The Evolutionary Ecology of Autotomy in Leaf-Footed Bugs (Insecta: Hemiptera: Coreidae)

THE EVOLUTIONARY ECOLOGY OF AUTOTOMY IN LEAF-FOOTED BUGS (INSECTA: HEMIPTERA: COREIDAE)

ZACHARY EMBERTS

A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA

2019

To my family

ACKNOWLEDGMENTS

I would like to thank my husband, Cody Coyotee Howard, for his support throughout this endeavor. I would also like to thank my advisors, Colette M. St. Mary and Christine W. Miller, for their guidance along the way. Finally, I would like to thank those who have assisted, encouraged, and challenged me to become the scientist that I am today.

TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ...... 4

LIST OF TABLES ...... 8

LIST OF FIGURES ...... 10

LIST OF ABBREVIATIONS ...... 12

ABSTRACT ...... 14

CHAPTER

1 THE ECOLOGY AND EVOLUTION OF AUTOTOMY...... 16

A Working Definition of Autotomy ...... 18 Evolution of the Autotomy Phenotype ...... 19 Evolutionary Origins and Losses of Autotomy ...... 19 Elaboration of Autotomizable Appendages...... 22 Variation in the Costs and Benefits of Autotomy ...... 25 Benefits of Autotomy ...... 25 Costs of Autotomy ...... 31 Economic Theory of Autotomy: Predicting When an Individual Should Autotomize ...... 38 Implications of Autotomy on Organismal and Environmental Interactions ...... 41 Predator–prey Interactions ...... 41 Intraspecific Competition ...... 43 Movement and Habitat Selection ...... 44 Applications of Autotomy Research ...... 45 Future Directions ...... 46 Conclusions ...... 47

2 COREIDAE (INSECTA: HEMIPTERA) LIMB LOSS AND AUTOTOMY ...... 53

Methods ...... 54 Results ...... 56 Missing Legs in Wild Populations ...... 56 Experimental Induction of Autotomy ...... 56 Discussion ...... 57 Autotomy and Sexually Selected Weaponry...... 58 Autotomy and Natural Selection ...... 59

3 AUTOTOMIZING INJURED LIMBS INCREASES SURVIVAL ...... 69

Methods ...... 71

Study Organism ...... 71 Study Design ...... 72 Insect rearing ...... 72 Experiment 1 ...... 73 Experiment 2 ...... 74 Data and Statistical Analyses ...... 75 Experiment 1 ...... 75 Experiment 2 ...... 76 Results ...... 77 Experiment 1 ...... 77 Experiment 2 ...... 78 Discussion ...... 79

4 COREIDAE THAT AUTOTOMIZE THEIR SEXUALLY-SELECTED HIND LEGS HAVE DECREASED FIGHTING ABILITY AND MATING SUCCESS ...... 95

Methods ...... 98 Rearing of Study Species ...... 98 Experimental Design ...... 99 Ethical Note ...... 101 Statistical Analyses ...... 101 Results ...... 103 Effects of Losing a Weapon on Fighting Ability and Behavior ...... 104 Effects of Losing a Weapon on Mating Success ...... 105 Effects of Losing One Limb Versus Two ...... 106 Discussion ...... 106

5 THE EVOLUTION OF AUTOTOMY IN LEAF-FOOTED BUGS ...... 118

Methods ...... 122 Behavioral and Morphological Data ...... 122 Molecular Data and Sequence Alignment ...... 123 Phylogeny and Divergence Time Estimation ...... 125 Statistical Analyses ...... 128 Results ...... 129 Phylogenetic Relationships ...... 129 Dating Analyses ...... 130 Autotomy ...... 131 Discussion ...... 132

APPENDIX

A STOCHASTIC CHARACTER SIMULATIONS REVEAL THAT AUTOTOMY AND AUTOTOMIZABLE LIMB ELABORATIONS HAVE EVOLVED MULTIPLE TIMES THROUGHOUT ANIMALIA ...... 155

B BODY SIZE, LATITUDE, AND THE PRESENCE OF ENLARGED HIND LEGS ALL EXPLAIN VARIATION IN THE RATE AT WHICH SPECIES AUTOTOMIZE ACROSS THIS CLADE, REGAURDLESS OF OUR MEASURE OF CENTRAL TENDENCEY ...... 158

C OUR DATING ANALYSES REVEALED YOUNGER AGE ESTIMATES THAN PREVIOUSLY REPORTED ...... 160

D BOTH MALE AND FEMALE LEAF-FOOTED BUG ANCESTORS AUTOTOMIZED THEIR HIND LEGS SLOWLY ...... 161

LIST OF REFERENCES ...... 162

BIOGRAPHICAL SKETCH ...... 185

LIST OF TABLES

Table page

2-1 Percentage of missing limbs by limb location...... 61

2-2 Data and analyses of wild caught specimens...... 62

2-3 Experimental autotomy data and analyses for ‘one-hour escape from entrapment’ scenario...... 64

2-4 Experimental autotomy data and analyses for ‘60 seconds escape from entrapment’ scenario...... 65

3-1 A taxonomic overview of autotomy, including anecdotal evidence of autotomy in response to injury...... 85

3-2 Experiment 1 – Developmental differences between autotomy, injury, and our control (no autotomy/ no injury)...... 86

3-3 Experiment 1 – Developmental differences between self-autotomizing and retaining an injured limb...... 87

3-4 Narnia femorata regenerative capabilities...... 88

3-5 The effect of autotomizing or retaining an injured limb on development...... 89

4-1 Narnia femorata fighting tactics...... 111

4-2 Results of Contrast 2 – the comparison between males missing middle legs and intact males...... 112

4-3 The number of limbs lost affected grappling behavior, but not the other response variables...... 113

5-1 An OU model of trait evolution best explains how the latency to autotomize evolved...... 139

5-2 Our model selection criterion (AICc) determined that the best model should include body size, distance from the equator (Latitude), and the presence of enlarged hind legs (Costly)...... 140

5-3 Our model selection criterion (AICc) determined that the best model should include body size, distance from the equator (latitude), and the presence of enlarged hind legs (Costly)...... 141

5-4 Best model for the mean female data...... 142

5-5 Best model for the mean male data...... 143

5-6 Best model for the mean male data, minus Pephricus paradoxus...... 144

LIST OF FIGURES

Figure page

1-1 Autotomy has evolved independently multiple times throughout Animalia...... 49

1-2 Autotomizable appendages are often elaborate such as the A) brightly colored tail of Morethia ruficauda or B) the elongated tail of Lialis burtonis...... 50

1-3 There are multiple benefits associated with autotomy...... 51

1-4 Under the economic theory of autotomy A), an individual should perform the action that is the least costly...... 52

2-1 Representative hind legs of each species in this study, to scale...... 66

2-2 Coreids can autotomize limbs to escape fouled molts...... 67

2-3 Percentage of wild caught coreid adults observed missing legs...... 68

3-1 A juvenile Narnia femorata...... 90

3-2 The right hind leg of a juvenile N. femorata, depicting the location of each injury site...... 91

3-3 Experiment 1 – Contrast of treatments to investigate the effects of autotomy and injury on the proportion of individuals (± SE) surviving to adulthood...... 92

3-4 Experiment 1 – Effect of injury location on autotomy and survival...... 93

3-5 Experiment 2 – Proportion of individuals (± SE) that survived in each treatment based on their autotomy behavior...... 94

4-1 Two Narnia femorata males posteriorly aligned abdomen to abdomen, ready to engage in a grapple...... 114

4-2 The effects of losing a weapon on fighting ability and mating success...... 115

4-3 Lack of plasticity in fighting behavior...... 116

4-4 Body size explains variation in A) contest duration for subordinate males, but not B) dominant ones...... 117

5-1 Comparing ancestral state reconstructions when assuming an Ornstein- Uhlenbeck (OU) model of trait evolution to a Brownian Motion (BM) model of trait evolution for the all data combined dataset...... 145

5-2 RAxML best tree with bootstrap values labeled at the nodes...... 146

5-3 Dated BEAST tree with median node ages labeled and bars denoting the 95% highest probability density interval...... 147

5-4 Dated TreePL tree with median node ages labeled and error bars that show the range of age estimates across 100 bootstrap trees...... 148

5-5 The ancestor of leaf-footed bugs autotomized their hind limbs slowly...... 149

5-6 Species that are smaller and closer to the equator autotomize more quickly, and the degree to which having an enlarged hind femur influences the mean latency to autotomize is sex and size specific...... 151

5-7 Small males with enlarged hind legs found near the equator autotomize quickly...... 152

5-8 Ancestral state reconstruction for the mean latency to autotomize assuming Brownian Motion (BM) with tip labels...... 153

5-9 Degrees from equator, our latitudinal gradient, had a bimodal distribution...... 154

A-1 A visual representation of a single stochastic character simulation for the ability to autotomize ...... 157

LIST OF ABBREVIATIONS

AICc Corrected Akaike Information Criterion

BM Brownian Motion bp Base pairs

C Celsius

D Day df Degrees of freedom

EB Early Burst

ESS Effective sample size

GLM Generalized linear model

GLMM Generalized linear mixed model

GTR General Time Reversible

HC Hight costs hr Hour hrs Hours

L Light

LC Low costs

MCC Maximum clade credibility

MCMC Markov chain Monte Carlo mm Millimeters mya Million years ago

N North n Sample size ng nanograms

OU Ornstein-Uhlenbeck

pGLM phylogenetic generalized linear model

PW Pronotal width s seconds

SE Standard error

T Time

TE Target enrichment

TE-TD Touch-down

UCE Ultraconserved elements v Version

W West

µL Microliters

Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy

THE EVOLUTIONARY ECOLOGY OF AUTOTOMY IN LEAF-FOOTED BUGS (INSECTA: HEMIPTERA: COREIDAE)

Zachary Emberts

December 2019

Chair: Colette M. St. Mary Cochair: Christine W. Miller Major: Zoology

Predation has acted as a strong selective pressure, which has resulted in a remarkable amount of morphological and behavioral diversity. Autotomy, a phenomenon in which organisms literally drop appendages in order to escape predation, is an extreme example showcasing the extent to which lineages have evolved in order to survive predatory attacks. Here, I investigate the fitness consequences and evolutionary history of this extreme trait. I start by reviewing the current state of knowledge about the ecology and evolution of autotomy, where I identify clear directions for future research. In my second chapter, I introduce my study system, leaf-footed bugs (Insecta: Hemiptera: Coreidae). These insects can autotomize all of their legs, but most frequently autotomize their hind legs. Hind legs in this clade can serve a variety of ecological functions, and once they are dropped they cannot be regenerated, which is one of many factors that make this an ideal clade to investigate fundamental question about autotomy. In my third and fourth chapters respectively, I show that autotomy can increase survival and that it can also reduce an organism’s future mating success. For my final chapter, I took an across species comparative

approach to investigate the evolution of autotomy within leaf-footed bugs and allies. I found that the ancestor of leaf-footed bugs autotomized, but did so slowly (> 15 min); rapid autotomy (< 2 min) then arose multiple times. The ancestor likely used slow autotomy to reduce the cost of injury or to escape non-predatory entrapment, but could not use autotomy to escape predation. This result suggests that autotomy to escape predation is a co-opted benefit (i.e., an exaptation), revealing one way that sacrificing a limb to escape predation may arise. In addition to identifying the origins of rapid autotomy, I also show that across species variation in the rates of autotomy can be explained by body size, distance from the equator, and enlargement of the autotomizable appendage.

CHAPTER 1 THE ECOLOGY AND EVOLUTION OF AUTOTOMY*

A lizard dropping its tail to escape predation dramatically illustrates the importance of survival in the context of natural selection, which is why autotomy (i.e. self-induced loss of a body part) has become a textbook example of an anti-predatory trait (Goodenough et al. 2009). This example, however, does not capture the full diversity of organisms that use autotomy, the variety of body parts that can be dropped, nor the subtler nuances of autotomy, such as the range of ecological contexts under which autotomy can occur. In addition to lizards, autotomy has been observed in several other lineages, including cephalopods, arthropods, and salamanders (Wake and

Dresner 1967, Fleming et al. 2007, Bush 2012). Phylogenetic comparative analyses – including analyses conducted for this chapter – reveal that autotomy has multiple evolutionary origins, as well as losses (Zani 1996). The diversity of organisms that autotomize is mirrored in a diverse array of autotomizable appendages. Examples include tails, claws, and legs (Wake and Dresner 1967, Fleming et al. 2007). These autotomizable structures are often more elaborate (e.g. larger in size, have more conspicuous coloration) than homologous structures that cannot be dropped (Arnold

1984, Fleming et al. 2013), which suggests that having the ability to drop a limb may facilitate trait diversification. The first of three objectives of this chapter is to synthesize our current understanding of the evolution of autotomy and the elaboration of autotomizable appendages (i.e. synthesize the evolution of the autotomy phenotype).

* This chapter is reprinted with permission from Emberts, Z., I. Escalante, and P. W. Bateman. 2019. The ecology and evolution of autotomy. Biological Reviews

In addition to being an extreme anti-predatory trait, autotomy can be beneficial in a variety of other contexts, and sacrificing a limb may not always be as costly as it is often assumed. Autotomy is also used to escape non-predatory entrapment (Maginnis

2008), reduce the cost of injury (Emberts et al. 2017a), and increase an individual’s reproductive success. Some male spiders, for example, have been observed autotomizing part of their intromittent organs (an external structure used to transfer sperm) inside a female’s reproductive tract (Fromhage and Schneider 2006, Uhl et al.

2010), which can ultimately increase the male’s reproductive success (Snow et al.

2006). Complete loss of these intromittent organs may result in the male becoming functionally sterile, which highlights the extreme costs that can be associated with autotomy. However, these costs can vary dramatically from one organism to the next.

For example, harvestmen can lose up to two legs before experiencing any notable costs to locomotion (Guffey 1999). The second objective of this chapter is to synthesize our current understanding of the costs and benefits associated with autotomy. Given such variation in the cost to benefit ratio, we also modified the economic theory of escape

(Ydenberg and Dill 1986) to generate an economic theory of autotomy, which makes predictions about when an individual should autotomize.

Loss of an autotomizable appendage can also affect how an individual interacts with other organisms and its environment. For example, male Iberian rock lizards

(Iberolacerta monticola) reduce their home-range size following tail autotomy, which has implications for reproductive opportunity (Salvador et al. 1995). Loss of an autotomizable appendage can also affect an organism’s foraging behavior and success.

Asian shore crabs (Hemigrapsus sanguineus) missing both of their chelipeds, for

example, have a slower feeding rate and consume smaller prey items than do intact individuals (Davis et al. 2005). This altered pattern of foraging behavior and feeding success highlights that autotomy may have a cascading effect on community dynamics, especially when one considers that 16% of the observed population is missing at least one cheliped. The final objective of this chapter is to synthesize our current understanding of the ecological implications of autotomy.

This review chapter does not provide a comprehensive taxonomic overview of autotomy nor does it directly discuss autotomy in relation to regeneration. We intentionally avoided these topics because we believe that they have already received a good amount of attention. For taxonomic specific reviews of autotomy see McVean

(1975), Arnold (1984), Juanes and Smith (1995), Wilkie (2001), Fleming et al. (2007,

2013), Bateman and Fleming (2009), and Higham et al. (2013). For previous reviews on the relationship between autotomy and regeneration see Wilkie (2001), Maginnis (2006) and Bely and Nyberg (2010). Because autotomy and regeneration are often coupled, we do consider some implications of regeneration. However, it is important to note that regeneration does not always succeed autotomy, and in many cases the loss of an appendage is permanent.

A Working Definition of Autotomy

One of the largest obstacles surrounding the study of autotomy is defining it.

Since it was first coined (Fredericq 1883), ‘autotomy’ has generally been used to denote a conscious decision on the part of an organism to drop an appendage, usually at a specific plane and often as a defense against predators. Herein, we simply define autotomy as the self-controlled loss of a body part at a predetermined breakage location. Therefore, as long as detachment consistently occurs at one, or in some cases

multiple, predetermined fracture planes across a population, we consider the self- controlled loss of a body part to be autotomy. To differentiate autotomy from molting/shedding, we also specify that autotomy should not be restricted to a certain climatic season or transitional period during development. We find this purely descriptive definition to be suitable for numerous reasons. First, autotomy has additional benefits beyond that of escaping predation, thus whether an individual uses autotomy in an anti-predatory context should not be incorporated into the definition. Second, by removing adaptive benefits from the definition, we can also begin to move away from a purely adaptive perspective and therefore broaden our scope, and hopefully our understanding, of autotomy. This definition, for example, allows us to investigate other adaptive [e.g. increased reproductive success (Ghislandi et al. 2015)] and non-adaptive

[e.g. phylogenetic inertia (Van Sluys et al. 2002)] hypotheses for observing self-induced appendage loss.

Evolution of the Autotomy Phenotype

Evolutionary Origins and Losses of Autotomy

The number and diversity of species that can undergo autotomy is quite remarkable. This diversity alone suggests that autotomy has evolved independently multiple times. In fact, it is widely accepted that autotomy has more than one evolutionary origin (McVean 1975). However, no study has provided evidence for this claim. This discrepancy is likely due to the absence of a study that has mapped autotomy onto a phylogenetic tree of Animalia, which, to gain a broad overview of autotomy’s evolutionary history, we present in Fig. 1-1 (see Appendix A). Although it is difficult to calculate the exact number of times autotomy has evolved independently, identifying whether autotomy has more than one evolutionary origin is relatively simple.

Therefore, we conducted stochastic character simulations to estimate the number of times autotomy has evolved and found that autotomy has at least nine independent origins across Animalia (Appendix A). Despite having multiple origins, much remains unknown about the evolution of autotomy as there have been no studies that have implemented experimental evolution and there have been only a handful of studies that have explicitly investigated autotomy within a phylogenetic framework (Zani 1996,

Mueller et al. 2004; implicit studies include Bateman and Fleming 2008, Pafilis et al.

2009). Of the phylogenetic studies, autotomy is always found to be the ancestral character of the investigated clade, but there are often multiple secondary losses [i.e. there are species that are unable to autotomize nested within the clade (Zani 1996,

Mueller et al. 2004)]. As a result, most research on the evolution of autotomy has investigated factors that correlate with its secondary loss.

Two factors observed to correlate with a reduction in the ability to autotomize across species are (1) body size, and (2) the function of the autotomizable body part, but there is conflicting evidence. In orthopterans, larger species take longer to autotomize their hind limbs (Bateman and Fleming 2008). This across-species correlation, however, does not hold for lizards (Zani 1996), as there is no correlation between the frequency of tail loss in natural populations – a fair proxy for the latency to autotomize in this clade (Pafilis et al. 2009) – and body size (Zani 1996). Since adult orthopterans cannot regenerate their autotomized body part, while lizards can, these differing patterns could potentially be explained by regeneration. With regards to the function of the autotomizable body part, the inability to autotomize a tail correlates with the tail’s function in salamanders (Mueller et al. 2004). For example, salamanders that

have tails that serve an important function in locomotion (e.g. swimming) are less likely to have the ability to drop their tail (Mueller et al. 2004). Although, across other taxa, the functional value of the autotomizable appendage does not always predict the ease of autotomy (Arnold 1984, Zani 1996, Fleming et al. 2013). In addition to these two factors, secondary loss of autotomizable body parts has been hypothesized to correlate negatively with other anti-predatory traits (Arnold 1984, Bateman and Fleming 2008), such as the ability to actively fight back against a predator (Fleming et al. 2013), and a decrease in predatory pressure (Arnold 1984, Pafilis et al. 2009).

The evolutionary origins of autotomy are unknown and have received relatively little attention. Hypothetically, there are at least two possible evolutionary routes to sacrificing a body part to escape predation. First, the ‘intermediate step’ hypothesis proposes that autotomy first arose to reduce the cost of an injured limb or body part, and was then co-opted to escape predation (McVean 1982, Wasson et al. 2002). This hypothesis predicts that when autotomy first arose it took a long time for appendage loss to occur, but organisms maintained this trait because autotomizing a body part slowly came with benefits (e.g. reducing the cost of injury, escaping non-predatory entrapment). Once organisms had the ability to drop a body part, selection could then act on the speed at which the body part was dropped, which, under this hypothesis, eventually led to organisms being able to drop their body parts quickly enough to escape the grasp of a predator. A natural alternative to this hypothesis is one in which autotomizing a limb quickly enough to escape predation evolves without an intermediate-latency step. We refer to this as the ‘fast latency’ hypothesis. Under this hypothesis, organisms first evolve an anti-predatory trait that deflects attacks towards a

specific portion of the body (e.g. false heads, brightly-colored tails) or they simply have an appendage that is disproportionally attacked. Selection could then drive such an appendage to be removed with ease (i.e. easy to fracture, but the fracturing would not be under the control of the individual), eventually resulting in the organisms being able to drop their own appendage in a rapid manner (i.e. individual control of removing the appendage). Both hypotheses largely overlook the morphological component of autotomy – the fracture plane – and future work on the evolution of autotomy should consider this component as well.

Elaboration of Autotomizable Appendages

Once an organism has evolved the ability to drop a body part, selection may act on the appendage to increase the efficacy (i.e. increase its benefit) or efficiency (i.e. mitigate its cost) of autotomy. This selection could explain why some autotomizable body parts are so elaborate. Elaboration of autotomizable appendages includes bright coloration (Fig. 1-2A; Arnold 1984), elongation (Fig. 1-2B; Fleming et al. 2013, Barr et al. 2018), and post-autotomy appendage movement (Dial and Fitzpatrick 1983).

Understanding the evolutionary pressures that promote and constrain these patterns of elaboration remains an exciting avenue of research.

One well-studied example of autotomy efficacy is the bright coloration of some lizard tails (Clark and Hall 1970, Arnold 1984, Castilla et al. 1999, Hawlena et al. 2006,

Watson et al. 2012) – an appendage that can be regenerated. Such conspicuous coloration has been shown to divert predator attacks towards the autotomizable body part (Clark and Hall 1970, Cooper and Vitt 1985, Watson et al. 2012, Bateman et al.

2014, Fresnillo et al. 2015), which likely increases an individual’s ability to survive a predation event. In some cases, however, this bright coloration also increases detection

by predators (Bateman et al. 2014, Fresnillo et al. 2015, Nasri et al. 2018, but see

Watson et al. 2012), making them risky decoys. Traits that help misdirect attacks towards autotomizable body parts can also be behavioral. Salamanders (Ducey et al.

1993) (which can regenerate), lizards (Minton 1966, Congdon et al. 1974, Arnold 1984,

Mori 1990) (which can regenerate), and true bugs (Emberts et al. 2016) (which cannot regenerate) have all been observed waving their autotomizable body parts when predators are near. When individuals perform this behavior, predators are more likely to strike their autotomizable body part (Ducey et al. 1993, Telemeco et al. 2011). Thus, predator deflection is one factor likely promoting and/or maintaining the elaboration of autotomizable body parts.

After being autotomized, some body parts will move in a wiggling, thrashing, or violently twitching fashion. This trait increases predator distraction time, ultimately increasing the efficacy of autotomy by providing the individual that autotomized the appendage with more time to escape (Dial and Fitzpatrick 1983). Post-autotomy appendage movement has evolved independently multiple times (Appendix A) and is observed in lizards (Higham and Russell 2010), salamanders (Labanick 1984), and arachnids [harvestmen (Miller 1977, Roth and Roth 1984), spiders (Johnson and Jakob

1999) and scorpions (Mattoni et al. 2015); of which some can regenerate and others cannot]. In some species, there is also inter-population variation in the vigor of post- autotomy appendage movement, which correlates with local predatory pressure

(Cooper et al. 2004; suggested in Cromie and Chapple 2012, Otaibi et al. 2017).

However, in other species, the duration of post-autotomy body-part movement is highly conserved (Pafilis et al. 2005, Pafilis et al. 2008, Pafilis et al. 2009). Therefore, the role

that current predatory pressure has in maintaining the intensity and duration of post- autotomy body-part movement remains unclear. Nonetheless, there is still strong evidence to suggest that the movement of autotomizable body parts has been maintained due to its ability to distract predators.

In addition to traits that can increase the success of autotomy, there are also traits that can help mitigate the costs. For example, lizards (Etheridge 1967, Haacke

1975, Haacke 1976) (which can regenerate), squid (Bush 2012) (which can regenerate), and scorpions (Mattoni et al. 2015) (which cannot regenerate) have evolved multiple autotomy fracture planes. Having multiple fracture planes along an autotomizable body part allows individuals to minimize the amount of body part that is sacrificed during autotomy: this is referred to as the economy of autotomy (Arnold 1984). Losing a smaller portion of the body part decreases both the short-term and long-term costs associated with appendage loss (e.g. reduced costs associated with regenerating a smaller appendage; Cromie and Chapple 2013). However, losing a smaller portion of the body part may also decrease the efficacy of autotomy. Shorter autotomized tail segments, for example, have reduced post-autotomy movement [e.g. distance moved

(Cooper and Smith 2009)], which may decrease the amount of time that an autotomized body part can distract a predator. Although, to date, there is no evidence to suggest that the amount of tail autotomized by lizards influences an individual’s probability of survival

(Cromie and Chapple 2013). The number of autotomy fracture planes in a lizard tail also varies both within and across species (Haacke 1975, Haacke 1976, Winchester and

Bellairs 1977, Arnold 1984, Gilbert et al. 2013, LeBlanc et al. 2018), and much of the variability within species has been attributed to ontogeny, with age decreasing the

number of fracture planes that an individual has (Arnold 1984). Across-species variability, however, has received far less attention, despite being an ideal topic to understand the role that current and historical pressures have had in influencing autotomizable body parts. For example, as lizard tails become relatively longer/larger they should cost more to regenerate (Bateman and Fleming 2009), thus, future studies should investigate whether obligatory regenerative costs associated with tail autotomy can explain variation in the number of fracture planes across species.

The patterns of body-part elaboration above suggest that the ability to drop an appendage may be a key innovation, facilitating trait diversification associated with that body part. Here, we postulate that autotomy facilitates the evolution of coloration, morphological shape elaboration, and behavior. Phylogenetic comparative analyses in lizards, for example, have found that the ability to autotomize tails evolved before tail coloration (Murali et al. 2018). This suggests that the ability to autotomize promotes the elaboration of the autotomy-related phenotypes, in this case conspicuous tail coloration

(Murali et al. 2018). However, the possibility of the opposite evolutionary pathway – that body-part elaboration promotes the evolution of autotomy – cannot be excluded. It should also be emphasized that non-adaptive hypotheses cannot be excluded either.

Future studies should explore these avenues more thoroughly.

Variation in the Costs and Benefits of Autotomy

Benefits of Autotomy

‘Autotomy’, as used by Fredericq (1883), attempts to describe the phenomenon in which some animals quickly drop part of their body in a defensive, anti-predatory, manner. However, since then we have gained a better understanding of self-induced appendage loss. For example, we now know that autotomy has additional benefits

beyond that of escaping predation (Fig. 1-3; Maginnis 2008, Emberts et al. 2017a). To understand how autotomy is selected for and maintained throughout Animalia we need a thorough understanding of all the benefits of autotomy. For this review chapter, we place the benefits of autotomy into four broad categories: (1) escaping predation; (2) escaping non-predatory entrapment; (3) reducing the cost of injury; and (4) increasing reproductive success. It is important to note that these benefits are not mutually exclusive and that selection may be maintaining autotomy within a population due to any combination of these categories.

Escaping predation is the most investigated benefit of autotomy. Several studies have shown that individuals are more vulnerable to predation once they have dropped their autotomizable body part (e.g. Congdon et al. 1974, Stoks 1998, Downes and Shine

2001, Bateman and Fleming 2006b). This result is likely to be due to several factors.

Most notably, individuals without their autotomizable body part cannot re-autotomize that appendage as a last-ditch effort to escape predation. This is particularly relevant for animals that do not regenerate appendages. However, the observed differences in vulnerability could also be due to predators differentially attacking individuals that are missing their autotomizable appendage (i.e. predator preference, but see Congdon et al. 1974, Lancaster and Wise 1996, Stoks 1998). Moreover, individuals missing body parts often have reduced locomotor capabilities (Fleming and Bateman 2007, Maginnis

2006, Fleming et al. 2007, but see Lu et al. 2010), which would inhibit their ability to flee. No study has teased these factors apart. To date, the degree to which autotomy itself increases an individual’s probability of escaping a predation event has yet to be demonstrated. Such a study would require manipulating an individual’s ability to

autotomize without removing the autotomizable body part and then comparing differences in escape ability between those that can autotomize and those that cannot, ideally using natural predators. Despite the lack of direct, experimental studies, there is still strong evidence to suggest that autotomy has an anti-predatory benefit.

Other evidence for autotomy having an anti-predatory benefit includes anecdotal observations, the speed at which autotomy is induced, and correlations between autotomy and local predation pressure. Autotomy of a body part during a predation event has been observed in numerous taxa [e.g. lizards (Congdon et al. 1974); spiders

(Punzo 1997); crustaceans (Lawton 1989, Wasson et al. 2002); earthworms (Sugiura

2010); salamanders (Labanick 1984)] and the most parsimonious explanation is that autotomy occurs to escape predation. Additionally, the speed at which organisms autotomize a body part has been used, implicitly, as evidence for an anti-predatory benefit. This can be seen in studies where researchers terminate autotomy trials after

10–120 s (Easton 1972, Cooper et al. 2004) if autotomy has yet to occur. Presumably these time points were selected because predation events often occur rapidly, and being unable to autotomize quickly suggests that autotomy would fail to serve as an escape mechanism. Correlations between predation pressure and the latency or frequency to autotomize across populations also suggest that autotomy in such taxa is primarily used to escape a predation event (e.g. Pafilis et al. 2009, Brock et al. 2015).

On the other hand, lack of such correlations suggests that autotomy is not exclusively used to escape predation. On some predator-free islands, for example, gecko populations have higher rates of autotomy than in mainland populations, which

suggests that autotomy functions in other contexts, in this case, to reduce the cost of injury from intraspecific competition (Itescu et al. 2017).

Traits associated with escaping are often associated with escaping predation, but there are also scenarios in which autotomy would benefit individuals that need to escape from non-predatory entrapment (e.g. Maginnis 2008, Hodgkin et al. 2014).

Within arthropods, non-predatory entrapment often manifests itself in the form of a bad molt. All arthropods go through multiple molting episodes during development, and in some cases throughout adult life. During this process, limbs can become stuck (Fig. 1-

3B). To avoid entrapment and potential death, individuals may simply autotomize these limbs. Autotomy to escape a bad molt has been observed in coreids (Emberts et al.

2016), walking sticks (Maginnis 2008), spiders (Foelix 1996), and decapods (Wood and

Wood 1932). In some cases, evading entrapment requires autotomizing several limbs.

The crab Carcinus maenas, for example, has been observed autotomizing up to three legs to escape a fouled molt (Wood and Wood 1932). Another non-predatory entrapment scenario can include getting stuck in tree sap/resin (hypothesized in

Maginnis 2008), although in certain contexts this may be considered predatory entrapment as well [e.g. arthropods caught by carnivorous sundews (Cross and

Bateman 2018)].

Autotomy can also be used to reduce the cost of (externally induced) injury, which can occur from predatory encounters or intraspecific competition. Possessing an injured limb makes an individual susceptible to blood loss and infection, as does autotomy (Slos et al. 2009, Yang et al. 2018). However, in terms of survival, injury is more costly than autotomy (Emberts et al. 2017a, Yang et al. 2018). This cost

differential makes it possible for individuals to gain a survival benefit by autotomizing their injured body parts, and experimental manipulations have found support for this hypothesis (Emberts et al. 2017a). The survival difference that comes with autotomizing injured body parts at a predetermined breakage plane is likely due to (1) a reduction in the amount of blood that is lost, and/or (2) having a less-compromised immune system.

Previous research has found support for both these mechanisms: blood loss following autotomy is negligible (Wake and Dresner 1967, Foelix 1996, Lesiuk and Drewes 1999,

Wilkie 2001) and the immune system of recently autotomized individuals is less compromised than those that have been injured (Yang et al. 2018). Envenomation is another source of injury, during which the predator punctures the outer layer of its prey and injects it with a toxin to debilitate or kill. If envenomation occurs on an autotomizable body part, individuals of several taxa have been observed autotomizing these compromised appendages, resulting in their survival [crabs (Muscatine and

Lenhoff 1974); spiders (Eisner and Camazine 1983); grasshoppers (Ortego and Bowers

1996)]. In addition to the survival benefit that comes with autotomizing injured body parts, there may be other benefits. For example, autotomy of injured body parts may reduce the metabolic cost of possessing a non-functional appendage. The number and diversity of organisms that have been observed autotomizing an injured body part suggests that this is a widespread benefit [e.g. lizards (Elwood et al. 2012), sea stars

(Glynn 1982, Bingham et al. 2000, Ramsay et al. 2001), true bugs (Emberts et al.

2017a) and crabs (McVean 1975)].

Autotomy can also be used to increase an individual’s reproductive success. In several species of arachnids, for example, males will autotomize their intromittent

organs inside a female’s reproductive tract (Fromhage and Schneider 2006, Uhl et al.

2010). These intromittent organs are not regenerated, but such genital mutilation is generally considered to be advantageous in these species because of low female encounter rates (Uhl et al. 2010). In some cases, the autotomized limb functions as a copulatory plug, temporarily preventing other males from mating with the female. In other cases, the autotomized structure plugs the female’s sperm storage site

(Berendonck and Greven 2000, Snow et al. 2006). In either case, successfully plugging the reproductive tract can increase a male’s fertilization success (Snow et al. 2006).

Autotomy can also be used as a component of nuptial gifts (Ghislandi et al. 2015). In

Pisaura mirabilis, a nuptial-gift-giving spider, males have been observed autotomizing their limbs and including them in their nuptial gifts, making the gifts larger (Ghislandi et al. 2015). Males that provide females with larger gifts mate for a longer duration and have higher fertilization success (Stalhandske 2001). Since autotomy can be costly, especially in the case of autotomizing intromittent organs (i.e. complete loss may result in the male becoming functionally sterile, but see Snow et al. 2006), future studies should investigate the scenarios in which individuals decide to autotomize their limbs in these mating contexts.

Given that autotomy has multiple benefits, it is important that we avoid the assumption that anti-predation is the sole, or even the primary, benefit of autotomy.

Future studies should explicitly test for, and, ideally, characterize which benefits apply to specific species. Moreover, should autotomy be beneficial in multiple contexts, studies should seek to approximate the ecological relevance of each benefit. In so doing, we would gain a better understanding of how autotomy is selected for and maintained

within populations. One way to test amongst different autotomy benefits is through predator exclusion. Using this method Maginnis (2008) found that approximately 50% of total limbs lost in the stick insect Didymuria violescens was not due to predation, and postulates that these limbs were autotomized to escape non-predatory entrapment.

Costs of Autotomy

The idea that losing body parts comes with negative consequences is intuitive and has been widely studied (reviewed in Maginnis 2006, Fleming et al. 2007). One reason that the negative consequences of autotomy have been so well studied is simply logistical feasibility. Researchers can control for aspects such as the time since autotomy, the conditions under which appendage loss occurred, and the force applied to release the limb. Despite being well studied, more recent research on the costs of autotomy has increased the resolution, explanatory power, and implications of these consequences. For example, studies have begun to follow autotomized individuals for longer periods of time, which has allowed researchers to investigate whether the costs of autotomy are mitigated (e.g. via regeneration or muscular compensation) or maintained over time (e.g. Lin et al. 2017). Since the costs of autotomy and regeneration have been thoroughly reviewed previously (Maginnis 2006, Fleming et al.

2007, Bateman and Fleming 2009, Higham et al. 2013), we aim here to build upon these previous reviews by summarizing new findings and novel approaches to studying the negative consequences of autotomy.

Fundamentally, fitness costs of autotomy can be seen as either direct costs to survival or direct costs to reproduction. Even though these categories can overlap or be correlated, we use them here to convey efficiently the new available information. Below we discuss how more proximate costs contribute to these categories.

Costs of autotomy on locomotion have been studied extensively. Leg and tail loss have been shown to reduce the locomotor performance (including velocity, acceleration, and endurance), as well as stability and control (maneuverability) of an animal running, jumping, flying or swimming (reviewed in Maginnis 2006, Fleming et al.

2007). This reduction in locomotion can compromise the chances of successfully escaping a future encounter with a predator, which indirectly compromises survival, as well as other life-history traits, such as foraging and reproduction. Recent research has built upon these findings by investigating the kinematic and morphological mechanisms that decrease locomotor performance after autotomy. For instance, Jagnandan and

Higham (2017) experimentally demonstrated that changes in the locomotion of lizards after tail loss were due to the absence of lateral undulations of the tail, rather than the loss of body mass per se or the anterior shift in the center of mass. Additionally, in

Anolis carolinensis, tail autotomy affected their in-air stability while jumping (Gillis et al.

2009). Context- and substrate-dependent effects of autotomy on locomotion have also been demonstrated. Examples include the width of the surface on which A. carolinensis could run (Hsieh 2016), the degree of surface incline while moving in cellar spiders

Pholcus manueli (Gerald et al. 2017) and fiddler crabs Uca pugilator (Gerald and

Thiesen 2014), as well as the three-dimensional substrate complexity which affects movement in the harvestman taxa Leiobunum (Houghton et al. 2011), Holmbergiana weyemberghi (Escalante et al. 2013), and Prionostemma (Domínguez et al. 2016).

Lastly, some studies have included the long-term monitoring of animals post-autotomy.

Testing locomotor performance repeatedly has allowed us to ask questions about potential recovery from autotomy, as well as the influence of regeneration. For instance,

limb kinematics and ground reaction forces changed immediately after tail autotomy in the leopard gecko (Eublepharis macularius), but these geckos recovered to initial pre- autotomy levels over the course of 22 weeks (Jagnandan et al. 2014) as the tail was regenerated. A similar pattern of recovery after regeneration was observed in the lacertid lizard Psammodromus algirus (Zamora-Camacho et al. 2016). In Anolis carolinensis some individuals recovered initial in-air stability over the course of five weeks post tail autotomy (Kuo et al. 2012). Finally, Prionostemma harvestmen recovered pre-autotomy locomotor performance in a much shorter time frame of 24 h (I.

Escalante, unpublished data). This latter example highlights the ways animals can recover locomotor performance in the absence of regeneration, mostly by modifying kinematic features of movement.

Recent work has also studied the costs of autotomy on physiology and the energetics of locomotion, with some studies showing costs and others finding no costs.

For example, autotomy was associated with short-term changes in cardiac output in blue crabs Calinectes sapidus (Mcgaw 2006), an increase in the metabolic costs of locomotion (CO2 emissions) in crickets Gryllus bimaculatus (Fleming and Bateman

2007), and an increase in standard metabolic rate in lizards Liolaemus belli (Naya et al.

2007). Some studies, however, have found no costs of autotomy on resting metabolic rates. For instance, Fleming et al. (2009) found lower CO2 production during exercise by geckos (Lygodactylus capensis) after autotomy, which was potentially associated with the loss of tissue. Starostová et al. (2017) found no differences in the resting metabolic rates of intact and autotomized geckos (Paroedura picta) immediately after autotomy or over the 22-week regeneration period. Additionally, the temperature

regulation of lizards (Psammodromus algirus) did not change after tail autotomy

(Zamora-Camacho et al. 2015). One additional physiological approach to quantifying the costs of autotomy involves metabolites, and cellular and histological processes. For instance, biochemical changes at the cellular level were studied in the Chinese mitten crab (Eriocheir sinensis), where Yang et al. (2018) found an increase in the concentration of several metabolic compounds after induced cheliped autotomy. This was suggested to be an efficient response to trauma when compared to regular ablation, as autotomized individuals recovered initial levels earlier than ablated individuals (Yang et al. 2018). Regeneration after autotomy also has costs (Maginnis

2006). In lizards, regenerated tails have different lipid and protein content, as well as greater amounts of skeleton and muscle, than original tails (Boozalis et al. 2012,

Russell et al. 2015). This redistribution of resources, particularly protein, has been shown to affect the digestive performance (gut passage time) in Podarcis erhardii lizards (Sagonas et al. 2017), and shell growth in Satsuma caliginosa land snails (Hoso

2012). Finally, the nervous system and associated histological mechanics of regeneration after autotomy were recently investigated in Coscinasterias muricata sea stars (Byrne et al. 2019). The authors showed that glia-like cells and the rapid arrival of migratory cells through haemal and coelomic compartments suggest that the autotomy plane is adapted to promote wound healing and regeneration (Byrne et al. 2019). These studies used careful experimental designs to control for the potential confounding factors of stress and injury. Nonetheless, we consider it imperative that future research explicitly aims to tease apart the actual effect of autotomy or regeneration, rather than

the injury, recovery, experience or even compensation, on these physiological and biochemical proxies.

A novel approach to the costs of autotomy has been studying its effect on disease and parasite loads. For instance, a reduced immune response and antioxidant defenses were recorded in Lestes viridis damselfly larvae after lamellae autotomy, an appendage used for locomotion and breathing (Slos et al. 2009). Additionally, increased mortality attributed to parasitoids after autotomy was found in two Parapodisima grasshopper species (Miura and Ohsaki 2015). In three species of Sceloporus lizards, individuals with a regenerated tail had higher ectoparasite loads than did intact individuals (Argaez et al. 2018), although it is challenging to tease apart the effects of autotomy from regeneration in this case.

Another new focus has been correlating autotomy with behavioral syndromes (or personalities), as well as with potential compensatory strategies. For example, in Cuban anole lizards (Anolis sagrei), individuals with a higher tendency to explore (‘bold’) were more likely to autotomize their tail than ‘shy’ individuals (Kuo et al. 2015). A similar pattern was found in damselfly larvae of Ischnura pumilio, in which individuals with increased risk-taking behavior also had a higher probability of autotomizing their caudal lamellae (Delnat et al. 2017). These findings support the hypothesis that animals rely on autotomy as a defense at the individual level with much context-based variation.

Additionally, these findings suggest ways that animals may incorporate a cost/benefit when deciding to induce autotomy

Long-term and interdisciplinary approaches to studying autotomy are also emerging. For instance, a seven-year long mark–recapture study coupled with statistical

modelling quantified the costs of autotomy on survival in a lizard (Takydromus viridipunctatus) population. Because of the study’s design, these predatory pressures could be attributed to specific bird species (Lin et al. 2017). These new results provide a compelling and complete multi-component approach to understanding the costs of autotomy.

Fewer studies have explored the direct costs of autotomy on reproduction than on survival. However, recently, important contributions have been made in understanding how autotomy affects several traits and stages associated with animal reproduction. Regarding initial stages, male fiddler crabs (Uca mjoebergi) that fully regenerated their major claw after autotomy were less likely to hold and defend territories than intact individuals (Reaney et al. 2008). Moreover, autotomy, with or without regeneration, has repeatedly been found to decrease an individual’s probability of winning intraspecific fights in the context of reproduction (Smith 1992, Martín and

Salvador 1993, Abello et al. 1994, Reaney et al. 2008, Daleo et al. 2009, Wada 2016,

Yasuda and Koga 2016, Emberts et al. 2018). In many cases these autotomizable appendages are used directly during intraspecific interactions (e.g. a crab’s claw) so the loss of the appendage comes with direct costs (Abello et al. 1994). However, in other cases, the autotomizable appendage is not directly involved in agonistic interactions

(e.g. a lizard’s tail), but its absence still decreases the individual’s probability of winning

(Martín and Salvador 1993). These patterns could be due to other marginal costs associated with losing an appendage, such as reduced locomotive ability and/or higher predatory risk aversion, which could be enough to result in a decrease in fighting ability.

Alternatively, if size is used to assess fighting ability and the presence of the

autotomizable appendage makes an individual appear larger, autotomy may compromise social/fighting status (Fox et al.1990).

In addition to these fighting costs, autotomy has been shown to influence courting and mating behavior. Behavioral compensation in courting effort post-autotomy has been recorded in male Dianemobius nigrofasciatus crickets, which increase their calling behavior after limb autotomy (Matsuoka et al. 2011). On the other hand, the loss of even one of the two pedipalps (appendages used for courtship and sperm transfer) in males reduced the intensity of courtship of the wolf spider Pardosa milvina (Lynam et al.

2006). Limb-autotomized male Menochilus sexmaculatus ladybird beetles experienced a delayed mating start and duration (Shandilya et al. 2018). Moreover, mating success was found to be lower in autotomized males of the wolf spider Schizocosa ocreata

(Taylor et al. 2008), as well as in the cactus bug Narnia femorata (Emberts et al. 2018).

In the cricket Gryllus bimaculatus experimental pairings when the female was missing either a middle or hind leg were less likely to transfer sperm (Bateman and Fleming

2006a). This was likely due to the inability of the female to mount the male properly.

In terms of fecundity and offspring survival, few studies have been able to provide evidence of the effects of autotomy. In the ladybird beetle Menochilus sexmaculatus, egg sacs fertilized by limb-autotomized males had lower fecundity and a smaller per cent of egg viability than those fertilized by intact individuals (Shandilya et al. 2018). Whether this pattern is mediated by female choice or is a byproduct of male condition is unknown, making this a topic that deserves further investigation. On the other hand, males that autotomized as juveniles (without the ability to regenerate)

produced more offspring than intact males in the cactus bug N. femorata (Joseph et al.

2018).

Overall, recent findings indicate that autotomy can have direct costs on one or more stages of animal reproduction. The absence of effect on a single trait or stage does not necessarily indicate that reproduction is unaffected in that species.

Consequently, we urge researchers to cover more than one stage in each taxon (i.e. access to mates, courtship, mating, and fecundity) to better understand if missing body parts compromises any measure of fitness.

Economic Theory of Autotomy: Predicting When an Individual Should Autotomize

Theoretical investigations of the decisions individuals make to escape a potentially agonistic interaction are common. However, most of this work investigates flight initiation distance (FID) in response to an approaching predator (Ydenberg and Dill

1986, Cooper and Frederick 2007, Cooper and Frederick 2010) and few theoretical studies consider autotomy, despite previous calls for action (Juanes and Smith 1995).

Theoretical models, even heuristic ones, are useful because they allow us to generate specific predictions and require us to dictate specific underlying assumptions, which provides directions for future empirical work.

Here, we modify the economic theory of escape to generate an economic theory of autotomy. In ‘the economics of fleeing from predators’, Ydenberg and Dill (1986) develop two hypotheses governing when an individual should flee. The first is based purely on detection of a predator, and predicts that an individual flees as soon as the predator is detected, whereas the second hypothesis is one of choice: given that the individual detects the predator it then chooses whether it should stay or flee. Modified for an autotomy framework, the first hypothesis would predict that individuals initiate

autotomy as soon as they are entrapped. The second hypothesis is still one of choice; given that the individual is entrapped it then chooses whether to struggle (e.g. kick, fight back, release chemical defenses) or autotomize to escape (discussed in Wasson and

Lyon 2005). For both hypotheses we can generate predictions about when an individual should autotomize.

The first hypothesis predicts that individuals should initiate autotomy as quickly as they can. Therefore, any differences in the latency to autotomize should be explained by differences in the amount of time it takes an individual to physically perform autotomy. Factors that contribute to an individual’s ability to perform autotomy include:

(1) the amount of force required to break an autotomy fracture plane, and (2) the amount of time it takes an individual to generate that amount of force. Thus, this hypothesis can be considered one of morphological and physiological constraints.

Previous studies have investigated the amount of force it takes to break an autotomy fracture plane in both vertebrates [lizards (Fox et al. 1994, Fox et al. 1998)] and invertebrates [damselflies (Gleason et al. 2014), starfish (Marrs et al. 2000), crabs

(Prestholdt et al. 2018)]. However, these studies often fail to identify the amount of time it takes an organism to generate the same amount of force and assume that the amount of force required to perform autotomy positively correlates with the latency to autotomize. Future work should test this assumption.

It is important to note that the no-choice hypothesis does not take context into consideration. Consequently, this hypothesis predicts that an individual should autotomize a limb entrapped by a predator just as quickly as a limb stuck in tree sap.

Morphology and physiology certainly contribute to the amount of time it takes an

individual to autotomize a limb; however, what differentiates the no-choice hypothesis from the behavioral-choice hypothesis is that the latter predicts that behavior contributes to most of the variation in the latency to autotomize. Under the behavioral-choice hypothesis, context matters. For example, how important is the limb to an individual’s future reproductive success and survival, will the limb be regenerated, and how dangerous is the current situation (e.g. entrapped in tree sap versus a predator)?

Through an economic lens, determining when an individual should behaviorally autotomize simply becomes a cost/benefit analysis. That is to say, at any given moment

(in this case, time, t; Fig. 1-4) an individual should perform the action that is the least costly. Therefore, an individual should not autotomize if the cost of autotomy is greater than the cost of retaining the compromised limb (Fig. 1-4; time < t). However, once the cost of autotomy is less than the cost of retaining a compromised limb (Fig. 1-4A; time > t), then the individual should readily autotomize. Consequently, this economic model predicts that the latency to autotomize is positively correlated with the cost of autotomy

(given that the benefits are the same; Fig. 1-4B), which formalizes multiple hypotheses found throughout the autotomy literature (Wood and Wood 1932, Robinson et al. 1970,

Arnold 1984, Guffey 1998, Fox et al. 1998, Pears et al. 2018). Note that if an organism is capable of regeneration, then the costs of autotomy include the costs and benefits of regeneration. Thus, depending on whether regeneration can reduce the net cost of autotomy or exaggerates the net cost of autotomy (which can potentially occur under obligatory regeneration, as discussed in Maginnis 2006) this economic model would predict the organism to autotomize more quickly or more slowly, respectively.

A key assumption of this economic model is that the probability of successfully using autotomy decreases with time (Fig. 1-4). For example, we assume that an individual who autotomizes their limb within 1 s has a better probability of surviving a predation event than an individual who waits 60 s. Although we believe this to be a safe assumption, the shape of the relationship between latency to autotomize and the probability of successfully using autotomy remains unclear (e.g. linear, decelerating power, accelerating power, or logistic). Another assumption of this economic model is that all individuals perceive that their limb is compromised (e.g. entrapped, injured) instantaneously. Future research should investigate the validity of these assumptions.

We also encourage others to develop and expand upon this and/or other autotomy models to help us understand what drives behavioral variation in the latency to autotomize both within and among species.

Implications of Autotomy on Organismal and Environmental Interactions

Predator–prey Interactions

When autotomy is used in an anti-predatory context, it influences the current predator–prey interaction. Most notably, once an organism autotomizes, it presents the predator with several new choices. The predator could choose to (1) release the autotomizable body part and continue pursuing the prey, (2) handle and consume the autotomized body part, then continue pursuing the prey, or (3) handle and consume the autotomized body part and not continue to pursue the prey. Much remains unknown about these choices because studies often focus on the prey species (e.g. did the prey use autotomy, did autotomy enable escape?). However, a handful of studies have provided some insights into the predator’s actions. When investigating the role of post- autotomy appendage movement, Dial and Fitzpatrick (1983) showed that a predatory

cat ignored the autotomized tails of Anolis carolinensis, but attacked (i.e. handled) the autotomized tails of Scincella lateralis. Both species exhibit post-autotomy appendage movement, but the movement exhibited by S. lateralis is more vigorous, and ultimately more successful at distracting the predator (Dial and Fitzpatrick, 1983). In predator–prey interactions between different species of crabs, approximately 85% of the crab predators handled and consumed the autotomized claw of their prey instead of continuing pursuit (Wasson et al. 2002). In the remaining 15%, the predator spent time handling the autotomized appendage, but continued pursuit of the prey (Wasson et al.

2002). Moreover, in predation trials between scorpions (predator) and wolf spiders

(prey), scorpions will consume autotomized limbs (Punzo 1997). Consumption of the autotomized appendage is often implicit in autotomy studies (e.g. Congdon et al. 1974), but future studies should explicitly state whether the appendage was consumed and whether the predator continued to pursue their prey. Future studies should also investigate the factors that contribute to these predator decisions. Some factors that could potentially influence a predator’s decision include, the size of the autotomizable appendage (i.e. the meal at hand), the predator’s level of hunger, and the distance an organism flees after autotomizing their limb. In addition to autotomy influencing the current predator–prey interaction, the absence of an autotomizable appendage can also influence future predator–prey interactions. For example, individuals without their autotomizable body parts are less successful at escaping predators (e.g. Congdon et al.

1974, Stoks 1998, Downes and Shine 2001, Bateman and Fleming 2006b, as discussed in Section IV.1). However, despite predators being more successful at capturing organisms missing an autotomizable appendage, there is no evidence to suggest that

predators differentially pursue such prey (Congdon et al. 1974, Lancaster and Wise

1996, Stoks 1998).

Loss of an autotomizable appendage can also affect an individual’s foraging behavior and feeding success (i.e. the influence of autotomy on predator–prey interactions when the organism that autotomizes is also a predator). This has mostly been studied in crustaceans because their autotomizable claws are directly associated with foraging, and these studies have frequently found that the loss of a claw reduces foraging efficiency (Smith and Hines 1991, Davis et al. 2005, Patterson et al. 2009,

Flynn et al. 2015, but see Smith and Hines 1991, de Oliveira et al. 2015). Moreover, autotomized crustaceans feed upon smaller and more easily attainable prey (e.g. prey with reduced shell thickness) when compared to intact individuals (Davis et al. 2005,

Flynn et al. 2015). These altered foraging patterns also extend beyond crustaceans.

Keeled earless lizards, Holbrookia propinqua, without their tail decrease their foraging effort (Cooper 2003), damselfly larvae without their autotomizable tail lamellae are less successful at capturing prey (Stoks 1998), and wolf spiders missing a leg prey upon smaller organisms (Brueseke et al. 2001). These altered patterns of foraging behavior and feeding success highlight that autotomy can potentially have a cascading effect on population and community dynamics, but the magnitude of these effects remain largely unquantified. Even less is known about the implications of regeneration on these effects, and future work is merited.

Intraspecific Competition

Not only can autotomy result from intraspecific competition (Van Buskirk and

Smith 1991, Juanes and Smith 1995, Itescu et al. 2017), but the previous loss of an autotomizable appendage can also alter the dynamics of these interactions. Several

studies have found that the absence of an autotomizable appendage decreases an individual’s probability of winning intraspecific fights (discussed in Section IV.2).

Although a decrease in fighting ability after autotomy is clear, there is variation in the degree to which organisms alter their behaviors associated with these interactions.

Some studies have found that organisms missing an appendage change their fighting behavior by (1) avoiding antagonistic interactions, or (2) altering how they engage in these interactions. In fiddler crabs, for example, males that have autotomized their major claw search for vacant mating burrows instead of fighting for occupied ones

(Booksmythe et al. 2010). Moreover, other crustaceans have been observed implementing fighting tactics that conceal the loss of their weapon (O’Neill and Cobb

1979, Berzins and Caldwell 1983). However, some studies have found that organisms missing an autotomizable appendage behave in a similar manner to their intact counterparts (Maginnis et al. 2015, Yasuda and Koga 2016); for example, in the leaf- footed cactus bug N. femorata, individuals missing their weapons are just as likely to engage in intraspecific interactions as are intact individuals, and weaponless individuals still try to use their weapon, behaving as if the weapon was still present (Emberts et al.

2018). The factors that drive such contrasting responses across species remains unclear.

Movement and Habitat Selection

Movement and habitat choice can have numerous consequences on population and community dynamics, including foraging and reproduction, and organisms missing an autotomizable appendage often use their habitat differently (e.g. Martín et al. 1992,

Houghton et al. 2011). Most notably, they are more risk averse in their habitat use. This often manifests itself as a reduction in the amount of time spent in open areas (Martín et

al. 1992, Salvador et al. 1995, Stoks 1999, Downes and Shine 2001, Cooper 2003,

Cooper 2007, Bateman and Fleming 2006c, Cooper and Wilson 2010). However, a few studies have found no notable difference in exposure time following autotomy [lizards

(McConnachie and Whiting 2003); crickets (Matsuoka et al. 2011)]. Loss of an autotomizable appendage also appears to decrease overall activity in some species

(Salvador et al. 1995, Martín and Salvador 1997, Downes and Shine 2001, Cooper

2007), but other studies have also found that this is not consistent across taxa (Cooper

2003, McConnachie and Whiting 2003). This across- and within-species variation in activity levels following autotomy might be explained by context dependency. The absence of an autotomizable appendage – the caudal lamellae – in damselfly larvae, for example, does not influence activity levels when a predator is present, but in the absence of predation individuals that have lost their autotomizable appendage show decreased activity (Stoks 1998). Loss of an autotomizable tail also decreases home- range size in both the male Iberian rock lizard and the long-tailed lizard (Salvador et al.

1995, Martín and Salvador 1997), which affects the number and identity of conspecifics that an individual interacts with. Future studies should continue to investigate the effects of autotomy on population and community dynamics, while also considering the implications of regeneration.

Applications of Autotomy Research

The study of autotomy has provided insights into fishery management, robotics, and conservation biology. Of all the animals that can autotomize, Crustacean autotomy currently has the most commercial applications. To provide one example, approximately

10.5 million stone crabs (Menippe) are caught, declawed, and then released in Florida

(United States) each year (Muller et al. 2006). Because stone crabs can regenerate

their claws (Savage and Sullivan 1978), this practice has the potential to create a sustainable fishery. However, research has shown that manually removing both claws reduces laboratory survival rates by almost 50% (Davis et al. 1978). Furthermore, inducing claw autotomy, as opposed to manually declawing, reduces stress responses and feeding suppression (Patterson et al. 2007). Thus, inducing autotomy of a single claw may be a more sustainable practice. Autotomy has also been considered in robotics as well. For instance, Wilshin et al. (2018) studied the postural and kinematic adjustments of wolf spiders after autotomy and outlined how such knowledge could be used to improve robotic design. Moreover, engineers have started incorporating similar biomimetic and bio-inspired designs for limb loss compensation into their research

(Cully et al. 2015). In terms of conservation biology, autotomy has the potential to show how animals are affected and respond to environmental change. For example, research on damselfly larvae has explored the effects of pesticides, changing temperatures, and competition on the incidence of autotomy (Op de Beeck et al. 2018, Janssens et al.

2018). Future studies on autotomy should continue to outline how the work can be translated into other areas of research, particularly those with more direct applications.

Future Directions

Despite being studied for over a century, fundamental questions about autotomy remain unanswered. For example, how does autotomy evolve? Future studies should investigate how populations can go from being unable to drop their appendage to being able to drop their appendage quickly enough to escape the grasp of a predator. One way to approach this is to use phylogenetic comparative methods to estimate the ancestral rate of autotomy at its origins. If the ancestral rate of autotomy is slow it would support the intermediate-step hypothesis, whereas if the rate of autotomy was rapid it

would support the fast-latency hypothesis. However, there is currently no evidence to support either hypothesis and these hypotheses largely ignore how the morphological component – the autotomy fracture plane – evolves. We have also provided evidence that post-autotomy appendage movement, a form of autotomizable limb elaboration, has evolved independently multiple times (Appendix A). Such convergent evolution suggests that autotomizable limbs are under selection to increase the efficacy or efficiency of autotomy. Thus, it might be most biologically meaningful to investigate the evolution of autotomy and autotomizable limbs simultaneously, which we collectively refer to as the autotomy phenotype.

In addition to questions about evolution, the effects of autotomy on population and community dynamics should also be investigated more explicitly. For example, is a reduction in home-range size a common consequence of autotomy? If so, what is the magnitude of this reduction and what are its repercussions? Future works should also explicitly investigate how autotomy influences predator–prey interactions, specifically when the organism that autotomizes is also a predator. We noted that appendage loss influences an organism’s prey preference in some cases (i.e. autotomized individuals take smaller prey and prey that is easier to handle), but it is unclear if this change has any effect on community dynamics. For example, does having predators that have lost their autotomizable limb alter prey-capture rates, and thus community-wide dynamics?

Conclusions

Conclusion 1. There are multiple independent origins of autotomy, as well as secondary losses, throughout Animalia. Two main hypotheses have been put forward to explain the origins of sacrificing a limb to escape predation, the intermediate-step hypothesis and the fast-latency hypothesis.

Conclusion 2. Autotomizable appendages are often elaborate. Examples of autotomizable limb elaboration include bright coloration, elongation, and post-autotomy limb movement. Such elaboration has likely been selected for and/or maintained to increase the efficacy or efficiency of autotomy.

Conclusion 3. There are multiple benefits associated with autotomy, including: escaping predation, escaping non-predatory entrapment, reducing the cost of injury, and increasing reproductive success. Costs of autotomy also vary among organisms.

Given this variation, we modified the economic theory of escape to generate the economic theory of autotomy, which makes predictions about when an individual should autotomize.

Conclusion 4. The loss of an autotomizable appendage can have a diversity of consequences on population and community dynamics. Organisms missing their autotomizable appendage generally have decreased foraging ability, are less successful at winning intraspecific fights, and are more risk averse in their habitat choice.

Conclusion 5. Future research on autotomy should focus on understanding how the autotomy phenotype evolves, demonstrating the species-specific benefits and costs associated with autotomy, discerning the ecological relevance of those costs and benefits (e.g. the proportion of autotomy events used to escape predation versus another benefit), and quantifying the effects of appendage loss (via autotomy) on population and community dynamics.

y E

c A n l a a

C i n a t c o d

o n e h s t

B p o p e a u r h p l e r r l e l o i o t y o o d i a o c o r m h g a d t a a z e c i n a o M a n h s C a N c a r o i e U r t r r th r a h B o t C o n a o in e c i h r p h a P t m h o t s l a r a a e l d G h o a ty ch t a ra H o a l e e r P if m d th i at a da ch a n ti o G ec Ec rd n h at ho ino a rt de O da X rm mi en at ye aco a ic elo D ora mo oph rph lid a Sca Cni da dari atoi a Nem P lacoz rada oa Tardig

Porifera Onychophora

Ctenophora Arthropoda

Figure 1-1. Autotomy has evolved independently multiple times throughout Animalia. Stochastic character simulations conducted for this chapter revealed that autotomy has at least nine independent origins (see Appendix A). Each lineage shown in red on this phylogenetic tree, which is modified from Dunn et al. (2014), has at least one species that can induce autotomy. Silhouette images, modified from www.phylopic.org, are visual aids for selected organisms that autotomize, as well as their autotomizable appendages. Silhouette images are not to scale.

A B

Figure 1-2. Autotomizable appendages are often elaborate such as the A) brightly colored tail of Morethia ruficauda or B) the elongated tail of Lialis burtonis. Yet, much remains unknown about the evolution of these autotomizable appendages. Photos courtesy of Damian Lettoof.

A B

Figure 1-3. There are multiple benefits associated with autotomy. Survival benefits include escaping predation, escaping non-predatory entrapment, and reducing the cost of injury. A) A jumping spider holding the autotomized limb of an orthopteran. Photo courtesy of Ummat Somjee. B) A coreid entrapped in its previous molt. Photos courtesy of author.

) e

A m B

(

m o t t

t HC

o t N t

a L

Time Cost of autotomy

Figure 1-4. Under the economic theory of autotomy A), an individual should perform the action that is the least costly. The dashed line represents the net cost of retaining a compromised limb over time, whereas the solid line represents the net cost of autotomizing a compromised limb over time. An individual should refrain from autotomizing if the cost of autotomy is greater than the cost of retaining the compromised limb (time < t). However, once the cost of autotomy is less than the cost of retaining a compromised limb (time > t), then the individual should readily autotomize. B) This model predicts that there should be a positive correlation between the cost of autotomy and the latency to autotomize, if the benefits of autotomy are held constant. tLC is the predicted latency to autotomize when autotomy costs are low; tHC is the predicted latency to autotomize when autotomy costs are high. Note that when it is more costly to autotomize, an organism should wait longer before dropping their appendage.

CHAPTER 2 COREIDAE (INSECTA: HEMIPTERA) LIMB LOSS AND AUTOTOMY*

Throughout the animal kingdom organisms can be observed releasing limbs (i.e. autotomizing) to avoid predation and entrapment, increasing their immediate survival

(Maginnis 2006, Fleming et al. 2007). However, autotomizing can also have long-term consequences related to locomotion (Bateman and Fleming 2005), reproduction (Uetz et al. 1996), and longevity (Figiel et al. 1995). Still, the ability to escape by releasing a limb appears to be strongly selected for, as autotomy has independently evolved multiple times (Wasson et al. 2002, Fleming et al. 2007). Understanding autotomy’s taxonomic diversity can provide key insights into how this extreme trait has evolved and can help us to understand the consequences of autotomy on life history, morphology, and behavior. While recent reviews have highlighted some of the taxa in which autotomy occurs (Fleming et al. 2007), gaps in our knowledge limit our understanding of autotomy’s diversity. In an attempt to fill one of these gaps, we investigate whether insects within the Coreidae clade possess the ability to autotomize their legs.

Coreids (Hemiptera: Coreidae), commonly referred to as leaf-footed bugs, are known for their diverse hind leg morphology (Baranowski and Slater 1986, Brailovsky

2006, Emlen 2008), which can include enlarged femurs, tibial flags, and spines (Fig. 2-

1). In some cases, elaborate legs are weapons (Mitchell 1980, Fujisaki 1981, Miyatake

1993, Eberhard 1998, Procter et al. 2012), but in others, their only known function is for locomotion. The elaboration and unique function of hind legs in some species, but not in

* This chapter is reprinted with permission from Emberts, Z., C. M. St. Mary, and C. W. Miller. 2016. Coreidae (Insecta: Hemiptera) limb loss and autotomy. Annals of the Entomological Society of America 109: 678-683.

others, make this group of particular interest in the evolution of autotomy. For instance, losing a leg in some species may mean the loss of a sexually selected weapon (Mitchell

1980, Fujisaki 1981, Miyatake 1993, Eberhard 1998, Procter et al. 2012), and may thus have drastic effects on mating success (Smith 1992). Therefore, identifying patterns in the ability to autotomize can provide important insights into the evolution of coreid hind leg function.

Previously, missing limbs have been noted in wild coreid populations (Eberhard

1998). However, the prevalence of limb loss has yet to be quantified. Thus, our first aim was to identify the frequency of limb loss across multiple species of wild coreid populations. Furthermore, since other Hemipterans have been observed to autotomize limbs (Bulliére and Bulliére 1985, Fleming et al. 2007), our second aim was to identify whether coreids possessed this ability as well. Personal observations had revealed that males and females, for all of the investigated species, had limbs missing at the trochanter-femur joint (Z.E., personal observation). Therefore, we hypothesized that all nine species possessed the ability to autotomize and would do so at this joint.

Methods

In an attempt to capture some of the morphological, and potentially behavioral, diversity found within the coreid clade we investigated the ability to autotomize across multiple genera (Euthochtha galeator Fabricius, Anasa andresii Guérin-Méneville,

Anasa tristis DeGeer, Narnia femorata Stål, Chelinidea vittiger McAtee, Leptoglossus phyllopus Linnaeus, Acanthocephala declivis Say, Acanthocephala terminalis Dallas, and Acanthocephala femorata Fabricius). Coreid adults were collected opportunistically throughout north central Florida in 2015, from June to September. Upon collection, limb presence and limb location (i.e. front, middle, hind) was recorded for each individual. A

sample of intact individuals, n=20 per species (10 males and 10 females), were then used to experimentally quantify autotomy, in each case within 48 hours of capture.

Autotomy induction tests were performed by gripping the insect’s hind right femur with forceps, for up to an hour, while the insect was in contact with a piece of wood (38 x 44 x 305mm). By holding their leg, individuals could only escape by releasing their trapped limb. If individuals autotomized, the location at which autotomy occurred was recorded.

Historically, the term autotomy has been used in a variety of contexts that reflect the manner in which the appendage is lost (Wood and Wood 1932, Maginnis 2006,

Fleming et al. 2007). Here, we define autotomy as limb loss that (1) morphologically occurs at a specific breakage plane, and is (2) behaviorally used to escape. Therefore, to identify the presence of autotomy in a species, at least two individuals would have to drop their entrapped hind leg. While one individual would sufficiently demonstrate the ability to induce limb loss, a second individual allows us to confirm whether limb loss is occurrent at a specific breakage plane.

In addition to autotomizing limbs to escape from predators, within arthropods, autotomy can also occur to escape from a fouled molt (Wood and Wood 1932, Smith

1992, Foelix 1996, Maginnis 2008). In the lab, limb loss is still observed in coreids, although it is not as common as limb loss in wild populations (Z.E., unpublished data).

Furthermore, when lab reared individuals are missing a limb; the missing limb is often found stuck within their previous molt (Fig. 2-2), suggesting that coreids, like other arthropods, also use autotomy to escape a fouled molt. However, it remains unclear how quickly an individual will autotomize a limb in this scenario. Nonetheless, since limb

loss in coreids is also thought to be due to predation, we also noted whether or not each individual autotomized within the first 60 seconds.

To analyze our data we used Fisher’s exact tests or, if appropriate (i.e. if the expected values for each category were 5 or more), a chi-square test. All of our analyses were conducted using R statistical software, version 3.1.2 (R Core Team

2014).

Results

Missing Legs in Wild Populations

In each species of coreid, we observed individuals (at least 1 male and 1 female) with missing legs in wild populations (Fig. 2-3). However, the frequency of missing leg varied across species (Fisher’s exact test, P = 0.002), as did the position (i.e. front, middle, and hind) of the missing legs. In general, the missing legs were disproportionately those in the hind leg position, which is also the location of the greatest limb elaborations in the species sampled (Table 2-1). L. phyllopus had the highest rate of limb absence at 21.5%, while E. galeator had the lowest observed rate at

7.9%. All of the species investigated, except A. terminalis, E. galeator, and A. tristis, had at least one individual missing two limbs, while N. femorata, C. vittiger, and L. phyllopus all had one individual missing three limbs (Fig. 2-3). We did not observe any individual to be missing four or more limbs. Furthermore, we observed a significant sex difference in the number of individuals missing limbs for A. declivis (Fisher’s exact test, P = 0.015).

Experimental Induction of Autotomy

We found that all 9 species had the ability to autotomize and did so at the trochanter-femur joint (the proximal break for all legs featured in Fig. 2-1). Behaviorally, autotomy also appeared to be stereotypic; individuals would raise their abdomen and

then drop it, shearing their limb from their body. Although autotomy was stereotypic and widespread, species varied in their expression. Remarkably, in our ‘one-hour escape from entrapment’ scenario, autotomy was observed for all 20 individuals we tested in A. andresii, A. tristis, N. femorata, C. vittiger, and L. phyllopus. While we did not observe autotomy in every individual for the other four species, our criterion of two individuals autotomizing per species was well exceeded (Table 2-3). Interestingly, autotomy was not observed for males (but it was for females) in A. terminalis and A. declivis during our

‘one-hour escape from entrapment’ scenario. Additionally, we also observed a significant sex difference in the number of individuals that experimentally autotomized for all of the Acanthocephala spp. (Table 2-3). Furthermore, two or more individuals of each species autotomized a limb at the trochanter-femur joint within 60 seconds (Table

2-4). However, in this scenario, we did not observe any sex differences in the number of individuals that experimentally autotomized (Table 2-4).

Discussion

We found that all nine species of coreids autotomized, and these results were consistent whether we looked at autotomy within 60 seconds or within an hour. While the ability to autotomize was found within each species, we also observed sex differences in the propensity to autotomize for A. terminalis, A. femorata, and A. declivis. In fact, we did not experimentally observe a single male autotomizing for A. terminalis nor A. declivis. However, wild caught specimens revealed missing limbs for both males and females for every species investigated. Additionally, in wild coreid populations, the frequency of missing limbs varied across species (Fig. 2-3), within species (e.g. sex difference in A. declivis; Fig. 2-3), and at the individual level (i.e. by

limb location; Table 2-1). These results set the stage for further study of autotomy in this clade and for comparative work with other clades.

Autotomy and Sexually Selected Weaponry

Some coreids (e.g. N. femorata, A. femorata, and A. declivis), with sexually dimorphic hind femurs, have been observed using their legs in intraspecific competitions over territories and females (Mitchell 1980, Fujisaki 1981, Miyatake 1993, Eberhard

1998, Procter et al. 2012), suggesting that their hind legs are sexually selected weapons. Thus, by demonstrating that coreids can autotomize their legs, we have also demonstrated that at least some coreids can drop their sexually selected weapons. The ability to drop a sexually selected weapon is not uncommon: cervids shed their antlers

(Goss 1983), antilocaprids drop their horns (Kitchen 1974), and crustaceans can release their claws (Smith 1992). However, the previously mentioned taxa will regenerate their weapons if given another breeding season and/or have molts remaining in their life cycle. Adult coreids, like all hemimetabolous insects, have completed development and cannot regrow lost structures. Therefore, when a coreid adult autotomizes its weapon, the weapon is permanently lost. Sexually selected weapons are usually important to an individual's reproductive success (Anderson 1994), making a permanent loss of a weapon counter-intuitive and particularly costly to overall fitness. Thus, while the animal may use autotomy to survive an otherwise fatal event, it may not be able to successfully secure any future reproductive fitness.

If these weapons are indeed important for reproductive success, then selection should act on individuals to retain them. One way weapons could be retained is through differential autotomy. Previously, in the field cricket Gryllus bimaculatus De Geer, a female’s mating history was shown to affect the willingness to autotomize a limb that

helps with locating mates, suggesting that the reproductive cost of losing a limb can affect the propensity to autotomize (Bateman and Fleming 2006a). In principle, a similar difference in reproductive cost might explain why we did not observe any males in A. terminalis or A. declivis autotomizing. However, it is also possible that our observational period was not long enough to detect their ability to autotomize (e.g. males might autotomize after an hour). Still, our observations demonstrate that there is at least a sex difference in the propensity to autotomize in all three Acanthocephala spp. Future research should investigate the degree to which the ability and timing of autotomy correlates with sexual dimorphism and the strength of sexual selection for weaponry across the coreid clade and across other groups of animals.

Autotomy and Natural Selection

The elaborate hind legs found in many coreid species are not exclusively used as weapons in male-male competitions. Certainly, they have a locomotory function. In addition, they may reduce predation in some species and in some situations. For example, the passiflora bug, Anisoscelis flavolineata Blanchard, has brightly colored legs that extend in response to disturbances (C.W.M., personal observation). These large legs may play a role in survival by directing predators to attack their legs, as opposed to their body. This risky-decoy hypothesis (Bateman et al. 2014) becomes even more plausible now that we have confirmed that coreids autotomize. Alternatively, these elaborate hind legs could play a role in conspecific signaling. Nonetheless, since elaborate hind legs in coreids are disproportionally absent from wild populations, relative to other legs (Table 2-1), our data suggests that elaborate hind legs may be more costly to maintain. Thus, it is also possible that autotomy evolved as a cost reducing mechanism (Møller 1996). Future research should explore the role autotomy

has had in driving the form of coreid hind legs, as well as more broadly, identifying how selection differentially acts on autotomizable limbs. For example, to what degree does having the ability to autotomize a limb change the limbs form? Does having the ability to drop these limbs allow them to become more elaborate and conspicuous, or less?

Additionally, future scientific gains are likely to be made by investigating autotomy during juvenile development, where species may possess the ability to regenerate their lost limb/weapon or reallocate the energy elsewhere.

Table 2-1. Percentage of missing limbs by limb location. We compared our observed wild caught data against the null hypothesis that leg absences should be evenly distributed amongst leg positions (hind, middle, and front). In general, missing legs were not evenly distributed, as some species with elaborated hind legs were disproportionally missing hind legs.

MISSING LEG POSITION (%) species n Hind leg Middle leg Front leg P

ELABORATED HIND LEGS A. declivis 8 87.5 12.5 0.0 0.084 A. femorata 13 69.2 30.8 0.0 0.053 A. terminalis 9 77.8 22.2 0.0 0.088 N. femorata 41 75.6 19.5 4.9 < 0.001 L. phyllopus 56 26.8 42.9 30.4 0.565 E. galeator 5 60.0 20.0 20.0 1.000 C. vittiger 29 69.0 20.7 10.3 0.019 TOTAL 161 57.1 29.8 13.0 < 0.001

NON-ELABORATED LEGS A. andresii 33 45.5 42.4 12.1 0.119 A. tristis 2 50.0 0.0 50.0 1.000 TOTAL 35 45.7 40.0 14.3 0.169

Table 2-2. Data and analyses of wild caught specimens.

total number Percentage of Percentage Sex difference in the > 3 1 leg 2 legs of individuals individuals of number of Statistical Analysis Species Sex legs n missing missing with missing missing at least individuals individuals missing used to determine P missing leg(s) one leg with all legs limb(s) (P) A. declivis F 5 1 0 6 18 33.3 66.7 - A. declivis M 1 0 0 1 25 4.0 96.0 - A. declivis ALL 6 1 0 7 43 16.3 83.7 0.015 Fisher's Exact Test A. femorata F 8 0 0 8 42 19.0 81.0 - A. femorata M 3 1 0 4 48 8.3 91.7 - A. femorata ALL 11 1 0 12 90 13.3 86.7 0.238 Chi-squared Test A. terminalis F 7 0 0 7 30 23.3 76.7 - A. terminalis M 2 0 0 2 16 12.5 87.5 - A. terminalis ALL 9 0 0 9 46 19.6 80.4 0.463 Fisher's Exact Test N. femorata F 15 3 0 18 113 15.9 84.1 - N. femorata M 13 2 1 16 146 11.0 89.0 - N. femorata ALL 28 5 1 34 259 13.1 86.9 0.323 Chi-squared Test L. phyllopus F 12 4 0 16 89 18.0 82.0 - L. phyllopus M 29 2 1 32 134 23.9 76.1 - L. phyllopus ALL 41 6 1 48 223 21.5 78.5 0.377 Chi-squared Test E. galeator F 3 0 0 3 22 13.6 86.4 - E. galeator M 2 0 0 2 41 4.9 95.1 - E. galeator ALL 5 0 0 5 63 7.9 92.1 0.333 Fisher's Exact Test C. vittiger F 13 0 1 14 152 9.2 90.8 - C. vittiger M 11 1 0 12 155 7.7 92.3 - C. vittiger ALL 24 1 1 26 307 8.5 91.5 0.797 Chi-squared Test

Table 2-2. Continued A. andresii F 12 2 0 14 81 17.3 82.7 - A. andresii M 13 2 0 15 81 18.5 81.5 - A. andresii ALL 25 4 0 29 162 17.9 82.1 1.000 Chi-squared Test A. tristis F 1 0 0 1 12 8.3 91.7 - A. tristis M 1 0 0 1 11 9.1 90.9 - A. tristis ALL 2 0 0 2 23 8.7 91.3 1.000 Fisher's Exact Test

Table 2-3. Experimental autotomy data and analyses for ‘one-hour escape from entrapment’ scenario.

Number of Sex difference in the number of Species Sex n individuals that individuals that experimentally autotomized autotomized (P)

A. declivis M 10 0 - A. declivis F 10 5 - A. declivis ALL 20 5 0.0325 A. femorata M 10 2 - A. femorata F 10 8 - A. femorata ALL 20 10 0.0230 A. terminalis M 10 0 - A. terminalis F 10 9 - A. terminalis ALL 20 9 0.0001 N. femorata M 10 10 - N. femorata F 10 10 - N. femorata ALL 20 20 1.0000 L. phyllopus M 10 10 - L. phyllopus F 10 10 - L. phyllopus ALL 20 20 1.0000 E. galeator M 10 2 - E. galeator F 10 7 - E. galeator ALL 20 9 0.0698 C. vittiger M 10 10 - C. vittiger F 10 10 - C. vittiger ALL 20 20 1.0000 A. andresii M 10 10 - A. andresii F 10 10 - A. andresii ALL 20 20 1.0000 A. tristis M 10 10 - A. tristis F 10 10 - A. tristis ALL 20 20 1.0000

Table 2-4. Experimental autotomy data and analyses for ‘60 seconds escape from entrapment’ scenario.

Number of Sex difference in the number of Species Sex n individuals that individuals that experimentally autotomized autotomized (P)

A. declivis M 10 0 - A. declivis F 10 2 - A. declivis ALL 20 2 0.4737 A. femorata M 10 0 - A. femorata F 10 4 - A. femorata ALL 20 4 0.0867 A. terminalis M 10 0 - A. terminalis F 10 2 - A. terminalis ALL 20 2 0.4737 N. femorata M 10 10 - N. femorata F 10 10 - N. femorata ALL 20 20 1.0000 L. phyllopus M 10 7 - L. phyllopus F 10 6 - L. phyllopus ALL 20 13 1.0000 E. galeator M 10 2 - E. galeator F 10 2 - E. galeator ALL 20 4 1.0000 C. vittiger M 10 7 - C. vittiger F 10 3 - C. vittiger ALL 20 10 0.1789 A. andresii M 10 7 - A. andresii F 10 4 - A. andresii ALL 20 11 0.3698 A. tristis M 10 3 - A. tristis F 10 5 - A. tristis ALL 20 8 0.6499

Figure 2-1. Representative hind legs of each species in this study, to scale. Morphologically, across the coreid clade, hind legs vary in size, shape, and color. Hind legs can also vary between males (right) and females (left). To create the leg images above, legs were removed at the joint between the trochanter and femur.

Figure 2-2. Coreids can autotomize limbs to escape fouled molts. An adult Narnia femorata with its missing leg in its terminal exuviae. Photo courtesy of author.

Figure 2-3. Percentage of wild caught coreid adults observed missing legs. Exact numbers, proportions, and analyses performed can be found in Table 2-2.

CHAPTER 3 AUTOTOMIZING INJURED LIMBS INCREASES SURVIVAL*

Sacrificing a limb by self-amputation (i.e. autotomy) has evolved throughout the animal kingdom despite the enormous costs associated with this behavior (Table 3-1).

Minimally, the cost of self-amputation includes the potential loss of blood or comparable bodily fluid (Lawrence 1992, Foelix 1996, Wilkie 2001, Fleming et al. 2007) and the potential for infection (Fleming et al. 2007, Slos et al. 2009, Bely and Nyberg 2010).

Additionally, autotomy comes with costs that coincide with the function of the lost limb

(Maginnis 2006, Fleming et al. 2007, Tsurui et al. 2014, Domínguez et al. 2016). These costs can be especially substantial should they decrease an individual’s future reproductive success (Smith 1992, Maginnis 2006, Fleming et al. 2007). Yet, given autotomy’s evolutionary persistence (McVean 1982, Fleming et al. 2007, Zani 1996), the benefits must outweigh the substantial costs (Arnold 1984). Thus, to better understand how this extreme trait evolves, we must identify the adaptive benefits of self- induced limb loss.

One benefit of autotomy is its ability to help an individual escape predation. In this context, individuals use autotomy to break free from a predator’s grasp and, in some cases, to distract the predator. Predator distraction occurs when the predator spends time handling and/or consuming an autotomized limb as oppose to trying to catch the surviving individual. Post-autotomy tail movement (observed in some lizards and salamanders) exemplifies this benefit, as autotomized tails that wiggle have been

* This chapter is reprinted with permission from Emberts, Z., C. W. Miller, D. Kiehl, and C. M. St. Mary. 2017. Cut your losses: self-amputation of injured limbs increases survival. Behavioral Ecology 28: 1047- 1054.

shown to increase predator handling and consumption time, thereby allowing the individual more time to escape (Dial and Fitzpatrick 1983). While the means of predator escape can vary, the ultimate benefit has been demonstrated in numerous taxa, including lizards (Congdon et al. 1974, Downes and Shine 2001), starfish (Bingham et al. 2000), decapods (Lawton 1989, Wasson et al. 2002), spiders (Punzo 1997,

Brueseke et al. 2001), and crickets (Bateman and Fleming 2006b). However, it is important to recognize that autotomy has additional benefits beyond that of escaping predation.

Other benefits of autotomy include escaping non-predatory entrapment (Foelix

1996) and reducing the cost of envenomation (Eisner and Camazine 1983, Ortego and

Bowers 1996). Non-predatory entrapment is frequently observed in arthropods who undergo a complex molting process (Robinson et al. 1991, Juanes and Smith 1995,

Foelix 1996, Johnson and Jakob 1999, Maginnis 2008). During this process, limbs, especially elaborated and elongated ones, may get stuck and autotomy provides a viable option for escaping (Maginnis 2008). Moreover, there are also benefits of autotomy that are not related to survival. For instance, self-amputated limbs can be used as copulatory plugs to increase a male’s reproductive success (Knoflach and van

Harten 2001, Knoflach 2002, Fromhage and Schneider 2006, Snow et al. 2006, Nessler et al. 2007, Uhl et al. 2009). Therefore, when considering the evolution of autotomy in a broader context, it is critically important to separate autotomy from the assumption that its sole adaptive function is to escape predation.

Another hypothesized benefit of autotomy, one that has gone untested, is that autotomy can limit damage associated with wounded body parts. In other words, if injury

occurs on an autotomizable limb it is hypothesized that individuals can self-amputate

(autotomize) the injured limb to reduce the cost of the injury. This hypothesis is largely inspired by physiological and behavioral observations. Physiologically, self-induced injuries (i.e. injuries induced through autotomy) are thought to quickly heal (Wake and

Dresner 1967, Foelix 1996, Wilkie 2001). Therefore, if an externally-induced injury was severe, self-amputating the injured limb may reduce the loss of blood and the chance of infection, ultimately increasing survival. Still, despite the anecdotal but taxonomically widespread observance of this behavior (Table 3-1), the benefit of autotomizing in response to injury has yet to be investigated.

For this behavior to be beneficial, the cost of the injury has to exceed the cost of autotomy. Thus, our first aim is to investigate whether injury has a higher survival and/or developmental cost than autotomy. Then, we experimentally investigate whether injured individuals can reduce this cost differential by autotomizing their damaged limb (i.e. reduce the cost of injury via autotomy).

Methods

Study Organism

To investigate whether autotomy can indeed reduce the cost of injury we used the leaf-footed cactus bug, Narnia femorata (Insecta: Hemiptera: Coreidae; Figure 3-1).

Previously, N. femorata has been shown to use autotomy to escape from entrapment

(Emberts et al. 2016). Furthermore, the behavior normally occurs within 60 seconds suggesting that individuals could also use this trait to escape from predation (Emberts et al. 2016). However, the role autotomy plays in reducing the cost of injury remains unclear.

In Hemipterans, potential responses to limb injury include autotomizing the injured limb or (retaining and) regenerating it, but not both (Luscher 1948, Shaw and

Bryant 1974). In the case of autotomy, the limb is dropped at the trochanter-femur joint

(Luscher 1948, Emberts et al. 2016), a location from which regeneration has not been shown to occur (Luscher 1948, Shaw and Bryant 1974). The alternative, regeneration, is only available to individuals who retain (i.e. do not autotomize) their damaged limb and have molts remaining to regrow the lost structure (i.e. juveniles). Furthermore, the regenerative capabilities are quite limited as juveniles have only been shown to partially regenerate their tibia and tarsi (Luscher 1948, Shaw and Bryant 1974). Consequently, injury location may factor into an individual’s decision to autotomize or retain an injured limb. Additionally, for N. femorata, the loss of a male’s hind leg may have costly implications for reproductive success, as males use their hind legs in intrasexual competition (Procter et al. 2012). Thus, our study takes sex, injury location, and the ability to regenerate into consideration.

Study Design

Insect rearing

For our experiments we used first-generation lab-reared individuals. The populations were founded in November of 2015 with 29 mating pairs collected from Live

Oak, Florida (30.26°N, -83.18°W). Individuals were reared in deli cups containing Opuntia mesacantha subsp. lata cladodes (cactus pads) and fruit collected from the same location throughout the experiment. Before experimentation, individuals were reared with siblings in a greenhouse (set temperature: 21-32°C, and photoperiod:

14:10 hr L:D) until their second instar.

Experiment 1

To investigate how injury and/or autotomy affects survival and development (e.g. time to reach adulthood, regeneration, and terminal body size) we randomly assigned second instar juveniles with all of their legs to one of six treatments (final sample sizes ranged from 19 to 25 per treatment): (1) control (no injury/no autotomy), (2) experimentally induced autotomy at the trochanter-femur joint, (3) incision (i.e. cut completely through the leg) at the trochanter-femur joint (henceforth referred to as Injury

A to reflect that this experimentally induced injury occurred at the same location as self- induced autotomy), (4) incision at the femur-tibia joint (henceforth referred to as Injury

1), (5) incision through the middle of the tibia (henceforth referred to as Injury 2), and (6) incision at the tibia-tarsus joint (henceforth referred to as Injury 3; Figure 3-2). We only used individuals with all of their legs because limb loss has been shown to affect the propensity to autotomize additional limbs (Bateman and Fleming 2005). Autotomy was induced by gripping the insect’s right hind femur with reverse-action forceps while the insect was in contact with a piece of wood (38 x 44 x 305mm; Emberts et al. 2016). For a comparative baseline, individuals in the control (no injury/no autotomy) treatment underwent a sham autotomy protocol (i.e. their legs were held for a shorter amount of time (one second) with reverse-action forceps), but were not induced to autotomize. For the remaining treatments (e.g. injury A, 1, 2, and 3), injury was induced with iridectomy scissors at the specified location following previous regeneration protocols in other species (Luscher 1948, Shaw and Bryant 1974). After an individual’s respective procedure, it was moved into its own deli cup and placed into an incubator

(temperature: 32°C, photoperiod: 14:10 hr L:D). As an ethical note, animals were treated as humanely as possible by inducing injury in a disinfected environment and by

providing them with species-specific optimal living conditions. Individuals were monitored daily (maximum 32 days) for developmental rate (i.e. molt timing), limb loss, and death. Upon becoming an adult, individuals were sexed, and their body and legs were photographed using a Canon EOS 50D digital camera attached to a Leica M165 C dissecting microscope. Pronotal width (PW; a body size metric) and hind leg length were measured to the nearest micrometer using ImageJ v1.46.

Experiment 2

To experimentally test if autotomizing an injured leg increases survival, we randomly assigned juveniles (second instars) into one of three treatment groups in which the right hind leg was (1) experimentally induced to autotomize (n = 38), (2) injured then experimentally induced to autotomize (n = 39), or (3) injured without experimental autotomy (n = 40). This experiment involved two stages separated by one hour. First, injury was induced in treatments 2 and 3 at the femur-tibia joint with disinfected iridectomy scissors. Individuals in treatment 1 were handled in the same manner, but injury was not induced. If unplanned autotomy occurred due to handling, the individual was removed from the experiment and replaced. In the second stage (one hour later), individuals in treatment 1 and 2 were induced to autotomize their right hind leg with reverse-action forceps, while individuals in treatment 3 underwent a sham autotomy protocol, as detailed above for Experiment 1. After treatment manipulations, each individual was placed into a separate deli cup with O. mesacantha subsp. lata fruit and water and moved into an incubator (temperature: 32°C, photoperiod: 14:10 h L:D).

Individuals were checked at 12 hr intervals over 48 hrs and survival and limb loss were recorded.

Data and Statistical Analyses

Experiment 1

To investigate the effect injury has on survivorship (live/die) we conducted planned contrasts in the context of a binary, generalized linear mixed model with family as a random factor (GLMM; logit-link function assuming a binomial distribution). Since we hypothesized that injury would have a negative effect on survivorship regardless of injury site, we contrasted all of our injury treatments (injury A, 1, 2, and 3) against the control treatment (treatment 1). Using this same approach (i.e. binary GLMM using contrasts), we also investigated the effect autotomy has on survivorship by contrasting the autotomy treatment versus the control treatment. Finally, we contrasted autotomy and injuries, to evaluate whether their effects differed. Comparable analyses were done, using a GLMM with contrasts (identity-link function assuming a Gaussian distribution), to investigate whether injury and/or autotomy affected the number of days to reach adulthood or terminal body size.

To investigate whether autotomizing an injured limb resulted in higher survivorship we compared the survival of those that autotomized their injured limbs versus those that retained them using a binary GLMM (logit-link function with assumed binomial distribution) for all three injury treatments combined (injury 1, 2, and 3). We excluded injury A from this and subsequent analyses because we could not determine whether individuals in this treatment autotomized their injured limb; as injury was induced at the same location where autotomy occurs. Similarly, to see whether autotomizing an injured limb affects the time to reach adulthood and/or terminal body size, we compared these metrics for individuals that retained their injured limbs versus those that autotomized their injured limb using a GLMM (identity-link function assuming

a Gaussian distribution). We also investigated, using a binary GLMM (logit-link function with assumed binomial distribution), whether injury site or sex could explain any variation in the propensity to autotomize.

By using landmark locations on the legs (e.g. joints), and by measuring leg length from these landmarks, we were also able to quantify regenerative ability. If there was no growth beyond an injury site, then we classified the injury as non-regenerative.

Experiment 2

Our goal with the second experiment was to compare the probability of survival for those retaining injured limbs, autotomizing injured limbs, and autotomizing uninjured limbs. However, injured individuals in the ‘retain injured limb treatment’ cannot be prevented from self-autotomizing their damaged limbs. Thus, we proceeded with a series of analyses designed to test differences across and within our treatment groups.

For these tests we used binary GLMMs (logit-link function with assumed binomial distribution) and, where relevant, used contrasts in our models to compare groups of interest for which we had developed a priori hypotheses of their relationships.

Across both experiments, sample sizes varied because we were unable to retrieve some individuals (n=15) at the end of the experiment. While there are multiple reasons that might explain how these individuals were lost, we believe that most of these individuals became buried in the soil upon death, making it extremely challenging to locate them. Still, we took the most conservative approach and excluded these individuals from our analyses. However, even making the reasonable assumption that all 15 of these individuals died, excluding or including these individuals had no effect on our conclusions.

Results

Experiment 1

We first compared survival for injured insects (Injury locations A, 1, 2, & 3), insects experiencing experimentally induced autotomy, and individuals in our control group. We found that injured insects had approximately 25% lower survival on average than those that were experimentally induced to autotomize (GLMM with a priori contrasts: 2 = 5.43, df = 1, P = 0.020, Figure 3-3). Insects that were not injured and not experimentally induced to autotomize (control group) did not differ in survival relative to the injured insects (binary GLMM with a priori contrasts: 2 = 1.64, df = 1, P = 0.200,

Figure 3-3) or those experiencing induced autotomy (binary GLMM with a priori contrasts: 2 = 0.48, df = 1, P = 0.476, Figure 3-3). When we compared terminal body size and the number of days it took to reach adulthood across our contrasted treatments we did not find any significant differences (Table 3-2).

For all of the injury treatments where autotomy was possible (Injury 1, 2, and 3), a large fraction (50.7%) of individuals responded to their injury by autotomizing. In general, individuals who autotomized their injured limb had higher survival than those that retained their injured limb (binary GLMM: 2 = 5.67, df = 1, P = 0.017), but the location of the injury also tended to affect this benefit (binary GLMM: 2 = 5.53, df = 2, P

= 0.063). In particular, autotomy of limbs injured at the femur-tibia joint (injury 1) and tibia (injury 2) led to higher survival while autotomy after an injury on the tibia-tarsus joint (injury 3) did not (Figure 3-4).

Injury location also had an effect on the propensity to autotomize (GLMM: 2 =

23.03, df = 2, P < 0.001; Figure 4). Specifically, individuals injured at the tibia-tarsus

joint (injury 3) were significantly less likely to autotomize than individuals injured at the femur-tibia joint (injury 1; GLMM: 2 = 14.79, df = 1, P < 0.001) and individuals injured through their tibia (injury 2; GLMM: 2 = 20.04, df = 1, P < 0.001; Figure 3-4). Sex did not explain any of the variation in the expression of autotomy (GLMM: 2 = 0.07, df = 1,

P = 0.793).

Of those individuals that retained their injured limb, less than half (40.6%) survived to adulthood. Compared to those that autotomized their injured limbs, those that retained them required fewer days to reach adulthood (Table 3-3) and showed some form of regeneration. Individuals who retained their injured limb in the injury 3 treatment regenerated their first tarsal segment, while individuals in the injury 2 treatment regenerated both their tibia and their first tarsal segment. The regenerated tarsi in both treatments were 55% shorter than our control (Table 3-4). Thus, we conclude that N. femorata has partial regenerative capabilities. None of the individuals that retained their injured limb in the injury 1 treatment survived to adulthood (0 out of

4), therefore we were unable to quantify the potential for regeneration from this injury location.

Experiment 2

Experiment 2 involved three treatment groups: (1) experimentally induced autotomy of a non-injured limb, (2) experimentally induced autotomy of an injured limb, and (3) injured without experimental autotomy. When we compared treatment 2 to treatment 3, we did not find that experimental autotomy of injured limbs significantly increased survival (contrasted GLMM: 2 = 1.58, df = 1, P = 0.209). However, over half

(57.50%) of the individuals in treatment 3 self-autotomized their injured limb. This

difference in behavior allowed us to additionally consider the behavioral decision to self- autotomize an injured limb. Those that self-autotomized their injured limb had higher survivorship than those that maintained their injured limb (GLMM: 2 = 6.10, df = 1, P =

0.014; Figure 3-5). Furthermore, those that self-autotomized their injured limbs and those that were experimentally induced to autotomize their injured limbs had similar survivorship (contrasted GLMM: 2 < 0.001, df = 1, P = 0.985; Figure 3-5), as we hypothesized. Thus, we compared all the individuals that autotomized their injured limb, whether it was self-induced or experimentally induced, to those individuals who maintained their injured limb; we found that those who autotomized their injured limb had significantly higher survival than those who did not (contrasted GLMM: 2 = 4.02, df

= 1, P = 0.045; Figure 3-5).

Discussion

Here, we have shown that autotomy can reduce the cost of injury. Autotomy after injury has been observed across taxa (Table 1), but the benefits of the behavior have only been assumed, not tested (Savory 1928, Lewis 1981, Glynn 1982, Bulliére and

Bulliére 1985, Johnson and Jakob 1999, Bingham et al. 2000, Ramsay et al. 2001).

Thus, this study is the first to provide evidence of a novel benefit of autotomy — reducing the (survival) cost of injury. Other widespread benefits of autotomy include escaping predation (Congdon et al. 1974, Carlberg 1986, Lawton 1989, Punzo 1997,

Bingham et al. 2000, Brueseke et al. 2001, Downes and Shine 2001, Sword 2001,

Wasson et al. 2002, Bateman and Fleming 2006b) and escaping non-predatory entrapment (Robinson et al. 1991, Juanes and Smith 1995, Foelix 1996, Johnson and

Jakob 1999, Maginnis 2008). Our results, and others, highlight that there are multiple

benefits of autotomy, which may select for and maintain the trait. As this evidence grows it becomes crucial that we abandon the assumption that autotomy’s sole, or even primary, adaptive benefit is escaping predation. By doing so, we stand to gain a more comprehensive understanding of how such an extreme trait evolves.

Another major implication of this study is that autotomy is less costly than injury, but only with respect to survival as we did not find injury to have an effect on the time to reach adulthood nor terminal body size. Regardless, these results highlight that autotomy and injury should not be considered synonymous. Specifically, autotomy is self-induced, or self-controlled (Fleming et al. 2007), injury. Recognizing this distinction is vital to understanding how autotomy can reduce the costs of injury. To elaborate, injuries, whether self-induced or externally induced, can result in blood loss and infection, both of which may ultimately result in death. Thus, there should be selection to minimize these effects, such as selection on an immune system (Medzhitov and

Janeway 1997, Cooper and Alder 2006, Cerenius and Söderhäll 2011). However, what differentiates self-controlled injury from externally induced injury is that self-controlled injury can consistently occur at a very precise location. This consistency allows selection, over time, to potentially act on morphology to reduce the risk of infection and the loss of blood. Consequently, self-induced injury at a pre-determined breakage plane may be less severe than externally induced injury. Previous studies have noted that self-induced injuries (i.e. due to autotomy) quickly seal and result in negligible amounts of blood loss (Wake and Dresner 1967, Foelix 1996, Wilkie 2001). However, the differences in blood loss and immune response between autotomy and externally

induced injury have yet to be measured. Although we have not quantified these effects, the differences in survival observed in our study likely result from such differences.

The benefit of autotomizing an injured limb appears to vary by injury location. For example, we did not find a survival benefit of autotomy if the injury occurred at the tibia- tarsus joint, our most distal injury. This lack of a benefit might explain why insects experiencing injuries at that site rarely autotomized their limb. By retaining their injured limb, individuals could regenerate part of their tarsus and have a resulting hind leg that was only 19% shorter than our comparative baseline (Table 3-4). This pattern highlights the likely trade-off individuals face between autotomizing an injured limb and retaining it.

That is to say, when autotomizing a limb does not increase survival (i.e. the severity of the injury is minimal), few individuals should autotomize their limbs. However, when autotomy does increase survival (i.e. the injury is severe) individuals should readily drop their limb, even though it comes with the cost of permanently losing their leg.

While injury location influenced the tendency to autotomize injured limbs, sex did not. In N. femorata male hind legs have been shown to function as sexually-selected weapons (Procter et al. 2012, Nolen et al. 2017). Thus, the permanent loss of a male hind leg potentially comes with a larger cost than the loss of a female hind leg. In previous studies, when the costs and benefits of autotomy differ between the sexes there is often a corresponding difference in the propensity to autotomize (Fox et al.

1998, Wasson and Lyon 2005). For our study, however, there are several possible explanations for why we did not observe such differences. First, it is possible that the loss of a hind leg comes with an equal cost to males and females, as males may compensate, behaviorally (e.g. Berzins and Caldwell 1983) and/or morphologically (e.g.

Simmons and Emlen 2006), for the loss of their weapon, an intriguing future direction.

Second, it is also possible that there was a sex difference in the propensity to autotomize (that occurred on the scale of seconds, minutes, or hours), but because of our experimental design we were unable to detect the difference. Still, even if such a difference existed, sex did not ultimately affect whether or not an individual autotomized their injured limb.

Additionally, in our study system, the consequences of autotomy and regeneration are not confounded as individuals may only autotomize or regenerate their injured limb, but not both. In other arthropods, autotomy (at a preformed breakage plane) often precedes regeneration. Therefore, in some instances, it can be challenging to differentiate consequences of regeneration from consequences of autotomy. One of these challenges is determining whether regeneration and/or autotomy alters developmental time. In arthropods, regeneration (preceded by autotomy) is often shown to increase the amount of time it takes to develop (Maginnis 2006). However, it is possible that this developmental delay is a consequence of autotomy, not regeneration.

With our study species, we are presented with a unique opportunity to separately investigate the consequences of autotomy and (independently) the consequences of regeneration. In N. femorata, autotomy had no effect on the number of days it took to reach adulthood (Table 2, Autotomy vs. Control). However, regeneration did. When comparing those that regenerated their injured limbs (without being preceded by autotomy) to those that autotomized their injured limbs (without being followed by regeneration) we found that individuals who regenerated had shorter intermolt intervals

(i.e. they developed from 3rd instars to adults more quickly). In Hemipterans, and other

arthropods, regeneration coincides with molting. Consequently, by decreasing intermolt intervals an individual may be able to replace its missing limb more quickly (Maginnis

2006). These results could be interpreted to mean that regeneration accelerates development in N. femorata. However, it is also important to note that the dataset from which we drew these conclusions inherently excluded individuals that did not survive to adulthood, and thereby disproportionally excluding individuals that retained their injured limb (Figure 3-4A). Thus, these results could also reflect that only quickly developing individuals can retain an injured limb (with subsequent regeneration) and survive until adulthood.

Our second experiment, although it did not fully demonstrate cause and effect, provides further support that autotomy of injured limbs increases survival. As with experiment 1, we found a positive association between self-autotomizing injured limbs and survival. This result could reflect (as we have postulated) that autotomizing injured limbs increases survival. However, because this result is correlative, it could also suggest that high quality individuals (i.e. those predisposed to higher survival) are more likely to autotomize their injured limb. Thus, in our second experiment, we induced autotomy to directly investigate these alternatives. We found that experimentally inducing autotomy had the same effect as self-induced autotomy. This similarity suggests that the patterns of survivorship we observed are not due to variation in individual quality, but instead stem from autotomy of injured limbs; thereby strongly supporting the hypothesis that autotomy can indeed reduce the cost of injury.

In conclusion, the results of this study are the first to provide evidence that autotomy can reduce the cost of injury. Specifically, here, we observed a survival

difference between individuals that autotomized their injured limb and those that retained it. Furthermore, in our second experiment, we observed this survival difference just 48 hours post-injury, suggesting a relatively immediate benefit to autotomizing injured limbs. However, it is also possible that autotomizing injured limbs comes with long-term benefits too. For example, if a limb is severely damaged, an individual may be able to reduce the metabolic cost of carrying around a lame limb by autotomizing it.

Moreover, if the species can regenerate, autotomizing a lame limb may promote the growth of a new, functional one. Such benefits of autotomizing injured limbs are not necessarily mutually exclusive alternatives. Instead, we hypothesize that these potential benefits may additionally contribute to the selection and maintenance of autotomy.

Thus, to get a better understand how this extreme trait evolves, we must continue to identify the adaptive benefits of self-inducing limb loss.

Table 3-1. A taxonomic overview of autotomy, including anecdotal evidence of autotomy in response to injury. Autotomizable Autotomy Autotomy in Phyla Group Citations appendage to escape response to injury Bickell-Page and Coelenterata Jellyfish Tentacles Yes - Mackie 1991 Nudibranchs Cerata Yes - Marin and Ros 2004 Mollusca Bivalves Tentacles Yes - Donovan et al. 1994 Squid Tentacles Yes - Bush 2012 Annelida Earthworms Tail Yes - Fiore et al. 2004 Legs, Savory 1928, Punzo Spiders Pedipalps Yes Yes 1997 Scorpions Tail Yes - Mattoni et al. 2015 Centipedes Legs Yes Yes Lewis 1981 McVean 1973, Arthopoda Crabs Claws, Legs Yes Yes McVean 1982 Bateman and Fleming Crickets Legs Yes - 2006a Luscher 1948, Emberts True Bugs Legs Yes Yes et al. 2016 Sea stars Arms Yes Yes Glynn 1982 Echinodermata Brittlestars Arms Yes - Wilkie 2001 Elwood et al. 2012, Lizards Tail Yes Yes Congdeon et al. 1974 Wake and Dresner Chordata Salamanders Tail Yes Yes 1967 Mice Tail skin Yes - Shargal et al. 1999

Table 3-2. Experiment 1 – Developmental differences between autotomy, injury, and our control (no autotomy/ no injury). 2 df P Days until adulthood Autotomy vs. Control 0.285 1 0.593 Injury vs. Control 1.265 1 0.261 Autotomy vs. Injury 3.37 1 0.067

Terminal body size (PW) Autotomy vs. Control 1.2 1 0.274 Injury vs. Control 0.992 1 0.319 Autotomy vs. Injury 0.223 1 0.637

Table 3-3. Experiment 1 – Developmental differences between self-autotomizing and retaining an injured limb. We investigated how injury location, the decision to autotomize, and their interaction affected the number of days it took a juvenile to reach adulthood and terminal body size. Means and standard errors are reported in Table 3-5. 2 df P Days until adulthood Autotomy 9.085 1 0.003 Injury location 3.445 2 0.179 Autotomy x Injury location 16.404 1 < 0.001

Terminal body size (PW) Autotomy 3.554 1 0.059 Injury location 1.350 2 0.509 Autotomy x Injury location 4.171 1 0.041

Table 3-4. Narnia femorata regenerative capabilities. The mean and standard error of adult (i.e. “final”) legs are provided for each leg segment and the total leg length. Since individuals in the injury 2 and 3 treatment both had their tarsus completely removed, tarsus growth in these treatments indicates some regenerative capabilities. None of the individuals that retained their injured limb in the injury 1 treatment survived to adulthood, thus we were unable to quantify the regenerative capability for this injury location. Injury 2 Injury 3 Control (Treatment 1) Hind Femur Length (mm) 6.31 ± 0.13 6.08 ± 0.10 6.54 ± 0.15 Hind Tibia Length (mm) 4.06 ± 0.34 5.43 ± 0.15 6.80 ± 0.15 Hind Tarsus Length (mm) 0.86 ± 0.08 0.87 ± 0.06 1.91 ± 0.04 Total Leg Length (mm) 11.22 ± 0.49 12.37 ± 0.25 15.25 ± 0.33

Table 3-5. The effect of autotomizing or retaining an injured limb on development. Means and standard error reported for terminal body size (PW) and the number of days it took to reach adulthood. Behavioral Terminal body Days until decision regarding size (PW) adulthood injured limb Autotomize 4.116 ± 0.116 15.053 ± 0727 All Injuries Retain 4.450 ± 0.115 11.667 ±0.710 Autotomize 4.098 ± 0.174 15.000 ± 1.035 Injury 1 Retain N/A N/A Autotomize 4.213 ± 0.150 14.200 ± 0.611 Injury 2 Retain 4.353 ± 0.301 13.667 ± 1.856 Autotomize 3.140 ± N/A 24.000 ± N/A Injury 3 Retain 4.479 ± 0.128 11.000 ± 0.645

Figure 3-1. A juvenile Narnia femorata. Photo courtesy of Christine Miller.

Figure 3-2. The right hind leg of a juvenile N. femorata, depicting the location of each injury site.

0.9

0.8

v i

v 0.7

u S

0.6

n = 25 n = 23 n = 86 0.5 No Autotomy Autotomy Injury Treatment

Figure 3-3. Experiment 1 – Contrast of treatments to investigate the effects of autotomy and injury on the proportion of individuals (± SE) surviving to adulthood.

Autotomized Injured Limb Retained Injured Limb A 0.9

0.8

0.7

0.6

a v

i 0.5

r u

S 0.4

0.3

0.2

0.1 n = 15 n = 4 n = 15 n = 8 n = 3 n = 20 0.0 Injury 1 Injury 2 Injury 3 Treatment 0.9 B 0.8

0.7

0.6 y

m 0.5

t o

t 0.4

u A 0.3

0.2

0.1 n = 19 n = 23 n = 23 0.0 Injury 1 Injury 2 Injury 3 Treatment

Figure 3-4. Experiment 1 – Effect of injury location on autotomy and survival. A) depicts the proportion of individuals (± SE) that survived based on their behavioral decision to autotomize or retain their injured limb for each injury location. B) illustrates the variation in the proportion of individuals (± SE) that autotomized at each injury location. Individuals in the injury 3 treatment had a significantly lower propensity to autotomize than individuals in the Injury 1 and Injury 2 treatments. Furthermore, autotomizing limbs injured at the Injury 3 location did not increase survival.

0.7

0.6

0.5

l 0.4

r u

S 0.3

0.2

0.1

0.0 n = 38 n = 39 n = 23 n = 17 No Injury/ Injury/ Injury/ Injury/ Induced Autotomy Induced Autotomy Self Autotomy No Autotomy

Treatment

Figure 3-5. Experiment 2 – Proportion of individuals (± SE) that survived in each treatment based on their autotomy behavior. In treatments where autotomy was experimentally induced, individuals did not have a behavioral choice. However, when only injury was induced, an individual could have self autotomized or retained (no autotomy) the injured limb.

CHAPTER 4 COREIDAE THAT AUTOTOMIZE THEIR SEXUALLY-SELECTED HIND LEGS HAVE DECREASED FIGHTING ABILITY AND MATING SUCCESS*

Many of the large and extravagant weapons found throughout the animal kingdom are products of sexual selection, including the antlers of deer and the horns of beetles (Emlen 2008). In these species, males use their weapons in direct, intrasexual competitions to gain access to females and territories (Anderson 1994). Males with larger weapons, as predicted by theory, are more likely to win these contests (fighting asymmetry; Parker 1974), resulting in greater access to mates (Anderson 1994). Still, despite the benefits of possessing a large sexually-selected weapon, these structures are sometimes lost. Since weapons are functional traits – being used in male-male combat – they can break if they exceed their mechanical limit (McCullough 2014,

McCullough et al. 2014). In addition to breaking, some species drop (i.e. autotomize) their weapons to escape life-threatening situations (e.g. fiddler crabs, Hoadley 1937; leaf-footed bugs, Emberts et al. 2016). In either case, the loss of the weapon is permanent if the individual is unable to regenerate a new one. Previous research has shown that the loss of a weapon decreases an individual’s fighting ability (O’Neill and

Cobb 1979, Berzins and Caldwell 1983, Neil 1985, Smith 1992, Abello et al. 1994,

Yasuda et al. 2011, Yasuda and Koga 2016). As a result, it is often assumed that the permanent loss of a sexually-selected weapon dramatically reduces, or even eliminates, an individual’s ability to secure matings. However, such an assumption fails to consider the possibility that an individual may behaviorally compensate for their missing weapon.

* This chapter is reprinted with permission from Emberts, Z., C. M. St. Mary, T. J. Herrington, and C. W. Miller. 2018. Males missing their sexually selected weapon have decreased fighting ability and mating success in a competitive environment. Behavioral Ecology and Sociobiology 72: 81.

Consider cases in which males are unable to develop (relatively) large weapons.

Many of these males are still able to secure some degree of mating success, and often do so by altering their behavior (Gross 1996). For example, in the dung beetle

Onthophagus acuminatus, weaponless males secure mating opportunities by using a sneaking behavior that allows them to avoid direct male-male contests (Emlen 1997).

On the other hand, instead of avoiding contests, some small males actively engage in combat against larger opponents (Morris et al. 1995). Although slightly counter intuitive, theory predicts that using more aggressive tactics (e.g. initiating contests, persisting in contests), under certain scenarios, can result in smaller males winning (Morrell et al.

2005). Thus, given this potential for behavioral compensation, the aim of our study was to determine the effects of permanent weapon loss on fighting strategy, fighting ability, and mating success.

To investigate this aim we used the leaf-footed cactus bug Narnia femorata Stål

(Hemiptera: Coreidae; Figure 4-1). Males in this species possess sexually-selected hind legs, which they use to compete against other males over access to mates and resources (Procter et al. 2012, Nolen et al. 2017). During combat, males use their enlarged, sexually-dimorphic, hind legs to kick and squeeze their opponents (Table 4-

1), enabling them to physically remove other males from a territory (Nolen et al. 2017).

Larger males have relatively larger weapons (Allen and Miller 2017) and are more likely to win male-male contests (Procter et al. 2012). Therefore, there is strong evidence to suggest that the hind legs of N. femorata are sexually-selected weapons (Procter et al.

2012, Nolen et al. 2017). Despite their use as weapons, males can autotomize their hind legs; and once the leg is dropped it is permanently lost, as N. femorata cannot

regenerate the limb (Emberts et al. 2016). When limbs are lost through autotomy, individuals in other species are frequently observed altering their behavior to compensate for the lost limb (Herreid and Full 1986, Stoks 1999, Bateman and Fleming

2006c, Cooper 2007, Oliveira et al. 2015). Consequently, weapon loss via autotomy provides an excellent opportunity to determine the extent to which individuals can behaviorally respond to the permanent loss of a weapon.

There are three types of compensatory strategies that males without their weapons might employ to secure access to females. Like males that develop small weapons, males that lose their weapons may establish some degree of dominance by

(1) increasing their aggressiveness. This strategy could manifest itself through an increase in contest initiation, contest persistence, or both. Alternatively, individuals that lose their weapon(s) may be able to establish dominance by (2) changing the fighting tactics that they use. Such a strategy has previously been observed in both decapods

(O’Neill and Cobb 1979) and stomatopods (Berzins and Caldwell 1983). In N. femorata, weaponless males could implement this strategy by decreasing their use of fighting tactics that require both hind legs (grappling; Table 4-1). Finally, males may still secure access to females while (3) avoiding direct male-male competition, which would suggest that weaponless males implement an alternative reproductive tactic (i.e. a condition- dependent tactic). In our study, we examined the behavior and success of males with and without their weapons to evaluate these three alternatives.

Even if a male acquires access to a female, mating success is still not guaranteed because females could choose not to mate. In N. femorata, after a male initiates a mating, the female then determines whether the mating will occur because, if

she is unreceptive, she can simply keep her genital plates closed (Gillespie et al. 2014,

Cirino and Miller 2017). Consequently, an investigation of male mating success would be incomplete without also considering female choice. In many species, females assess the quality of males before deciding whether to mate (e.g. Hamilton and Zuk 1982, Hill

1991). Since autotomy is a form of injury (i.e. self-induced injury, Emberts et al. 2017a), females may differentially mate with intact males over males with missing limbs.

Therefore, to investigate the effects of weapons loss on mating success, we need to determine whether weaponless males can access females and whether females are receptive to those males.

Since weapon loss is both permanent and frequently observed in the wild

(Emberts et al. 2016, Emberts et al. 2017a), we hypothesized that weaponless N. femorata males would alter their behavior to compensate for their missing weapons.

Nonetheless, since autotomy is a form of injury, we also predicted that males missing their weapons would have lower mating success, as females should prefer intact males over those with missing limbs.

Methods

Rearing of Study Species

We used first generation lab-reared individuals for our experiment. Populations were founded in July of 2015 with wild-caught N. femorata, which were collected from

Live Oak, Florida (30.26°N, -83.18°W). Narnia femorata were reared in deli cups (top diameter: 118mm, bottom diameter: 85mm, height: 148mm) containing Opuntia mesacantha ssp. lata cladodes (cactus pads) and fruit collected from the same location throughout the experiment. Nymphs were reared in groups of 5 – 10 and checked daily, until their terminal molt. Upon becoming adults, individuals were placed into their own

deli cup. Each adult was then given 21 days to reach sexual maturity. Individuals that had been in the adult stage longer than 28 days were not used.

Experimental Design

Unlike the sexually-selected horns of some beetles, the sexually-selected trait in

N. femorata (i.e. their hind legs) is not just limited to male-male combat, but also serves a locomotive function. Thus, it would be imprudent for us to ignore the concurrent negative effects that losing a leg might also have on fighting and mating outcomes. To address this complexity, we induced limb loss in both mid legs (i.e. a non-sexually- selected leg) and hind legs (i.e. a sexually-selected leg) to investigate the effects of losing a sexually-selected weapon on fighting ability and mating success. Specifically, we randomly placed focal males into one of five treatments (n=48 per treatment): all legs present, missing a single mid leg, missing a single hind (i.e. sexually-selected) leg, missing two mid legs, and missing two hind (i.e. sexually-selected) legs. Then, each focal individual was randomly assigned to an unrelated male with all of his legs, henceforth, the rival male. Thus, each trial consisted of a unique, random, intact male

(the rival) and a focal male. One day before fights occurred, we marked the focal and rival males with paint pens (Elmer’s Paint Markers) to distinguish them from one another. Thus, it was not possible to record data blind because our study involves following a focal individual. Also at this time we induced autotomy in focal males according to treatment. Autotomy was induced by gripping the leg with reverse action forceps for a few seconds, while the insect was in contact with a piece of wood

(Emberts et al. 2016). On the day of experimentation, paired focal and rival males were placed into an arena which consisted of a deli cup containing a single planted O. mesacantha ssp. lata cactus with an attached fruit. Once both males were added, we

began behavioral observations, live, for the next 4 hrs. Since behavioral interactions are both stereotypic and relatively rare, we had a single observer conducted up to 10 trails concurrently.

During behavioral observation, we quantified whether fighting occurred, the fighting tactics used, and ultimately assigned the dominant individual on the basis of these observations. To do this, we defined the following intrasexual interactions: displaying, charging, kicking, wrapping, and grappling (Table 4-1). If any of these defined tactics (e.g. charging, kicking, grappling, etc.) resulted in an opponent retreat, then the aggressor was awarded a single dominance point. After experimentation, following the same protocol as Procter et al. (2012), the number of dominance points accrued by each male (focal and rival) was summed and the individual with the greatest number of dominance points was considered the dominant male. In the case of ties, dominance was not assigned.

Two hours into the live observation we added an unrelated, sexually-mature female and began to quantify intersexual interactions (specifically, mating and mating attempts), while continuing to evaluate dominance, for the final 2 hrs. Mating in N. femorata is highly stereotypic. Males will initiate mating by mounting a female and aligning his genitals with hers (i.e. male mating attempt). Once aligned, he then presses his intromittent organ against her genital plates. Should the female be receptive, she will open her genital plates, at which point the male will dismount and turn himself away from the female while maintaining genital contact (i.e. mating). If the female is unreceptive, she can flee, kick, and/or simply keep her genital plates closed. In all cases, the male would be unsuccessful in mating. Therefore, we considered female

100

rejection to occur if a male mounted a female (i.e. attempted mating), but was unsuccessful at mating (Gillespie et al. 2014, Cirino and Miller 2017).

Following each trial, all individuals were photographed using a digital camera

(Canon EOS 50D) attached to a dissecting microscope (Leica M165 C). Then, body size (pronotal width) was measured to the nearest micrometer using ImageJ software

(v. 1.46).

Ethical Note

Narnia femorata naturally autotomize their front, mid, and hind legs to escape predation and entrapment (Emberts et al. 2016, Emberts et al. 2017a). In fact, up to

13% of individuals in the wild are missing at least one limb, with some missing up to three (Emberts et al. 2016). In the lab, autotomy is never forced; we simply initiate self- induced autotomy by holding an individual’s limb for a few seconds. Like autotomy, male-male competition is also a naturally occurring behavior in N. femorata. However, to reduce competitive interactions, adult males were only paired during experimental trials. Finally, all individuals were treated as humanely as possible by providing them with ideal rearing conditions throughout the study.

Statistical Analyses

To investigate the effects of limb loss on dominance and fighting behavior we conducted planned, a priori, contrasts in the context of generalized linear models

(GLMs). Since we specifically hypothesized that the loss of the hind, sexually-selected legs would affect fighting ability and behavior we conducted three contrasts: Contrast 1

– individuals missing hind leg(s) versus individuals with all of their legs, Contrast 2 – individuals missing mid leg(s) versus individuals with all of their legs, and Contrast 3 – individuals missing hind leg(s) versus individuals missing mid leg(s). In all of our

101

models, we also included body size (i.e. pronotal width of focal individual) and the difference in body size between the focal and rival as covariates. Both traits were included as covariates because previous work in this species has shown that body size and the difference in body size can have a significant effect on fighting outcomes and/or behaviors (Procter et al. 2012, Nolen et al. 2017).

First, to investigate whether limb loss affects the propensity to engage in male- male contests, we compared the frequency of fighting encounters (i.e. whether at least one fighting behavior, including a display, occurred per pair, yes/no) across the contrasted treatments using a logit-link function and assuming a binomial distribution

(i.e. a GLM for binary data). Then, given that there was a fighting encounter, we investigated whether weapon loss affected fighting outcomes by comparing dominance

(yes/no) of the focal male across the contrasted treatments. Using this same subset of the data, we also investigated whether limb loss affected male competitive behavior.

Specifically, we asked whether the propensity to grapple (yes/no) or the propensity to charge (yes/no) varied across the contrasted treatments. These two tactics were explicitly chosen because they can easily be identified by an observer even if a male is missing both of his hind legs. For example, due to the stereotypic nature of a grapple

(Figure 4-1), an observer can determine if a weaponless male is attempting to grapple when the weaponless male aligns abdomen to abdomen with his rival. Finally, we also investigated whether limb loss affected fighting persistence (i.e. the number of fighting behaviors that occurred per pair) across our contrasted treatments using a log-link function and assuming a Poisson distribution. Having a measure of an individual’s decisions to flee (i.e. persistence) helps us understand the extent to which autotomy

102

affects the decision making process during fights. We also explicitly investigated the correlation between body size and fighting persistence (i.e. contest duration) for winners and losers (Arnott and Elwood 2009) to place the behavior of N. femorata in the context of previous literature on give-up decisions in male-male combat.

Next, we conducted similar analyses (i.e. GLMs with contrasts) to investigate the effects of limb loss on mating success (mated yes/no). Then, to better understand any differences in mating success, we investigated whether the loss of hind legs and/or dominance affected male mating attempts (mount yes/no) and/or female choice

(accept/reject).

Although differences in losing one limb versus losing two is not the main focus of this manuscript, to ease curiosities and develop hypotheses, we also investigated whether the number of limbs lost affected fighting ability and mating success. For these analyses we used GLMs with contrasts, but contrasted those missing two limbs against those missing just one. All analyses were conducted in R v3.3.1 (R Core Team 2016).

Results

To gain a better understanding of how weapon loss affects fighting ability and mating success, first, we describe the outcomes and behaviors of the focal individuals with all of their legs to establish a comparative baseline. Recall that focal males were always paired with an intact (non-autotomized) rival. Of the 48 focal males with all their legs, 18 (38%) had at least one fighting encounter. Of those 18, the focal individual established dominance in 10 (56%; Figure 4-2). Furthermore, 5 of the 18 focal males

(28%) engaged in grappling at least once, while 2 of the 18 charged at least once (11%;

Figure 4-3). With regards to mating, 25 (52%) of the focal males attempted to mate (i.e.

103

mounted), but only 20 (42%) mated (Figure 4-2). Therefore, females rejected 20% of the males that attempted to mate.

When we contrasted focal individuals missing their mid leg(s) against focal individuals with all of their legs (Contrast 2) we found no evidence to suggest that the loss of a mid leg affected any of the investigated response variables (Figure 4-2; Figure

4-3; Table 4-2).

Effects of Losing a Weapon on Fighting Ability and Behavior

The loss of a sexually-selected weapon affected a male’s dominance, but not fighting behavior. We found that focal males missing their sexually-selected hind legs had a lower probability of establishing dominance (9%) compared to those missing mid

2 legs (45%; Contrast 3: 1 = 11.113, p < 0.001) and those with all of their legs (56%;

2 Contrast 1: 1 = 17.546, p < 0.001; Figure 4-2). Nonetheless, loss of hind legs did not have a significant effect on fighting behavior. Individuals missing hind legs were just as likely to engage in male-male competition as individuals with all of their legs (Contrast 1:

2 2 1 = 0.362, p = 0.547) and individuals missing mid legs (Contrast 3: 1 = 1.132, p =

0.287; Figure 4-3). Furthermore, the proportion of individuals that used a grapple, a fighting tactic that incorporates both hind legs, did not vary across the contrasted

2 2 treatments (Contrast 1: 1 = 1.115, p = 0.291; Contrast 3: 1 = 2.482, p = 0.115; Figure

4-3). The proportion of focal individuals that used a charge was similar as well (Contrast

2 2 1: 1 = 0.010, p = 0.920; Contrast 3: 1 = 0.038, p = 0.845; Figure 4-3). Moreover, there was no observable difference in fighting persistence. Males missing their sexually- selected hind legs were just as likely to persist in fights as males missing mid legs

2 2 (Contrast 3: 1 = 2.413, p = 0.120) and males with all of their legs (Contrast 1: 1 =

104

0.405, p = 0.524). Fighting persistence did however positively correlate with body size

2 for subordinate males (1 = 37.56, p < 0.001), although the relationship between

2 persistence and body size was not present for males that became dominant (1 = 1.456, p = 0.228; Figure 4-4).

Effects of Losing a Weapon on Mating Success

Loss of a sexually-selected weapon also reduced a male’s probability of mating.

2 Focal males missing hind legs mated less than those missing mid legs (Contrast 3: 1 =

2 7.003, p = 0.008) and those who had all their legs (Contrast 1: 1 = 6.128, p = 0.013;

Figure 4-2). Still, approximately 25% of males missing hind legs secured a mating

(compared to approximately 40% for those missing their mid legs and those with all their legs).

A decrease in mating success could reflect that weapon loss (1) decreased male mating attempts or (2) increased female rejection, or both. In our study, there was no evidence to suggest that weapon loss increased female rejection (i.e. there was no

2 variation among treatments in female choice; Contrast 3: 1 = 1.412, p = 0.235;

2 Contrast 1: 1 = 0.353, p = 0.552). Instead, we found that the loss of sexually-selected hind legs reduced a male’s mating attempts compared to those who lost mid legs

2 2 (Contrast 3: 1 = 6.001, p = 0.014) and those who had all of their legs (Contrast 1: 1 =

6.272, p = 0.012; Figure 4-2), as only 34% of these males attempted to mate compared to approximately 50% of males in other treatments.

As hypothesized in Nolen et al. (2017), we also found that dominance explained

2 variation in attempted mating (1 = 6.881, p = 0.009) and tended to explain variation in

2 mating success (1 = 3.523, p = 0.061) when we considered all competing males

105

together. Therefore, our mating observations could be a result of dominance interactions, not the loss of weaponry per se. Thus, we statistically controlled for dominance to tease apart the direct effects of weapon loss on male mating behavior and success by adding fighting status as a covariate into our models. When doing so,

2 weapon loss, itself, did not influence attempted matings (Contrast 1: 1 = 0.404, p =

2 2 0.525; Contrast 3: 1 = 2.121, p = 0.145) nor mating success (Contrast 1: 1 = 0.147, p

2 = 0.701; Contrast 3: 1 = 2.538, p = 0.111).

Effects of Losing One Limb Versus Two

The number of limbs lost (i.e. 1 limb versus 2) did not have an effect on the investigated outcomes (i.e. dominance and mating success), but we did find that

2 individuals missing two limbs were less likely to grapple than individuals missing one (1

= 3.920, p = 0.048; Table 4-3).

Discussion

Males who lost their sexually-selected weapons were more likely to lose contests. Despite this handicap, males missing their weapons were just as likely to engage in male-male competition, persist once the competition started, and even used fighting tactics that appear to require both hind legs. These results reveal a surprising lack of plasticity in N. femorata fighting behavior. Similarly, loss of weaponry, per se, did not affect a male’s mating behavior. However, subordinate males were less likely to attempt to mate (see also Nolen et al. 2017) and males without their weapons were more likely to be subordinate. Consequently, weapon loss influenced a male’s pursuit of females, which ultimately decreased male mating success in a competitive environment.

These results support the assumption that the permanent loss of a sexually-selected

106

weapon reduces an individual’s ability to secure matings. However, our results refute the claim that the loss of a sexually-selected weapon removes an individual from the mating pool, as 25% of the males who lost their weapons still secured a mating.

Compensatory behavior has frequently been observed following autotomy

(Herreid and Full 1986, Stoks 1999, Bateman and Fleming 2006c, Cooper 2007,

Oliveira et al. 2015), and likely mitigates the costs associated with losing a limb. As such, we hypothesized that weaponless males would alter their behavior to compensate for their missing weapon(s). We found that males without their weapons behaved in the same manner as males that possessed their weapons, highlighting that compensatory behavior does not always succeed autotomy.

There are several hypotheses that might explain the lack of observed fighting plasticity. It is possible that weaponless males did not alter their behavior because they were unable to perceive their (new) fighting ability prior to interactions. This hypothesis is pertinent because we only investigated the males’ first fighting opportunity post autotomy. Moreover, since our results suggest that N. femorata make their fighting decisions following pure-self-assessment models (Figure 4-4, Arnott and Elwood 2009), like the ‘energetic war of attrition’ (Payne and Pagel 1996), it makes sense that dropping a weapon would not necessarily alter a male’s first fighting interaction. Fighting decisions under pure self-assessment models are based on a cost threshold that is established before fighting even begins and does not change during an interaction. In such cases, we would not expect to observe a change in fighting behavior until subsequent fighting interactions. Alternatively, male N. femorata may primarily use their body size, as opposed to their weapon size, to assess their fighting ability, as is

107

commonly found across species (Anderson 1994). Under this scenario, the loss of a hind leg may not alter a male’s decision to engage, persist, and grapple in competitions because their body size is unaffected (discussed in Yasuda and Koga 2016). It is also possible that males accurately assess their fight ability, but did not alter their behavior because the cost of fighting is low, while the benefits are high. In such instances, weaponless males would still be predicted to engage in male-male competitions (Morrell et al. 2005). Finally, weaponless males in their natural environment may be able to avoid conflict by seeking territories that are unoccupied (Booksmythe et al. 2009).

Under this scenario, there would be little to no selection for males to evolve compensatory fighting behaviors. Still, in this study, males missing weapons were able to secure dominance 9% of the time (as compared to ~50% for males that possessed their weapons) without changing any of the investigated fighting behaviors.

Success in male-male competition may not completely determine mating opportunity. Therefore, we also investigated how weapon loss affected a male’s mating behavior. Like fighting behavior, weapon loss, per se, did not affect mating behavior.

However, dominance did (see also Nolen et al. 2017). Compared to a subordinate male, a dominant male was twice as likely to attempt a mating. Since weaponless males were less likely to be dominant, they were also less likely to attempt a mating. Therefore, weapon loss appears to indirectly influence mating behavior. We should take some caution in this assertion because statistically controlling for dominance also reduced our sample size (from n=238 to n=88) since we were unable to assess dominance in every trial. Regardless, losing a weapon ultimately reduces a male’s mating opportunities in a competitive environment.

108

Female mating behavior could have also affected a male’s mating success because females could choose not to mate. In our study, neither the presence of weaponry nor dominance status affected female choice. However, females clearly make mating decisions in N. femorata (e.g. Gillespie et al. 2014), and limb loss provides an ideal opportunity to understand whether injury and asymmetry effect female choice. In several species, females are more likely to mate with males that have symmetrical traits

(e.g. zebra finch, swordtail fish, Swaddle and Cuthill 1994, Morris 1998). In general, male N. femorata (with all their legs) are fairly symmetric (Allen and Miller, unpublished data). However, up to 9% are missing a single leg in the wild (Emberts et al. 2016), making them radically asymmetric. When we investigated female choice for symmetrical males (loss of two legs versus one from the same position), we found no evidence that females differentially mated with symmetric males over asymmetric ones. Besides asymmetry, females may also assess the health and/or quality of males before deciding whether to mate (e.g. Hamilton and Zuk 1982, Hill 1991), and since autotomy is a form of injury (i.e. self-induced injury, Emberts et al. 2017a), females may discriminate against males with autotomized limbs. However, N. femorata females did not differentially discriminate amongst males from our different treatments.

In summary, weapon loss decreased both fighting ability and mating success, as weaponless males did not behaviorally compensate for their missing limbs. These intriguing results suggest directions for future investigation. Future studies should determine whether the differences in mating success, such as those observed here, result in similar differences in reproductive success. The distinction between reproductive success and mating success is particularly important when you consider

109

that both the focal and rival male mated with the female in 11% of the trials.

Consequently, post copulatory competition could play a major role in determining an individual’s fitness. For example, weaponless/subordinate males may allocate more sperm per mating, potentially increasing their relative reproductive success (Joseph et al. 2018, Somjee et al. 2018). Future studies should also determine whether there are temporal or experiential consequences to permanently losing a weapon. It is possible, for example, that previous fighting experiences affect the subsequent fighting outcomes and behaviors of weaponless males (Hsu et al. 2006). These additional studies would ultimately contribute to our understanding of the costs associated with permanently losing a sexually-selected weapon.

110

Table 4-1. Narnia femorata fighting tactics. These behaviors correspond with coreid fighting tactics already defined in the literature (Nolen et al. 2017), with one exception; we altered the definition of a wrap to include squeezes that only use a single leg (previous definitions required the use of both). Furthermore, we distinguish between stereotypic wrapping, termed grappling (Figure 4-1), and non-stereotypic wrapping. We differentiated grappling from wrapping because grappling provides an ideal fighting tactic to investigate behavioral responses to losing a weapon since it involves both hind legs and, due to its stereotypic nature, is easy to identify regardless of leg presence (Figure 4-1).

Fighting Tactic Definition

A raising and widening of the hind limb(s) in the direction of the opponent. This Displaying tactic does not require the opponents to physically touch.

A swift and direct movement toward the opponent, which ending in a touching- Charging interaction. Note that this is the only tactic that does not require hind legs.

A rapid extension of a hind leg in the direction of the opponent that finishes with Kicking physical contact. The aggressor may lunge at his opponent in tandem with the kick in order to ensure contact.

When an individual grips and/or squeeze his opponent with one or two hind Wrapping legs.

A stereotypic behavior in which males aligned abdomen to abdomen to wrap Grappling one another with both legs.

111

Table 4-2. Results of Contrast 2 – the comparison between males missing middle legs and intact males. Loss of middle leg(s) did not affect any of the investigated outcomes and behaviors relative to intact (non-autotomized) individuals. Variables 2 df p

Mounting 0.246 1 0.620

Rejection 0.087 1 0.768

Mating 0.102 1 0.750

Dominance 1.649 1 0.199

Engage in Fight 0.074 1 0.786

Grapple 0.048 1 0.828

Charge 0.002 1 0.962

Persistence 0.331 1 0.565

112

Table 4-3. The number of limbs lost affected grappling behavior, but not the other response variables. Variables 2 df p

Mounting 0.950 1 0.330

Rejection 0.682 1 0.409

Mating 1.334 1 0.248

Dominance 1.669 1 0.197

Engage in Fight 0.922 1 0.337

Grapple 3.92 1 0.048

Charge 0.006 1 0.939

Persistence 1.223 1 0.269

113

Figure 4-1. Two Narnia femorata males posteriorly aligned abdomen to abdomen, ready to engage in a grapple. Photo courtesy of Christine Miller.

114

Figure 4-2. The effects of losing a weapon on fighting ability and mating success. Loss of a weapon in N. femorata significantly decreases a male’s fighting ability (i.e. dominance), mating attempts (i.e. mounting), and mating success (i.e. mated). Loss of a mid leg had no effect on these factors (Table 4-1). Females did not discriminate against males across the contrasted treatments. Predicted proportions and standard errors displayed were derived from a generalized linear model with a logit-link function and the assumption of a binomial distribution taking difference in body size into consideration. This simplified model was used to help visualize the data. All models were constrained within the bounds of the x-axis (i.e. difference in body size) to avoid extrapolation. Only a subset of the data was used to investigate fighting ability because we were unable to assess dominance for each trial.

115

Figure 4-3. Lack of plasticity in fighting behavior. Loss of a sexually-selected weapon did not affect a male’s propensity to engage in a fight nor did it affect a male’s fighting tactic. Displayed is the proportion of focal individuals that engaged in fighting, grappling, and charging ± SE.

116

25 25 A B

20 20

e c

c 15 15

r e

e 10 10

P P

5 5

0 0

3.0 3.5 4.0 4.5 5.0 3.0 3.5 4.0 4.5 5.0 Body Size (PW, mm) Body Size (PW, mm)

Figure 4. Body size explains variation in contest duration for subordinate males (A), but not dominant ones (B). Predicted Figure t4re-n4.d l in Bodyes and s tsizeandar dexplains errors were dvariationerived from ginen eA)rali zcontested linear m odurationdels with a l oforg-lin subordinatek function and the males,assumptio n of a Poissobutn di snottribu tB)ion. dominantSolid trend line ones.s include Predictedall data points, wtrendhile th elines dashed and trend standardline excludes oerrorsur outlie rwere (Persist ence = 66, Body Size = 4.41 mm; data point not displayed). Including or excluding the outlier did not affect our conclusions. Persisderivedtence is the from total n ugeneralizedmber of fighting i nlinearteraction modelss over the 4 with hour d aur alogtion.- link function and the assumption of a Poisson distribution. Solid trend lines include all data points, while the dashed trend line excludes our outlier (Persistence = 66, Body Size = 4.41 mm; data point not displayed). Including or excluding the outlier did not affect our conclusions. Persistence is the total number of fighting interactions over the 4 hour duration.

117

CHAPTER 5 THE EVOLUTION OF AUTOTOMY IN LEAF-FOOTED BUGS

Predation can impose a strong selective pressure. As a result, animals have evolved extraordinary defenses (i.e. anti-predatory traits) that reduce their vulnerability, contributing to the morphological and behavioral diversity we see in organisms today

(Caro 2005, Ruxton et al. 2018). One of the most extreme forms of anti-predatory defense is autotomy, where individuals literally sacrifice part of their body while attempting to escape. A lizard dropping its tail in response to a predator attack is an iconic example (Arnold 1984, Bateman and Fleming 2009). However, autotomy also occurs in a diversity of other organisms: squid amputate their arms (Bush 2012), harvestmen release their legs (Guffey 1999), and salamanders drop their tails

(Maiorana 1977). Despite having multiple origins (Emberts et al. 2019), fundamental questions about the evolution of autotomy remain unanswered. Most notably, how does sacrificing a limb to escape predation evolve, and what factors promote and constrain autotomy’s evolution?

Autotomy is predominately thought of as an anti-predatory trait, but there are two additional survival benefits associated with autotomy: reducing the cost of injury

(Emberts et al. 2017a) and escaping non-predatory entrapment (Maginnis 2008).

Several species have been observed autotomizing limbs that have been severely damaged (e.g. after intraspecific competition or a failed predation event), and this reduces mortality (Emberts et al. 2017a). Escape from non-predatory entrapment

(Maginnis 2008) may also be a widespread benefit, especially in the Ecdysozoa (i.e. arthropods, nematodes, and allies; Fleming et al. 2007, Hodgkin et al. 2014). Within this

118

clade, non-predatory entrapment frequently manifests itself as entrapment in a fouled molt, which an organism can escape through autotomy (Wood and Wood 1932).

Although escaping non-predatory entrapment and reducing the cost of injury are rarely discussed, these benefits may be crucial to understanding the evolution of autotomy because these two benefits do not require individuals to autotomize quickly. In fact, autotomy can occur several minutes, and potentially even hours, after an incident occurs (i.e. entrapment, injury) and still be beneficial (Emberts et al. 2017a). On the other hand, individuals need to be able to drop their limbs quickly to use autotomy in an anti-predatory capacity. As a result, it has been hypothesized that escaping predation is not the ancestral benefit of autotomy (McVean 1982). Instead, McVean (1982) hypothesized that sacrificing a limb to escape predation is a co-opted benefit, and that autotomy to reduce the cost of injury is an ancestral state (i.e. benefits associated with autotomizing slowly are an intermediate step in the evolution of much more rapid autotomy). Alternatively, autotomy’s evolution may be exclusively associated with dropping a limb quickly enough to escape predation (i.e. the fast latency hypothesis;

Emberts et al. 2019). Note that the fast latency hypothesis states that organisms transition from being unable to autotomize to being able to autotomize their limb quickly

(i.e. there is no intermediate step with regards to the rate of autotomy). The first aim of this study is to investigate these two alternative hypotheses for the origins of rapid autotomy.

The second aim of our study is to investigate the ecological and morphological factors associated with autotomy’s evolution. Three major factors that are thought to influence autotomy are (1) predation, (2) the costs associated with autotomizing, and (3)

119

body size. Autotomy is not an effective strategy against all predator classes. For example, tail autotomy in some lizards does not defend against predation by falcons, but it is effectively used against teiids (i.e. a predatory lizard) and snakes (Medel et al.

1988). Similar patterns have also been found in crickets, where autotomy is more effective against mice than against skinks (Bateman and Fleming 2006b). As a result, the predatory selective pressure acting on autotomy depends on the abundance of certain predator species. One broad-scale metric that has been used to capture this predatory pressure is predator diversity, which has been shown to positively correlate with the ease of tail autotomy in lizards (Cooper et al. 2004, Brock et al. 2015). While predation is thought to promote autotomy, the cost of losing the autotomizable limb, even temporarily, potentially constrains it. This constraint is most discernable when a sexually selected trait is on an autotomizable appendage, as autotomy results in the costly loss of a trait that is important for reproductive success (discussed in Wasson and

Lyon 2005, Emberts et al. 2016). Finally, a major morphological factor that is thought to influence autotomy’s evolution is an organisms’ body size. Across orthopterans (i.e. grasshoppers, katydids, and allies), for example, larger species autotomize more slowly

(Bateman and Fleming 2008). The ease of autotomy might decrease as body size increases because of morphological constraints (discussed in Bateman and Fleming

2008) or differences in predatory pressure associated with an organism’s size

(Rememel et al. 2011).

Leaf-footed bugs and allies (Insecta: Hemiptera: Coreidae + Alydidae; henceforth referred to as leaf-footed bugs; see Forthman et al. 2019) are an ideal clade to investigate the evolution of autotomy. These insects autotomize their legs to escape

120

predation (Z.E. pers. obs.), to escape non-predatory entrapment (Emberts et al. 2016), and to reduce the cost of injury (Emberts et al. 2017a). Species within this clade also show substantial variation in the latency to autotomize their legs (Emberts et al. 2016).

In some species, autotomy occurs quickly enough to escape predation (<60 seconds;

Embers et al. 2016), while in other species autotomy takes more than fifteen minutes (Z.

Emberts, unpublished data). Additionally, the autotomizable legs of leaf-footed bugs come in a variety of forms and serve a variety of functions. Several species have hind legs that resemble their front and mid legs, and legs that take this form are thought to only serve a locomotive function. However, other species have hind legs with enlarged femurs that are costly to develop and maintain for both males and females (Joseph et al. 2018, Somjee et al. 2018a, Somjee et al. 2018b, Miller et al. 2019). Given these costs, enlarged hind legs likely have a function beyond locomotion (e.g. sexually selected weapons, Miyatake 1997, Mitchell 1980, Eberhard 1998, Miller and Emlen

2010, Miller et al. 2016, Emberts et al. 2017b). Finally, leaf-footed bugs vary dramatically in body size and have a cosmopolitan distribution.

To investigate the evolution of autotomy in this clade we quantified the latency to autotomize in 59 species of leaf-footed bugs from around the world and conducted phylogenetic comparative analyses. We hypothesized that predatory pressure promotes autotomy, while the cost of autotomy and body size constrains it. Predator diversity and abundance increases towards the equator (Jeanne 1979, Schemske et al. 2009), as do anti-predatory traits (Møller and Liang 2013, Diaz et al. 2013, Samia et al. 2015, Levin and York 1978 and Laurila et al. 2008). Thus, we hypothesized that leaf-footed bugs closer to the equator should autotomize more quickly. We also hypothesized that

121

species with enlarged hind legs should be less willing to release them because loss of these legs has been shown to decrease an organism’s future mating success in some species (Emberts et al. 2018). Finally, we hypothesized that larger taxa have a slower latency to autotomize, following similar patterns observed in other insect clades (e.g. orthopterans, Bateman and Fleming 2008).

Methods

Behavioral and Morphological Data

To investigate the evolution of autotomy in leaf-footed bugs, we collected 59 species from the wild, plus one Lygaeidae and two Rhopalidae outgroup taxa (n=62).

The species used for this study represent almost all the leaf-footed bug species that we could find at our field sites in Singapore, Panama, Australia, eSwatini (previously

Swaziland), South Africa, and the United States. Within each species, we aimed to induce autotomy of hind legs in 30 female and 30 male individuals (sample size mean =

20, median = 10). However, it is important to note that some species are represented by a single individual. Autotomy was induced by following a previously established protocol

(Emberts et al 2016) where we gripped the insect’s leg with constant pressure (reverse action) forceps while the insect’s other legs were in contact with a piece of wood. Time to autotomize was recorded using a stopwatch. Since autotomy did not always occur, behavioral trials were terminated after an hour and these individuals were recorded as taking 3600 s to autotomize. Autotomy trials were conducted between the hours of 7:00

– 23:00 and at a temperature of 27 ± 4C. We considered a species capable of autotomizing if a single individual dropped their leg in the one-hour escape from entrapment scenario or if at least one wild caught individual was missing a limb at their autotomy fracture plane (i.e. the morphological plane at which autotomy occurs, which

122

is at the trochanter-femur joint in this clade). For species capable of autotomy, which was every species investigated for this study, we quantified the mean and median time to autotomize (two measures of central tendency) and assigned these values for the purposes of trait mapping. All autotomy data was square root transformed to better fit model assumptions (e.g. normality, homoscedasticity). Each individual was also photographed with a scale bar using a digital camera (Canon EOS 50D) so we could measure pronotal width, a body size proxy (Procter et al. 2012), to the nearest micrometer using ImageJ v 1.46 (Abràmoff et al. 2004).

Molecular Data and Sequence Alignment

Of our 62 taxa, 27 had already been sequenced and aligned (Kieran et al. 2019,

Forthman et al. 2019).1 For the remaining 35 taxa, genomic DNA was extracted following Forthman et al. (2019).1 Isolated DNA was visualized using 1% agarose gel electrophoresis and quantified using a Qubit 2.0 fluorometer. Samples with DNA concentrations greater than 20 ng/µL were normalized to 10–20 ng/µL. A Biorupter

UCD-300 sonication device (4–10 cycles of 30 s) or a Covaris M220 Focused- ultrasonicator (20–60 s) was used to fragment high molecular weight samples into 200–

1000 bp fragments. For library construction, we used a modified KAPA Hyper Prep Kit protocol, which included the use of iTru universal adapter stubs and 8 bp dual-indexes

(Glenn et al. 2016).

For target enrichment, we used a custom myBaits kit (Arbor Biosciences) that subsampled ultraconserved element (UCE) baits designed by Faircloth (2017). Some

1 Forthman, M., C. W. Miller, and R. T. Kimball. In review. Phylogenomics of the leaf-footed bug subfamily Coreinae (Hemiptera: Coreidae): applicability of ultraconserved elements at shallower depths.

123

samples were subjected to the target enrichment (TE) protocol outlined by Forthman et al. (2019) while others were subjected to a touch-down (TE-TD) approach.1 During touch-down target enrichment, probes were initially hybridized with library pools at 65°C for 18 hours followed by 60°C for 18 hours, but this hybridization procedure was terminated prematurely. At the recommendation of Arbor Biosciences, we added additional baits (2.75 µL) to these samples and re-ran the touch-down hybridization protocol to completion. Enriched pools were quantified using a Qubit 2.0 fluorometer and pooled in equimolar ratios prior to sequencing on an Illumina HiSeq3000 lane

(2x100) at the University of Florida’s Interdisciplinary Center for Biotechnology

Research (ICBR).

Sequence reads were demultiplexed and adapters were trimmed from raw sequence reads with illumiprocessor (Faircloth 2013, Bolger et al. 2014). Duplicate reads were filtered using PRINSEQ-lite v 0.20.4 (Schmieder and Edwards 2011).

Remaining reads were error-corrected with QuorUM v 1.1.0 (Marçais et al. 2015) and de novo assembled in Trinity v 2.8.3 (Grabherr et al. 2011) using default settings.

Contigs were matched to UCE probes using PHYLUCE v 1.5.0 (Faircloth 2016). Loci were individually aligned in PHYLUCE using default settings and were internally trimmed using trimAl (Capella-Gutierrez et al. 2009). Locus alignments with at least

70% of taxa (567 loci) were concatenated into a supermatrix. PartitionFinder v 2.1.1

(Lanfear et al. 2016) was used to select the best-fit partitioning scheme (‘rcluster” algorithm; Lanfear et al. 2014) with individual loci treated as data blocks and branch

124

lengths unlinked. All models under the “raxml” option were examined, with model selection based on the corrected Akaike Information Criterion (AICc).

Phylogeny and Divergence Time Estimation

Phylogenetic reconstruction using maximum likelihood was performed using

RAxML v 8.2.3 (Stamatakis 2014). We used GTRGAMMA, performed a partitioned analysis using the output from PartitionFinder (-q), and ran 500 rapid bootstraps (-f a).

Divergence time estimation was implemented in BEAST v 1.10.4 (Suchard et al. 2018).

Due to the large phylogenomic dataset, we ran our 567 loci through SortaDate (Smith et al. 2018) in order to reduce alignment complexity and to identify those loci most suitable for dating analysis. Upon completion, we chose the best 50 loci sorted first by bipartition, then root-to-tip variance, then tree length. Adding additional loci has been shown to have a negligible impact on the accuracy of dating analyses (Zheng and

Wiens 2015). Priors for our BEAST analysis included: GTR + gamma nucleotide substitution model, an uncorrelated relaxed clock, a yule speciation process, constrained species relationships in accordance with our maximum likelihood tree, and four fossil calibration points (see below). We executed five independent MCMC chains of 300 million generations and sampled every 10000 generations. To assess stationarity, effective sample size, and appropriate burn-in for each individual chain, we inspected each using Tracer v 1.7.1 (Rambaut et al. 2018). Based on visual inspection of the MCMC chains, we determined that four of five the chains required 25% burn-in and one of the five chains required 45% burn-in. These respective burn-in percentages were used when combining both the log files and tree files. The maximum clade credibility (MCC) tree was summarized from the remaining sampled trees using

TreeAnnotator v 1.10.4, and visualized using FigTree v 1.4.3 (Rambaut 2017).

125

In addition to our BEAST dating analysis, we also conducted a dating method that uses penalized likelihood as implemented in treePL (Smith and O’Meara 2012), which allowed us to obtain a rapid and reliable validation of our BEAST output, the latter of which was later found to have convergence issues. Fossil placement followed the

BEAST analysis with the exception that stem fossils were placed on the nodes immediately preceding their respective stems in the treePL analysis (see below). For this dating method, we first generated 100 bootstrap replicates from our molecular dataset and built a RAxML tree for each one. When building these trees, we used the topology recovered from the best maximum likelihood tree (Fig. 5-4) as a constraint to ensure consistent fossil placement for our proceeding step. Next, we dated each of these 100 trees using treePL. We performed one priming step followed by ten cross- validation procedures to ensure convergence on the same smoothing parameter (in our case 0.1). The resulting 100 time-calibrated trees were summarized using

TreeAnnotator with no burn-in. We direct the readers to Lu et al. (2018) and Li et al.

(2019) for further justification of this method.

For our fossil calibrations, we focused on fossils assigned to the extant genera represented in our phylogeny (n=9) to help ensure accurate identification and placement. Descriptions for each of these fossils were then examined to independently assess their taxonomic assignment. We were confident that four of these fossils were correctly assigned to the Coreoidea superfamily. The fossil described as Jadera interita

Cockerell, 1909 is indeed a species of Rhopalidae, but, we could not confidently assign it to either Rhopalinae or Serinethinae, the two subfamilies. Therefore, we assigned it to the stem of Rhopalidae. The fossil described as Spartocera insignis Heer, 1853 has

126

large antenniferous tubercles that are close together and a robust body, superficially resembling Spartocera batatas. Since we only have two representatives of Spartocera in our phylogeny and because we could not confidently say that it is sister to either of the two species we sampled, we placed this fossil on the stem leading to this clade. The fossil taxon Homoeocerus attenuatus Zhang, Sun, & Zhang, 1994 was appropriately assigned to the Homoeocerus genus given its simple hind legs, short square head with clypeus deflexed between the antenniferous tubercles, antennal segment IV not distinctly longer than segment I, and the humeral angles of the pronotum not prominently expanded, among other traits. Since Homoeocerus is not monophyletic

(see results), we assigned the fossil to the stem of the clade that minimally included all

Homoeocerus taxa. Finally, the fossil described as Alydus pulchellus Heer, 1853 is referenced as being most similar to the extant species Hyalymenus tarsatus (Fabricius,

1803) (=Alydus recurvis Herrich-Schäffer, 1846). Indeed, the description of its curved hind tibia and humeral spines on the pronotum, along with other characters, suggests that this fossil taxon should be assigned to Hyalymenus and not Alydus. Since we only have a single extant representative of Hyalymenus in our phylogeny, we placed this fossil at the crown node of Alydinae.

Alydus pulchellus was collected in Baden-Württemberg, Germany and S. insignis was collected in Rabodoj, Croatia (Heer 1853). Both collecting locations can be placed in the Sarmatian Stage (11.6–12.7mya; Harzhauser and Piller 2004) using stratigraphic dating (Heer 1853). Jadera interita was collected in the Green River Formation

(Cockerell 1909), which has been estimated to be 48.5–53.5mya by argon isotope dating (Smith et al 2003). Homoeocerus attenuatus was collected in the Shanwang

127

Formation (Zhang et al. 1994), which has been estimated to be 17.3-21.0mya by argon isotope dating (He et al. 2011). We added these four fossil calibrations into both the

Bayesian (BEAST) and penalized likelihood (treePL) dating analyses. In BEAST, we applied a lognormal distribution with a mean of five and a standard deviation of one to each of the fossil calibrations discussed. These distributions were then offset by the youngest fossil age reported (i.e. A. pulchellus = 11.6mya, S. insignis = 11.6mya, J. interita = 48.5mya, and H. attenuates = 17.3mya).

Statistical Analyses

To investigate the evolutionary origins of sacrificing a limb to escape predation, we conducted ancestral state reconstructions on the latency to autotomize, which was square root transformed. Before conducting these analyses, we investigated whether

Brownian Motion (BM), Ornstein-Uhlenbeck (OU), Early Burst, or white-noise was the best model for our data using geiger v 1.2.2 (Harmon et al. 2008). Corrected Akaike

Information Criterion (AICc) always identified the best model as an OU model of trait evolution (Table 5-1). We then estimated the ancestral rate of autotomy for the ancestor of all leaf-footed bugs assuming an OU model as implemented in phytools v 0.6-60

(Revell 2012). However, we also estimated the ancestral rate of autotomy for the ancestor of all leaf-footed bugs assuming a BM model of trait evolution because current

OU implementations do not specify 95% confidence intervals that enable hypothesis testing. The two models produced quantitatively similar results (BM node estimate =

0.9452*(OU node estimate) + 2.886, R2 = 0.989; Figure 5-1). Since autotomy needs to occur quickly to have an anti-predatory benefit and since 120 s is the longest autotomy cutoff that has previously been considered for this capacity (Cooper et al. 2004,

Emberts et al. 2019), we reason that if our ancestral state was below 120 s and our

128

95% confidence interval excluded 120 s our results would strongly support the fast latency hypothesis. Alternatively, if our ancestor was above 120 s, and our 95% confidence interval excluded 120 s, our results would strongly support the intermediate step hypothesis.

To investigate the factors that contribute to variation in the latency to autotomize across species, we conducted a phylogenetic generalized linear model (pGLM) assuming an OU model of trait evolution as implemented in phylolm v 2.6 (Ho and Ané

2014). We included a body size proxy (i.e. mean pronotal width; continuous), degrees from equator (i.e. the location of the population sampled for this study, continuous), presence of enlarged hind legs (binary), and all pairwise interactions into our full model.

Then, we used AICc to reduce model parameters and identify the best model. In addition to using mean latency to autotomize, which we report here, all statistical analyses were also conducted on median latency to autotomize, which produced qualitatively similar results (Appendix B; Table 5-2). We also separated our data by sex and reran all statistical analyses to determine whether the importance of our variables was sex specific.

Results

Phylogenetic Relationships

Our maximum likelihood tree found high support (bootstraps ≥ 90) for all identified relationships (Figure 5-2). These relationships were largely consistent with those previously published given our taxon sampling. For example, our tree also recovered non-monophyly of the Meropachyinae and Coreinae subfamilies, the

129

Anisoscelini and Hypselonotini tribes, and the Leptoglossus genus.1 However, our increased taxon-sampling of Homoeocerus and inclusion of Mictis revealed the non- monophyly of these genera. Specifically, Prismatocerus auriculatus was nested within

Homoeocerus, and Anoplocnemis was nested within Mictis.

Dating Analyses

Our BEAST dating analysis did not reach an appropriate effective sample size

(i.e. ESS >200), despite running for 1,500,000,000 generations. However, visual inspection of the trace suggested that our runs had reached stationarity (i.e. evenness across the trace after burn-in and a unimodal distribution). Our BEAST analysis placed the origins of the Coreoidea between 51.46–54.90 mya (median 53.06 mya), Coreidae +

Alydidae between 24.02–49.24 mya (median 33.53 mya), Coreidae between 23.11–

40.21 mya (median 28.57 mya), and Alydidae between 14.07-27.84 mya (median 18.26 mya; Figure 5-3). Our treePL dating analysis was largely congruent with the BEAST analysis and placed the origins of Coreoidea at 48.50 mya (median 48.50 mya),

Coreidae + Alydidae between 32.78–35.22 mya (median 34.00 mya), Coreidae between

27.81–29.61 mya (median 28.49 mya), and Alydidae between 28.12–29.90 mya

(median 29.10 mya; Figure 5-4). Given that both analyses produced similar results

(treePL node age = 0.9446*(BEAST node age) + 3.0734, R2 = 0.821) and that our

BEAST analysis did not reach an appropriate ESS, all subsequent analyses used the dated tree from treePL (see Appendix C for discussion of dating analyses).

130

Autotomy

The ancestor of leaf-footed bugs autotomized their hind limbs slowly, with an estimated mean latency to autotomize of 1,186 s (back transformed from 34.434 √s) under an OU model of trait evolution (Figure 5-5). Under BM, the ancestor of leaf-footed bugs was estimated to have a mean latency to autotomize of 1,293 s (back transformed from 35.954 √s) with a 95% confidence interval that ranges from 137–3,624 s (back transformed from 11.706-60.202 √s). From this slow ancestral rate, autotomizing quickly

(< 120s) arose nine times (Figure 5-5). Three of these transitions occurred on internal branches, and six of these transitions occurred on terminal branches. These results support the intermediate step hypothesis and reject the fast-latency hypothesis. When we separated our data by sex and re-ran our analyses, males and females returned qualitatively similar results, but our confidence in those results varied (Appendix D).

Body size, distance from the equator, and the presence of enlarged hind legs all influenced the mean latency to autotomize (Table 5-3; Figure 5-6A, B). Smaller species, and those closer to the equator, autotomized their limbs more quickly (respectively, t=2.932, p=0.005 and t=2.359, p=0.022), as hypothesized (Figure 5-6A). However, contrary to our hypothesis, species with enlarged hind legs autotomized more quickly for their given size and locality (t=-2.090, p=0.041; Figure 5-6B). When analyzing all the data, none of the pairwise interactions were included in the best model (Table 5-3).

Females returned qualitatively similar results when we separated our data by sex and reran our analyses (Figure 5-6C, D; Tables 5-3, 5-4). The male data, however, was a bit more nuanced (Figure 5-6E, F). The best model for the male data included all three main effects as well as the pairwise interactions that included body size (Tables 5-3, 5-

5). Visual inspection of the male data (Figure 5-6E, F) highlighted the potential for our

131

best model to be driven by a single data point, Pephricus paradoxus. Thus, we removed this data point and reran our analysis to determine the sensitivity of the male results.

We found that removing this data point resulted in the interaction between body size and enlarged hind legs being dropped from the best model, but the interaction between body size and latitude remained (Figure 5-7; Tables 5-2, 5-6).

Discussion

Our study clearly indicates that sacrificing a limb to escape predation evolved via an intermediate step in leaf-footed bugs. All our analyses supported the conclusion that the leaf-footed bug ancestor took more than 15 minutes to autotomize hind legs. At this rate, the ancestor probably used autotomy to reduce the cost of injury or to escape a fouled molt. While the ancestor of leaf-footed bugs autotomized slowly, we found that

20% (12 out of 62) of species autotomize quickly enough to use autotomy in an anti- predatory capacity (i.e. a mean latency to autotomize below 120 s). These results suggest that autotomy to escape predation is a co-opted benefit (i.e. exaptation), revealing one way that sacrificing a limb to escape predation may arise. Our results also highlight that dropping a limb slowly is a pervasive trait found throughout the clade, as more than 60% of the investigated species had a mean latency to autotomize that was greater than 15 minutes. Since we have shown that dropping a limb slowly can be an integral step in the evolution of sacrificing a limb to escape predation, future work should continue to quantify the fitness consequences of autotomizing slowly across other clades.

Latitude, body size, and the presence of enlarged hind legs all influence the rate at which species autotomize in this clade. In addition to these main effects, our data from male specimens also reveal the possibility of a body size by latitude interaction as

132

well as a body size by enlarged hind leg interaction. These results provide intriguing insights into the evolution of autotomy.

Species closer to the equator autotomize more quickly, which suggests that leaf- footed bugs are more likely to use autotomy in an anti-predatory capacity when predator diversity and abundance is highest. This result is consistent with other studies that have found a correlation between the strength of predatory pressure and the ease of tail autotomy in lizards (Cooper et al. 2004, Brock et al. 2015, Lin et al. 2017, but see Itescu et al. 2017). Moreover, this result agrees with other latitudinal studies, in which anti- predatory behaviors are heightened closer to the equator (Møller and Liang 2013, Diaz et al. 2013, Samia et al. 2015). However, because previous studies have mostly investigated flight initiation distance (i.e. FID; Møller and Liang 2013, Diaz et al. 2013,

Samia et al. 2015, but Laurila et al. 2008 noted differences in other behaviors), our results highlight the possibility of a more general pattern, that anti-predatory behavior becomes more relaxed as organisms get farther away from the equator. The idea that anti-predatory traits correlate with latitudinal gradients is not new (Schemske et al.

2009), but previous work has mostly focused on morphology (e.g. Vermeij 1978, Palmer

1979).

We also found that larger leaf-footed bugs autotomize more slowly, which follows a similar pattern observed in another insect clade – orthopterans (i.e. grasshoppers, katydids, and allies; Bateman and Fleming 2008). Although this pattern has been observed before, much remains unknown about the mechanisms that drive this association. There are a few notable differences between larger species of insects when compared to smaller species of insects that could be driving this trend. First,

133

larger species of insects face different predatory pressures; larger individuals face more predation by birds and less predation by invertebrates (Remmel et al. 2011). If hind leg autotomy is a less effective anti-predatory strategy against birds, for example, then predatory pressure may not be selecting for a faster rate of autotomy in larger taxa.

Second, as insects get larger, the relative energetic costs of flight may not increase in direct proportion to size (Harrison and Roberts 2000; Niven and Scharlemann 2005).

Under such a scenario, larger insects might rely more on their legs for locomotion, increasing the indirect cost of leg autotomy and constraining its evolution. Finally, larger individuals likely have larger autotomy fracture planes (Z. E. unpublished data), which may make it harder for them to autotomize (i.e. a morphological constraint). Although body size has been shown to constrain autotomy in orthopterans, and now coreids, it does not influence autotomy in lizards (Zani 1996). Since lizards can regenerate, while orthopterans and coreids cannot, these differing patterns might be explained by regeneration, which potentially eliminates some constraints on the evolution of autotomy. However, insects and lizards vary in many ways, including the appendage that they autotomize, and the differences in the association between body size and autotomy could be due to any number of factors.

The influence of our latitudinal gradient on the latency to autotomize also becomes less informative as species get larger. This trend is clearly driven by the males

(Figure 5-6C and 2E, Table 5-5 and 5-6). Such a pattern could be a result of males in larger species being less willing to use autotomy in an anti-predatory capacity (i.e. larger species autotomize more slowly). As a result, distance from the equator (i.e. a proxy for the strength of predatory pressure) simply explains less variation in the latency

134

to autotomize as taxa get larger. However, this hypothesis itself does not explain why this interaction is sex specific. One possible explanation is that males and females face different predatory pressure. In the sweet potato bug, Physomerus grossipes – a species included in this study – females guard their eggs against parasitoids

(Hemmingsen 1947). Egg guarding in this species confines a female to a single location during reproduction, which may increase the female’s chances of becoming prey. We noticed that it made collecting them much easier. Most of these females (70%) autotomized their hind legs within 120s, despite their large size (mean PW of 6.02mm).

Few males of this species (13%), however, autotomized as quickly. If similar sex differences in predatory pressure occur in other large species, this could explain why distance from the equator (i.e. a proxy for the strength of predatory pressure) influences latency to autotomize in larger females, but not males.

We also found that coreid species with enlarged hind legs autotomized more quickly, contrary to our hypothesis. Despite the likely costs associated with losing these legs, the benefits of autotomy must outweigh the costs in these species. Rapid autotomy of enlarged hind legs may be adaptive for several reasons. First, since the autotomizable appendages are larger, they are likely to have a higher probability of being grabbed by a predator. This would increase the selective pressure for these limbs to be dropped quickly. Second, autotomizing larger hind legs may increase the efficacy of autotomy. After an individual successfully uses autotomy in an anti-predatory context, the predator then needs to decide whether to continue pursuit of their prey. By autotomizing a larger proportion of their body, prey may ensure that the predators are more content with their meal at hand, and hence reduce the frequency with which the

135

predator continues pursuit. Third, autotomy in insects is commonly used to escape a fouled molt, a form of non-predatory entrapment. Since these hind legs are elaborated, they may be more susceptible to entrapment during the molting process (Maginnis

2008). None of these benefits are mutually exclusive and any combination of them could outweigh the costs of losing an enlarged hind leg.

Finally, the inclusion of a body size by enlarged hind leg interaction as an explanatory variable for our male autotomy data is the result of a single unique species,

Pephricus paradoxus. Despite male autotomy in Pe. paradoxus being a deviant data point, it likely captures biological reality. Much remains unknown about the biology of

Pe. paradoxus, but this Phyllomorphini species is related to Phyllomorpha laciniata, the golden egg bug. Golden egg bugs are both morphologically and behaviorally unique.

Most notably, this species, together with Pe. paradoxus, has a flared and enlarged pronotum. Because of this unique morphology, pronotal width is likely a poor proxy for their body size relative to other coreids. In addition to this morphological difference, Ph. laciniata also exhibits paternal care (Reguera and Gomendio 1999, Gomendio and

Reguera 2001), which may increase a male’s chances of being captured by a predator.

This could create increased selection for an anti-predator defense that enables males to escape the grasp of a predator, such as autotomy.

The methodology we employed to investigate the evolution of autotomy has strengths as well as limitations. We initially believed that our sampling, which included two species from Rhopalidae and one species from Lygaeidae (taxa outside of leaf- footed bugs), would recover the evolutionary origins of autotomy. However, every taxon met our criteria for having the ability to autotomize. This could have been problematic if

136

we found support for the fast latency hypothesis (i.e. if the leaf-footed bug ancestor had been reconstructed as autotomizing in less than 120 s) because it would have been unclear whether a more distant ancestor could have autotomized slowly. In that case, we would have been unable to reject the intermediate step hypothesis. Another potential issue with our taxon sampling was that it led to discontinuity in our sampling of distance from the equator. Although having taxa evenly sampled across our latitudinal gradient would have been ideal, our bimodal sampling is not problematic for our statistical analyses (i.e. our analyses met linear model assumptions) nor our biological interpretation (i.e. there were species close to the equator and far from the equator).

Additionally, we do not know the mechanistic causes of the patterns that we have shown here. For example, we found that distance from the equator explained variation in the latency to autotomize and we postulate that this is driven by predatory pressure.

However, it is possible that other factors that correlate with latitude may be responsible for this trend, such as seasonality of temperature. Now that these associations have been identified, more focused studies should identify the mechanism behind them.

Finally, to test our alternative hypotheses on the origins of rapid autotomy, we had to assign a cutoff value; in this case, we selected 120s. We a priori selected this time point because it is the longest time cutoff that has previously been used (Cooper et al. 2004,

Emberts et al. 2019) and it made biological sense given the predators and mobility of leaf-footed bugs. Even if one decided that a lower threshold such as 60s or 10s would be more appropriate for this clade, our results would still qualitatively hold since our

95% confidence intervals excluded these values.

137

Autotomy is one of the most extreme forms of anti-predatory defense, dramatically illustrating the importance of survival in the context of natural selection.

Nonetheless, fundamental questions of how this extreme trait evolves have remained unanswered. Here, we show that some leaf-footed bugs evolved the ability to autotomize their limbs rapidly enough to escape predation via an intermediate latency step. However, this is just one possible pathway by which rapid autotomy may evolve and future studies should investigate how autotomy has evolved in other clades (e.g. walking sticks, salamanders, decapod crustaceans, spiders, harvestmen). Moreover, future studies should seek to investigate the evolutionary origins of autotomy itself as opposed to the origins of autotomizing quickly. Studies that investigate the evolutionary morphology of autotomy fracture planes and/or the evolutionary physiology of autotomy would provide valuable insights. Our study also highlights the possibility of a broad latitudinal pattern – that anti-predatory behaviors become more relaxed farther away from the equator. Future work should continue to investigate this association with autotomy, as well as other anti-predatory behaviors, such as vigilance and death feigning. Such studies will continue to shed light on the factors that contribute to the behavioral diversity in anti-predator responses observed throughout Animalia.

138

Table 5-1. An OU model of trait evolution best explains how the latency to autotomize evolved. Note lowest AICc value in each column. AICc = Corrected Akaike Information Criterion, BM = Brownian Motion, OU = Ornstein-Uhlenbeck, EB = Early Burst, white = white-noise AICc all Model AICc male data AICc female data data white 548.1 510.2 529.2 BM 568.0 518.0 551.2 OU 540.1 499.5 525.0 EB 570.2 520.3 553.4

139

Table 5-2. Our model selection criterion (AICc) determined that the best model should include body size, distance from the equator (Latitude), and the presence of enlarged hind legs (Costly). When we subsetted our data to only include males, the best model also included the pairwise interactions with body size. However, this result was sensitive to a single data point (Pephricus paradoxus). Note lowest AICc value in each column. AICc = Corrected Akaike Information Criterion. AICc all ΔAICc all AICc male ΔAICc male AICc female ΔAICc female Model data data data data data data Latitude 536.3 5.2 495.0 2.5 520.5 8.3 Costly 540.7 9.6 501.0 8.5 523.6 11.4 Body Size 536.0 4.9 497.4 4.9 522.1 9.9 Body Size+Latitude 532.9 1.8 494.1 1.6 518.4 6.2 Body Size+Latitude+Body Size:Latitude 534.3 3.2 494.2 1.7 520.5 8.3 Latitude+Costly 537.0 5.9 497.0 4.5 518.8 6.6 Latitude+Costly+Latitude:Costly 538.5 7.4 498.6 6.1 520.6 8.4 Costly+Body Size 534.6 3.5 498.8 6.3 517.5 5.3 Costly+Body Size+Costly:Body Size 536.4 5.3 497.8 5.3 519.7 7.5 Latitude+Costly+Body Size 531.1 0.0 495.6 3.1 512.2 0.0 Latitude+Costly+Body Size+Costly:Body Size 532.8 1.7 493.7 1.2 514.4 2.2 Latitude+Costly+Body Size+Latitude:Costly 532.7 1.6 497.3 4.8 513.9 1.7 Latitude+Costly+Body Size+Body Size:Latitude 532.2 1.1 495.6 3.1 514.4 2.2 Latitude+Costly+Body Size+Costly:Body Size+Latitude:Costly 534.4 3.3 494.8 2.3 516.3 4.1 Latitude+Costly+Body Size+Costly:Body Size+Body Size:Latitude 533.9 2.8 492.5 0.0 516.7 4.5 Latitude+Costly+Body Size+Latitude:Costly+Body Size:Latitude 534.5 3.4 497.9 5.4 516.3 4.1 Latitude+Costly+Body Size+Latitude:Costly+Body Size:Latitude+Costly:Body Size 536.1 5.0 496.2 3.7 518.8 6.6

140

Table 5-3. Our model selection criterion (AICc) determined that the best model should include body size, distance from the equator (latitude), and the presence of enlarged hind legs (Costly). Note that when we subsetted our data to only include males, the best model also included the pairwise interactions with body size. Note lowest AICc value in each column. AICc = Corrected Akaike Information Criterion AICc male AICc male AICc all AICc all AICc male AICc male data minus data minus AICc female AICc female data Model data data data data Pephricus Pephricus data mean median mean median mean median paradoxus paradoxus mean median Latitude 536.3 545.3 495.0 509.7 484.9 500.2 520.5 533.7 Costly 540.7 549.4 501.0 513.5 491.1 504.2 523.6 531.9 Body Size 536.0 545.1 497.4 507.5 484.9 495.1 522.1 532.1 Body Size+Latitude 532.9 542.9 494.1 506.4 481.2 493.8 518.4 531.3 Body Size+Latitude+Body Size:Latitude 534.3 544.4 494.2 506.4 481.0 493.4 520.5 533.1 Latitude+Costly 537.0 546.3 497.0 511.8 486.6 502.2 518.8 529.6 Latitude+Costly+Latitude:Costly 538.5 547.7 498.6 513.3 488.0 503.6 520.6 531.2 Costly+Body Size 534.6 544.8 498.8 509.1 484.9 495.6 517.5 524.9 Costly+Body Size+Costly:Body Size 536.4 545.2 497.8 509.1 487.1 497.8 519.7 527.1 Latitude+Costly+Body Size 531.1 542.4 495.6 508.2 481.0 494.3 512.2 523.2 Latitude+Costly+Body Size+Costly:Body Size 532.8 542.4 493.7 507.6 483.1 496.6 514.4 525.5 Latitude+Costly+Body Size+Latitude:Costly 532.7 543.9 497.3 509.8 482.4 495.6 513.9 524.5 Latitude+Costly+Body Size+Body Size:Latitude 532.2 544.0 495.6 507.9 480.3 493.3 514.4 525.0 Latitude+Costly+Body Size+Costly:Body 534.4 543.8 494.8 508.5 484.5 497.9 516.3 526.9 Size+Latitude:Costly

Latitude+Costly+Body Size+Costly:Body 533.9 543.8 492.5 506.4 482.2 495.6 516.7 527.4 Size+Body Size:Latitude Latitude+Costly+Body Size+Latitude:Costly+Body 534.5 546.0 497.9 510.3 482.7 495.7 516.3 526.8 Size:Latitude Latitude+Costly+Body Size+Latitude:Costly+Body 536.1 545.8 496.2 509.7 484.7 498.1 518.8 529.3 Size:Latitude+Costly:Body Size

141

Table 5-4. Best model for the mean female data. Estimated Standard t p Coefficient Error Intercept 7.24364 7.54845 0.95962 0.341447 Latitude 0.63452 0.22876 2.77373 0.007553 Body 2.94169 0.94109 3.12584 0.002831 Size Costly -16.80564 5.12578 -3.27865 0.001812

142

Table 5-5. Best model for the mean male data. Estimated Standard t p Coefficient Error Intercept 10.143178 15.851696 0.63988 0.525115 Latitude 1.483707 0.553177 2.682154 0.009833 Body Size 2.698446 3.145863 0.857776 0.395029 Costly -31.93852 12.66494 -2.521806 0.014841 Latitude:Body -0.175682 0.097263 -1.80626 0.07678 Size Costly:Body Size 5.551336 2.429915 2.28458 0.026531

143

Table 5-6. Best model for the mean male data, minus Pephricus paradoxus. Estimated Standard t p Coefficient Error Intercept -6.470711 14.53343 -0.445229 0.658038 Latitude 1.37477 0.536069 2.564539 0.013318 Body Size 7.520353 2.763235 2.721576 0.008867 Costly -10.397952 5.961791 -1.744099 0.087167 Latitude:Body -0.16018 0.094221 -1.700055 0.095212 Size

144

Figure 5-1. Comparing ancestral state reconstructions when assuming an Ornstein- Uhlenbeck (OU) model of trait evolution to a Brownian Motion (BM) model of trait evolution for the all data combined dataset. The OU estimates slightly quicker rates of autotomy (based on mean latency to autotomize) at ancestral nodes.

145

Figure 5-2. RAxML best tree with bootstrap values labeled at the nodes.

146

Figure 5-3. Dated BEAST tree with median node ages labeled and bars denoting the 95% highest probability density interval.

147

Figure 5-4. Dated TreePL tree with median node ages labeled and error bars that show the range of age estimates across 100 bootstrap trees. An absent error bar at a node means that all 100 bootstrap trees converged on the same age for that node.

148

Figure 5-5. The ancestor of leaf-footed bugs autotomized their hind limbs slowly. From there, rapid autotomy evolved nine times (white dashes). Leaf-footed bugs that autotomize slowly likely use(d) autotomy to reduce the cost of injury or to escape non-predatory entrapment, but not to escape predation. This suggests that autotomy to escape predation is a co-opted benefit (i.e. exaptation), revealing one way that sacrificing a limb to escape predation may arise. A visualization of our ancestral state reconstruction under Brownian Motion (BM) illustrates how latency to autotomize likely evolved in this clade. Leaf- footed bug drawings are modified from Distant, 1893. See Figure 5-8 for a visualization of our ancestral state reconstruction that includes tip labels.

149

A B 0.5

3000

) 30300000 0.4 latitude

s (

z i

m 20

u o

u 10 as.factor(SSW)

A 2000

2000 o

u 2000

T a

t n

Degrees froe m

t y

asa .factor(latitude.2)

a c

3000 eqlatitudeuator L

n n

n 0

a t 30e Enlarged e 30

a 1000 M

l 1000

1M 000 e

1 hind legs

i n

a 20

m 20 NA

e o

t No

M t

u 1100 A

2000 Yes o

T 0 0

0 y c 2.5 5.0 7.5 10.0 12.5

n 2.5 5.0 7.5 10.0 12.5 e

t 2.5 5.0 Body 7Siz.5e 10.0 12as.5 .factor(latitude.2)2.5 5.0 Body Siz7.5e 10.0 12.5

a L Body size (pronotal width, mm) Body size (pronotal width, mm)

n 0 a e 1000 C D M 1

) NA

(

z 3000 i 30300000 latitude

m 30

z z

0 i

i t

m 20

2.5 5.0 7.5 10.0 12.5 t

u u

o 10 as.factor(SSW)

t A

Body Size

2000 2000

2000 o

T T

n e

n 1

e t

t as.factor(latitude.2)

n n

n 0

e e e 1000 1000 M

1M 000 1

l NA

e F 00 0 55.0 1100.0 55.0 1100.0 Body Size Body Size Body size (pronotal width, mm) Body size (pronotal width, mm)

E F

)

(

e 3000

3000 3000 z

i latitude

m e

e 30

t t

t 20

u u

as.factor(SSW)

o 20200000 200010

c c

n n

e 1

t L

asL .factor(latitude.2)

n e e 0

a 10100000 M

M 1000 e

a M

00 0

22.5.5 55.0.0 77.5.5 110.00.0 112.52.5 22.5.5 55.0.0 77.5.5 110.00.0 112.52.5 Body Size Body Size Body size (pronotal width, mm) Body size (pronotal width, mm)

150

Figure 5-6. Species that are smaller and closer to the equator autotomize more quickly, and the degree to which having an enlarged hind femur influences the mean latency to autotomize is sex and size specific. Circle coloration corresponds to distance from the equator, in degrees, from which species were collected (A, C, E). Since degrees from equator had a bimodal distribution (Figure 5-9) we plotted linear regressions associated with each mode to help visualize the data. The solid line includes species relatively far from the equator (>25 degrees), whereas the dashed line includes species closer to the equator (<10 degrees). Note that for a given body size, species closer to the equator often autotomize more quickly. Open triangles and the corresponding solid line regressions denote presence of enlarged hind legs, while closed triangles and dashed line regressions correspond to the absence of enlarged hind legs (B, D, E). Note that for a given body size, species with enlarged hind legs generally autotomize more quickly when analyzing all the data (B) and female only data (D). However, for the male only data (E), there is an interaction between body size and the presence of enlarged hind legs. This interaction is strongly driven by a single data point (Table 5-2; Figure 5-7). Untransformed autotomy data was used in this figure to aid data interpretation.

151

A B

)

(

30300000 e e

z 3000 z

i latitude

m m

e 30

t o

t 20

o t

u as.factor(SSW)

20 200000 10

o o

o 2000

c n

Degrees frc om

n a

3000 e eqlatitudeuator 1

e L

t as.factor(latitude.2)

n Enlarged

l a

30 L

e 30 0 n 1000 hind legs

10M 00

e a

n 1000

e i

20a 1 m

2e 0 No

l o

t a

u 1100 Yes

A M

2000 o

T 0

0 0

y c

n 2.5 5.0 7.5 10.0 12.5

e 2.5 5.0 7.5 10.0 12.5

t 2.5 5.0 Body 7Siz.5e 10.0 1as2.5.factor(latitude2.2).5 5.0 7.5 10.0 12.5 a

L Body Size

n Body size (pronotal width, mm) 0 Body size (pronotal width, mm) a e 1000 M Figure 5-7. Small males with enlarged hind legs1 found near the equator autotomize quickly. Here, we visualize the maleNA autotomy data while removing a single data point (Pephricus paradoxus, which has a mean latency to autotomize of 97 s, a body size of 10.3 mm, was collected 26.3 degrees away from equator, 0 and does not have enlarged hind legs). Notice that the interactions are not as 2.5 5.0 7.5 10.0 12.5 distinctBody Sizaftere removing our P. paradoxus data point when compared to Figure 5-6E and F. Circle coloration corresponds to distance from the equator, in degrees, from which species were collected (A). Open triangles and the corresponding solid lined regressions denote presence of enlarged hind legs, while closed triangles and dashed lined regressions correspond to the absence of enlarged hind legs (B). Untransformed autotomy data was used in this figure to aid data interpretation.

152

Oncopeltus cingulifer Harmostes serratus Jadera haematoloma Mutusca brevicornis Leptocorisa acuta Stenocoris tipuloides Stenocoris furcifera Stenocoris filiformis Hyalymenus longispinus Melanacanthus margineguttatus Melanacanthus scutella ris Riptortus pedestris Hamedius incarnatus Neomegalotomus rufipes Alydus pilosulus Piezogaster calcarator Thasus neocalifornicus Spartocera batatas Spartocera fusca Physomerus grossipes Acanthocoris sp. Acanthocoris scaber Homoeocerus marginellus Prismatocerus auriculatus Homoeocerus angulatus Pephricus paradoxus Cletomorpha nyasana Cletus binotulatus Cletus sp. Elasmopoda alata Mictis profana Mictis longicornis Anoplocnemis curvipes Anoplocnemis phasianus Althos obscurator Zicca taeniola Euthochtha galeator Merocoris typhaeus Chariesterus antennator Hypselonotus lineatus Hypselonotus punctiventris Acanthocephala declivis Acanthocephala thomasi Acanthocephala femorata Acanthocephala terminalis Chelinidea vittiger Anisoscelis alipes Chondrocera laticornis Anasa scorbutica Anasa andresii Anasa tristis Leptoscelis quadrisignatus Leptoscelis tricolor Narnia femorata Phthiacnemia picta Leptoglossus gonagra Leptoglossus fulvico rnis 1 latency to autotomize ( √s) 60 Leptoglossus ashmeadi Leptoglossus corculus Leptoglossus phyllopus fast slow Leptoglossus oppositus autotomy autotomy Leptoglossus zonatus Figure 5-8. Ancestral state reconstruction for the mean latency to autotomize assuming Brownian Motion (BM) with tip labels.

153

Figure 5-9. Degrees from equator, our latitudinal gradient, had a bimodal distribution. Therefore, we categorized species that were collected within 10 degrees of the equator as species close to the equator and those farther than 25 degrees away from the equator as species far from the equator. This categorization was only used to help visualize the potential interaction between body size and degrees from the equator in Figure 5-6, and not used in statistical analyses.

154

APPENDIX A STOCHASTIC CHARACTER SIMULATIONS REVEAL THAT AUTOTOMY AND AUTOTOMIZABLE LIMB ELABORATIONS HAVE EVOLVED MULTIPLE TIMES THROUGHOUT ANIMALIA

To determine whether autotomy has evolved multiple times throughout Animalia, we mapped the ability to autotomize on to the 2008 Dunn and colleagues (Dunn et al

2008) animal tree of life and simulated stochastic character histories. We were specifically interested in estimating the number of times autotomy evolved at the phyla level. Therefore, if at least one species was known to autotomize in a phylum (Arnold

1984, Johnson 1988, Dejean et al. 1999, Fleming et al. 2007, Urata et al. 2012, Boll et al. 2015, Hodgkin et al. 2017), every species used to represent that phylum was denoted as having the ability to autotomize. We considered this to be a conservative approach because we were interested in determining whether autotomy evolved more than once (i.e. we wanted to conservatively estimate the minimum number of times autotomy evolved).

We used the Dunn et al. (2008) molecular dataset (https://treebase.org/treebase- web/search/study/matrix.html?id=2020) to reconstruct their tree because the exact phylogenetic trees (i.e. a tree that included both branch lengths and relationships) reported in their paper were not publicly available. Specifically, we ran a maximum likelihood analysis assuming a PROTGAMMAWAG model of amino acid change and constrained the analysis with the Dunn et al. (2008) publicly accessible cladogram

(https://treebase.org/treebase-web/search/study/trees.html?id=2020). This analysis was conducted using RAxML (Stamatakis 2014). The phylogenetic tree that we generated was similar to the RAxML phylogram presented in Dunn et al. (2008) figure 1. However, there were a few differences. Our analysis found Saccharomyces cerevisiae sister to all

155

others, while Dunn et al. found S. cerevisiae and Cryptococcus neoformans sister to all others. Additionally, the branch length associated with these two taxa differed. Since these two species were both outgroup species, we simply pruned them from our tree before we ran our subsequent analyses (Figure A-1).

To get an estimate of the number of times autotomy (a binary, discrete character) evolved, we ran 1000 stochastic character simulations (i.e. stochastic character mapping) using the all-rates-different (ARD) model of trait evolution. This analysis was conducted with the phytools package (Revell 2012) in R (R Core Team 2018). Every simulation estimated autotomy to have independently evolved more than once (range: 9

– 19), with an average of 11.917 origins.

We also investigated whether the elaboration of autotomizable appendages had multiple independent origins within Animalia. Specifically, we investigated whether post- autotomy appendage movement had independently evolved multiple times. If at least one species was known to have post-autotomy appendage movement within a phylum, every species used to represent that phylum was denoted as having post-autotomy appendage movement. We took this approach because we were interested in conservatively estimating the minimum number of times post-autotomy appendage movement evolved. Once presence/absence of this character was assigned, we ran another 1000 stochastic character simulations (i.e. stochastic character mapping) using the all-rates-different (ARD) model of trait evolution in phytools (Revell 2012). Every simulation estimated post-autotomy appendage movement to have independently evolved more than once (range: 2 – 5), with an average of 2.03 origins.

156

Figure A-1. A visual representation of a single stochastic character simulation for the ability to autotomize

157

APPENDIX B BODY SIZE, LATITUDE, AND THE PRESENCE OF ENLARGED HIND LEGS ALL EXPLAIN VARIATION IN THE RATE AT WHICH SPECIES AUTOTOMIZE ACROSS THIS CLADE, REGAURDLESS OF OUR MEASURE OF CENTRAL TENDENCEY

The main text reports analyses conducted on mean latency to autotomize. Here, we report the results of our analyses conducted on median latency to autotomize, which produced qualitatively similar results.

Our best model (i.e. lowest AICc model) when analyzing all the data included body size, distance from the equator, and the presence of enlarged hind legs. Smaller species, and those closer to the equator, autotomized their limbs more quickly

(respectively, estimated coefficient= 2.940, t=2.464, p=0.017 and estimated coefficient=

0.495, t=2.106, p=0.040). Moreover, species with enlarged hind legs autotomized more quickly for their given size and locality, but this effect was not statistically significant

(estimated coefficient= -11.572, t=-1.672, p=0.100). A competing model, which had the same AICc score, included these three main effects as well as the interaction between body size and the presence of enlarged hind legs (Table S2).

Body size, distance from the equator, and the presence of enlarged hind legs all influenced median latency to autotomize when analyzing only the female data. Smaller species autotomized more quickly (estimated coefficient= 3.359, t=1.143, p=0.005), those farther from the equator tended to autotomize more slowly (estimated coefficient=

0.503, t=-1.974, p=0.053), and species with enlarged hind legs autotomized more rapidly (estimated coefficient= -23.120, t=-3.506, p=0.001). There were no competing models for the female data (Table S2).

Like the mean male data, the median male data was a bit more nuanced. Our best male model included the following variables: latitude, body size, the presence of

158

enlarged hind legs, a body size by latitude interaction, as well as a body size by enlarged hind leg interaction (Table S2). However, visual inspection of the median data also suggested that the body size by enlarged hind leg interaction was strongly driven by a deviant data point, Pephricus paradoxus, and removal of this species resulted in the interaction being dropped from the best model (Table S2). Under this model, neither enlarged hind legs nor the interaction between body size and distance from the equator were significant (respectively, estimated coefficient= -10.370, t=-1.525, p=0.133 and estimated coefficient= -0.184, t=-1.746, p=0.087). However, as main effects, both distance from the equator and body size were (respectively, estimated coefficient=

1.420, t=2.362, p=0.022 and estimated coefficient= 9.150, t=2.953, p=0.005).

It is worth noting that presence of enlarged hind legs was included as an explanatory variable in all the best models, but its significance varied, which is likely a result of difference in effect sizes observed between the different data sets. The difference in effect size may be explained by the fact that enlarged male hind legs are more costly to lose when compared to enlarged female hind legs.

159

APPENDIX C OUR DATING ANALYSES REVEALED YOUNGER AGE ESTIMATES THAN PREVIOUSLY REPORTED

Our BEAST dating analysis placed the origins of the Coreoidea between 51.46–

54.9 mya (median 53.06 mya) and Coreidae + Alydidae between 24.02–49.24 mya

(median 33.53 mya). Whereas our treePL dating analysis placed Coreoidea at 48.5 mya

(median 48.5 mya) and Coreidae + Alydidae between 32.7785–35.2195 mya (median

33.9972 mya). Both dating analyses identify these clades as originating more recently than previously reported (Wang et al. 2016, Li et al. 2017, Johnson et al. 2018, Liu et al.

2019). Wang et al (2016) found Coreoidea originated around 153-166 mya (median

160) and Coreidae + Alydidae around 112-154mya (median 138). This study (Wang et al. 2016) used BEAST to date their tree and included 15 fossil calibrations. Li et al

(2017) estimated the origin of Coreoidea to be 157 mya (range 143-168) and Liu et al

(2019) estimated the Coreoidea origin to be 162 mya (range 139-179); both studies dated their trees with PhyloBayes. Finally, Johnson and colleagues (2018) estimated

Coreoidea to have originated around 93 mya and Coreidae + Alydidae around 72 mya using MCMCTree (Bayesian) molecular dating analysis as implemented in PAML.

Differences between our age estimates and those previously reported are most likely a result of using different fossil calibrations. None of the fossil calibrations in these other studies included Coreidae or Alydidae fossils, while our study included three. Moreover, differences in the specific dating analyses and their specified parameters, taxon sampling, and differences in how the molecular data were obtained likely contribute to variation in the age estimates as well.

160

APPENDIX D BOTH MALE AND FEMALE LEAF-FOOTED BUG ANCESTORS AUTOTOMIZED THEIR HIND LEGS SLOWLY

Ancestral state reconstruction using only the male data estimates the ancestor of leaf-footed bugs to have a mean latency to autotomize of 1,552 s (back transformed from 39.396 √s assuming an OU model of trait evolution and 39.398 √s assuming a BM model) with a 95% confidence interval that ranges from 289–3,820 s (back transformed from 16.993–61.804 √s assuming a BM model of trait evolution). With only the female data the ancestor of leaf-footed bugs is estimated to have a mean latency to autotomize of 947 s (back transformed from 30.779 √s) assuming an OU model of trait evolution and 1,024 s (back transformed from 31.995 √s) when assuming a BM model of trait evolution. The 95% confidence interval for the leaf-footed bug ancestor using the female data ranges from 29–3,437 s (back transformed from 5.361–58.629 √s) under BM.

161

LIST OF REFERENCES

Abello, P., C. G. Warman, D. G. Reid, and E. Naylor. 1994. Chela loss in the shore crab Carcinus maenas (Crustacea: Brachyura) and its effect on mating success. Mar. Biol. 121:247–252.

Abràmoff, M. D., P. J. Magalhães, and S. J. Ram. 2004. Image processing with ImageJ. Biophotonics international 11:36–42.

Allen, B.J. and J. S. Levinton. 2007. Costs of bearing a sexually selected ornamental weapon in a fiddler crab. Funct. Ecol. 21, 154–161.

Allen, P.E., and C. W. Miller. 2017. Novel host plant leads to the loss of sexual dimorphism in a sexually selected male weapon. Proc. R. Soc. Lond. B 284:20171269.

Argaez, V., I. Solano-Zavaleta, and J. J. Zúñiga-Vega. 2018. Another potential cost of tail autotomy: tail loss may result in high ectoparasite loads in Sceloporus lizards. AMRE 39:191–202.

Anderson, M. 1994. Sexual Selection. Princeton University Press, Princeton, NJ

Arnold, E. N. 1984. Evolutionary aspects of tail shedding in lizards and their relatives. J. Nat. Hist. 18:127–169.

Arnott, G., R. W. 2009. Assessment of fighting ability in animal contests. Anim. Behav. 77:991–1004.

Baranowski, R. M., and J. A. Slater. 1986. Coreidae of Florida (Hemiptera: Heteroptera). Arthropods Florida Neighboring L. Areas.

Barr, J. I., R. Somaweera, S. S. Godfrey, and P. W. Bateman. 2018. Increased tail length in the King’s skink, Egernia kingii (Reptilia: Scincidae): an anti-predation tactic for juveniles? Biol. J. Linnean Soc. doi: 10.1093/biolinnean/bly196.

Bateman, P. W., and P. A. Fleming. 2005. Direct and indirect costs of limb autotomy in field crickets, Gryllus bimaculatus. Anim. Behav. 69: 151–159.

Bateman, P. W., and P. A. Fleming. 2006a. Sex and the single (-eared) female: leg function, limb autotomy and mating history trade-offs in field crickets (Gryllus bimaculatus). Biol. Lett. 2: 33–35.

Bateman, P. W., and P. A. Fleming. 2006b. Increased Susceptibility to Predation for Autotomized House Crickets (Acheta domestica). Ethology 112:670–677.

162

Bateman, P.W., and P. A. Fleming. 2006c. Sex, intimidation and severed limbs: The effect of simulated predator attack and limb autotomy on calling and emergence behaviour in the field cricket Gryllus bimaculatus. Behav. Ecol. Sociobiol. 59:674–681.

Bateman, P. W., and P. A. Fleming. 2008. An intra- and interspecific study of body size and autotomy as a defense in Orthoptera. J. Orthoptera Res. 17:315–320.

Bateman, P. W., and P. A. Fleming. 2009. To cut a long tail short: a review of lizard caudal autotomy studies carried out over the last 20 years. Zoology 277:1–14.

Bateman, P. W., P. A. Fleming, and B. Rolek. 2014. Bite me: Blue tails as a “risky- decoy” defense tactic for lizards. Curr. Zool. 60:333–337.

Bely, A. E., and K. G. Nyberg. 2010. Evolution of animal regeneration: re-emergence of a field. Trends Ecol. Evol. 25:161–170.

Berendonck, B. E., and H. A. Greven. 2000. Morphology of female and male genitalia of Latrodectus revivensis Shulov, 1948 (Araneae, Theridiidae) with regard to sperm priority patterns. European arachnology 19:157–167.

Berzins I.K., and R. L. Caldwell. 1983. The effect of injury on the agonistic behavior of the Stomatopod, Gonodactylus bredini (manning). Mar. Behav. Physiol. 10:83– 96.

Bolger, A. M., M. Lohse, and B. Usadel. 2014. Trimmomatic: A flexible trimmer for Illumina Sequence Data. Bioinformatics.

Boll, P. K., I. Rossi, S.V. Amaral, and A. Leal-Zanchet. 2015. A taste for exotic food: Neotropical land planarians feeding on an invasive flatworm. PeerJ 3:e1307.

Booksmythe, I., R. N. C. Milner, M. D. Jennions, and P. R. Y. Backwell. 2009. How do weaponless male fiddler crabs avoid aggression? Behav. Ecol. Sociobiol. 64:485–491.

Bickell-Page, L.R., and G. O. Mackie. 1991. Tentacle Autotomy in the Hydromedusa Aglantha digitale (Cnidaria): An Ultrastructural and Neurophysiological Analysis. Proc. R. Soc. Lond. B 331:155–170.

Bingham, B., J. Burr, and H. Head. 2000. Causes and consequences of arm damage in the sea star Leptasterias hexactis. Can. J. Zool. 78:596–605.

Brailovsky, H. 2006. a New Genus and Species of Amorbini From Papua New Guinea (Heteroptera: Coreidae: Coreinae), With a Key To the Known Genera. J. New York Entomol. Soc. 114: 72–77.

163

Brock, K. M., P. A. Bednekoff, P. Pafilis, and J. Foufopoulos. 2015. Evolution of antipredator behavior in an island lizard species, Podarcis erhardii (Reptilia: Lacertidae): The sum of all fears? Evolution 69:216–231.

Brueseke, M. A., A. L. Rypstra, S. E. Walker, and M. H. Persons. 2001. Leg Autotomy in the Wolf Spider Pardosa milvina: A Common Phenomenon with Few Apparent Costs. Am. Midl. Nat. 146:153–160.

Bulliére, D., and F. Bulliére. 1985. Regeneration, pp. 371–424. In Compr. Insect Physiol. Biochem. Pharmacol. 2. Pergamon Press.

Bush, S. L. 2012. Economy of arm autotomy in the mesopelagic squid Octopoteuthis deletron. Mar. Ecol. Prog. Ser. 458:133–140.

Capella-Gutiérrez, S., Silla-Martínez, J. M., and Gabaldón, T. 2009. trimAl: a tool for automated alignment trimming in large-scale phylogenetic analyses. Bioinformatics 25:1972-1973.

Carlberg U. 1986. Thanatosis and autotomy as defence in Baculum sp. 1 (Insecta: Phasmida). Zool. Anz. 217:39–53.

Caro, T. 2005. Antipredator defenses in birds and mammals. University of Chicago Press.

Castilla, A. M., A. Gosá, P. Galán, and V. Pérez-Mellado. 1999. Green tails in lizards of the genus Podarcis: do they influence the intensity of predation? Herpetologica 530–537.

Cerenius, L., and K Söderhäll. 2011. Coagulation in invertebrates. J. Innate Immun. 3:3–8.

Cirino, L.A., and C. W. Miller. 2017. Seasonal effects on the population, morphology and reproductive behavior of Narnia femorata (Hemiptera: Coreidae). Insects 8:13

Clark, D. R., and R. J. Hall. 1970. Function of the blue tail-coloration of the five-lined skink (Eumeces fasciatus). Herpetologica 271–274.

Cockerell, T. D. 1909. Fossil insects from Colorado. The Entomologist 42:170–274.

Congdon, J. D., L. J. Vitt, and W. W. King. 1974. Geckos: adaptive significance and energetics of tail autotomy. Science 184:1379–1380.

Cooper, W. E. 2003. Shifted balance of risk and cost after autotomy affects use of cover, escape, activity, and foraging in the keeled earless lizard (Holbrookia propinqua). Behav. Ecol. Sociobiol. 54:179–187.

164

Cooper, W.E. 2007. Compensatory changes in escape and refuge use following autotomy in the lizard Sceloporus virgatus. Can. J. Zool. 85:99–107.

Cooper, W. E., and W. G. Frederick. 2007. Optimal flight initiation distance. J. Theor. Biol. 244:59–67.

Cooper, W. E., and W. G. Frederick. 2010. Predator lethality, optimal escape behavior, and autotomy. Behav. Ecol. 21:91–96.

Cooper, W. E., V. Prez-Mellado, and L. J. Vitt. 2004. Ease and effectiveness of costly autotomy vary with predation intensity among lizard populations. J. Zool. 262:243–255.

Cooper, W. E., and C. S. Smith. 2009. Costs and economy of autotomy for tail movement and running speed in the skink Trachylepis maculilabris. Can. J. Zool 87:400–406.

Cooper, W. E., and L. J. Vitt. 1985. Blue tails and autotomy: enhancement of predation avoidance in juvenile skinks. Zeitschrift für Tierpsychologie 70:265–276.

Cooper, W. E., and D. S. Wilson. 2010. Longer Hiding Time in Refuge Implies Greater Assessed Risk After Capture and Autotomy in Striped Plateau Lizards (Sceloporus virgatus). Herpetologica 66:425–431.

Cromie, G. L., and D. G. Chapple. 2012. Impact of tail loss on the behaviour and locomotor performance of two sympatric Lampropholis skink species. PLoS ONE 7:e34732.

Cromie, G. L., and D. G. Chapple. 2013. Is partial tail loss the key to a complete understanding of caudal autotomy? Austral. Ecol. 38:452–455.

Cross, A. T., and P. W. Bateman. 2018. How dangerous is Drosera ? Limb autotomy increases passive predation risk in crickets. J. Zool. 306:217–222.

Cully, A., J. Clune, D. Tarapore, and J.-B. Mouret. 2015. Robots that can adapt like animals. Nature 521:503–507.

Daleo, P., T. Luppi, A. M. Casariego, M. Escapa, P. Ribeiro, P. Silva, and O. Iribarne. 2009. The effect of size and cheliped autotomy on sexual competition between males of the mud crab Cyrtograpsus angulatus Dana. Mar. Biol. 156:269.

Davis, J. L. D., N. A. Dobroski, J. T. Carlton, J. Prevas, S. Parks, D. Hong, and E. Southworth. 2005. Autotomy in the Asian Shore Crab (Hemigrapsus sanguineus) in a Non-Native Area of Its Range. J. Crustac. Biol. 25:655–660.

Dejean, A., B. Schatz, J. Orivel, and G. Beugnon. 1999. Prey capture behavior of Psalidomyrmex procerus (Formicidae: Ponerinae), a specialist predator of earthworms (Annelida). Sociobiology 34:545–554.

165

Delnat, V., S. Debecker, and R. Stoks. 2017. Integrating trait multidimensionality, predation and autotomy to explain the maintenance of boldness. Anim. Behav. 130:97–105. de Oliveira, D. N., R. A. Christofoletti, and R. E. Barreto. 2015. Feeding behavior of a crab according to cheliped number. PLoS ONE 10:e0145121.

Dial, B. E., and L. C. Fitzpatrick. 1983. Lizard Tail Autotomy: Function and Energetics of Postautotomy Tail Movement in Scincella lateralis. Science 219:391–393.

Díaz, M., A. P. Møller, E. Flensted-Jensen, T. Grim, J. D. Ibáñez-Álamo, J. Jokimäki, G. Markó, and P. Tryjanowski. 2013. The geography of fear: a latitudinal gradient in anti-predator escape distances of birds across Europe. PLoS ONE 8:e64634.

Distant, W. L. 1880-1893. Biologia Centrali-Americana. Insecta. Rhynchota. Hemiptera- Heteroptera. Vol. I. London

Dixon, K. A. 1989. Effect of leg type and sex on autotomy in the Texas bush katydid, Scudderia texensis. Can. J. Zool. 67:1607–1609.

Domínguez, M., I. Escalante, F. Carrasco-Rueda, C. E. Figuerola-Hernández, M. Marta Ayup, M. N. Umaña, D. Ramos, A. González-Zamora, C. Brizuela, W. Delgado, and J. Pacheco-Esquivel. 2016. Losing legs and walking hard: effects of autotomy and different substrates in the locomotion of harvestmen in the genus Prionostemma. J. Arachnol. 44:76–82.

Donovan D.A., J. P. Elias, and J. Baldwin. 2004. Swimming behavior and morphometry of the file shell Limaria fragilis. Mar. Freshw. Behav. Physiol. 37:7–16.

Downes, S., and R. Shine. 2001. Why does tail loss increase a lizard’s later vulnerability to snake predators? Ecology 82:1293–1303.

Ducey, P. K., E. D. Brodie, and E. A. Baness. 1993. Salamander tail autotomy and snake predation: role of antipredator behavior and toxicity for three neotropical bolitoglossa (caudata: plethodontidae). Biotropica 25:344.

Dunn, C. W., G. Giribet, G. D. Edgecombe, and A. Hejnol. 2014. Animal phylogeny and its evolutionary implications. Annu. Rev. Ecol. Evol. Syst. 45, 371-395.

Dunn, C. W., A. Hejnol, D. Q. Matus, K. Pang, W. E. Browne, S. A. Smith, E. Seaver, G. W. Rouse, M. Obst, G. D. Edgecombe, M. V. Sørensen, S. H. D. Haddock, A. Schmidt-Rhaesa, A. Okusu, R. M. Kristensen, W. C. Wheeler, M. Q. Martindale, and G. Giribet. 2008. Broad phylogenomic sampling improves resolution of the animal tree of life. Nature 452:745.

Easton, D. M. 1972. Autotomy of walking legs in the Pacific Shore crab Hemigrapsus oregonesis. Mar. Freshw. Behav. Phy. 1:209–217.

166

Eberhard, W. G. 1998. Sexual Behavior of Acanthocephala declivis guatemalana (Hemiptera: Coreidae) and the Allometric Scaling of their Modified Hind Legs. Ann. Entomol. Soc. Am. 91:863–871.

Eisner, T., and S. Camazine. 1983. Spider leg autotomy induced by prey venom injection: An adaptive response to “pain”? PNAS 80:3382–3385.

Ellington, C. P. 1991. Limitations on animal flight performance. J. Exp. Biol. 160:71–91.

Elwood, C., J. Pelsinski, and B. Bateman. 2012. Anolis sagrei (Brown Anole). Voluntary autotomy. Herpetol. Rev. 43:642–642.

Emberts, Z., I. Escalante, and P. W. Bateman. 2019. The ecology and evolution of autotomy. Biol. Rev. doi: 10.1111/brv.12539.

Emberts, Z., C. W. Miller, D. Kiehl, and C. M. St Mary. 2017a. Cut your losses: self- amputation of injured limbs increases survival. Behav. Ecol. 28:1047–1054.

Emberts, Z., C. W. Miller, D. Li, W. S. Hwang, and C. M. St Mary. 2017b. Multiple male morphs in the leaf‐footed bug Mictis longicornis (Hemiptera: Coreidae). Entomol. Sci. 20: 396–401.

Emberts, Z., C. M. St Mary, T. J. Herrington, and C. W. Miller. 2018. Males missing their sexually selected weapon have decreased fighting ability and mating success in a competitive environment. Behav. Ecol. Sociobiol. 72:81.

Emberts, Z., C. M. St Mary, and C. W. Miller. 2016. Coreidae (Insecta: Hemiptera) limb loss and autotomy. Ann. Entomol. Soc. Am. 109:678–683.

Emlen, D. 1997. Alternative reproductive tactics and male-dimorphism in the horned beetle Onthophagus acumiatus (Coleoptera: Scarabaeidae). Behav. Ecol. Sociobiol. 41:335–341.

Emlen, D. 2008. The Evolution of Animal Weapons. Annu. Rev. Ecol. Evol. Syst. 39: 387–413.

Escalante, I., A. Albín, and A. Aisenberg. 2013. Lacking sensory (rather than locomotive) legs affects locomotion but not food detection in the harvestman Holmbergiana weyenberghi. Can. J. Zool. 91:726–731.

Etheridge, R. 1967. Lizard Caudal Vertebrae. Copeia 1967:699.

Faircloth, BC. 2013. illumiprocessor: a trimmomatic wrapper for parallel adapter and quality trimming.

Faircloth, B. C. 2016. PHYLUCE is a software package for the analysis of conserved genomic loci. Bioinformatics. 32:786–788.

167

Faircloth, B. C. 2017. Identifying conserved genomic elements and designing universal bait sets to enrich them. Methods Ecol. Evo. 8:1103–1112.

Figiel, C. R., G. L. Miller, and E. J. Meeman. 1995. The Frequency of Chela Autotomy and Its Influence on the Growth and Survival of the Crayfish Procambarus clarkii (Girard, 1852) (Decapoda, Cambaridae). Crustaceana. 68: 472–483.

Fiore, C., J. L. Tull, S. Zehner, and P. K. Ducey. 2004. Tracking and predation on earthworms by the invasive terrestrial planarian Bipalium adventitium (Tricladida, Platyhelminthes). Behav. Processes 67:327–334.

Fleming, P. A., and P. W. Bateman. 2007. Just drop it and run: the effect of limb autotomy on running distance and locomotion energetics of field crickets (Gryllus bimaculatus). J. Exp. Biol. 210:1446–1454.

Fleming, P. A., D. Muller, and P. W. Bateman. 2007. Leave it all behind: a taxonomic perspective of autotomy in invertebrates. Biol. Rev. Camb. Philos. Soc. 82:481– 510.

Fleming, P. A., L. E. Valentine, and P. W. Bateman. 2013. Telling tails: selective pressures acting on investment in lizard tails. Physiol. Biochem. Zool. 86:645– 658.

Fleming, P. A., L. Verburgt, M. Scantlebury, K. Medger, and P. W. Bateman. 2009. Jettisoning ballast or fuel? Caudal autotomy and locomotory energetics of the Cape dwarf gecko Lygodactylus capensis (Gekkonidae). Physiol. Biochem. Zool. 82:756–765.

Flynn, P.S.T., C. L. Mellish, T. R. Pickering, and P. A. Quijón. 2015. Effects of claw autotomy on green crab (Carcinus maenas) feeding rates. Journal of Sea Research 103, 113–119.

Foelix, R. F. 1996. Biology of Spiders. Oxford University Press, New York.

Forthman, M., C. W. Miller, and R. T. Kimball. 2019. Phylogenomic analysis suggests Coreidae and Alydidae (Hemiptera: Heteroptera) are not monophyletic. Zool. Scr. 48:520–534.

Fox, S. F., J. M. Conder, and A. E. Smith. 1998. Sexual Dimorphism in the Ease of Tail Autotomy: Uta stansburiana with and without Previous Tail Loss. Copeia 1998:376.

Fox, S. F., N. A. Heger, and L. S. Delay. 1990. Social cost of tail loss in Uta stansburiana: lizard tails as status-signalling badges. Anim. Behav. 39:549–554.

Fox, S. F., S. Perea-Fox, and R. C. Franco. 1994. Development of the tail autotomy adaptation in lizards under disparate levels of predation at high and low elevations in Mexico. Southwest. Nat. 311–322.

168

Fredericq, L. 1883. About autotomy or mutilation by reflex as a defense mean of animals. Archives de Zoologie Expérimentale et Générale 1, 413–426

Fresnillo, B., J. Belliure, and J. J. Cuervo. 2015. Red tails are effective decoys for avian predators. Evol. Ecol. 29:123–135.

Fromhage, L., and J. M. Schneider. 2006. Emasculation to plug up females: the significance of pedipalp damage in Nephila fenestrata. Behav. Ecol. 17:353–357.

Fujisaki, K. 1981. Studies on the mating system of the winter cherry bug, Acanthocoris sordidus Thunberg (Heteroptera: Coreidae) II. Harem defence polygyny. Res. Popul. Ecol. (Kyoto). 23: 262–279.

Gerald, G. W., and K. E. Thiesen. 2014. Locomotor hindrance of carrying an enlarged sexually selected structure on inclines for male fiddler crabs. J. Zool. 294:129– 138.

Gerald, G. W., M. M. Thompson, T. D. Levine, and K. M. Wrinn. 2017. Interactive effects of leg autotomy and incline on locomotor performance and kinematics of the cellar spider, Pholcus manueli. Ecol. Evol. 7:6729–6735.

Ghislandi, P., T. Bilde, and C. Tuni. 2015. Extreme Male Mating Behaviours: Anecdotes in a Nuptial Gift-Giving Spider. Arachnology 16:273–275.

Gilbert, E. A. B., S. L. Payne, and M. K. Vickaryous. 2013. The anatomy and histology of caudal autotomy and regeneration in lizards. Physiol. Biochem. Zool. 86:631– 644.

Gillis, G. B., L. A. Bonvini, and D. J. Irschick. 2009. Losing stability: tail loss and jumping in the arboreal lizard Anolis carolinensis. J. Exp. Biol. 212:604–609.

Gleason, J. E., D. S. Fudge, and B. W. Robinson. 2014. Eco-mechanics of lamellar autotomy in larval damselflies. J. Exp. Biol. 217:185–191.

Gillespie, S.R., S. M. Tudor, A. J. Moore, and C. W. Miller. 2014. Sexual selection is influenced by both developmental and adult environments. Evolution 68:3421– 3432.

Glenn, T. C., R. A. Nilsen, T. J. Kieran, J. W. Finger, T. W. Pierson, K. E. Bentley, S. L. Hoffberg, S. Louha, F. J. García-De León, M. A. d. R. Portilla, K. D. Reed, J. L. Anderson, J. K. Meece, S. E. Aggery, R. Rekaya, M. Alabady, M. Bélanger, K. Winker, and B. C. Faircloth. 2016. Adapterama I: universal stubs and primers for thousands of dual-indexed Illumina libraries (iTru & iNext). BioRxiv 049114.

Glynn, P. W. 1982. Acanthaster population regulation by a shrimp and a worm. Internat Coral Reef Symp, Manilla, Philippines.

169

Gomendio, M., and P. Reguera. 2001. Egg carrying in the golden egg bug (Phyllomorpha laciniata): parental care, parasitism, or both? Reply to Kaitala et al. Behav. Ecol. 12:369–373.

Goodenough, J., B. McGuire, and E. Jakob. 2009. Perspectives on animal behavior. John Wiley & Sons.

Grabherr, M. G., B. J. Haas, M. Yassour, J. Z. Levin, D. A. Thompson, I. Amit, X. Adiconis, L. Fan, R. Raychowdhury, Q. Zhang, Z. Chen, E. Mauceli, N. Hacohen, A. Gnirke, N. Rhind, F. di Palma, B. W. Birren, C. Nusbaum, K. Lindbald-Toh, N. Friedman, and A. Regev. 2011. Trinity: reconstructing a full-length transcriptome without a genome from RNA-Seq data. Nature Biotechnology, 29, 644.

Gross, M.R. 1996. Alternative reproductive strategies and tactics: diversity within sexes. Trends Ecol. Evolut. 11:92–98.

Goss, R. J. 1983. Deer antlers: regeneration, function and evolution. Academic Press.

Guffey, C. 1998. Leg autotomy and its potential fitness costs for two species of harvestmen (Arachnida, Opiliones). J. Arachnol. 296–302.

Guffey, C. 1999. Costs associated with leg autotomy in the harvestmen Leiobunum nigripes and Leiobunum vittatum (Arachnida: Opiliones). Can. J. Zool. 77:824– 830.

Haacke, W. D. 1975. The burrowing geckos of southern Africa, 1 (Reptilia: Gekkonidae). Annals of the Transvaal Museum 29:197–243.

Haacke, W. D. 1976. The burrowing geckos of southern Africa, 4 (Reptilia: Gekkonidae). Annals of the Transvaal Museum 30:53–70.

Hamilton, W.D., and M. Zuk. 1982. Heritable true fitness and bright birds: A role for parasites? Science 218:384–387.

Harmon, L. J., J. T. Weir, C. D. Brock, R. E. Glor, and W. Challenger. 2008. GEIGER: investigating evolutionary radiations. Bioinformatics 24:129–131.

Harrison, J. F., and S. P. Roberts. 2000. Flight respiration and energetics. Annu. Rev. Physiol. 62:179–205.

Harzhauser, M., and W. E. Piller. 2004. Integrated stratigraphy of the sarmatian (Upper Middle Miocene) in the western Central Paratethys. Stratigraphy 1:65–86.

Hawlena, D., R. Boochnik, Z. Abramsky, and A. Bouskila. 2006. Blue tail and striped body: why do lizards change their infant costume when growing up? Behav. Ecol. 17:889–896.

170

He, J., X. Wu, and Y. Cheng. 2016. Effects of limb autotomy on growth, feeding and regeneration in the juvenile Eriocheir sinensis. Aquaculture 457:79–84.

Heer, O. 1853. Die Insektenfauna der Tertiärgebilde von Oeningen und von Radoboj in Croatien. Leipzig,W. Engelmann.

Hemmingsen, A. M. 1947. Plant bug guarding eggs and offspring and shooting anal jets (Physomerus grossipes F., Coreidae). Entomol. Medd. 25.

Herreid, C.F., and R. J. Full. 1986. Locomotion of Hermit Crabs (Coenobita compressus) on Beach and Treadmill. J. Exp. Biol. 120:283–296.

He, H., C. Deng, Y. Pan, T. Deng, Z. Luo, J. Sun, and R. Zhu. 2011. New 40Ar/39Ar dating results from the Shanwang Basin, eastern China: Constraints on the age of the Shanwang Formation and associated biota. Phys. Earth Planet. Inter. 187:66–75.

He, J., X. Wu, and Y. Cheng. 2016. Effects of limb autotomy on growth, feeding and regeneration in the juvenile Eriocheir sinensis. Aquaculture 457, 79–84.

Higham, T. E., and A. P. Russell. 2010. Flip, flop and fly: modulated motor control and highly variable movement patterns of autotomized gecko tails. Biol. Lett. 6:70–73.

Higham, T. E., A. P. Russell, and P. A. Zani. 2013. Integrative biology of tail autotomy in lizards. Physiol. Biochem. Zool. 86:603–610.

Hill, G.E. 1991. Plumage coloration is a sexually selected indicator of male quality. Nature 350:337–339.

Hoadley, L. 1937. Autotomy in the brachyuran, Uca pugax. The Biological Bulletin 73:155–163.

Hodgkin, J., L. C. Clark, and M. J. Gravato-Nobre. 2014. Worm-stars and half-worms: Novel dangers and novel defense. Worm 3:e27939.

Ho, L. T., and C. Ané. 2014. A linear-time algorithm for Gaussian and non-Gaussian trait evolution models. Syst. Biol. 63:397–408.

Houghton, J. E., V. R. Townsend, and D. N. Proud. 2011. The ecological significance of leg autotomy for climbing temperate species of harvestmen (arachnida, opiliones, sclerosomatidae). Southeast. Nat.10:579–590.

Hsieh, S. T. T. 2016. Tail loss and narrow surfaces decrease locomotor stability in the arboreal green anole lizard (Anolis carolinensis). J. Exp. Biol. 219:364–373.

Hsu, Y., R. L. Earley, and L. L. Wolf. 2006. Modulation of aggressive behaviour by fighting experience: mechanisms and contest outcomes. Biol. Rev. 81:33–74.

171

Ijspeert, A. J. 2014. Biorobotics: using robots to emulate and investigate agile locomotion. Science 346:196–203.

Itescu, Y., R. Schwarz, S. Meiri, and P. Pafilis. 2017. Intraspecific competition, not predation, drives lizard tail loss on islands. J. Anim. Ecol. 86:66–74.

Jagnandan, K., and T. E. Higham. 2017. Lateral movements of a massive tail influence gecko locomotion: an integrative study comparing tail restriction and autotomy. Sci. Rep. 7:10865.

Jagnandan, K., A. P. Russell, and T. E. Higham. 2014. Tail autotomy and subsequent regeneration alter the mechanics of locomotion in lizards. J. Exp. Biol. 217:3891– 3897.

Janssens, L., W. Verberk, and R. Stoks. 2018. A widespread morphological antipredator mechanism reduces the sensitivity to pesticides and increases the susceptibility to warming. Sci. Total Environ. 626:1230–1235.

Jeanne, R. L. 1979. A latitudinal gradient in rates of ant predation. Ecology 60:1211.

Johnson, A. S. 1988. Hydrodynamic study of the functional morphology of the benthic suspension feeder Phoronopsis viridis (Phoronida). Marine Biology 100:117–126.

Johnson, K. P., C. H. Dietrich, F. Friedrich, R. G. Beutel, B. Wipfler, R. S. Peters, J. M. Allen, M. Petersen, A. Donath, K. K. O. Walden, A. M. Kozlov, L. Podsiadlowski, C. Mayer, K. Meusemann, A. Vasilikopoulos, R. M. Waterhouse, S. L. Cameron, C. Weirauch, D. R. Swanson, D. M. Percy, and K. Yoshizawa. 2018. Phylogenomics and the evolution of hemipteroid insects. PNAS 115:12775– 12780.

Johnson, S. A., and E. M. Jakob. 1999. Leg autotomy in a spider has minimal costs in competitive ability and development. Anim. Behav. 57:957–965.

Joseph, P. N., Z. Emberts, D. A. Sasson, and C. W. Miller. 2018. Males that drop a sexually selected weapon grow larger testes. Evolution 72:113–122.

Juanes, F., and L. D. Smith. 1995. The ecological consequences of limb damage and loss in decapod crustaceans: a review and prospectus. J. Exp. Mar. Biol. Ecol. 193:197–223.

Kieran, T. J., E. Gordon, M. Forthman, R. Hoey-Chamberlain, R. T. Kimball, B. C. Faircloth, C. Weirauch, and T. C. Glenn. 2019. Insights from an ultraconserved element bait set designed for hemipteran phylogenetics integrated with genomic resources. Mol. Phylogenet. Evol. 130: 297–303

Kitchen, D. W. 1974. Social Behavior and Ecology of the Pronghorn. Wildl. Monogr. 3– 96.

172

Knoflach, B. 2002. Copulation and emasculation in Echinotheridion gibberosum (Kulczynski, 1899) (Araneae, Theridiidae). Eur. Arachnol. 2000:139–144.

Knoflach, B., and A. van Harten. 2001. Tidarren argo sp. nov. (Araneae: Theridiidae) and its exceptional copulatory behaviour: emasculation, male palpal organ as a mating plug and sexual cannibalism. J. Zool. 254:449–459.

Kuo, C.-Y., G. B. Gillis, and D. J. Irschick. 2012. Take this broken tail and learn to jump: the ability to recover from reduced in-air stability in tailless green anole lizards [Anolis carolinensis (Squamata: Dactyloidae)]. Biol. J. Linn. Soc. Lond. 107:583– 592.

Kuo, C.-Y., D. J. Irschick, and S. P. Lailvaux. 2015. Trait compensation between boldness and the propensity for tail autotomy under different food availabilities in similarly aged brown anole lizards. Funct. Ecol. 29:385–392.

Labanick, G. M. 1984. Anti-predator effectiveness of autotomized tails of the salamander Desmognathus ochrophaeus. Herpetologica 110–118.

Lancaster, D. L., and S. E. Wise. 1996. Differential response by the ringneck snake, Diadophis punctatus, to odors of tail-autotomizing prey. Herpetologica 98–108.

Lanfear, R., B. Calcott, D. Kainer, C. Mayer, and A. Stamatakis. 2014. Selecting optimal partitioning schemes for phylogenomic datasets. BMC Evol. Biol. 14:82.

Lanfear, R., P. B. Frandsen, A. M. Wright, T. Senfeld, and B. Calcott. 2016. PartitionFinder 2: new methods for selecting partitioned models of evolution for molecular and morphological phylogenetic analyses. Mol. Biol. Evol. 34:772–773.

Laurila, A., B. Lindgren, and A. T. Laugen. 2008. Antipredator defenses along a latitudinal gradient in Rana temporaria. Ecology 89:1399–1413.

Lawrence, J. 1992. Arm loss and regeneration in Asteroidea (Echinodermata). In: Scalera-Liaci L, Canicatti C, editors. Echinoderm Research. Rotterdam, the Netherlands: A.A. Balkema. p. 39–52.

Lawton, P. 1989. Predatory interaction between the brachyuran crab Cancer pagurus and decapod crustacean prey. Mar. Ecol. Prog. Ser. 52:169–179.

LeBlanc, A. R. H., M. J. MacDougall, Y. Haridy, D. Scott, and R. R. Reisz. 2018. Caudal autotomy as anti-predatory behaviour in Palaeozoic reptiles. Sci. Rep. 8:3328.

Lesiuk, N. M., and C. D. Drewes. 1999. Autotomy reflex in a freshwater oligochaete, Lumbriculus variegatus (Clitellata: Lumbriculidae). Pp. 253–261 in B. M. Healy, T. B. Reynoldson, and K. A. Coates, eds. Aquatic Oligochaetes. Springer Netherlands, Dordrecht.

173

Levin, D. A., and B. M. York. 1978. The toxicity of plant alkaloids: an Ecogeographic perspective. Biochem. Syst. Ecol. 6:61–76.

Lewis, J.G.E. 1981. The Biology of Centipedes. Cambridge: Cambridge University Press.

Li, H., J. M. Leavengood, E. G. Chapman, D. Burkhardt, F. Song, P. Jiang, J. Liu, X. Zhou, and W. Cai. 2017. Mitochondrial phylogenomics of Hemiptera reveals adaptive innovations driving the diversification of true bugs. Proc. R. Soc. Lond. B 284.

Li, H.-T., T.-S. Yi, L.-M. Gao, P.-F. Ma, T. Zhang, J.-B. Yang, M. A. Gitzendanner, P. W. Fritsch, J. Cai, Y. Luo, H. Wang, M. van der Bank, S.-D. Zhang, Q.-F. Wang, J. Wang, Z.-R. Zhang, C.-N. Fu, J. Yang, P. M. Hollingsworth, M. W. Chase, D. E. Soltis, P. S. Soltis, and D.-Z. Li. 2019. Origin of angiosperms and the puzzle of the Jurassic gap. Nature plants 5:461

Lin, J. W., Y. R. Chen, Y. H. Wang, K. C. Hung, and S. M. Lin. 2017. Tail regeneration after autotomy revives survival: a case from a long-term monitored lizard population under avian predation. Proc. R. Soc. Lond. B 284:20162538.

Liu, Y., H. Li, F. Song, Y. Zhao, J. Wilson, and W. Cai. 2019. Higher‐level phylogeny and evolutionary history of Pentatomomorpha (Hemiptera: Heteroptera) inferred from mitochondrial genome sequences. Syst. Entomol.

Lu, H.-L., G.-H. Ding, P. Ding, and X. Ji. 2010. Tail Autotomy Plays No Important Role in Influencing Locomotor Performance and Anti-Predator Behavior in a Cursorial Gecko. Ethology.

Lu, L.-M., L.-F. Mao, T. Yang, J.-F. Ye, B. Liu, H.-L. Li, M. Sun, J. T. Miller, S. Mathews, H.-H. Hu, Y.-T. Niu, D.-X. Peng, Y.-H. Chen, S. A. Smith, M. Chen, K.-L. Xiang, C.-T. Le, V.-C. Dang, A.-M. Lu, P. Soltis, D. Soltis, J.-H. Li, and Z.-D. Chen. 2018. Evolutionary history of the angiosperm flora of China. Nature 554:234–238.

Luscher, M. 1948. The regeneration of legs in Rhodnius prolixus (Hempiptera). J. Exp. Biol. 25:334–343.

Lynam, E. C., J. C. Owens, and M. H. Persons. 2006. The Influence of Pedipalp Autotomy on the Courtship and Mating Behavior of Pardosa milvina (Araneae: Lycosidae). J. Insect Behav. 19:63–75.

Maginnis, T. L. 2006. The costs of autotomy and regeneration in animals: a review and framework for future research. Behav. Ecol. 17:857–872.

Maginnis, T. L. 2008. Autotomy in a stick insect (Insecta: Phasmida): predation versus molting. Fla. Entomol. 91:126–127.

174

Maginnis, T. L., M. Isikbay, S. G. Degerstedt, T. J. Luethke, and M. Ortman. 2015. The effects of cheliped autotomy and regeneration on aggression in purple shore crabs (Hemigrapsus nudus). Mar. Freshw. Behav. Physiol. 48:241–251.

Maiorana, V. C. 1977. Tail autotomy, functional conflicts and their resolution by a salamander. Nature 295:533–535.

Marçais, G., J. A. Yorke, and A. Zimin. 2015. QuorUM: an error corrector for Illumina reads. PLoS One 10:e0130821.

Marín, A., and J. Ros. 2004. Chemical defenses in Sacoglossan Opisthobranchs: Taxonomic trends and evolutive implications. Sci. Mar. 68:227–241.

Marrs, J., I. C. Wilkie, M. Sk�ld, W. M. Maclaren, and J. D. McKenzie. 2000. Size- related aspects of arm damage, tissue mechanics, and autotomy in the starfish Asterias rubens. Mar. Biol. 137:59–70.

Martin, J., and A. Salvador. 1993. Tail loss reduces mating success in the Iberian rock- lizard, Lacerta monticola. Behav. Ecol. Sociobiol. 32.

Martín, J., and A. Salvador. 1997. Effects of tail loss on the time-budgets, movements, and spacing patterns of Iberian rock lizards, Lacerta monticola. Herpetologica 117–125.

Martín, J., A. Salvador, and J. Martin. 1992. Tail Loss Consequences on Habitat Use by the Iberian Rock Lizard, Lacerta monticola. Oikos 65:328.

Matsuoka, N., M. Miyakawa, and M. Ishihara. 2011. Effect of hind-limb autotomy on calling and hiding behavior in the band-legged ground cricket, Dianemobius nigrofasciatus. J. Ethol. 29:209–213.

Mattoni, C. I., S. García-Hernández, R. Botero-Trujillo, J. A. Ochoa, A. A. Ojanguren- Affilastro, R. Pinto-da-Rocha, and L. Prendini. 2015. Scorpion sheds “tail” to escape: consequences and implications of autotomy in scorpions (Buthidae: Ananteris). PLoS ONE 10:e0116639.

McCambridge, C., J. T. A. Dick, and R. W. Elwood. 2016. Effects of Autotomy Compared to Manual Declawing on Contests between Males for Females in the Edible Crab Cancer pagurus: Implications for Fishery Practice and Animal Welfare. J. Shellfish Res. 35:1037–1044.

McConnachie, S., and M. J. Whiting. 2003. Costs associated with tail autotomy in an ambush foraging lizard, Cordylus melanotus melanotus. Afr. Zool. 38:57–65.

McCullough, E. L. 2014. Mechanical limits to maximum weapon size in a giant rhinoceros beetle. Proc. R. Soc. Lond. B 281:20140696.

175

McCullough, E.L., B. W. Tobalske, and D. J. Emlen. 2014. Structural adaptations to diverse fighting styles in sexually selected weapons. PNAS 111:14484–14488.

Mcgaw, I. J. 2006. Cardiovascular and respiratory responses associated with limb autotomy in the blue crab. Mar. Freshw. Behav. Physiol. 39:131–141.

McVean, A. 1973. Autotomy in Cavcinus maenas ( Decapoda: Crustacea ). J. Zool. Proc. Zool. Soc. London 169:349–364.

McVean, A. 1975. Autotomy. Comparative Biochemistry and Physiology Part A: Physiology 51:497–505.

McVean, A. 1982. Autotomy. In D. Bliss (ed. The Biology of Crustacea), pp. 107–132. Academic Press, New York

Medel, R. G., J. E. Jiménez, S. F. Fox, F. M. Jaksić, J. E. Jimenez, and F. M. Jaksic. 1988. Experimental evidence that high population frequencies of lizard tail autotomy indicate inefficient predation. Oikos 53:321.

Medzhitov, R., and C. A. J. Janeway. 1997. Innate immunity: Minireview the virtues of a nonclonal system of recognition. Cell 91:295–298.

Miller, C. W., and D. J. Emlen. 2010. Across- and Within-Population Differences in the Size and Scaling Relationship of a Sexually Selected Trait in Leptoscelis tricolor (Hemiptera: Coreidae). Ann. Entomol. Soc. Am. 103:209–215.

Miller, C. W., P. N. Joseph, R. Kilner, and Z. Emberts. 2019. A weapons-testes tradeoff in males is amplified in female traits. Proc. R. Soc. Lond. B 286:20190906.

Miller, C. W., G. C. McDonald, and A. J. Moore. 2016. The tale of the shrinking weapon: seasonal changes in nutrition affect weapon size and sexual dimorphism, but not contemporary evolution. J. Evol. Biol. 29:2266–2275.

Miller, P. L. 1977. Neurogenic pacemakers in the legs of Opiliones. Physiol. Entomol. 2:213–224.

Minton, S. A. 1966. A contribution to the herpetology of West Pakistan. Bulletin of the AMNH; v. 134, article 2.

Mitchell, P. L. 1980. Combat and Territorial Defense of Acanthocephala femorata (Hemiptera: Coreidae). Ann. Entomol. Soc. Am. 73:404–408.

Miura, K., and N. Ohsaki. 2015. The cost of autotomy caused by the parasitoid fly Blaesoxipha japonensis (Diptera: Sarcophagidae): an interspecific comparison between two sympatric grasshopper host species. Ecol. Res. 30:33–39.

176

Miyatake, T. 1993. Male-male aggressive behavior is changed by body size difference in the leaf-footed plant bug, Leptoglossus australis, Fabricius (Heteroptera: Coreidae). J. Ethol. 11:63–65.

Miyatake, T. 1997. Functional morphology of the hind legs as weapons for male contests in Leptoglossus australis (Heteroptera: Coreidae). J. Insect Behav. 10:727–735.

Møller, A. P. 1996. The cost of secondary sexual characters and the evolution of cost- reducing traits. Ibis 138: 112–119.

Møller, A. P., and W. Liang. 2013. Tropical birds take small risks. Behav. Ecol. 24:267– 272.

Mori, A. 1990. Tail vibration of the Japanese grass lizard Takydromus tachydromoides as a tactic against a snake predator. J. Ethol. 8:81–88.

Morrell, L.J., J. Lindström, and G. D. Ruxton. 2005. Why are small males aggressive? Proc. R. Soc. Lond. B 272:1235–1241.

Morris, M. R. 1998. Female preference for trait symmetry in addition to trait size in swordtail fish. Proc. R. Soc. Lond. B 265:907–911.

Morris, M.R., L. Gass, and M. J. Ryan. 1995. Assessment and Individual Recognition of Opponents in the Pygmy Swordtails Xiphophorus nigrensis and X. Multilineatus. Behav. Ecol. Sociobiol. 37:303–310.

Mueller, R. L., J. R. Macey, M. Jaekel, D. B. Wake, and J. L. Boore. 2004. Morphological homoplasy, life history evolution, and historical biogeography of plethodontid salamanders inferred from complete mitochondrial genomes. PNAS 101:13820–13825.

Murali, G., S. Merilaita, and U. Kodandaramaiah. 2018. Grab my tail: evolution of dazzle stripes and colourful tails in lizards. J. Evol. Biol. 31:1675–1688.

Muscatine, L., and H. M. Lenhoff. 1974. Coelenterate Biology, Reviews and New Perspectives.

Nasri, I., F. Hamza, J. Belliure, and S. Selmi. 2018. Tail conspicuousness and antipredatory behaviour in Bosk’s fringe-toed lizard (Acanthodactylus boskianus). Ethology Ecology & Evolution 30:319–330.

Naya, D. E., C. Veloso, J. L. P. Muñoz, and F. Bozinovic. 2007. Some vaguely explored (but not trivial) costs of tail autotomy in lizards. Comp Biochem Physiol, Part A Mol Integr Physiol 146:189–193.

Neil, S.J. 1985. Size Assessment and Cues: Studies of Hermit Crab Contests. Behaviour 92:22–38.

177

Nentwig, W. 1982. Why do only certain insects escape from a spider’s web? Oecologia 53:412–417.

Nessler S.H., G. Uhl, and J. M. Schneider. 2007. Genital damage in the orb-web spider Argiope bruennichi (Araneae: Araneidae) increases paternity success. Behav. Ecol. 18:174–181.

Niven, J. E., and J. P. W. Scharlemann. 2005. Do insect metabolic rates at rest and during flight scale with body mass? Biol. Lett. 1:346–349.

Nolen, Z.J., P. E. Allen, and C. W. Miller. 2017. Seasonal resource value and male size influence male aggressive interactions in the leaf footed cactus bug, Narnia femorata. Behav. Processes 138:1–6.

O’Neill, D.J., J. S. Cobb.1979. Some factors influencing the outcome of shelter competition in lobsters (Homarus americanus). Mar. Behav. Physiol. 6:33–45.

Oliveira, D.N. de, R. A. Christofoletti, and R. E. Barreto. 2015. Feeding Behavior of a Crab According to Cheliped Number. PLoS One 10:Article No e0145121.

Op de Beeck, L., J. Verheyen, and R. Stoks. 2018. Competition magnifies the impact of a pesticide in a warming world by reducing heat tolerance and increasing autotomy. Environ. Pollut. 233:226–234.

Ortego, F., and W. S. Bowers. 1996. Induction of autotomy in the American bird grasshopper Schistocerca americana (Drury) by the ecdysone agonist RH-5849 and investigation of its mode of action. Experientia 52:42–50.

Otaibi, B. W., Q. K. Johnson, and B. J. Cosentino. 2017. Postautotomy tail movement differs between colour morphs of the red-backed salamander (Plethodon cinereus). AMRE 38:395–399.

O’Neill, D. J., and J. S. Cobb. 1979. Some factors influencing the outcome of shelter competition in lobsters (Homarus americanus). Mar. Behav. Physiol. 6:33–45.

Pafilis, P., J. Foufopoulos, N. Poulakakis, P. Lymberakis, and E. D. Valakos. 2009. Tail shedding in island lizards [Lacertidae, Reptilia]: decline of antipredator defenses in relaxed predation environments. Evolution 63:1262–1278.

Pafilis, P., V. Pérez-Mellado, and E. Valakos. 2008. Postautotomy tail activity in the Balearic lizard, Podarcis lilfordi. Naturwissenschaften 95:217–221.

Pafilis, P., E. D. Valakos, and J. Foufopoulos. 2005. Comparative postautotomy tail activity in six Mediterranean lacertid lizard species. Physiol. Biochem. Zool. 78:828–838.

Palmer, A. R. 1979. Fish predation and the evolution of gastropod shell sculpture: experimental and geographic evidence. Evolution 33:697–713.

178

Parker, G. 1974. Assessment strategy and the evolution of animal conflicts. J. Theor. Biol. 47:223–243.

Patterson, L., J. T. A. Dick, and R. W. Elwood. 2009. Claw removal and feeding ability in the edible crab, Cancer pagurus: Implications for fishery practice. Appl. Anim. Behav. Sci. 116:302–305.

Payne, R. J. H., and M. Pagel. 1996. Escalation and time costs in displays of endurance. J. Theor. Biol. 183:185–193.

Pears, J. B., Z. Emberts, and P. W. Bateman. 2018. The Scent of Danger: the Impact of Predator Chemical Cues on Emergence from Refuge and Willingness to Autotomize Limbs in the House Cricket (Acheta domesticus). J. Insect Behav. 31:416–426.

Prestholdt, T. E., S. M. Rager, J. N. Williams, and C. L. Weilhoefer. 2018. The force needed for leg autotomy in the crab Hemigrapsus nudus. Mar. Freshw. Behav. Physiol. 51:175–181.

Procter, D. S., A. J. Moore, and C. W. Miller. 2012. The form of sexual selection arising from male-male competition depends on the presence of females in the social environment. J. Evol. Biol. 25:803–812.

Punzo, F. 1997. Leg autotomy and avoidance behavior in response to a predator in the wolf spider, Schizocosa avida (Araneae, Lycosidae). J. Arachnol. 202–205.

R Core Team. 2014. R: A Language and Environment for Statistical Computing. Vienna, Austria.

R Core Team. 2016. R: A Language and Environment for Statistical Computing. Vienna, Austria.

R Core Team. 2018. R: A Language and Environment for Statistical Computing. Vienna, Austria.

Rambaut, A. 2017. FigTree-version 1.4. 3, a graphical viewer of phylogenetic trees.

Rambaut, A., A. J. Drummond, D. Xie, G. Baele, and M. A. Suchard. 2018. Posterior summarization in bayesian phylogenetics using tracer 1.7. Syst. Biol. 67:901– 904.

Ramsay, K., M. Bergmann, L. O. Veale, C. A. Richardson, M. J. Kaiser, S. J. Vize, and S. W. Feist. 2001. Damage, autotomy and arm regeneration in starfish caught by towed demersal fishing gears. Mar. Biol. 138:527–536.

Reaney, L. T., R. N. C. Milner, T. Detto, and P. R. Y. Backwell. 2008. The effects of claw regeneration on territory ownership and mating success in the fiddler crab Uca mjoebergi. Anim. Behav. 75:1473–1478.

179

Reguera, P., and M. Gomendio. 1999. Predation costs associated with parental care in the golden egg bug Phyllomorpha laciniata (Heteroptera: Coreidae). Behav. Ecol. 10:541–544.

Remmel, T., J. Davison, and T. Tammaru. 2011. Quantifying predation on folivorous insect larvae: the perspective of life-history evolution. Biol. J. Linnean Soc. 104:1–18.

Revell, L. J. 2012. phytools: an R package for phylogenetic comparative biology (and other things). Methods Ecol. Evo. 3:217–223.

Robinson, M. H., L. G. Abele, and B. Robinson. 1970. Attack autotomy: a defense against predators. Science 169:300–301.

Roth, V. D., and B. M. Roth. 1984. review of appendotomy in spiders and other arachnids. Bulletin-British Arachnological Society.

Ruxton, G. D., W. L. Allen, T. N. Sherratt, and M. P. Speed. 2018. Avoiding Attack. Oxford University Press.

Salvador, A., J. Martin, and P. López. 1995. Tail loss reduces home range size and access to females in male lizards, Psammodromus algirus. Behav. Ecol. 6:382– 387.

Samia, D. S. M., A. P. Møller, D. T. Blumstein, T. Stankowich, and W. E. Cooper. 2015. Sex differences in lizard escape decisions vary with latitude, but not sexual dimorphism. Proc. R. Soc. Lond. B 282.

Savory, T.H. 1928. The Biology of Spiders. New York: Macmillan.

Schemske, D. W., G. G. Mittelbach, H. V. Cornell, J. M. Sobel, and K. Roy. 2009. Is There a Latitudinal Gradient in the Importance of Biotic Interactions? Annu. Rev. Ecol. Evol. Syst. 40:245–269.

Schmieder, R., and R. Edwards. 2011. Quality control and preprocessing of metagenomic datasets. Bioinformatics, 27:863–864.

Shandilya, A., G. Mishra, and Omkar. 2018. Leg impairment affects the reproductive attributes of a ladybird, Menochilus sexmaculatus (Fabricius). J. Asia Pac. Entomol. 21:797–806.

Shargal, E., L. Rath-Wolfson, N. Kronfeld, and T. Dayan. 1999. Ecological and histological aspects of tail loss in spiny mice (Rodentia: Muridae, Acomys) with a review of its occurrence in rodents. J. Zool. 249:187–193.

Shaw, V.K., and P. J. Bryant. 1974. Regeneration of appendages in the large milkweed bug, Oncopeltus fasciatus. J. Insect Physiol. 20:1849–1857.

180

Simmons, L.W., and D. J. Emlen. 2006. Evolutionary trade-off between weapons and testes. PNAS 103:16346–16351.

Slos, S., M. De Block, and R. Stoks. 2009. Autotomy reduces immune function and antioxidant defence. Biol. Lett. 5:90–92.

Smith, D. L., and A. H. Hines. 1991. The effect of cheliped loss on blue crab Callinectes sapidus Rathbun foraging rate on soft-shell clams Mya arenaria L. J. Exp. Mar. Biol. Ecol. 151:245–256.

Smith, L. D. 1992. The impact of limb autotomy on mate competition in blue crabs Callinectes sapidus Rathbun. Oecologia 89:494–501.

Smith, M. E., B. Singer, and A. Carroll. 2003. 40Ar/39Ar geochronology of the Eocene Green River Formation, Wyoming. Geol. Soc. Am. Bull. 115:549–565.

Smith, S. A., and B. C. O’Meara. 2012. treePL: divergence time estimation using penalized likelihood for large phylogenies. Bioinformatics 28:2689–2690.

Smith, S. A., J. W. Brown, and J. F. Walker. 2018. So many genes, so little time: A practical approach to divergence-time estimation in the genomic era. PLoS ONE 13:e0197433.

Snow, L. S. E., A. Abdel-Mesih, and M. C. B. Andrade. 2006. Broken Copulatory Organs are Low-Cost Adaptations to Sperm Competition in Redback Spiders. Ethology 112:379–389.

Somjee, U., C. W. Miller, N. J. Tatarnic, and L. W. Simmons. 2018a. Experimental manipulation reveals a trade-off between weapons and testes. J. Evol. Biol. 31:57–65.

Somjee, U., H. A. Woods, M. Duell, and C. W. Miller. 2018b. The hidden cost of sexually selected traits: the metabolic expense of maintaining a sexually selected weapon. Proc. R. Soc. Lond. B 285.

Stalhandske, P. 2001. Nuptial gift in the spider Pisaura mirabilis maintained by sexual selection. Behav. Ecol. 12:691–697.

Stamatakis, A. 2014. RAxML version 8: a tool for phylogenetic analysis and post- analysis of large phylogenies. Bioinformatics 30:1312-1313.

Starostová, Z., L. Gvoždík, and L. Kratochvíl. 2017. An energetic perspective on tissue regeneration: The costs of tail autotomy in growing geckos. Comp. Biochem. Physiol., Part A Mol. Integr. Physiol. 206:82–86.

Stoks, R. 1998. Effect of lamellae autotomy on survival and foraging success of the damselfly Lestes sponsa (Odonata: Lestidae). Oecologia 117:443–448.

181

Stoks, R. 1999. Autotomy shapes the trade-off between seeking cover and foraging in larval damselflies. Behav. Ecol. Sociobiol. 47:70–75.

Suchard, M. A., P. Lemey, G. Baele, D. L. Ayres, A. J. Drummond, and A. Rambaut. 2018. Bayesian phylogenetic and phylodynamic data integration using BEAST 1.10. Virus Evol. 4:vey016.

Sugiura, S. 2010. Prey preference and gregarious attacks by the invasive flatworm Platydemus manokwari. Biol. Invasions 12:1499–1507.

Swaddle, J.P., and I. C. Cuthill. 1994. Female zebra finches prefer male with symmetric chest plumage. Proc. R. Soc. Lond. B 258:267–271.

Sword, G.A. 2001. Tasty on the outside, but toxic in the middle: Grasshopper regurgitation and host plant-mediated toxicity to a vertebrate predator. Oecologia 128:416–421.

Taylor, P. W., J. A. Roberts, A. E. Wignall, and G. W. Uetz. 2008. Foreleg Autotomy Reduces Mating Success of Male Schizocosa ocreata Wolf Spiders. J. Insect Behav. 21:193–202.

Telemeco, R. S., T. A. Baird, and R. Shine. 2011. Tail waving in a lizard (Bassiana duperreyi) functions to deflect attacks rather than as a pursuit-deterrent signal. Anim. Behav. 82:369–375.

Tsurui, K., S. Narita, Y. Iwatani, and A. Honma. 2014. Change in take-off elevation angle after limb autotomy mitigates the reduction in jumping distance in rice grasshoppers Oxya yezoensis. Entomol. Sci. 17:181–190.

Tummon Flynn, P. S., C. L. Mellish, T. R. Pickering, and P. A. Quijón. 2015. Effects of claw autotomy on green crab (Carcinus maenas) feeding rates. J. Sea Res. 103:113–119.

Uetz, G. W., W. J. McClintock, D. Miller, E. I. Smith, and K. K. Cook. 1996. Limb regeneration and subsequent asymmetry in a male secondary sexual character influences sexual selection in wolf spiders. Behav. Ecol. Sociobiol. 38: 253–257.

Uhl, G., S. H. Nessler, and J. M. Schneider. 2010. Securing paternity in spiders? A review on occurrence and effects of mating plugs and male genital mutilation. Genetica 138:75–104.

Urata, M., S. Iwasaki, and S. Ohtsuka. 2012. Biology of the swimming acorn worm Glandiceps hacksi from the Seto Inland Sea of Japan. Zool. Sci. 29:305–311.

Van Buskirk, J., and D. C. Smith. 1991. Density-Dependent Population Regulation in a Salamander. Ecology 72:1747–1756.

182

Van Sluys, M., D. Vrcibradic, and C. F. D. Rocha. 2002. Tail loss in the Syntopic Lizards Tropidurus itambere (Tropiduridae) and Mabuya frenata (Scincidae) in Southeastern Brazil. Stud. Neotrop. Fauna Environ. 37:227–231.

Vermeij, G. J. 1978. Biogeography and adaptation: patterns of marine life. Harvard University Press.

Wada, T. 2016. Size-assortative mating and arm loss in the wild shallow-water octopus Abdopus sp. (Cephalopoda: Octopodidae). J. Nat. Hist 51:1–10.

Wake, D. B., and I. G. Dresner. 1967. Functional morphology and evolution of tail autotomy in salamanders. J. Morphol. 122:265–305.

Wang, Y., Y. Cui, D. Rédei, P. Baňař, Q. Xie, P. Štys, J. Damgaard, P. Chen, W. Yi, Y. Wang, K. Dang, C. Li, and W. Bu. 2016. Phylogenetic divergences of the true bugs (Insecta: Hemiptera: Heteroptera), with emphasis on the aquatic lineages: the last piece of the aquatic insect jigsaw originated in the Late Permian/Early Triassic. Cladistics 32:390–405.

Wasson, K., and B. E. Lyon. 2005. Flight or fight: flexible antipredatory strategies in porcelain crabs. Behav. Ecol. 16:1037–1041.

Wasson, K., B. E. Lyon, and M. Knope. 2002. Hair-trigger autotomy in porcelain crabs is a highly effective escape strategy. Behav. Ecol. 13:481–486.

Watson, C. M., C. E. Roelke, P. N. Pasichnyk, and C. L. Cox. 2012. The fitness consequences of the autotomous blue tail in lizards: an empirical test of predator response using clay models. Zoology 115:339–344.

Wilkie, I. C. 2001. Autotomy as a prelude to regeneration in echinoderms. Microsc. Res. Tech. 55:369–396.

Wilshin, S., P. S. Shamble, K. J. Hovey, R. Harris, A. J. Spence, and S. T. Hsieh. 2018. Limping following limb loss increases locomotor stability. J. Exp. Biol. 221.

Wilson, A. D. M., E. M. Whattam, R. Bennett, L. Visanuvimol, C. Lauzon, and S. M. Bertram. 2010. Behavioral correlations across activity, mating, exploration, aggression, and antipredator contexts in the European house cricket, Acheta domesticus. Behav. Ecol. Sociobiol. 64:703–715.

Winchester, L., and A. d’A. Bellairs. 1977. Aspects of vertebral development in lizards and snakes. J. Zool. 181:495–525.

Wood, F. D., and H. E. Wood. 1932. Autotomy in decapod Crustacea. J. Exp. Zool. 62:1–55.

183

Yang, X.-Z., C. Zhang, G.-Y. Huang, M.-J. Xu, Y.-X. Cheng, Z.-G. Yang, Q. Zhang, and Y.-Y. Wang. 2018. Cellular and biochemical parameters following autotomy and ablation-mediated cheliped loss in the Chinese mitten crab, Eriocheir sinensis. Dev. Comp. Immunol. 81:33–43.

Yasuda, C. I., and T. Koga. 2016. Do weaponless males of the hermit crab Pagurus minutus give up contests without escalation? Behavior of intruders that lack their major cheliped in male–male contests. J. Ethol. 34:249–254

Yasuda, C., Y. Suzuki, and S. Wada. 2011. Function of the major cheliped in male-male competition in the hermit crab Pagurus nigrofascia. Mar. Biol. 158:2327–2334.

Ydenberg, R. C., and L. M. Dill. 1986. The Economics of Fleeing from Predators. Adv. Stud. Behav. 16:229–249.

Zamora-Camacho, F. J., S. Reguera, and G. Moreno-Rueda. 2015. Does tail autotomy affect thermoregulation in an accurately thermoregulating lizard? Lessons from a 2200-m elevational gradient. J. Zool. 297:204–210.

Zani, P. A. 1996. Patterns of caudal-autotomy evolution in lizards. J. Zool. 240:201–220.

Zhang, C., X.-Z. Yang, M.-J. Xu, G.-Y. Huang, Q. Zhang, Y.-X. Cheng, L. He, and H.-Y. Ren. 2018. Melatonin Promotes Cheliped Regeneration, Digestive Enzyme Function, and Immunity Following Autotomy in the Chinese Mitten Crab, Eriocheir sinensis. Front. Physiol. 9:269.

Zhang, J., B. Song, and X. Zhang. 1994. Miocene insects and spiders from Shanwang, Shandong. Ke xue chu ban she : Xin hua.

Zheng, Y., and J. J. Wiens. 2015. Do missing data influence the accuracy of divergence-time estimation with BEAST? Mol. Phylogenet. Evol. 85:41–49.

184

BIOGRAPHICAL SKETCH

Zachary Emberts started his career as a behavioral ecologist at the University of

Minnesota where he investigated mating behavior in field crickets with Marlene Zuk and

Susan Balenger. After obtaining his undergraduate degree he worked at the Sarasota

Dolphin Research Program, Duke University Marine Lab, and Walt Disney World

Resorts studying dolphin communication with a variety of scientists, including Goldie

Phillips, Doug Nowacek, and Andy Stamper. Zachary then went on to receive his Ph.D. from the University of Florida. In his free time, Zachary flies kites, collects magnets, and is known to exaggerate his hobbies. He is also the first and favorite son of Bryan and

Melissa Emberts.

185