190

APPENDIX FOR CHAPTER 5

We investigated how leg removal affected vulnerability to aeshnid predators. We collected 136 P. longipennis larvae, and randomly assigned them to either a control treatment, where no leg was removed, or a treatment in which we removed one leg below the tibia. Among the larvae in the leg-removal treatment, we then randomly assigned each larva to one of six treatments, corresponding to which of its six legs was to be removed (Fig. S5.2). On either the day of (36 trials) or the day after leg removal (32 trials), we placed the larva, as well as an unmarked larva, into a plastic container (34.6 cm L x 21 cm W x 12.4 cm H) filled with 3 L of aged water. To provide additional structure for the larvae, we taped six lengths of bamboo (length) facing down from the lip of the container into the water. We added one ultimate instar Anax junius larva to the container one hour after we added the P. lonipennis larvae. We starved A. junius larvae for 48 hours prior to all trials. After 24 hours, we removed all surviving P. longipennis larvae, and scored which, if any, had been consumed.

We then tested for differences in survival between marked and unmarked larvae, and for differences in survival among the leg-removal treatments. We also tested for differences in survival between larvae that were marked either on the day of the trial or

! the day before the trial, but no effects were detected (day: �! = 1.69, P = 0.194; day x

! leg-removal interaction: �! = 2.54, P = 0.639), and we do not consider it further. We first analyzed survival differences between marked and unmarked larvae using a generalized linear mixed-effects model with each larva’s survival (0 = consumed, 1 = not consumed) as the response, and treatment as a fixed effect. We included the trial number as a random effect to control for the non-independence of the larvae in the same trial. We next used a 191 generalized linear model to test whether survival differed among larvae that had different legs removed. The survival of each larva (as above) was treated as the response variable and the leg removed (Fig. S5.2) was included as the explanatory variable. As trials were only conducted with one marked and one unmarked larva, two marked larvae were never in the same trial, and we therefore did not need to include trial as a random effect for this analysis. We tested the significance of each term using likelihood ratio tests of models with and without the term.

While we found that marked larvae had 17.6% lower survival than unmarked larvae

! (odds ratio ± SE: 2.11 ± 0.75, �! = 4.45, P = 0.035), there were no significant differences

! in survival among larvae with different legs removed (�! = 2.84, P = 0.724; Table S5.9).

These results did not change if we analyzed only the 44 trials in which at least one larva

! was consumed (marked vs unmarked �! = 7.02, P = 0.008; among leg-removal

! treatments: �! = 8.57, P = 0.128). Based on these results, we chose to mark larvae by removing legs 1, 2, or 5 (Fig. S5.2).

192

Table S5.1. Tukey post hoc tests for survival differences among the immune treatments in pools with (a) or without aeshnids (b), and survival differences for each immune treatment between pools with and without aeshnids (c). Contrasts between levels are given by the odds ratio of survival ± SE.

a) with aeshnids Odds Ratio SE t P Sham - 12HR 2.021 0.808 1.759 0.184 Sham - 24HR 3.173 1.353 2.707 0.019 12HR - 24HR 1.570 0.691 1.025 0.561 b) without aeshnids Sham - 12HR 1.269 0.621 0.542 0.851 Sham - 24HR 0.666 0.302 -0.898 0.642 12HR - 24HR 0.514 0.239 -1.432 0.324 c) aeshnids - no aeshnids Sham 2.214 0.948 1.856 0.063 12HR 1.420 0.645 0.772 0.440 24HR 0.465 0.210 -1.698 0.090

193

Table S5.2. Tukey post hoc tests for differences in the probability of emergence among the immune treatments. Contrasts are given by the odds ratio of emergence ± SE.

Contrast Odds ratio SE t P Sham – 12HR 7.428 6.249 2.384 0.045 Sham – 24HR 8.380 7.057 2.525 0.031 12HR – 24HR 1.128 0.670 0.203 0.9775

194

Table S5.3. Tukey post hoc tests for differences in emergence day among the immune treatments. Contrasts between levels are given as the mean difference in days ± SE.

Contrast Estimate SE t P Sham – 12HR 0.512 1.389 0.367 0.928 Sham – 24HR -3.429 1.161 2.954 0.013 12HR – 24HR -2.919 1.488 -1.962 0.130

195

Table S5.4. Variation in size at emergence depended on an interaction between sex and immune treatment, but not on main effects of predator treatment, sex, emergence day, or any interactions. F-tests are derived from a linear mixed-effects model with the Kenward-

Roger denominator degrees of freedom approximation

Effect F df P Immune Treatment 2.32 2, 56.4 0.108 Predator Treatment 0.78 1, 8.6 0.400 Sex 0.05 1, 57.6 0.828 Emergence Day 0.91 1, 57.0 0.344 Immune x Predator 0.05 2, 55.5 0.947 Sex x Immune 4.91 2, 55.1 0.011 Sex x Predator 0.03 1, 57.8 0.860

196

Table S5.5. Tukey post hoc tests for sex-specific effects of immune treatment on adult size

a) Females Estimate SE t P Sham – 12HR -14.583 7.660 -1.904 0.147 Sham – 24HR -2.488 6.745 -0.368 0.928 12HR – 24HR 12.099 8.956 1.351 0.373 b) Males Sham – 12HR 9.718 6.114 1.589 0.258 Sham – 24HR 20.477 6.360 3.219 0.006 12HR – 24HR 10.759 6.262 1.718 0.207 c) Females - Males Sham -16.724 5.795 -2.886 0.006 12HR 7.576 8.129 0.932 0.355 24HR 6.237 6.774 0.921 0.361

197

Table S5.6. Variation in mass at emergence did not depend on immune treatment, predator, treatment, sex, or any interactions after accounting for variation in body size. F- tests are derived from a linear mixed-effects model with the Kenward-Roger denominator degrees of freedom approximation

Effect F df P Body size 6.57 1, 56.9 0.013 Immune Treatment 0.33 2, 56.0 0.719 Predator Treatment 2.52 1, 8.5 0.149 Sex 1.82 1, 52.5 0.183 Emergence Day 1.87 1, 55.9 0.177 Immune x Predator 0.44 2, 54.5 0.649 Sex x Immune 0.50 2, 54.1 0.612 Sex x Predator 3.60 1, 56.5 0.063

198

Table S5.7. The probability of a male having produced wing coloration on the third day after emergence did not depend on immune treatment, predator treatment, or their

2 interaction. χ tests are derived from likelihood ratio tests of models with versus without the effect.

Effect χ2 df P Immune Treatment 1.13 2 0.568 Predator Treatment 3.11 1 0.078 Immune x Predator 0.47 2 0.792

199

Table S5.8. Tukey post hoc tests for differences in the extent of wing coloration three days after emergence among the immune treatments. Contrasts between levels are given as mean difference in wing coloration ± SE, and are on the arcsine square-root scale.

Contrast Estimate SE t P Sham – 12HR 0.068 0.042 1.606 0.271 Sham – 24HR 0.137 0.039 3.467 0.008 12HR – 24HR 0.069 0.043 1.590 0.279

200

Table S5.9. Probability of survival (± 95% confidence intervals) with Anax junius for larvae that had one of their legs removed. Leg removal categories correspond to Figure

S5.2. Probabilities and confidence limits are those estimated from the model using all of the larvae described in the text.

Leg Removed N Probability of Survival 95% Confidence Intervals 1 12 0.500 0.244 - 0.756 2 12 0.667 0.376 - 0.869 3 12 0.333 0.131 - 0.624 4 11 0.545 0.268 - 0.797 5 11 0.545 0.268 - 0.797 6 10 0.500 0.225 - 0.775

201

40 Melanin Deposition

6 12 18 24 Immune−challenge duration (hours)

Figure S5.1. Melanin deposition on knotted implants increased significantly with the duration of immune challenge. Points represent individual larvae, and line is the fitted line from the mixed-effects model. While analyses were performed on the loge scale, all values are backtransformed here to improve visual clarity.

202

5 mm 1 6 2

4 3

Figure S5.2. Pachydiplax longipennis larva. The numbers correspond to the leg that we removed in the marking experiment (see also Table S5.9).

203

APPENDIX FOR CHAPTER 6

Testing within-year immune differences among species

We conducted a supplemental analysis to ensure that our results were not greatly affected by our choice to account for between-year variation using a mixed-effect model with year as a random effect. Specifically, we tested whether individuals from species with more wing coloration had weaker immune responses than those individuals who were sampled in the same year but were from species with less wing coloration.

To facilitate direct comparisons among individuals within a year, we z-transformed each individual’s immune response score relative to the other individuals that were sampled in that same year. We then used ANOVA to compare these relativized immune response scores among the six species. Because z-transforming within years ensures that an individual with the average immune response score for each year has the same score

(i.e. 0), we do not need to include year as a random effect in this model to account for average differences between years. Next, we tested for directional trends in the ordering of the species’ means using the same ordered heterogeneity test described in the main text.

We should find the same negative trend if individuals from species with more wing coloration had weaker immune responses than their counterparts from the same year that belonged to species with less wing coloration.

Species differed in these relativized immune response scores (F5,201 = 4.24, P = 0.001;

Table S6.1). Moreover, individuals from species with more wing coloration tended to have the weakest immune response scores within the year they were sampled

(phylogenetically controlled: OH = -0.789, P = 0.006; not phylogenetically controlled: 204

OH = -0.651, P = 0.013). Our results are therefore robust across two methods of accounting for year-to-year differences.

Testing for the effects of body condition

Within Pachydiplax longipennis, it was previously shown that an individual’s immune defense is positively associated its energetic resources (i.e. “body condition”, Chapter 4;

Moore & Martin 2018). If this pattern scales up to affect species-level variation, then it is possible that the differences in immune defense among species could be due to species- level differences in energetic reserves.

To test this proposition, we first estimated each individual’s available energetic reserves by quantifying a proxy for this characteristic: its residual mass from a loge-loge regression of mass on head size. Previous intraspecific work indicates that the residuals from a loge-loge regression of mass on head size are a strong estimate of variation in available energetic reserves. For example, this metric increases with food quantity received over short time periods, and it is significantly correlated with fat stores (Chapter

4; Moore & Martin 2018). To ensure that the slope of the relationship between mass and head size did not differ among species, we first ran a loge-loge regression with an individual’s mass as the response variable; and its head size, species identity, and the interaction of those two variables as explanatory variables. Species differed in mass relative to head size (F5,196 = 106.6, P < 0.001), but there was no significant interaction

(F5,196 = 2.0, P = 0.079). A single slope therefore best defines the relationship between mass and head size across all species. Additionally, the significant main effect of species suggests that species had differing average levels of energetic reserves in this study. Thus, 205 to estimate each individual’s energetic reserves, we used the residuals from a separate univariate loge-loge regression of mass on head size that included all individuals across all species. We did not include an explanatory variable for species in this analysis because we explicitly wanted to test if the greater residual mass of some species relative to others translated into relatively stronger immune defenses. While other species-specific could contribute to greater residual mass (e.g. cuticle mass), we expect that residual mass should nonetheless capture most of the variation associated with differences in energetic stores among species.

To test if species-level differences in this proxy for energetic reserves explain differences in immune defense, we conducted a linear mixed-effects model with an individual’s immune response score as the response; its species identity, its residual mass, and their interaction as fixed effects; and year as a random effect. If differences in residual mass among species drive differences in immune defense, then, when we include both residual mass and species into analyses of immune responses, we should find a significant effect of residual mass but not of species. However, while species remained a significant predictor of immune response score in this analysis (F5,193.1 = 2.9, P = 0.015), we found no interaction between species and residual mass (F5,193.0 = 1.4, P = 0.217), nor a main effect of residual mass (F1,193.0 = 0.1, P = 0.802). Moreover, the ordered heterogeneity tests of species’ mean immune responses, after accounting for the non- significant covariate of residual mass, continued to show a negative relationship with wing coloration (phylogenetically informed: OH = -0.788, P = 0.007; phylogenetically uninformed: OH = -0.567, P = 0.024). This indicates that species-level differences in energetic reserves are not driving the pattern described in the main text. 206

Table S6.1. Post hoc pairwise comparisons among species in the mean immune response.

Mixed-effects model compared among species’ mean immune responses including year of sample as a random effect. The z-score model tests for species-level immune differences within each year by z-transforming each individual’s immune defense relative to other individuals in the same year. Pairwise comparisons are Tukey’s HSD tests.

Species abbreviations: EC – Epitheca cynosura; ES – Erythemis simplicicollis; LI –

Leuccorhina intacta; LP – Libellula pulchella; TL – Tramea lacerata; PL – Pachydiplax longipennis.

Mixed-effects model z-score model C ontrast estimate SE t P estimate SE t P EC - ES -5.964 3.895 -1.53 0.645 -0.220 0.200 -1.10 0.883 EC - LI -1.362 7.498 -0.18 1.000 -0.240 0.336 -0.71 0.980 EC - LP -3.548 4.253 -0.83 0.961 -0.031 0.219 -0.14 1.000 EC - PL 9.762 3.701 2.64 0.093 0.444 0.190 2.34 0.183 EC - TL 21.097 7.498 2.81 0.059 0.825 0.336 2.45 0.144 ES - LI 4.602 7.779 0.59 0.992 -0.021 0.334 -0.06 1.000 ES - LP 2.416 4.133 0.59 0.992 0.189 0.215 0.88 0.952 ES - PL 15.726 3.764 4.18 0.001 0.664 0.185 3.58 0.006 ES - TL 27.061 7.779 3.48 0.008 1.044 0.334 3.13 0.025 LI - LP -2.186 7.965 -0.27 1.000 0.209 0.346 0.61 0.991 LI - PL 11.124 7.049 1.58 0.614 0.685 0.328 2.09 0.297 LI - TL 22.459 8.246 2.72 0.075 1.065 0.430 2.48 0.135 LP - PL 13.310 4.133 3.22 0.018 0.475 0.206 2.31 0.194 LP - TL 24.645 7.965 3.09 0.027 0.855 0.346 2.48 0.136 PL - TL 11.336 7.049 1.61 0.594 0.380 0.328 1.16 0.855

207

0.3 ● ●

● −0.4

●

−1.1 Immune Response PIC − Immune

−1.8 ●

−0.4 −0.2 0.0 0.2 PIC − Wing color

Figure S6.1. Relationship between wing coloration and immune response after controlling for the effects of shared evolutionary history among the species. Rather than the species mean, each circle represents the phylogenetic independent contrast for larval immune response and adult wing coloration.

208

BIBLIOGRAPHY

Abrams, P.A. 2000. The evolution of predator-prey interactions: theory and evidence.

Annu. Rev. Ecol. Syst. 31: 79-105.

Adamo, S.A., Roberts, J.L., Easy, R.H., Ross, N.W. 2008. Competition between immune

function and lipid transport for the protein apolipophorin III leads to stress-induced

immunosuppression in crickets. J. Exp. Biol. 211: 531-538.

Aguirre, J.D., Blows, M.W., & Marshall, D.J. 2014. The genetic covariance between life

cycle stages separated by metamorphosis. Proc. R. Soc. B. 281: 20141091.

Allen, P.E. & Miller, C.W. 2014. Novel host plant leads to the loss of sexual dimorphism

in a sexually selected male weapon. Proc. R. Soc. B 284: 20171269.

Alonso-Alvarez, C., Bertrand, S., Devevey, G., Gaillard, M., Prose, J., Faivre, B., &

Sorci, G. 2004. An experimental test of the dose-dependent effect of carotenoids

and immune activation on sexual signals and antioxidant activity. Am. Nat. 164:

651-659.

Anderson, C.N., Córdoba-Aguilar, A., Drury, J.P., & Grether, G.F. 2011. An assessment

of marking techniques for odonates in the family Calopterygidae. Entomol. Exp.

Appl. 141: 258-261.

Anderson, C.N. & Grether, G.F. 2010. Interspecific aggression and character

displacement of competitor recognition in Hetaerina damselflies. Proc. R. Soc. B

277: 549-555.

Anderson, T.L. & Semlistch, R.D. 2016. Top predators and habitat complexity alter an

intraguild predation module in pond communities. J. Anim. Ecol. 85: 548-588.

Andersson, M. 1994. Sexual Selection. Princeton University Press, Princeton, NJ USA. 209

Angilletta, Jr. M.J. 2006. Estimating and comparing thermal performance curves. J.

Therm. Biol. 31: 541-545.

Angilletta, Jr. M.J. 2009. Thermal adaptation: a theoretical and empirical synthesis.

Oxford University Press, Oxford, UK.

Ardia, D.R., Gantz, J.E., Schneider, B.C., & Strebel, S. 2012. Costs of immunity in

insects: an induced immune response increases metabolic rate and decreases

antimicrobial activity. Funct. Ecol. 26: 732-739.

Armitage, S.A.O.& Siva-Jothy, M.T. 2005. Immune function responds to selection for

cuticular colour in Tenebrio molitor. Heredity 94: 650-656.

Armitage S.A.O., Thompson, J.J.W., Rolff, J., & Siva-Jothy, M.T. 2003. Examining costs

of induced and constitutive immune investment in Tenebrio molitor. J. Evol. Biol.

16: 1038-1044.

Arnegard, M.E., McIntyre, P.B., Harmon, L.J., Zelditch, M.L., Crampton, W.G.R., Davis

J.K., et al. (2010) Sexual signal evolution outpaces ecological divergence during

electric fish species radiation. Am. Nat. 176: 335-356.

Arnold, S.J., Bürger, R., Hohenlohe, P.A., Ajie, B.C., & Jones, A.G. 2008.

Understanding the evolution and stability of the G-matrix. Evolution 62: 2451-2461.

Arnold, S.J. & Wade, M.J. 1984. On the measurement of natural and sexual selection –

theory. Evolution 38: 709-719.

Arnott, G. & Elwood, R.W. 2009. Assessment of fighting ability in animal contests. Anim.

Behav. 77: 991-1004.

Badyaev, A.V. 2004. Developmental persepective on the evolution of sexual ornaments.

Evol. Ecol. Res. 6: 1-17. 210

Barghi, N., Tobler, R., Nolte, V., Jakšić A., Mallard, F., Otte, K.A. et al. 2019. Genetic

redundancy fuels polygenic adaptation in Drosophila. PLoS Biol 17: e3000128.

Bates, D., Machler, M., Bolker, B., & Walker, S. 2015. lme4: Linear mixed-effects

models using Eigen and S4. R package version 1.0-6.

Benard, M.F. 2004. Predator-induced phenotypic plasticity in organisms with complex

life histories. Annu. Rev. Ecol. Evol. Syst. 34: 651-673.

Benke AC, & Benke SS. 1975. Comparative dynamics and life histories of coexisting

dragonfly populations. Ecology 56: 302-317.

Birkhead, T.R., Pellatt, E.J., Matthews, I.M., Roddis, N.J., Hunter, F.M., McPhie, F.,

Castillo-Juarez, H. 2006. Genic capture and the genetic basis of sexually selected

traits in zebra finch. Evolution 60: 2389-2398.

Bivand, R. & Yu, D. 2017. ‘spgwr’. Geographically weighted regression. R package

version 0.6-32.

Böhner. J. & Antonić, O. 2009. Land-surface parameters specific to topo-climatology. In:

Developments in Soil Science, (T. Hengle & H.I. Reuter, eds.). pp. 195-226.

Elsevier.

Bonduriansky, R. 2011. Sexual selection and conflict as engines of ecological

diversification. Am. Nat. 178: 729-745.

Bonduriansky, R. & Chenoweth, S.F. 2009. Intralocus sexual conflict. Trends Ecol. Evol.

24: 280-288.

Boughman, J.W. 2001. Divergent sexual selection enhances reproductive isolation in

sticklebacks. Nature, 411: 944-947. 211

Brashears, J., Aiello, A., & Seymoure, B.M. 2016. Cool bands: wing bands decrease rate

of heating, but not equilibrium temperature in Anartia fatima. J. Therm. Biol. 56:

100-108.

Brodie, E.D. III, Moore, A.J., & Janzen, F.J. 1995. Visualizing and quantifying natural

selection. Trends Ecol. Evol. 10: 313-318.

Brodin, T.D., Mikolajewski, D.J., & Johansson, F. 2006. Behavioural and life history

effects of predator diet cues during ontogeny in damselfly larvae. Oecologia 148:

162-169.

Brooks, R. 2000. Negative genetic correlation between male sexual attractiveness and

survival. Nature 406: 67-70.

Calsbeek, R., & Goedert, D. 2017. Performance tradeoffs, ontogenetic conflict, and

multisport athletes: how is an ironman triathlete like a frog? Integr. Comp. Biol. 57:

207-216.

Calsbeek, R., Smith, T.B. 2007. Probing the adaptive landscape using experimental

islands: density-dependent natural selection on lizard body size. Evolution 61:1052-

1061.

Caruso, C.M., Martin, R.A., Sletvold, N., Morrissey, M.B., Wade, M.J., Augustine, K.E.

et al. 2017. What are the environmental determinants of phenotypic selection? A

meta-analysis of experimental studies. Am. Nat. 190: 363-376.

Catania, S.V.L., & McCauley, S.J. 2015. Evaluating the use of coded-wire tags in

individually marking Odonata larvae. Can. Entomol. 148: 371-374

Chapman, T. 2006. Evolutionary conflicts of interest between males and females. Curr.

Biol. 16: R744-R754. 212

Chauhan, P., Hansson, B., Kraaijeveld, K., de Knijff, P., Svensson, E.I., & Wellenreuther,

M. 2014. De novo transcriptome of Ischnura elegans provides insights into sensory

biology, colour and vision genes. BMC Genomics 15: 808.

Chenoweth, S.F., Hunt, J., & Rundle, H.D. 2012. Analyzing and comparing the geometry

of individual fitness surfaces. In: The adaptive landscape in evolutionary biology

(E.I. Svensson & R. Calsbeek eds), pp. 126-149. Oxford University Press. Oxford,

UK.

Chippindale, A.K., Gibson, J.R., & Rice, W.R. .2001. Negative genetic correlation for

adult fitness between sexes reveals ontogenetic conflict in Drosophila. Proc. Natl.

Acad. Sci. USA 98: 1671-1675.

Clusella Trullas, S., Van Whyk, J.H., & Spotila, J.R. 2007. Thermal melanism in

ectotherms. J. Therm. Biol. 32: 235-245.

Collet, J., & Fellous, S. 2019. Do traits separated by metamorphosis evolve

independently? Concepts and methods. Proc. R. Soc. B 286: 20190445.

Contreras-Garduño, J., Buzatto, B.A., Serrano-Meneses, M.A., Nájera-Cordero, K., &

Córdoba-Aguilar, A. 2008. The size of the red wing spot of the American rubyspot

as a heightened condition-dependent ornament. Behav. Ecol. 19: 724-732.

Contreras-Garduño, J., Canales-Lazcano, J., & Córdoba-Aguilar, A. 2006. Wing

pigmentation, immune ability, fat reserves and territorial status in males of the

rubyspot damselfly, Hetaerina americana. J. Ethol. 24: 165-173.

Contreras-Garduño, J., Córdoba-Aguilar, A., Azpilicueta-Amorín, M., & Cordero-Rivera,

A. 2011. Juvenile hormone favors sexually-selected traits but impairs fat reserves

and abdomen mass in males and females. Evol. Ecol. 25: 845-856. 213

Corbet, P.S. 1999. Dragonflies: behavior and ecology of Odonata. Cornell University

Press, Ithaca, NY

Córdoba-Aguilar, A. 2002. Wing pigmentation in territorial male damselflies, Calopteryx

haemorrhoidalis: a possible relation to sexual selection. Anim. Behav. 63: 759-766.

Cornwallis, C.K. & Uller, T. 2010. Towards an evolutionary ecology of sexual traits.

Trends Ecol. Evol. 25: 145-152

Cotter, S.C., Myatt, J.P., Benskin, C.M.H., & Wilson, K. 2008. Selection for cuticular

melanism reveals immune function and life-history trade-offs in Spodoptera

littoralis. J. Evol. Biol. 21: 1744-1754.

Cotter, S.C., Simpson, S.J., Raubenheimer, D., & Wilson, K. 2010. Macronutrient

balance mediates trade-offs between immune function and life history traits. Funct.

Ecol. 25: 186-198.

Cotton, S., Folwer, K., & Pomiankowski, A. 2004. Do sexual ornaments demonstrate

heightened condition-dependent expression as predicted by the handicap

hypothesis? Proc. R. Soc. B 271: 771-783.

Cox, R.M. & Calsbeek, R. 2009. Sexually antagonistic selection, sexual dimorphism, and

the resolution of intralocus sexual conflict. Am. Nat. 173: 176-187.

Cox, R.M. & Calsbeek, R. 2015. Survival of the fattest? Indices of body condition do not

predict viability in the brown anole. Funct. Ecol. 29: 404-413.

Coyne, J. & Orr, H.A. 2004. Speciation. Sinauer Associates, Sunderland, MA.

Craig, J.K. & Foote, C.J. 2001. Countergradient variation and secondary sexual color:

phenotypic convergence promotes genetic divergence in carotenoid use between 214

sympatric anadromous and nonanadromous morphs of sockeye salmon

(Oncorhynchus nerka). Evolution 55: 380-391.

Crean, A.J., Monro, K., & Marshall, D.J. 2011. Fitness consequences of larval traits

persist across the metamorphic boundary. Evolution 65: 3079-3089.

Crone, E.E. Is survivorship a better fitness surrogate than fecundity? Evolution 55: 2611-

2614.

Crumrine, P.W. 2005. Size structure and substitutability in an odonate intraguild

predation system. Oecologia 145: 132-139.

Crumrine, P.W., Switzer, P.V., & Crowley, P.H. 2008. Structure and dynamics of

odonate communities: assessing habitat, responding to risk, and enabling

reproduction. In: Dragonflies & damselflies: model organisms for ecological and

evolutionary research (A. Córdoba-Aguilar A ed.), pp. 21-38. Oxford University

Press, Oxford, UK.

Cuthill, I.C., Allen, W.L., Arbuckle, K., Caspers, B., Chaplin, G., Hauber, M.E., et al.

2017. The biology of color. Science 357: eaan0221.

Dale, J., Dey, C.J., Delhey, K., Kempenaers, B., & Valcu, M. 2015. The effects of life

history and sexual selection on male and female plumage colouration. Nature 527:

367-370.

Darwin, C. 1871. The Descent of Man and Selection in Relation to Sex. John Murray,

London, UK.

Debecker, S., Sommaruga, R., Maes, T., & Stoks, R. 2015. Larval UV exposure impairs

adult immune function through a trade-off with larval investment in cuticular

melanin. Funct. Ecol. 29: 1292-1299. 215

Dillon, M.E., Wang, G., & Huey, R.B. 2010. Global metabolic impacts of recent climate

warming. Nature 467: 704-706.

Ducrest, A.-L., Keller, L., & Roulin, A. 2008. Pleiotropy in the melanocortin system,

coloration and behavioural syndromes. Trends Ecol. Evol. 23: 502-510.

Dugand, R.J., Tomkins, J.L., & Kennington, J. 2019. Molecular evidence supports a

genic capture resolution of the lek paradox. Nat. Comm. 10:1359.

Dugas, M.B., Moore, M.P., Martin, R.A., Richards-Zawacki, C.L., & Sprehn, C.G. 2016.

The pay-offs of maternal care increase as offspring develop, favouring extended

provisioning in an egg-feeding frog. J. Evol. Biol. 29: 1977-1985.

Duong, T.M. & McCauley, S.J. 2016. Predation risk increases immune response in a

larval dragonfly (Leucorrhinia intacta). Ecology 97: 1605-1610.

Ellers, J. & Boggs, C.L. 2002. The evolution of wing color in Colias butterflies:

heritability, sex linkage, and population divergence. Evolution 56: 836-840.

Ellers, J. & Boggs, C.L. 2003. The evolution of wing color: male mate choice opposes

adaptive wing color divergence in Colias butterflies. Evolution 57: 1100-1106.

Emlen, S.T. & Oring, L.W. 1977. Ecology, sexual selection, and the evolution of mating

systems. Science 197: 215-223.

Endler, J.A. 1980. Natural selection on color patterns in Poecilia reticulata. Evolution 34:

76-91.

Endler, J.A. 1992. Signals, signal conditions, and the direction of evolution. Am Nat. 139:

S125-S153. 216

Evison, S.E.F., Gallagher J.D., Thompson, J.J.W., Siva-Jothy, M.T., & Armitage, S.A.O.

2017. Cuticular colour reflects underlying architecture and is affected by a limiting

resource. J. Insect. Physiol. 98: 7-13.

Fedorka, K.M., Lee, V., Winterhalter, W.E. 2013. Thermal environment shapes cuticle

melanism and melanin-based immunity in the ground cricket Allonemobius socius.

Evol. Ecol. 27: 521-531.

Fedorka, K.M., Zuk, M., & Mousseau, T.A. 2004. Immune suppression and the cost of

reproduction in the ground cricket, Allonemobius socius. Evolution 58: 2478-2485.

Fellous, S. & Lazzaro, B.P. 2011. Potential for evolutionary coupling and decoupling of

larval and adult immune gene expression. Molec. Ecol. 20: 1558-1567.

Felsenstein, J. 1985. Phylogenies and the comparative method. Am. Nat. 125: 1-15.

Fick, S.E. & Hijmans, R.J. 2017. WorldClim 2: new 1-km spatial resolution climate

surfaces for global land areas. Int. J. Climat. 37: 4302-4315.

Fisher, R.A. 1930. The genetical theory of natural selection. Oxford University Press,

Oxford, UK.

Folstad, I. & Karter, A.J. 1992. Parasites, bright males, and the immunocompetence

handicap. Am. Nat. 139: 603-622.

Freda, P.J., Alex, J.T., Morgan, T.J., & Ragland, G.J. 2017. Genetic decoupling of

thermal hardiness across metamorphosis in Drosophila melanogaster. Integr. Comp.

Biol. 57: 999-1009.

Freitak, D., Ots, I., Vanatoa, A., & Hörak, P. 2003. Immune response is energetically

costly in white cabbage butterfly pupae. Biol. Lett. 270: S220-S222. 217

Fried, C.S. & May, M.L. 1983. Energy expenditure and food intake of territorial male

Pachydiplax longipennis (Odonata: Libellulidae). Ecol. Entomol. 8: 283-292.

Forbes, M.R. & Robb, T. 2008. Testing hypotheses about parasite-mediated selection

using odonate hosts. In: Dragonflies & Damselflies: Model organisms for

ecological and evolutionary research (A. Córdoba-Aguilar ed.), pp. 175-188.

Oxford University Press, Oxford, UK.

Futahashi, R. 2016. Color vision and color formation in dragonflies. Curr. Opin. Ins. Sci.

17: 32-39.

García-Berthou, E. 2001. On the misuse of residuals in ecology: testing regression

residuals vs. the analysis of covariance. J. Anim. Ecol. 70: 708-711.

Giery, S.T. & Layman, C.A. 2019. Ecological consequences of sexually selected traits:

an eco-evolutionary perspective. Quart. Rev. Biol. 94: 29-74.

Gilbert, R., Karp, R.D., & Uetz, G.W. 2016. Effects of juvenile infection on adult

immunity and secondary sexual characters in a wolf spider. Behav. Ecol. 27: 946-

954

Gillooly, J.F., Brown, J.H., West, G.B., Savage, V.M., & Charnov, E.L. 2001. Effects of

size and temperature on metabolic rate. Science 293: 2249-2251.

Gilmore, A. 2014. The effects of atrazine and predation risk on larval dragonflies

Ladonna deplanata (Odonata: libellulidae). MSc Thesis. Murray State University,

Murray, KY, USA.

Gomez-Mestre, I. & Buchholz, D.R. 2006. Developmental plasticity mirrors differences

among taxa in spadefoot toads linking plasticity and diversity. Proc. Natl. Acad. Sci.

USA 103: 19021-19026. 218

González-Santoyo, I. & Córdoba-Aguilar, A. 2012. Phenoloxidase: a key component of

the insect immune system. Entomol. Exp. Appl. 142: 1-16.

Gosler, A.G., Greenwood, J.J.D., & Perrins, C. 1995. Predation risk and the cost of being

fat. Nature 377: 621-623.

Graham, M.H. 2003. Confronting multicollinearity in ecological multiple regression.

Ecology 84: 2809-2815.

Grether, G.F. 1996a. Sexual selection and survival selection on wing coloration and body

size in the rubyspot damselfly Hetaerina americana. Evolution 50: 1939-1948.

Grether, G.F. 1996b. Intrasexual competition alone favors a sexually dimorphic ornament

in the rubyspot damselfly Hetaerina americana. Evolution 50: 1949-1957.

Grether, G.F. 1997. Survival cost of an intrasexually selected ornament in a damselfly.

Proc. R. Soc. B 264: 207-210.

Guillermo-Ferreira, R., Gorb, S.N., Appel, E., Kovalev, A., & Bispo, P.C. 2015. Variable

assessment of wing colouration in aerial contests of the red-winged damselfly

Mnesarete pudica (Zygoptera, Calopterygidae). Sci. Nat. 102: 13.

Hadfield, J.D. 2008. Estimating evolutionary parameters when viability selection is

operating. Proc. R. Soc. B 275: 723-734.

Halfwerk, W., Bot, S., Bukx, J., van der Velde, M., Komdeur, J., ten Cate, C., &

Slabbekoorn, H. 2011. Low-frequency songs lose their potency in noisy urban

conditions. Proc. Natl. Acad. Sci. USA 108: 14549-14554.

Haller, B.C. & Hendry, A.P. 2014. Solving the paradox of stasis: squashed stabilizing

selection and the limits of detection. Evolution 68: 483-500. 219

Hamon, T.R. 2005. Measurement of concurrent selection episodes. Evolution 59: 1096-

1103.

Hanley, D., Miller, N.G., Flockhart, D.T.T., & Norris, D.R. 2013. Forewing pigmentation

predicts migration disetance in wild-caught migratory monarch butterflies. Behav.

Ecol. 24: 1108-1113.

Hansen, T.F., Pélabon, C., & Houle, D. 2011. Heritability is not evolvability. Evol. Biol.

38: 258.

Harris, I., Jones, P.D., Osborn, T.J,. & Lister, D.H. 2013. Updated high-resolution grids

of monthly climatic observations – the CRU TS3.10 Dataset. Int. J. Climat. 34:

623-642.

Hassall, C. 2014. Continental variation in wing pigmentation in Calopteryx damselflies is

related to the presence of heterospecifics. PeerJ 2: e438.

Hegna, R.H. Nokelainen, O., Hegna J.R., & Mappes, J. 2013. To quiver or to shiver:

increased melanization benefits thermoregulation, but reduces warning signal

efficacy in the wood tiger moth. Proc. R. Soc. B 280: 20122812.

Heinen-Kay, J.L., Morris, K.E., Ryan, N.A., Byerley, S.L., Venezia, R.E., Peterson,

M.N., Langerhans, R.B. 2015. A tradeoff between natural and sexual selection

underlies diversification of a sexual signal. Behav. Ecol. 26: 533-542.

Hendry, A.P. & Gonzalez, A. 2008. Whither adaptation? Biol. Phil. 23: 673-699.

Hijmans, R.J., Cameron, S.E., Parra, J.L., Jones, P.G., & Jarvis, A. 2005. Very high

resolution interpolated climate surfaces for global land areas. Int. J. Climat. 25:

1965-1978. 220

Hill, G.E. 2011. Condition-dependent traits as signals of the functionality of vital cellular

processes. Ecol. Lett. 14: 625-634.

Hine, E., McGuigan, K., & Blows, M.W. 2011. Natural selection stops the evolution of

male attractiveness. Proc. Natl. Acad. Sci. USA 108: 3659-3664.

Hoffmann, A.A., Chown, S.L., & Clusella-Trullas, S. 2013. Upper thermal limits in

terrestrial ectotherms: how constrained are they? Funct. Ecol. 27: 934-949.

Hooper, R.E. & Siva-Jothy, M.T. 1996. Last male sperm precedence in a damselfly

demonstrated by RAPD profiling. Molec. Ecol. 5: 449-452.

Hooper, R.E., Tsubaki, T., & Siva-Jothy, M.T. 1999. Expression of a costly, plastic

secondary sexual trait is correlated with age and condition in a damselfly with two

male morphs. Physiol. Entomol. 24: 364-369.

Hopper, K.R., Crowley, P.H., & Kielman, D. 1996. Density dependence, hatching

synchrony, and within-cohort cannibalism in young dragonfly larvae. Ecology 77:

191-200.

Huey, R.B., Deutsch, C.A., Tewksbury, J.J., Vitt, L.J., Hertz, P.E., Pérez, H.J.Á., &

Garland, T. 2009. Why tropical lizards are vulnerable to climate warming. Proc. R.

Soc. B 276: 1939-1948.

Huey, R.B. & Kingsolver, J.G. 1989. The evolution of thermal sensitivity of ectotherm

performance. Trends Ecol. Evol. 4: 131-135.

IPCC. 2014. Climate Change 2013: the physical science basis. Working Group I

Contribution to the AR5. Cambridge University Press, Cambridge, NY USA.

Izzo, A.S. & Tibbetts, E.A. 2012. Spotting the top male: sexually selected signals in male

Polistes dominulus wasps. Anim. Behav. 83: 839-845. 221

Jacomb, F., Marsh, J., & Holman, L. 2016. Sexual selection expedites the evolution of

pesticide resistance. Evolution 70: 2746-2751.

Jacot, A., Scheuber, H., & Brinkhof, M.W.G. 2004. Costs of an induced immune

response on sexual display and longevity in field crickets. Evolution 58: 2280-2286.

Jacot, A., Scheuber, H., Kurtz, J., & Brinkhof, M.W.G. 2005. Juvenile immune status

affects the expression of a sexually selected trait in field crickets. J. Evol. Biol. 18:

1060-1068.

Jakob, E.M., Marshall, S.E., & Uetz, G.W. 1996. Estimating fitness: a comparison of

body condition indicies. Oikos 77: 61-67.

Janssens, L. & Stoks, R. 2014 Non-pathogenic aquatic bacteria activate the immune

system and increase predation risk in damselfly larvae. Freshw. Biol. 59: 417-426.

Janzen, F.J. & Stern, H.S. 1998. Logistic regression for empirical studies of multivariate

selection. Evolution 52: 1564-1571.

Jennions, M.D., Møller, A.P., & Petrie, M. 2001. Sexually selected traits and adult

survival: a meta-analysis. Quart. Rev. Biol. 76: 3-36.

Johnson, C. 1962. A study of territoriality and breeding behavior in Pachydiplax

longipennis Burmeister (Odonata: Libellulidae). Southwest. Nat. 7: 191-197.

Johnson, D.W. & Hixon, M.A. 2011. Sexual and lifetime selection on body size in a

marine fish: the importance of life-history trade-offs. J. Evol. Biol. 24: 1653-1663.

Junior, R.S.L. & Peixoto, P.E.C. 2013. Males of the dragonfly Diastatops obscura fight

according to predictions from game theory models. Anim. Behav. 85: 663-669. 222

Kapari, L., Haukioja, E., Rantala, M.J., & Ruuhola, T. 2006. Defoliating insect immune

defense interacts with induced plant defense during a population outbreak. Ecology

87: 291-296.

Kasumovic, M.M. 2013. The multidimensional consequences of the juvenile

environment: towards an integrative view of the adult phenotype. Anim. Behav. 85:

1049-1059.

Kaufman, S.D., Johnston, T.A., Leggett, W.C., Moles, M.D., Casselman, J.M., &

Schulte-Hostedde, A.I. (2007) Relationships between body condition indices and

proximate composition in adult walleyes. T. Am. Fish. Soc. 136: 1566-1576.

Kelly, C.D., Neyer, A.A., & Gress, B.E. 2014. Sex-specific life history responses to

nymphal diet quality and immune status in a field cricket. J. Evol. Biol. 27: 381-390.

Kelly, C.D., Tawes, B.R., & Worthington, A.M. 2014. Evaluating indices of body

condition in two cricket species. Ecol. Evol. 23: 4476-4487.

Kenward, M.G. & Roger. J.H. 1997. Small sample inference for fixed effects from

maximum likelihood. Biometrics 53: 983-997.

Khan, I., Agashe, D., & Rolff, J. 2017. Early-life inflammation, immune response and

ageing. Proc. R. Soc. B. 284: 20170125.

Kingsolver, J.G. 1983. Thermoregulation and flight in Colias butterflies: elevational

patterns and mechanistic limiations. Ecology 64: 534-545.

Kingsolver, J.G. 1995. Fitness consequences of seasonal polyphenism in western white

butterflies. Evolution 49: 942-954.

Kingsolver, J.G. & Buckley L.B. 2017. Evolution of plasticity and adaptive responses to

climate change along climate gradients. Proc. R. Soc. B 284: 20170386. 223

Kingsolver, J.G., Diamond, S.E., & Buckley, L.B. 2013. Heat stress and the fitness

consequences for terrestrial ectotherms. Funct. Ecol. 27: 1415-1423.

Kingsolver, J.G., Diamond, S.E., Siepelski, A.M., & Carlson, S.M. 2012. Synthetic

analyses of phenotypic selection in natural populations: lessons, limitations and

future directions. Evol. Ecol. 26: 1101-1118.

Kingsolver, J.G., Hoekstra, H.E., Hoekstra, J.M., Berrigan, D., Vignieri, S.N., Hill, C.E.,

Hoang, A., Giberg, P., & Beerli, P. 2001. The strength of phenotypic selection in

natural populations. Am. Nat. 157: 245-261.

Kitano, H. 2004. Biological robustness. Nat. Rev. Gen. 5: 826-837.

Koenig, W.D. 2008. Lifetime reproductive success and sexual selection theory. In:

Dragonflies & Damselflies: Model organisms for ecological and evolutionary

research (Córdoba-Aguilar, A. ed.), pp. 153-166. Oxford University Press, Oxford,

UK.

Koenig, W.D. & Albano, S.S. 1987. Lifetime reproductive success, selection, and the

opportunity for selection in the white-tailed skimmer Plathemis lydia (Odonata:

Libellulidae). Evolution 41: 22-36.

Kolker, M., Meiri, S., & Holzman, R. 2019. Prepared for the future: a strong signal of

evolution toward the adult benthic niche during the pelagic stage in Labrid fishes.

Evolution 73: 803-816.

König, C. & Schmid-Hempel, P. 1995. Foraging activity and immunocompetence in

workers of the bumble bee, Bombus terrestris L. Proc. R. Soc. B. 260: 225-227. 224

Koskimäki, J., Rantala, M.J., Taskinen, J., Tynkkynen, K., & Suhonen, J. 2004.

Immunocompetence and resource holding potential in the damselfly Calopteryx

virgo L. Behav. Ecol. 15: 169-173.

Kotiaho, J.S. 2001. Costs of sexual traits: a mismatch between theoretical considerations

and empirical evidence. Biol. Rev. 76: 365-376

Krams, I., Daukšte, J., Kivelniece, I., Kaasik, A., Krama, T., Freeberg, T.M., & Rantala,

M.J. 2013. Trade-off between cellular immunity and life span in mealworm beetles

Tenebrio molitor. Curr. Zool. 59: 340-346.

Kuchta, S.R. & Svensson, E.I. 2014. Predator-mediated natural selection on the wings of

the damselfly Calopteryx splendens: differences in selection among trait types. Am.

Nat. 184: 91-109.

Lailvaux, S.P. & Irschick, D.J. 2006. A functional perspective on sexual selection:

insights and future prospects. Anim. Behav. 72: 263-273.

Lande, R. 1980. Sexual dimorphism, sexual selection, and adaptation in polygenic

characters. Evolution 34: 292-305.

Lande, R. 1981. Models of speciation by sexual selection on polygenic traits. Proc. Natl.

Acad. Sci. USA 78: 3721-3725.

Lande, R. & Arnold, S.J. 1983. The measurement of selection on correlated characters.

Evolution 37: 1210-1226.

Langerhans, R.B. & DeWitt, T.J. 2004. Shared and unique features of evolutionary

diversification. Am. Nat. 164: 335-349.

Lawniczak, M.K.N., Barnes, A.I., Linklater, J.R., Boone, J.M., Wigby, S., & Chapman, T.

2007. Mating and immunity in invertebrates. Trends Ecol. Evol. 22: 48-55. 225

Le Rouzic, A., Hansen, T.F., Gosden, T.P., & Svensson, E.I. 2015. Evolutionary time-

series analysis reveals the signature of frequency-dependent selection on a female

mating polymorphism. Am. Nat. 185: E182-E196.

Leighton, G.R.M., Hugo, P.S., Roulin, A., & Amar, A. 2016. Just Google it: assessing the

use of Google Images to describe geographical variation in visible traits of

organisms. Methods Ecol. Evol. 7: 1060-1070.

Leinonen, T., O’Hara, R.B., Cano, J.M., & Merilä, J. 2008. Comparative studies of

quantitative trait and neutral marker divergence: a meta-analysis. J Evol Biol 21: 1-

17.

Lenth, R.V. 2016. Least-squares means: the R package lsmeans. J. Stat. Softw. 69: 1-33.

Lessells, C.M. & Boag, P.T. 1987. Unrepatable repeatabilities: a common mistake. Auk

104: 116-121.

Letsch, H., Gottsberger, B., & Ware, J.L. 2016. Not going with the flow: a

comprehensive time-calibrated phylogeny of dragonflies (Anisoptera: Odonata:

Insecta) provides evidence for the role of lentic habitats on diversification. Molec.

Ecol. 25: 1340-1353.

Lima, S.L. 1998. Nonlethal effects in the ecology of predator-prey interactions.

Bioscience 48: 25-34.

Lindstedt, C., Schroderus, E., Lindström, L., Mappes, T., & Mappes, J. 2016.

Evolutionary constraints of warning signals: a genetic trade-off between the

efficacy of larval and adult warning coloration can maintain variation in signal

expression. Evolution 70: 2562-2572. 226

Lorch, P.D., Proulx, S., Rowe, L., & Day, T. 2003. Condition-dependent sexual selection

can accelerate adaptation. Evo. Eco. Res. 5: 867-881.

Lowe, C.D., Harvey, I.F., Watts, P.C., & Thompson, D.J. 2009. Reproductive timing and

patterns of development for the damselfly Coenagrion puella in the field. Ecology

90: 2202-2212.

Maan, M.E. & Cummings, M.E. 2008. Female preferences for aposematic signal

components in a polymorphic poison frog. Evolution 62: 2334-2345.

Maan, M.E. & Seehausen, O. 2011. Ecology, sexual selection and speciation. Ecol. Lett.

14: 591-602.

MacColl, A.D.C. 2011. The ecological causes of evolution. Trends Ecol. Evol. 26: 514-

522.

Maklakov, A.A., Simpson, S.J., Zajitschek, F., Hall, M.D., Dessmann, J., Clissold, F., et

al. 2008. Sex-specific fitness effects of nutrient intake on reproduction and lifespan.

Curr. Biol. 7: 1062-1066.

Marchinko, K.B. 2009. Predation’s role in repeated phenotypic and genetic divergence of

armor in threespine stickleback. Evolution 63: 127-138.

Marden, J.H. 1995. Large-scale changes in thermal sensitivity of flight performance

during adult maturation in a dragonfly. J. Exp. Biol. 198: 2095-2102.

Marden, J.H. & Cobb, J.R. 2004. Territorial and mating success of dragonflies that vary

in muscle power output and presence of gregarine gut parasites. Anim. Behav. 68:

857-865.

Marshall, D.J., Burgess, S.C., & Connallon, T. 2016. Global change, life-history

complexity and the potential for evolutionary rescue. Evol. Appl. 9: 1189-1201. 227

Marshall, D.J. & Morgan, S.G. 2011. Ecological and evolutionary consequences of linked

life-history stages in the sea. Curr. Biol. 21: R718-R725.

Martin, R.A. & Pfennig, D.W. 2010. Field and experimental evidence that competition

and ecological opportunity promote resource polymorphism. Biol. J. Linn. Soc. 100:

73-88.

Martin, R.A., Riesch, R., Heinen-Kay, J.L., & Langerhans, R.B. 2014. Evolution of male

coloration during a post-pleistocene radiation of Bahamas mosquitofish (Gambusia

hubbsi). Evolution 68: 397-411.

May, M.L. 1976. Thermoregulation and adaptation to temperature in dragonflies

(Odonata: Anisoptera). Ecol. Monogr. 46: 1-32.

Maynard Smith, J. & Harper, D. 2003. Animal Signals. Oxford University Press, Oxford,

UK.

McCauley, S.J. 2010. Body size and social dominance influence breeding dispersal in

male Pachydiplax longipennis (Odonata). Ecol. Entomol. 35: 377-385.

McCauley, S.J., Davis, C.J., Relyea, R.A., Yurewicz, K.L., Skelly, D.K., & Werner, E.E.

2008. Metacommunity patterns in larval odonates. Oecologia 158: 329-342.

McKean, K.A., Yourth, C.P., Lazzaro, B.P., & Clark, A.G. 2008. The evolutionary costs

of immunological maintenance and deployment. BMC Evol. Biol. 8: 76.

McKean, K.A. & Nunney, L. 2008. Sexual selection and immune function in Drosophila

melanogaster. Evolution 62: 386-400.

McNamara, K.B., van Lieshout, E., Jones, T.M., & Simmons, L.W. 2013. Age-dependent

trade-offs between immunity and male, but not female, reproduction. J. Anim. Ecol.

82: 235-244. 228

McPeek, M.A., Grace, M., & Richardson, J.M.L. 2001. Physiological and behavioral

responses to predators shape the growth/predation risk trade-off in damselflies.

Ecology 82: 1535-1545.

McVey, M.E. & Smittle, B.J. 1984. Sperm precedence in the dragonfly Erythemis

simplicicollis. J. Insect. Physiol. 30: 619-628.

Miehls, A.J., Peacor, S.D., & McAdam, A.G. 2014. Gape-limited predators as agents of

selection on the defensive morphology of an invasive invertebrate. Evolution 68:

2633-2643.

Mikolajewski, D.J., De Block, M., & Stoks, R. 2015. The interplay of adult and larval

time constraints shapes species differences in larval life history. Ecology 96: 1128-

1138.

Mikolawjewski, D.J., Johansson, F., Wohlfahrt, B., & Stoks, R. 2006. Invertebrate

predation selects for the loss of a morphological antipredator trait. Evolution 60:

1306-1310.

Miles, M.C. & Fuxjager, M.J. 2018. Synergistic selection regimens drive the evolution of

display complexity in birds of paradise. J. Anim. Ecol. 87: 1149-1159.

Miller, C.W. & Svensson, E.I. 2014. Sexual selection in complex environments. Annu.

Rev. Entomol. 59: 427-445.

Mills, S.C., Hazard, L., Lancaster, L., Mappes, T., Miles, D., Oksanen, T.A., & Sinervo,

B. 2008. Gonadotropin hormone modulation of testosterone, immune function,

performance, and behavioral trade-offs among male morphs of the lizard Uta

stansburiana. Am. Nat. 171: 339-357. 229

Mitchell, K.A., Sinclair, B.J., & Terblanche, J.S. 2013. Ontogenetic variation in cold

tolerance plasticity in Drosophila: is the Bogert effect bogus? Naturwissenschaften

100: 281-284.

Mitchell, T.S., Warner, D.A., & Janzen, F.J. 2013. Phenotypic and fitness consequences

of maternal nest-site choice across multiple early life stages. Ecology 94: 336-345.

Mitchell-Olds, T. & Shaw, R.G. 1987. Regression analysis of natural selection: statistical

inference and biological interpretation. Evolution 41: 1149-1161.

Moczek, A.P. & Nijhout, H.F. 2003. Rapid evolution of a polyphenic threshold. Evol.

Dev. 5: 259-268.

Mojica, J.P. & Kelly, J.K. 2010. Viability selection prior to trait expression is an essential

component of natural selection. Proc. R. Soc. B 277: 2945-2950.

Monro, K. & Marshall, D.J. 2014. Faster is not always better: selection on growth rate

fluctuates across life history and environments. Am. Nat. 183: 798-809.

Moore, A.J. 1990. The evolution of sexual dimorphism by sexual selection: the separate

effects of intrasexual selection and intersexual selection. Evolution 44: 315-331.

Moore, M.P., Lis, C., Gherghel, I., & Martin, R.A. 2019. Temperature shapes the costs,

benefits, and geographic diversification of sexual coloration in a dragonfly. Ecol.

Lett. 22: 437-446.

Moore, M.P., Lis, C., & Martin, R.A. 2018a. Immune deployment increases larval

vulnerability to predators and inhibits adult life-history traits in a dragonfly. J. Evol.

Biol. 31: 1365-1376.

Moore, M.P., Lis, C., & Martin, R.A. 2018b. Larval body condition regulates predator-

induced life-history variation in a dragonfly. Ecology 99: 224-230. 230

Moore, M.P. & Martin, R.A. 2016. Intrasexual selection favours an immune-correlated

color ornament in a dragonfly. J. Evol. Biol. 29: 2256-2265.

Moore, M.P. & Martin, R.A. 2018. Trade-offs between larval survival and adult

ornament development depend on predator regime in a territorial dragonfly.

Oecologia 188: 97-106.

Moore, M.P., Riesch, R., & Martin, R.A. 2016. The predictability and magnitude of life-

history divergence to ecological agents of selection: a meta-analysis in livebearing

fishes. Ecol. Lett. 19: 435-442.Moran, N.A. 1994. Adaptation and constraint in the

complex life cycles of animals. Annu. Rev. Ecol. Evol. System. 25: 573-600.

Morehouse, N.I. 2014. Condition-dependent ornaments, life histories, and the evolving

architecture of resource-use. Integr. Comp. Biol. 54: 591-600.

Morehouse, N.I., Nakazawa, T., Booher, C.M., Jeyasingh, P.D. & Hall, M.D. 2010. Sex

in a material world: why the study of sexual reproduction and sex-specific traits

should become more nutritionally-explicit. Oikos 119: 766-778.

Moret, Y. & Schmid-Hempel, P. 2000. Survival for immunity: the price of immune

system activation for bumblebee workers. Science 290: 1166-1168.

Morey, S., & Reznick, D. 2000. A comparative analysis of plasticity in larval

development in three species of spadefoot toads. Ecology 81: 1736-1749.

Morrissey, M.B. & Sakrejda, K. 2013. Unification of regression-based methods for the

analysis of natural selection. Evolution 67: 2094-2100.

Morrissey, M.B. & Sakrejda, K. 2014. gsg: Calculation of selection coefficients. R

package version 2.0. http://CRAN.R-project.org/package=gsg

Morton, E.S. 1975. Ecological sources of selection on avian sounds. Am Nat 109: 17-34. 231

Nappi, A.J., & Vass, E. 1993. Melanogenesis and the generation of cytotoxic molecules

during insect cellular immune reactions. Pigment Cell Res. 6: 117-126.

Nowak, M.A., Boerlijst, M.C., Cooke, J., & Maynard Smith, J.M. 1997. Evolution of

genetic redundancy. Nature 388: 167-171.

Oke, K.B., Rolshausen, G., LeBlond, C., & Hendry, A.P. 2017. How parallel is parallel

evolution? A comparative analysis in fishes. Am. Nat. 190: 1-16.

Otti, O., Gantenbein-Ritter, I., Jacot, A., & Brinkhof, M.W.G. 2012. Immune response

increases predation risk. Evolution 66: 732-739.

Outomuro, D. & Ocharan, F.J. 2011. Wing pigmentation in Calopteryx damselflies: a role

in thermoregulation? Biol. J. Linn. Soc. 103: 36-44.

Pascoal, S., Cezard, T., Eik-Nes, A., Gharbi, K., Majewska, J., Payne, E., et al. 2014.

Rapid convergent evolution in wild crickets. Curr. Biol. 24: 1369-1374.

Patricelli, G.L. & Blickley, J.L. 2006. Avian communication in urban noise: causes and

consequences of vocal adjustment. The Auk 123: 639-649.

Paulson, D. 2009. Dragonflies and damselflies of the west. Princeton University Press,

Princeton, NJ, USA.

Paulson, D. 2012. Dragonflies and damselflies of the east. Princeton University Press,

Princeton, NJ, USA.

Persson L, De Roos A, Claessen D, Byström P, Lövgren J, Sjögren S, et al. 2003.

Gigantic cannibals driving a whole-lake trophic cascade. Proc. Natl. Acad. Sci. USA

100: 4035-4039. 232

Peig, J. & Green, A.J. 2009. New perspectives for estimating body condition from

mass/length data: the scaled mass index as an alternative method. Oikos 118: 1883-

1891.

Peig, J. & Green, A.J. 2010. The paradigm of body condition: a critical reappraisal of

current methods based on mass and length. Funct. Ecol. 24: 1323-1332.

Phillips, P.C. 1998. Genetic constraints at the metamorphic boundary: morphological

development in the wood frog, Rana sylvatica. J Evol Biol 11: 453-463.

Pinheiro, J., Bates, D., DebRoy, S., Sarkar, D., & R Core Team. 2015. nlme: Linear and

nonlinear mixed effects models. R package version 3.1-119

Punzalan, D., Cooray, M., Rodd, F.H., Rowe, L. 2008a. Condition dependence of

sexually dimorphic colouration and longevity in the ambush bug Phymata

americana. J. Evol. Biol. 21: 1297-1306.

Punzalan, D., Rodd, F.H., & Rowe, L. 2008b. Contemporary sexual selection on sexually

dimorphic traits in the ambush bug Phymata americana. Behav. Ecol. 19: 860-870.

Punzalan, D., Rodd, F.H., & Rowe, L. 2008c. Sexual selection mediated by the

thermoregulatory effects of male colour pattern in the ambush bug Phymata

americana. Proc. R. Soc. B 275: 483-492.

R Development Core Team. 2014. R: a language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria.

R Development Core Team. 2015. R: a language and environment for statistical

computing. R Foundation for Statistical Computing, Vienna, Austria. 233

Rantala, M.J. Honkavaara, J., & Suhonen, J. 2010. Immune system activation interacts

with territory-holding potential and increases predation of the damselfly Calopteryx

splendens by birds. Oecologia 163: 825-832.

Rantala, M.J., Kaskimäki, J., Taskinen, J., Tynkkynen, K., & Suhonen, J. 2000.

Immunocompetence, developmental stability and wingspot size in the damselfly

Calopteryx splendens L. Proc. R. Soc. B 267: 2453-2457.

Rantala, M.J. & Roff, D.A. 2007. Inbreeding and extreme outbreeding cause sex

differences in immune defence and life history traits in Epirrita autumnata.

Heredity 98: 329-336.

Rasband, W.S. 2012. ImageJ. Betheseda, MD: U.S. National Institutes of Health.

Reddiex, A.J., Gosden, T.P., Bonduriansky, R., & Chenoweth, S.F. 2013. Sex-specific

fitness consequences of nutrient intake and the evolvability of diet preferences. Am.

Nat. 182: 91-102.

Relyea, R.A. 2007. Getting out alive: how predators affect the decision to metamorphose.

Oecologia 152: 389-400.

Reznick, D., Nunney, L., & Tessier, A. 2000. Big houses, big cars, superfleas and the

costs of reproduction. Trends Ecol. Evol. 15: 421-425.

Rice, W.R. & Gaines, S.D. 1994. Extending nondirectional heterogeneity tests to evaluate

simply ordered alternative hypotheses. Proc. Natl. Acad. Sci. USA 91: 225-226.

Robey, C.W. 1975. Observations on breeding behavior of Pachydiplax longipennis

(Odonata: Libellulidae). Psyche 82: 89-96. 234

Robinson, M.R., Pilkington, J.G., Clutton-Brock, T.H., Pemberton, J.M., & Kruuk, L.E.B.

2008. Environmental heterogeneity generates fluctuating selection on a secondary

sexual trait. Curr. Biol. 18: 751-757.

Rodríguez, R.L., Boughman, J.W., Gray, D.A., Hebets, E.A., Höbel, G., & Symes, L.B.

2013. Diversification under sexual selection: the relative roles of mate preference

strength and the degree of divergence in mate preferences. Ecol. Lett. 16: 964-974.

Roff, D.A. & Fairbairn, D.J. 2007. The evolution of trade-offs: where are we? J. Evol.

Biol. 20: 433-447.

Rolff, J., Armitage, S.A.O, & Coltman, D.W. 2005. Genetic constraints and sexual

dimorphism in immune defense. Evolution 59: 1844-1850.

Rolff, J. & Siva-Jothy, M.T. 2003. Invertebrate ecological immunology. Science 301:

472-475.

Rosche, L., Semroc, J.M., & Gilbert, L.K. 2008. Dragonflies and damselflies of northeast

Ohio. Cleveland Museum of Natural History, Cleveland, OH, USA.

Roulin, A. 2016. Condition-dependence, pleiotropy and the handicap principle of sexual

selection in melanin-based colouration. Biol. Rev. 91: 328-348.

Rowe, L. & Houle D. 1996. The lek paradox and the capture of genetic variance by

condition dependent traits. Proc R. Soc. B 263: 1415-1421.

Rubenstein, D.R. & Hauber, M.E. 2008. Dynamic feedback between phenotype and

physiology in sexually selected traits. Trends Ecol. Evol. 23: 655-658.

Rundle, H.D., Chenoweth, S.F., & Blows, M.W. 2006. The roles of natural and sexual

selection during adaptation to a novel environment. Evolution 60: 2218-2225. 235

Sadd, B., Holman, L., Armitage, H., Lock, F., Marland, R., & Siva-Jothy, M.T. 2006.

Modulation of sexual signaling by immune challenged male mealworm beetles

(Tenebrio molitor, L.): evidence for terminal investment and dishonesty. J. Evol.

Biol. 19: 321-325.

Sadd, B.M. & Siva-Jothy, M.T. 2006. Self-harm caused by an insect’s innate immunity.

Proc. R. Soc. B. 273: 2571-2574.

Safran, R.J., Adelman, J.S., McGraw, K.J., & Hau, M. 2008. Sexual signal exaggeration

affects physiological state in male barn swallows. Curr. Biol. 18: R461-R462.

Samejima, Y. & Tsubaki, Y. 2010. Body temperature and body size affect flight

performance in a damselfly. Behav. Ecol. Sociobiol. 64: 685-692.

Schielzeth, H. 2010. Simple means to improve the interpretability of regression

coefficients. Methods Ecol. Evol. 1: 103-113.

Schluter, D & Nychka, D. 1994. Exploring fitness surfaces. Am. Nat. 143: 597-616.

Schluter, D., Price, T.D., & Rowe, L. 1991. Conflicting selection pressures and life

history trade-offs. Proc. R. Soc. B 246: 11-17.

Schultte-Hostedde, A.I., Zinner, B., Millar, J.S., & Hickling, G.J. 2005. Restitution of

mass-size residuals: validating body condition indices. Ecology 86: 155-163.

Schmid-Hempel, P. 2005. Evolutionary ecology of insect immune defenses. Annu. Rev.

Entomol. 50: 529-551.

Schwarzenbach, G.A. & Ward, P.I. 2006. Responses to selection on phenoloxidase

activity in yellow dung flies. Evolution 60: 1612-1621.

Searcy, W.A. & Nowicki, S. 2005. The evolution of animal communication: reliability

and deception in signaling systems. Princeton University Press, Princeton, NJ, USA. 236

Seehausen, O., van Alphen, J.J.M., & Witte, F. 1997. Cichlid fish diversity threatened by

eutrophication that curbs sexual selection. Science 277: 1808-1811.

Seppälä, O. 2015. Natural selection on quantitative immune defence traits: a comparison

between theory and data. J. Evol. Biol. 28: 1-9.

Servedio, M.R. & Boughman, J.W. 2017. The role of sexual selection in local adaptation

and speciation. Annu. Rev. Ecol. Evol. Syst. 48: 85-109.

Sheldon, B.C. & Verhulst, S. 1996. Ecological immunology: costly parasite defences and

trade-offs in evolutionary ecology. Trends Ecol. Evol. 11: 317-321.

Sherman, K.J. 1983. The adaptive significance of postcopulatory mate guarding in a

dragonfly, Pachydiplax longipennis. Anim. Behav. 31: 1107-1115.

Sherratt, E., Vidal-García, M., Anstis, M., & Keogh, J.S. 2017. Adult frogs and tadpoles

have different macroevolutionary patterns across the Australian continent. Nature

Ecol. Evol. 1: 1385.

Siepielski, A.M., Morrissey, M.B., Buoro, M., Carlson, S.M., Caruso, C.M., Clegg, S.M.,

et al. (2017) Precipitation drives global variation in natural selection. Science 355:

959-962.

Signor, S.A., Liu, Y., Rebeiz, M., Kopp, A. 2016. Genetic convergence in the evolution

of male-specific color patterns in Drosophila. Curr. Biol. 26: 2423-2433.

Sinclair, B.J., Williams, C.M., & Terblanche, J.S. 2012. Variation in thermal

performance among insect populations. Phys. Biochem. Zool. 85: 594-606.

Sinervo, B. & Calsbeek, R. 2003. Physiological epistasis, ontogenetic conflict and natural

selection on physiology and life history. Integr. Comp. Biol. 43: 419-430. 237

Siva-Jothy, M.T. 2000. A mechanistic link between parasite resistance and expression of

a sexually selected trait in a damselfly. Proc. R. Soc. B 267: 2523-2527.

Siva-Jothy, M.T., Moret, Y., & Rolff, J. 2005. Insect immunity: an evolutionary ecology

perspective. Adv. Insect Physiol. 32: 1-48.

Slos, S., De Meester, L., & Stoks, R. 2009. Food level and sex shape predator-induced

physiological stress: immune defence and antioxidant defence. Oecologia 161: 461-

467.

Smith, A.B., Littlewood, D.T.J., & Wray, G.A. 1995. Comparing patterns of evolution:

larval and adult life history stages and ribosomal RNA of post-Palaeozoic echinoids.

Phil. Trans. R. Soc. Lond. B 349: 11-18.

Sogard, S.M. 1997. Size-selective mortality in the juvenile stage of teleost fishes: a

review. B. Mar. Sci. 60: 1129-1157.

Srygley, R.B., Lorch, P.D., Simpson, S.J., & Sword, G.A. 2009. Immediate protein

dietary effects on movement and the generalised immunocomptence of migrating

Mormon crickets Anabrus simplex (Orthoptera: Tettigoniidae). Ecol. Entomol. 34:

663-668.

Start, D. 2018. Animal behavior and algal camouflage jointly structure predation and

selection. J. Evol. Biol. 31: 773-778.

Stearns, S.C. 1992. The evolution of life histories. Oxford University Press, Oxford, UK.

Stoehr, A.M. 2006. Costly melanin ornaments: the importance of taxon. Funct. Ecol. 20:

276-281.

Stoks, R. & Córdoba-Aguilar, A. 2012. Evolutionary ecology of odonata: a complex life

cycle perspective. Annu. Rev. Entomol. 57: 249-265. 238

Stoks, R., De Block, M., & McPeek, M.A. 2005. Alternative growth and energy storage

responses to mortality threats in damselflies. Ecol. Lett. 8: 1307-1316.

Stoks, R., De Block, M., Slos, S., Van Doorslaer, W., & Rolff, J. 2006. Time constraints

mediate predator-induced plasticity in immune function, condition, and life history.

Ecology 87: 809-815.

Stoks, R., Nystrom, J.L., May, M.L., & McPeek, M.A. 2005. Parallel evolution in

ecological and reproductive traits to produce cryptic damselfly species across the

Holarctic. Evolution 59: 1976-1988.

Strobbe, F., McPeek, M.A., De Block, M., De Meester, L., Stoks, R. 2009. Survival

selection on escape performance and its underlying phenotypic traits: a case of

many-to-one mapping. J. Evol. Biol. 22: 1172-1182.

Stuart-Fox, D., Newton, E., & Clusella-Trullas, S. 2017. Thermal consequences of colour

and near-infrared reflectance. Phil. Trans. R. Soc. B 372: 20160345.

Suhonen, J., Rantala, M.J., & Honkavaara, J. 2008. Territoriality in odonates. In:

Dragonflies & Damselflies: Model organisms for ecological and evolutionary

research (Córdoba-Aguilar, A. ed.). pp. 203-217. Oxford University Press, Oxford,

UK.

Sunday, J.M., Bates, A.E., & Dulvy, N.K. 2012. Thermal tolerance and the global

redistribution of animals. Nat. Clim. Change 2: 686-690.

Sunday, J.M., Bates, A.E., Kearney, M.R., Colwell, R.K., Dulvy, N.K., Longino, J.T., &

Huey, R.B. 2014. Thermal-safety margins and the necessity of thermoregulatory

behavior across latitude and elevation. Proc. Natl. Acad. Sci. USA 111: 5610-5615. 239

Svensson, E.I. 2019. Eco-evolutionary dynamics of sexual selection and sexual conflict.

Funct. Ecol. 33: 60-72.

Svensson, E.I. & Friberg, M. 2007. Selective predation on wing morphology in sympatric

damselflies. Am. Nat. 170: 101-112.

Svensson, E.I., Kristofferson, L., Oskarsson, K., & Bensch, S. 2004. Molecular

population divergence and sexual selection on morphology in the banded

demoiselle (Calopteryx splendens). Heredity 93: 423-433.

Svensson, E.I. & Sinervo, B. 2004. Spatial scale and temporal component of selection in

side-blotched lizards. Am. Nat. 163: 726-734.

Svensson, E.I. & Waller, J.T. 2013. Ecology and sexual selection: evolution of wing

pigmentation in Calopterygid damselflies in relation to latitude, sexual dimorphism,

and speciation. Am. Nat. 182: E174-E195.

Takahashi, Y., Kagawa, K., Svensson, E.I., & Kawata, M. 2014. Evolution of increased

phenotypic diversity enhances population performance by reducing sexual

harassment in damselflies. Nat. Commun. 5: 4468.

Talloen, W., Van Dyck, H., & Lens, L. 2004. The cost of melanization: butterfly wing

coloration under environmental stress. Evolution 58: 360-366.

Thompson, D.J., Hassall, C., Lowe, C.D., & Watts, P.C. 2011. Field estimates of

reproductive success in a model insect: behavioural surrogates are poor predictors

of fitness. Ecology 14: 905-913.

Tibbetts, E.A., Crocker, K., & Huang, Z.Y. 2016. Socially selected ornaments influence

hormone titers of signalers and receivers. Proc. Natl. Acad. Sci. USA 113: 8478-

8483. 240

Touchon, J.C., McCoy, M.W., Landberg, T., Vonesh, J.R., and Warkentin, K.M. 2015.

Putting µ/g in a new light: plasticity in life history switch points reflects fine-scale

adaptive responses. Ecology 96: 2192-2202.

Tracy, C.R., Turner, J.S., Bartholomew, G.A., Bennett, A., Billings, W.D., Chabot, B.F.

1982. What is physiological ecology? Bull. Ecol. Soc. Am. 63: 340-347.

True, J.R., Edwards, K.A., Yamamoto, D., & Carroll, S.B. 1999. Drosophila wing

melanin patterns form by vein-dependent elaboration of enzymatic prepatterns.

Curr. Biol. 9: 1382-1391.

Tsubaki, Y., Samejima, Y., & Siva-Jothy, M.T. 2010. Damselfly females prefer hot

males: higher courtship success in males in sunspots. Behav. Ecol. Sociobiol. 64:

1547-1554.

Tüzün, N., Op de Beeck, L., & Stoks, R. 2017. Sexual selection reinforces a higher flight

endurance in urban damselflies. Evol. Appl. 10: 694-703.

Tüzün, N, Op de Beeck, L., & Stoks, R. 2017. Data from: Sexual selection reinforces a

higher flight endurance in urban damselflies. Dryad Digital Repository.

Tynkkynen, K., Rantala, M.J., & Suhonen, J. 2004. Interspecific aggression and character

displacement in the damselfly Calopteryx splendens. J. Evol. Biol. 17: 759-767.

Uller, T., Isaksson, C., & Olsson, M. 2006. Immune challenge reduces reproductive

output and growth in a lizard. Funct. Ecol. 20: 873-879.

Urban, M.C., Bürger, R., & Bolnick, D.I. 2013. Asymmetric selection and the evolution

of extraordinary defences. Nat. Commun. 4: 2085.

Van Buskirk, J. 1987. Density-dependent population dynamics in larvae of the dragonfly

Pachydiplax longipennis: a field experiment. Oecologia 72: 221-225. 241

Van Buskirk, J., McCollum, S.A., & Werner, E.E. 1997. Natural selection for

environmentally induced phenotypes in tadpoles. Evolution 51: 1983-1992. van Nordwijk, A.J. & de Jong, G. 1986. Acquisition and allocation of resources: their

influence on variation in life history tactics. Am. Nat. 128: 137-142. van Tienderen, P.H. 2000. Elasticities and the link between demographic and

evolutionary dynamics. Ecology 81: 666-679. von Schantz, T., Bensch, S., Grahn, M., Hasselquist, D., & Wittzell, H. 1999. Good genes,

oxidative stress and condition-dependent sexual signals. Proc. R. Soc. B 266: 1-12.

Viechtbauer, W. 2010. Conducting meta-analyses in R with the metaphor package. J. Stat.

Soft. 36: 1-48.

Vieira, M.C. & Piexoto, P.E.C. 2013. Winners and losers: a meta-analysis of functional

determinants of fighting ability in arthropod contests. Funct. Ecol. 27: 305-313.

Vitousek, M.N., Steward, R.A., & Safran, R.J. 2013. Female plumage colour influences

seasonal oxidative damage and testosterone profiles in a songbird. Biol. Lett. 9:

20130539.

Vitousek, M.N., Zonana, D.M. & Safran, R.J. 2014. An integrative view of the signaling

phenotype: dynamic links between signals, physiology, behavior and social context.

Curr. Zool. 60: 739-754.

Waage, J.K. 1979a. Dual function of the damselfly penis: sperm removal and transfer.

Science 203: 916-918.

Waage, J.K. 1979b. Reproductive character displacement in Calopteryx (Odonata:

Calopterygidae). Evolution 33: 104-116.

Wade, M.J. & Kalisz, S. 1990. The causes of natural selection. Evolution 44: 1947-1955. 242

Waller, J. & Svensson, E.I. 2016. The measurement of selection when detection is

imperfect: how good are naïve methods? Methods Ecol. Evol. 7: 538-548.

Watkins, T.B. 2001. A quantitative genetic test of adaptive decoupling across

metamorphosis for locomotor and life-history traits in the pacific tree frog, Hyla

Regilla. Evolution 55: 1668-1677.

Watt, W.B. 1968. Adaptive significance of pigment polymorphisms in Colias butterflies.

I. Variation in melanin pigment in relation to thermoregulation. Evolution 22: 437-

458.

Watt, W.B. 1969. Adaptive significance of pigment polymorphisms in Colias butterflies.

II. Thermoregulation and photoperiodically controlled melanin variation in Colias

eurytheme. Proc. Natl. Acad. Sci USA 63: 767-774.

Weber, M.G., Mitko, L., Eltz, T., & Ramírez, S.R. 2016. Macroevolution of perfume

signaling in orchid bees. Ecol. Lett. 19: 1314-1323.

Wellborn, G.A. 1994. Size-biased predation and prey life histories: a comparative study

of freshwater amphipod populations. Ecology 75: 2104-2117.

Werner, E.E. & Gilliam, J.F. 1984. The ontogenetic niche and species interactions in

size-structured populations. Annu. Rev. Ecol. Syst. 15: 393-425.

West, P.M. & Packer, C. 2002. Sexual selection, temperature and the lion’s mane.

Science 297: 1339-1343.

Wiens, J.J. 2001. Widespread loss of sexually selected traits: how the peacock lost its

spots. Trends Ecol. Evol. 16: 517-523.

Wilbur, HM. 1980. Complex life cycles. Annu Rev. Ecol. Syst. 11: 66-93. 243

Wissinger, S.A. 1988a. Effects of food availability on larval development and inter-instar

predation among larvae of Libellula lydia and Libellula luctuosa (Odonata:

Anisoptera). Can. J. Zool. 66: 543-549.

Wissinger, S.A. 1988b. Spatial distribution, life history, and estimates of survivorship in

a fourteen-species assemblage of larval dragonflies (Odonata: Anisoptera). Freshw.

Biol. 20: 329-340.

Wissinger, S.A. 1992. Niche overlap and the potential for competition and intraguild

predation between size-structured populations. Ecology 73: 1431-1444.

Wissinger, S.A., Whiteman, H.H., Denoël, M., Mumford, M.L., & Aubee, C.B. 2010.

Consumptive and nonconsumptive effects of cannibalism in fluctuating age-

structured populations. Ecology 91: 549-559.

Wittkopp, P.J. & Beldade, P. 2009. Development and evolution of insect pigmentation:

genetic mechanisms and the potential consequences of pleiotropy. Semin. Cell Dev.

Biol. 20: 65-71.

Wollenberg Valero, K.C., Garcia-Porta, J., Rodríguez, A., Arias, M., Shah, A.,

Randrianiaiana R.D., et al. 2017. Transcriptomic and macroevolutionary evidence

for phenotypic uncoupling between frog life history phases. Nature Commun. 8:

15213.

Wood, C.W. & Brodie III, E.D. 2015. Environmental effects on the structure of the G-

matrix. Evolution 69: 2927-2940.

Yang, A.S. 2001. Modularity, evolvability, and adaptive radiations: a comparison of the

hemi- and holometabolous insects. Evol. Dev. 3: 59-72. 244

Yun, L., Chen, P.J., Kwok, K.E., Angell, C.S., Rundle, H.D., & Agrawal, A.F. 2018.

Competition for mates and the improvement of nonsexual fitness. Proc. Natl. Acad.

Sci. USA 115: 6762-6767.

Zamora-Camacho, F.J., Reguera, S., Rubiño-Hispán, M.V., & Moreno-Rueda, G. 2014a.

Effects of limb length, body mass, gender, gravidity, and elevation on escape speed

lizard Psammodromus algirus. Evol. Biol. 41: 509-517.

Zamora-Camacho, F.J., Reguera, S., Rubiño-Hispán, M.V. & Moreno-Rueda, G. 2014b.

Eliciting an immune response reduces sprint speed in a lizard. Behav. Ecol. 26: 115-

120.

Zeuss, D., Brandl, R., Brändle, M., Rahbek, C., & Brunzel, S. 2014. Global warming

favours light-coloured insects in Europe. Nature Commun. 5: 3874.

Zuk, M. & Kolluru, G.R. 1998. Exploitation of sexual signals by predators and

parasitoids. Quart. Rev. Biol. 73: 415-438.

Zuk, M., Rotenberry, J.T., & Tinghitella, R.M. 2006. Silent night: adaptive disappearance

of a sexual signal in a parasitized population of field crickets. Biol. Lett. 2: 521-524.

Zuk, M. & Stoehr, A.M. 2002. Immune defense and host life history. Am. Nat. 160: S9-

S22.