The Ecology and Endocrinology of Facultative Paedomorphosis

University of Mississippi eGrove

Electronic Theses and Dissertations Graduate School

2019

Extended Adolescence: The Ecology and Endocrinology of Facultative Paedomorphosis

Jason R. Bohenek University of Mississippi

Follow this and additional works at: https://egrove.olemiss.edu/etd

Part of the Biology Commons

Recommended Citation Bohenek, Jason R., "Extended Adolescence: The Ecology and Endocrinology of Facultative Paedomorphosis" (2019). Electronic Theses and Dissertations. 1554. https://egrove.olemiss.edu/etd/1554

This Dissertation is brought to you for free and open access by the Graduate School at eGrove. It has been accepted for inclusion in Electronic Theses and Dissertations by an authorized administrator of eGrove. For more information, please contact [email protected]. EXTENDED ADOLESCENCE: THE ECOLOGY AND ENDOCRINOLOGY OF

FACULTATIVE PAEDOMORPHOSIS

A Dissertation presented in partial fulfillment of requirements for the degree of Doctor of Philosophy in the Department of Biology The University of Mississippi

JASON R. BOHENEK

May 2019

ABSTRACT

Phenotypic plasticity is an adaptation to unpredictable environments whereby an organism of a single genotype may express more than one phenotype under differing environmental conditions. Phenotypic plasticity can manifest as polyphenisms, which is an extreme form of phenotypic plasticity that produces two or more discrete, alternative phenotypes.

The expression of alternative phenotypes is controlled by biotic and abiotic environmental factors, which variably affect the strength and direction of phenotypic outcomes. Using a model polyphenic salamander, I sought to understand the ecological and hormonal processes that regulate alternative phenotype expression. The mole salamander (Ambystoma talpoideum) and eastern newt (Notophthalmus viridescens) are facultatively paedomorphic, which is a polyphenism with two alternative adult phenotypes: paedomorphs and metamorphs.

Paedomorphs retain juvenile characteristics at sexual maturity (i.e., gills and an aquatic morphology), while metamorphs transition to terrestrial environments. The expressed phenotype depends on the environment context in which the larvae develop, with paedomorphosis often occurring under favorable aquatic conditions. I conducted a series of experiments to investigate the roles of population density, predator presence, hydroperiod, and stress hormones in regulating the expression of paedomorphosis. Results indicated the regulation of paedomorphosis through multiple ecological factors may be reducible to density-mediated effects, with a few notable exceptions. I also show that elevated stress hormones play a central role in regulating metamorphosis suggesting that all ecological factors affecting facultative paedomorphosis may funnel through a simple stress physiology framework. In conclusion, environmental factors

affecting this polyphenism may share a common thread of inducing a stress response that initiates metamorphosis, thereby regulating phenotype in the population.

iii LIST OF ABBREVIATIONS AND SYMBOLS

°C Degrees Celsius

α Significance level

AICc Sample size corrected Akaike Information Criterion

ANOVA Analysis of variance

BOBL Best of a bad lot hypothesis cm Centimeter

CORT Corticosterone, the primary amphibian stress hormone

CRH Corticotropin-releasing hormone df Degrees of freedom

DO Dissolved oxygen (mg/L)

DP Dimorphic paedomorph hypothesis

F Test statistic in ANOVA g Gram

GLMM Generalized linear mixed model

HPI Hypothalamic-pituitary-interrenal axis

L Liter

LMM Linear mixed model m Meter

P Probability value used in null hypothesis testing

PA Paedomorph advantage hypothesis

PERMANOVA Permutational multivariate analysis of variance pH Measure of acidity; concentration of hydrogen ions

SE Standard error t Test statistic for the t-test

T3 Thyroxine

T4 Triiodothyronine

TDS Total dissolved solids

TSH Thyroid stimulating hormone

TRC Tyson Research Center

UMFS University of Mississippi Field Station

v ACKNOWLEDGEMENTS

I would to thank Matthew Pintar, Lauren Eveland, Tyler Breech, and Reed Scott for assistance with field work. Also, I would like to thank Timothy Chavez, Rachel Kroeger,

Brandon McDaniel, Zachary Mitchell, Randi Robison, and Nishant Rout for their assistance with field work throughout the years. William J. Resetarits Jr. provided guidance and feedback on experimental designs, analyses, and manuscripts, as did Jason Hoeksema, Christopher Leary,

Howard Whiteman, and Gregg Davidson. This work would not be possible without support from the Tyson Research Center and Washington University in St. Louis, with special thanks to Kevin

Smith and Barbara Schall, for allowing us to use their facilities. Also, Beth Biro helped substantially with collecting at the Tyson Research Center. I would also like to thank the

University of Mississippi Graduate Student Council, University of Mississippi Field Station’s

Luther Knight Graduate Student Research Fund, the University of Mississippi Department of

Biology, and the Henry L. and Grace Doherty Foundation for financial assistance. Also, thank you to Scott Knight and the University of Mississippi Field Station staff for use of facilities and access to natural environments. Also, thank you to Rosemary Knapp for providing advice on running whole-body radioimmunoassay. Lastly, Lauren Fuller, B. Mikah, and B. Luxi provided emotional support. This research was approved by the University of Mississippi’s Institutional

Animal Care and Use Committee (protocol 14-028) and the Mississippi Department of Wildlife,

Fisheries, and Parks.

vi TABLE OF CONTENTS

Abstract ...... ii List of abbreviations and symbols ...... iv Acknowledgements ...... vi List of tables ...... ix List of figures ...... x

Chapter 1: Introduction ...... 1 Background ...... 2 Study site ...... 16 Mesocosms ...... 17 Study system ...... 18

Chapter 2: Initial density sets trajectory for density-dependent polyphenism thresholds ...... 21 Abstract ...... 22 Introduction ...... 24 Methods...... 27 Results ...... 30 Discussion ...... 34

Chapter 3: Are direct density cues, not resource competition, driving life history trajectories in a polyphenic salamander? ...... 38 Abstract ...... 39 Introduction ...... 41 Methods...... 46 Results ...... 52 Discussion ...... 57

Chapter 4: Ambystoma have convergent plastic responses among alternative habitat types separate by small spatial scales ...... 62 Abstract ...... 63 Introduction ...... 64 Methods...... 68 Results ...... 72 Discussion ...... 77

Chapter 5: Functional diversity of predators drive variable trait- and density-mediated effects on prey life history ...... 81 Abstract ...... 82 Introduction ...... 84

vii

Methods...... 87 Results ...... 93 Discussion ...... 98

Chapter 6: Glucocorticoids regulate life history strategies in a facultatively paedomorphic salamander ...... 103 Abstract ...... 104 Introduction ...... 106 Methods...... 110 Results ...... 115 Discussion ...... 120

List of references...... 126 Appendix ...... 141 Vita ...... 146

viii LIST OF TABLES

Chapter 3 Table A.1. Model comparison approach using AICc ...... 142 Table A.2. Complete statistical model summaries including random effects...... 143

LIST OF FIGURES

Chapter 1 Fig. 1.1. Continuous and discrete plastic traits ...... 3 Fig. 1.2. The evolution of reaction norms ...... 4 Fig. 1.3. Polyphenisms in a signal detection framework ...... 6 Fig. 1.4. The complex life cycle of facultatively paedomorphic salamanders ...... 8 Fig. 1.5. Growth in the Wilbur-Collins model ...... 9 Fig. 1.6. Body size in the Wilbur-Collins model ...... 10 Fig. 1.7. The Werner-Gilliam model ...... 11 Fig. 1.8. The Whiteman model for multiple hypotheses of facultative paedomorphosis ...... 12 Fig. 1.9. The complex life cycle of Notophthalmus viridescens ...... 20

Chapter 2 Fig. 2.1. Phenotype proportions across larval densities ...... 31 Fig. 2.2. The relationship between body size and mass ...... 32 Fig. 2.3. Mean body mass across larval densities ...... 32 Fig. 2.4. Mean body mass of each phenotype ...... 33 Fig. 2.5. Mean SVL across larval densities ...... 33

Chapter 3 Fig. 3.1. Phenotype proportions across each treatment ...... 53 Fig. 3.2. Body size and growth metrics ...... 55 Fig. 3.3. Mean snout-vent length over time ...... 56

Chapter 4 Fig. 4.1. Phenotype proportions across treatments ...... 73 Fig. 4.2. Mean body mass across treatments ...... 75 Fig. 4.3. Mean body condition across treatments ...... 75 Fig. 4.4. Mesocosm temperature time series across hydroperiod treatments ...... 76 Fig. 4.5. Mesocosm temperature variance across hydroperiod treatments ...... 76

Chapter 5 Fig. 5.1. Survival rates across treatments ...... 94 Fig. 5.2. Snout-vent length across treatments ...... 94 Fig. 5.3. Body condition index across treatments ...... 95 Fig. 5.4. Phenotype proportions across treatments ...... 96 Fig. 5.5. Principle component analysis of tail color traits ...... 97

Chapter 6 Fig. 6.1. Whole-body CORT concentrations across treatments ...... 116 Fig. 6.2. Phenotype proportions across CORT treatments ...... 117 Fig. 6.3. Morphometrics and life history variables ...... 118 Fig. 6.4. Body size in the Wilbur-Collins model ...... 119 Fig. 6.5. Endocrine model of facultative paedomorphosis ...... 122

CHAPTER 1:

INTRODUCTION

BACKGROUND

Phenotypic plasticity is a strategic response to spatial and temporal environmental heterogeneity where multiple behaviors or morphological/physiological states arise from a single genotype. Phenotypic plasticity is a form of intrapopulation biodiversity that manifests itself through various phenomena in a wide variety of organisms. Plasticity can be continuous, where a trait gradually changes with gradual changes in the environment (i.e., a linear reaction norm). Or plasticity can be extreme and produce two or more discrete, environmentally-cued alternative phenotypes in a population (West-Eberhard 1989) (Fig. 1.1). Discrete phenotypic plasticity is termed a “polyphenism” and follows a non-linear pattern because a switch occurs to an alternative phenotype at some point along an environmental gradient. Polyphenisms are potentially adaptive in that each alternative phenotype develops in order to maximize survival and reproductive output in particular environmental conditions (Moran 1992). In modern usage, polyphenisms are different from the similar “polymorphisms” (or “genetic polymorphisms”) because polymorphisms occur when alternative phenotypes are under genetic control, with no environmental effect. For example, Mendel’s classic study with pea plants considers only polymorphic traits.

Figure 1.1 Plastic traits can be either continuous (body size, left) or discrete (terrestrial and aquatic salamander phenotypes, right).

A reaction norm is the pattern of how phenotypes respond to environmental variation

(Fig 1.2) (Woltereck 1913). Through evolutionary processes, a continuous reaction norm can evolve into a discrete polyphenism if selection favors two phenotypic extremes over intermediate phenotypes. This process can also occur more easily if an environmental gradient itself is not continuous (i.e., discrete habitat types), thereby favoring highly divergent, discrete phenotypes.

Evolutionary processes can therefore shape and mold reaction norms making them more or less linear (Fig. 1.2). Ever since West-Eberhard (1989) rescued the idea of phenotypic plasticity from

Lamarckian evolution, it has been considered an important process of evolutionary change. This is especially true of polyphenisms whereby increasingly non-linear reaction norms are considered a crucial step in speciation because there is the potential for limited gene flow across different phenotypes (i.e., assortative mating among similar phenotypes) (Pfennig et al. 2010).

Reduced gene flow across populations exposed to different environmental pressures can potentially canalize the phenotypic response and fixate one of the alternative phenotypes in that population. Thus, plasticity can be lost altogether. In this respect, it is important to remember

phenotype is the sum of both environmental and genetic input and that nearly all traits fall somewhere between complete genetic control and complete environmental control (Pfennig et al.

2010).

Figure 1.2 The evolution of reactions norms, which shows the potential transition from a linear, continuous reaction norm (red) to a non-linear polyphenism (orange). As evolution progresses and alternative phenotypes are strongly favored in different environmental extremes, populations will diverge from a linear reaction norm (red) to slightly non-linear (green, blue) to a true polyphenism (orange).

Which phenotype occurs in which environment is critical to the fitness of polyphenic organisms because mismatches can be very costly (i.e., result in mortality). The fitness of each phenotype can be expressed as

f (p1, e1) > f (p2, e1) and f (p2, e2) > f (p1, e2) (1)

where phenotype 1 (p1) always has greater fitness in environment 1 (e1) and always has greater fitness than phenotype 2 (p2) in e1, and vice versa. However, in order to correctly match

phenotype to environment, a polyphenic organism must interpret available environmental information and then, using that information, select one of two phenotypic states (Moran 1992,

Getty 1996). If environmental cues are not correctly perceived, or are unreliable, phenotype mismatching can occur with potential fitness consequences (Eq. 1). This concept can be displayed in basic signal detection theory where phenotype mismatchings are considered errors

(miss or false alarm; Fig. 1.3).

There has been considerable focus on the environmental cues, both biotic and abiotic, that influence polyphenisms across a wide range of organisms: diet (Pfennig 1990, Moczek

1998, Michimae and Wakahara 2002), density or crowding effects (Uvarov 1921, Collins and

Cheek 1983, Harris 1987, Semlitsch 1987, Nijhout 2003, Pener and Simpson 2009, Cisse et al.

2015), relatedness (Pfennig and Collins 1993), tactile cues (Hoffman and Pfennig 1999,

Michimae et al. 2005), predation threat or cues thereof (Grunt and Bayly 1981, Lively 1986,

Harvell 1990, Jackson and Semlitsch 1993, McCollum and Buskirk 1996, Berenbaum and

Zangerl 1999, Brönmark et al. 1999, Gilbert 1999, Kuhlmann et al. 1999, Laforsch et al. 2009), pheromones (Nijhout 2003), food quantity and quality (Moczek 1998, Denoël and Poncin 2001,

Nijhout 2003), exogenous stress hormones (Maher et al. 2013), hydroperiod (Semlitsch 1987), photoperiod, and temperature (Nijhout 2003). The environment bombards organisms with these various cues. Parsing informative, reliable cues from uninformative, unreliable cues is an adaptive response that results in an assortment of strategies that reflect respective ecological contexts. Though much is known about the reaction norms involved in polyphenisms, little is known about cue modality or the mechanisms involved in acquisition of the cues that polyphenic organisms utilize to inform their polyphenisms.

Figure 1.3 Polyphenism in a signal detection framework. Organisms matching their phenotype correctly to the environmental cue will garner a “hit” or “correct rejection,” while errors will be incorrectly matched phenotypes, or “misses” or “false alarms.” When the cue is above the threshold, p2 will be expressed with some proportion making an error (“false alarm”) by expressing p1. When the cue is below the threshold, p1 will expressed, with some proportion making an error (“miss”) by expressing p2. The degree of overlap of the probability density curves determines cue reliability. More overlap means less reliability.

Salamanders (Caudata) are model organisms for studying plasticity because it is common within the order (Wilbur and Collins 1973). Many salamanders have complex, multi-stage life cycles. They hatch from eggs and develop as aquatic larvae that then metamorphose into an adult form. However, more than 10% of salamander species (57 species in 90% of recognized families) exhibit paedomorphosis, a heterochronic process where larvae eschew or delay metamorphosis in order to reproduce in a larval morphology (Gould 1977, Duellman and Trueb

1986, Whiteman 1994, Denoël et al. 2005) (Fig. 5). Furthermore, some species in the families

Salamandridae, Ambystomatidae, Dicamptodontidae, Hynobiidae and Plethodontidae are

facultatively paedomorphic, which is when paedomorphosis is essentially a polyphenism and is environmentally-cued with two discrete adult phenotypes: metamorphs and paedomorphs (Fig.

1.4) (Duellman and Trueb 1986, Semlitsch and Wilbur 1989, Whiteman 1994, Denoël et al.

2005). The resultant phenotype depends on the prevailing environmental conditions experienced by the developing larvae. Facultatively paedomorphic salamanders are a valuable model for aquatic ecologists since paedomorphosis is predicted to occur under relatively predictable and favorable aquatic conditions (Denoël and Ficetola 2014). This is referred to as the paedomorph advantage (PA) hypothesis (Wilbur and Collins 1973, Whiteman 1994). The PA hypothesis predicts that larvae in high quality aquatic habitats will optimize their growth rates by prolonging their aquatic state through paedomorphosis. Thus, these individuals are taking advantage of favorable aquatic conditions. Alternatively stated, PA predicts that metamorphosis occurs to avoid deteriorating aquatic conditions.

Figure 1.4 The complex life-history strategy of the mole salamander (Ambystoma talpoideum). Larvae can develop into either paedomorphs or terrestrial adults, both of which can reproduce. This strategy is also found in multiple species of salamanders from multiple continents and phylogenetic ancestry.

Many amphibians have complex, multistage life cycle that utilize temporary aquatic habitats for oviposition and larval development. In consequence, larval amphibians breed explosively when ponds fill, and then try to outpace pond desiccation by maximizing growth and development. Metamorphosing too early and obviating growth opportunities can result in small

body sizes and poor survival post-metamorphosis (Semlitsch et al. 1988). Delaying metamorphosis can result in direct mortality due to complete desiccation of the aquatic environment. Therefore, metamorphic timing and the information utilized to determine it are of paramount importance and have a large impact on survival and lifetime reproductive success.

The complexity of this decision was first described in the Wilbur-Collins model (Wilbur and

Collins 1973), which predicts that larval amphibians inform metamorphic timing based on bodily growth rates and environmental certainty (Fig. 1.4). Larval amphibians will remain in the aquatic habitat as long as body size (W) is increasing over time (dW/dt > g) and they remain under a maximal larval size (b+c). Additionally, the range of body sizes at which individuals metamorphose is expected to vary with the predictability of the environment. Highly predictable habitats (i.e., those with less temporal heterogeneity) are expected to select for less plasticity, as individuals can expect a fixed amount of time for their larval periods (e.g., permanent ponds). In unpredictable habitats, I expect more plasticity as individuals can take advantage of good growth conditions, or metamorphose at smaller body sizes if conditions rapidly deteriorate (Fig. 1.5).

Figure 1.5 Flow diagram depicting the Wilbur-Collins model (Wilbur and Collins 1973). Larval amphibians should initiate metamorphosis if body size (W) is greater than the species-specific lower limit of larval body size (b) and

growth (dW/dt > g) has stopped. Metamorphosis must initiate if W > b+c, which represents the species-specific upper limit of larval body size. Figure from Wilbur and Collins (1973).

Figure 1.6 Surface diagrams depicting the probability of metamorphosis (Pm) in uncertain and certain environments relative to growth rates (dW/dt) and W. “Certain” (highly predictable) environments (right) will narrow the range of body sizes at which metamorphosis can occur, while “uncertain” (highly unpredictable) environments (left) will broaden them. Figure from Wilbur and Collins (1973).

Whereas the Wilbur-Collins model focuses solely on growth opportunities of the aquatic environment, the Werner-Gilliam model (Werner and Gilliam 1984, Werner 1986) introduces growth and mortality rates of both the aquatic and terrestrial environment into the equation (Fig.

1.6). This model suggests that metamorphosis (ontogenetic niche shift from the aquatic to terrestrial environment) should occur at body sizes that maximize growth and minimize mortality

(µ). Thus, if the size-specific mortality/growth ratio in the aquatic environment (µa/ga) increases above the size-specific mortality/growth ratio in the terrestrial environment (µt/gt), then metamorphosis should occur (Fig. 1.6). These models have been highly influential in the

literature, but they do not consider the more complex scenario of paedomorphosis, where organisms may forego metamorphosis altogether.

Figure 1.7 Growth rates in the aquatic (ga) and terrestrial (gt) environments as a function of body size. Metamorphosis should occur at body size s’ to maintain maximum growth rates. However, when mortality (µ) in each environment is introduced as a ratio to growth (µ/g), optimal body size at metamorphosis delays to sopt in order to minimize risk of mortality. Figure from Werner (1986).

Paedomorphosis can increase lifetime reproductive success over metamorphosed individuals for several reasons. Paedomorphs can reach larger sizes faster (Healy 1973), thereby reproducing at a younger age (Healy 1974) and earlier (Semlitsch 1985, Scott 1993) relative to metamorphosed counterparts, thus improving their reproductive success (Roff 1992, Stearns

1992). Reproducing earlier in a season gives their offspring a competitive advantage over metamorphic offspring because of body size advantages that enable competitive dominance and intraguild predation (Anderson et al. 2013, 2017). Paedomorphs typically produce larger clutches due to larger body size (Rose and Armentrout 1976), but not always (Semlitsch 1985, Whiteman et al. 2012). It has also been hypothesized that paedomorphs should experience greater survivorship (Wilbur and Collins 1973), which can be conferred by avoiding desiccation and high mortality risk associated with metamorphosis and breeding pond migrations.

Figure 1.8 The three different hypotheses explaining facultative paedomorphosis. R, M* and P* are the minimum body sizes required for sexual maturity, metamorphosis, and PA paedomorphosis, respectively. P and M are the proportions of individuals that become paedomorphic and metamorphic, respectively. BOBL occurs when larval growth is poor and body sizes cannot reach M*, but are still greater than R. PA occurs to take advantage of favorable aquatic conditions and occurs when larval growth is so favorable that body sizes are greater than P*. DP explains paedomorphosis under BOBL and PA scenarios, but also explains metamorphosis in individuals at intermediate body sizes (i.e., larger than M*, but are still smaller than P*). DP individuals thus metamorphose to escape competition with the largest, most dominant larvae in the habitat. Figure from Whiteman (1994).

Though the PA hypothesis is the most studied and well-supported, paedomorphic salamanders do not always arise from the largest, fastest growing individuals (Whiteman et al.

2012). There are two other hypotheses explaining occurrence of facultative paedomorphs: best of a bad lot (BOBL) and dimorphic paedomorph (DP) hypotheses (Whiteman 1994). The best of a bad lot (BOBL) hypothesis explains that paedomorphosis can result from larvae that are too small for metamorphosis (did not reach a size threshold physiologically required for metamorphosis). Unable to metamorphose, these individuals maximize their fitness opportunities by reproducing at a young age and small body size (Doyle and Whiteman 2008, Fig. 1.8). The dimorphic paedomorph (DP) hypothesis encompasses both the BOBL and PA hypotheses occurring in the same population. The DP hypothesis explains that while the largest and small individuals become paedomorphs, the average-sized individuals metamorphose. These average- sized individuals are large enough for metamorphosis ( > M*), but are inferior competitors

compared to the largest individuals and thus metamorphose to escape competition (and possibly cannibalism; Fig. 1.8).

In facultatively paedomorphic salamanders, the expressed phenotype is environmentally cued and strongly depends upon aquatic conditions. Environmental cues such as conspecific density (Harris 1987, Semlitsch 1987), hydroperiod (Semlitsch 1987, Semlitsch et al. 1990), temperature (Sprules 1974) and food availability (Sprules 1974, Voss 1995, Ryan and Semlitsch

2003) have been shown to influence adult phenotype. Hydroperiod and conspecific density have been experimentally demonstrated as being two of the most important factors. Hydroperiod is a critical characteristic for organisms that utilize temporary aquatic habitats. Hydroperiod can affect larval development rate and adult phenotype, with metamorphs resulting from short hydroperiods and paedomorphs resulting from long/permanent hydroperiods (Wilbur and Collins

1973, Semlitsch 1987, Semlitsch and Wilbur 1988). Using hydroperiod as a cue informs larvae that pond drying is imminent and provides them with relative measure of time remaining to metamorphose and “escape.” There is at least some evidence that pond drying alone can accelerate development in western spadefoot toads (Spea hammondii) (Denver et al. 1998), but many correlated factors such as larval density, food levels and physical characteristics (e.g., pH, dissolved oxygen, temperature; Semlitsch and Gibbons 1985) change as a pond dries. It is difficult to segregate these effects in order to determine what information organisms are actually utilizing as cues.

Of those correlated factors, larval density has been experimentally demonstrated as an important factor affecting development. Multiple studies have shown facultative paedomorphosis to be negatively density-dependent, meaning that the frequency of paedomorphosis decreases as density increases (Harris 1987, Semlitsch 1987). The PA

hypothesis suggests that the density-dependent effect is best explained by a competitive model in which competition for food resources affects individual growth rates, which are used as a proxy for aquatic conditions (Wilbur and Collins 1973). Higher larval density increases intraspecific competition, resulting in poor aquatic conditions which elicits metamorphosis (Whiteman 1994).

However, this competitive model was shown to be insufficient by Semlitsch (1987). He crossed multiple levels of food resources, density, and hydroperiod and he concluded that food levels do not affect phenotype. Additionally, Richter et al. (2009) showed that density has an independent effect on metamorphic timing of southern leopard frogs (Lithobates sphenocephala). Tadpoles were reared at different densities, but individual growth rates were maintained via per capita food additions. Tadpoles reared in high densities had shorter larval periods with no cost of size at metamorphosis (Richter et al. 2009). These results, along with the findings of Semlitsch (1987), suggest that larval amphibians can assess density (either directly or indirectly) and that they utilize this information to make crucial metamorphic decisions. However, if food resources aren’t limited, then what drives this density-dependent response? What proximate mechanisms are involved?

I performed a series of outdoor field mesocosm experiments examining the role of conspecific density and associated factors on the expression of alternative phenotypes in facultatively paedomorphic salamanders. First, I studied the how an increasing gradient of conspecific density shapes the reaction norms of facultatively paedomorphic salamanders.

Second, I performed an experiment with caged larval salamanders to isolate the effects of food and crowding (agonistic interactions) from density cues to determine if the effects of density are chemically-mediated. Third, in a common garden experiment, I examined how reaction norms may diverge between populations with differential environmental pressures. Fourth, I studied the

role of predatory fish in affecting larval salamander density and thereby the expression of alternative phenotypes. Lastly, I tested whether facultative paedomorphosis is regulated through a stress response, whereby increased stress results in an ontogenetic transition to the terrestrial environment.

STUDY SITES

All studies were conducted at one of two locations — Tyson Research Center (TRC)

(38.5259° N, −90.5617° W) or the University of Mississippi Field Station (UMFS) (34.4295° N,

−89.3931° W). The Tyson Research Center is situated on Paleozoic carbonate that is located in the Missouri Ozark border region (Thom and Wilson 1980) and lies along the Meramec River in

St. Louis County, Missouri. It is 800 ha of fenced habitat, mostly comprised of oak and hickory forests, but it also includes numerous glades, ponds, old fields, and intermittent streams. The study conducted at TRC of Washington University was done in an old field.

UMFS is a 318 ha complex within the boundaries of Holly Springs National Forest in

Lafayette County, Mississippi. Situated in the Eocene hills of the interior Gulf Coastal Plain,

UMFS contains over 200 ponds of various sizes and hydroperiods (ranging from seasonal to semi-permanent to permanent), along with multiple streams, fields, and mixed forests of red maple, sweetgum, oak, and pine. Many of the current ponds were established as part of a fish hatchery that ceased operations prior to the University of Mississippi’s acquisition in 1985.

UMFS is home to 345 species of vascular plants, 16 frogs, 12 salamanders, 25 snakes, 10 turtles,

9 lizards, 55 butterflies, and 123 species of aquatic beetles (Keiser 1999, 2001, 2008, 2010, 2014,

Menon and Holland 2012, Pintar unpublished).

MESOCOSMS

Cattle tanks are large tubs used to provide drinking water to livestock and there is a long history of using them as experimental aquatic mesocosms (Morin 1981). They provide easy experimental replication while also allowing a great degree of control, which is not always possible when working in natural ponds across the landscape. I used 1200 L plastic tanks (1.8 m diameter, 50 cm depth, ARM-10138, Ace Roto-Mold, Hospers, Iowa, USA). A base layer of leaf litter was added to each tank to provide nutrients to the system and it also adds habitat heterogeneity that provides refuge. Mesocosms were filled with well water and then inoculated with concentrated pond water that contained naturally occurring microorganisms, algae, and zooplankton. They were covered with fiberglass screen lids (1.3 × 1.13 mm mesh) to prevent colonization and oviposition by other organisms. Mesocosms have been demonstrated to successfully mimic larger, natural habitats (Scott 1990, Pechmann 1994, Wilbur 1997).

STUDY SYSTEM

Eastern newts (Notophthalmus viridescens) are a polyphenic keystone species (Morin

1981) native to the Eastern United States. They have an extremely complex life history (Fig. 1.9, see early debates in Pope 1921, 1924, 1928, Noble 1926, 1929) that is subject to considerable environmental influence. Currently, their life cycle is oversimplified into four stages (egg

→larva → eft → adult). However, newt ontogeny has five distinct post-larval phenotypes, though not all populations express all phenotypes (Takahashi and Parris 2008). During the spring into early summer, adult newts deposit eggs in aquatic environments and those eggs hatch into aquatic larvae. The larvae grow and eventually develop into different phenotypes via multiple pathways (Fig. 1.9, Reilly 1987, Petranka 1998). Larvae can metamorphose into either one of two immature phenotypes, (1) terrestrial efts or (2) aquatic juveniles, and three reproductively mature phenotypes, (3) semi-aquatic adults, (4) branchiate adults (paedomorphosis through incomplete metamorphosis) or (5) complete neotenes (paedomorphosis with no degree of metamorphosis). The aquatic juvenile stage is transitional in nature and often overlooked, however it serves as an important avenue to bypass the eft stage when aquatic conditions permit.

Aquatic juveniles mostly differ from adults in size (though they also seem to have smaller tails), as they both assume the same morphology and breed in the following breeding season.

Paedomorphosis in Notophthalmus most often involves varying degrees of incomplete metamorphosis, with true neotenes being rare (Reilly 1987). Thus, most individuals are in a state of a gradual metamorphosis rather than in discrete metamorphic and neotenic states as is the case with many Ambystoma (Brandon and Bremer 1966, Reilly 1987).

Paedomorphosis in newts is characterized by retention of external gills, gill slits, and broad, compressed tailfins (7, Fig. 1.9). Newt metamorphosis is density-dependent. High initial larval density dramatically increases the proportion of efts (Harris 1987) and lowers the proportion of aquatic phenotypes: adults, paedomorphs and aquatic juveniles. Overall, previous studies on newts have shown that their life history polyphenism complies with the PA hypothesis

(Wilbur and Collins 1973, Whiteman 1994), especially if you consider that adults, paedomorphs, and aquatic juveniles have extensive niche overlap (Moran 1992).

The mole salamander (Ambystoma talpoideum) is a relatively small (8–12 cm), but stout, facultatively paedomorphic salamander. Its geographic range extends along the coastal plain of the Southeastern United States and north into the Central United States (Southern Illinois,

Western Kentucky, Southeastern Missouri) (Shoop 1964). At UMFS, adults migrate from terrestrial hibernacula to breed in ponds from late November to early February. Aquatic larvae feed and develop until late spring, when individuals first start metamorphosing. Metamorphosis has not been observed during the peak of summer, but individuals start metamorphosing again in later summer/early autumn. If larvae do not metamorphose in autumn, larvae may remain aquatic, overwinter in ponds and metamorphose in following season (Petranka 1998b). Larvae may also become breeding paedomorphic adults. Paedomorphs retain to ability to metamorphose in following seasons should aquatic conditions deteriorate (Fig. 1.7).

The studies at TRC (Chapter 3) utilized Eastern newts (Notophthalmus viridescens) as model organisms, while all studies at UMFS (Chapters 2, 4–6) utilized the mole salamander. The

Eastern newt subspecies used was the Central newt (Notophthalmus viridescens louisianensis), which ranges over much of the Midwest. Mole salamanders are not present at the TRC, but both species occur at UMFS.

Figure 1.9 The complex life cycle of Notophthalmus viridescens. (1) Eggs are laid singly in aquatic vegetation and hatch into (2) aquatic larvae. Larval development is plastic with four distinct post-larval phenotypes (Reilly 1987; Petranka 1998). Paedomorphs (3, 4) retain gills, compressed tail fins and aquatic lifestyles. However, the transition between the paedomorphic and metamorphic morphologies is a continuous reaction norm and may be arrested at any stage in the transition. Thus, paedomorph morphology can range from (3) retaining a full larval morphology with gill slits, large fleshy gills and large tail fins to (4) having no gill slits and partially resorbed gills and tail fins (Reilly 1987). Metamorphosed phenotypes include (5) aquatic juveniles, (6) terrestrial efts and (7) semi-aquatic adults. (5) Aquatic juveniles are larvae that have metamorphosed but have not made the transition to terrestrial habitats. They resemble small metamorphosed adults with smooth skin and compressed tailfins. (6) Terrestrial efts are larvae that have metamorphosed and made the transition to terrestrial habitats. They have dry, hydrophobic and often brightly colored skin, and tubular tails. While (5) aquatic juveniles can reach sexual maturity as soon as the next breeding season, (6) efts can take up to eight years to reach sexual maturity (Healy 1974). (7) Semi-aquatic adults are the metamorphosed adult life stage that seasonally migrates between terrestrial and aquatic habitats. Although seasonal migration is typical, both semi-aquatic adults and aquatic juveniles can forgeo migration and overwinter in ponds. Illustration by Tatiana Tushyna.

CHAPTER 2:

INITIAL DENSITY SETS TRAJECTORY FOR DENSITY-DEPENDENT POLYPHENISM

THRESHOLDS

ABSTRACT

Negative density-dependence is a common process in ecology where population growth, or related traits, is inhibited as a population increases in size. Organisms can respond to increases in population size with plastic phenotypic traits, such as dispersal or cannibalistic morphologies, or by changing their behavior. Polyphenisms are plastic responses to environmental variation in which organisms can express alternative, discrete phenotypes to maximize their fitness under different population densities. While population density is a continuously distributed in nature, polyphenisms are discrete responses, thus, organisms must utilize some threshold criterion at which they switch to an alternative phenotype. I used facultatively paedomorphic mole salamanders as model organisms and exposed larval individuals to a gradient of eight population densities in a replicated mesocosm experiment. Individuals could either metamorphose into terrestrial adults or remain aquatic as larvae or paedomorphic adults. As expected, body size negatively correlated with both initial larval density and survival rate. Both paedomorphosis and metamorphosis were also negatively density-dependent. The proportion of paedomorphic individuals at any given density was relatively constant (40-50%) until larval densities of ~23 individuals/m3, where I saw a sudden decline and a large increase in the number of individuals remaining as larvae. Rates of metamorphosis fell at even lower densities (~7 individuals/m3) and were largely replaced by individuals remaining as larvae. Current theory concerning thresholds of facultative paedomorphosis could not adequately explain all of our results. Since polyphenisms are thought to be early stages of speciation and density-dependent processes are

ubiquitous in nature, it is critical to understand how thresholds determine expression of alternative phenotypes.

INTRODUCTION

One frequent challenge for organisms is coping with density-dependence impacts resulting from large population size. Though often credited to Malthus (1798), the negative effects resulting from population growth that formed the basis of Darwin’s famous “struggle for existence” have been implicated as early as Aristotle (Historia Animalium IX). Density- dependence is now well established as an important driver of competition, population size, and community structure (Verhulst 1838, Pearl and Reed 1920, Volterra 1926, Lotka 1932, Gause

1934, Nicholson 1954, Hassell 1975, Cappuccino and Price 1995, Chesson 1996, Turchin 1999,

Hellriegel 2000, Brook and Bradshaw 2006, Johnson et al. 2012). Through a variety of mechanisms, density-dependence resulting from large population size can affect individual fitness components such as survival, growth rate, fecundity, and parasite load (Brockelman 1969,

Wilbur and Collins 1973, Hassell 1975, Wilbur 1987, 1977, Anderson and Gordon 1982,

Petranka 1989, Berven 1990, Turchin 1999, Altwegg 2003, Loman 2004, Davenport and

Chalcraft 2014). Classically, the negative effects of density-dependence are thought to arise from lower per capita resource levels resulting from higher competition over a fixed amount of resources (Nicholson 1933, Gause 1934, Elton and Nicholson 1942). However, density- dependence can result from alternative mechanisms whereby increased intraspecific aggression

(Petranka 1989, Semlitsch and Reichling 1989), cannibalism (Pfennig and Collins 1993, Walls

1998, Hoffman and Pfennig 1999), or stress levels (Glennemeier and Denver 2002) can have similar negative effects on growth, survival, and fecundity.

Phenotypic plasticity offers a unique response to the negative effects of high population density. To offset these effects, organisms may undergo changes in morphology (Dudley and

Schmitt 1996), develop a dispersal phenotype (Uvarov 1921, Harrison 1980, Applebaum and

Heifetz 1999, Cisse et al. 2015), or cue ontogenetic niche shifts (Collins and Cheek 1983, Harris

1987, Semlitsch 1987, Pfennig 1992, Newman 1994, Hoffman and Pfennig 1999). Since these changes can often be dramatic, such as metamorphosis into a new phenotype and/or dispersal to new habitats, density-dependence must exert significance selection pressure. The negative effects of density-dependence are typically continuous in nature, whereby increasing numbers of individuals result in incremental effects on growth, survival, and performance. However, some plastic responses are polyphenisms, which are environmentally-induced, discrete, alternative phenotypes that occur in a single population (West-Eberhard 1989, Moran 1992). With a binary response to continuous environmental variation, organisms must utilize a threshold criterion that triggers a switch to an alternative phenotype (Moran 1992, Getty 1996). Where this threshold lies is critical to determining population dynamics, as we should expect large shifts in dispersal or effects on fitness depending on how the polyphenism responds to changes in population density.

Larval amphibians are classic model organisms that have been used to study density- dependence, and their growth and body size at metamorphosis have been shown to be strongly density-dependent (Wilbur and Collins 1973, Semlitsch 1987). Since many amphibians breed in temporary ponds, which are only viable habitats for a short period of time, there is the potential for incredibly high population densities. Many amphibians have evolved plastic life history traits to compensate for such environmental variability by trading away size at metamorphosis for a shorter time to metamorphosis, thereby shortening the time spent in the larval period (Wilbur and

Collins 1973, Newman 1992, Day and Rowe 2002). However, when conditions are conducive to

growth and are relatively stable, some species of salamanders have evolved a unique plastic response to take advantage of the long-lasting, high quality aquatic conditions. These salamanders exhibit a polyphenism called facultative paedomorphosis, and while they can still metamorphose when conditions deteriorate, they can also forego metamorphosis and instead attain sexual maturity while retaining a juvenile morphology (larval, in this case; i.e., paedomorphosis) (Gould 1977, Pierce and Smith 1979).

Facultative paedomorphosis is environmentally cued, and three different hypotheses exist to explain the expression of the alternative phenotypes, metamorphs and paedomorphs (Fig. 1.3).

The “paedomorph advantage” (PA) hypothesis suggests that individuals metamorphose if conditions are poor, but remain paedomorphic if conditions are favorable. The “best of a bad lot”

(BOBL) hypothesis proposes that the smallest individuals do not reach the minimum size necessary for metamorphosis and thus remain paedomorpihc to maximize fitness opportunities.

However, both processes can occur simultaneously “dimorphic paedomorph” (DP) hypothesis, with intermediate-sized individuals metamorphosing because they are inferior competitors and potentially prey to the largest size classes (Whiteman 1994). Theoretical predictions would consider high larval density as a poor quality condition that produces small-bodied individuals and, thus, should induce metamorphosis. Low population densities are high quality conditions that should favor paedomorphosis. While the patterns are clear at the two extremes of the population density spectrum (Harris 1987, Semlitsch 1987), it is unclear how the expression of the alternative phenotypes changes over a density gradient. In order to understand how changes in larval density affect variation in phenotype expression, I conducted an experiment using a replicated, gradient mesocosm design with the mole salamander, Ambystoma talpoideum.

METHODS

Experimental Design

To test the effects of initial density on adult phenotype, I constructed an experimental array of forty 1.8 m diameter, 1200L cattle tanks (mesocosms; N = 40) in a mowed field at

UMFS. Mesocosms were filled with well-water and randomly received assigned aliquots of both zooplankton inocula (1.5 L) from a fishless pond (#61) and dry, hardwood leaf litter (2 kg).

Mesocosms were fitted with a window screen lid that closed it to colonization by other organisms. Mesocosms were assigned one of the following initial larval numbers (starting at 3 individuals followed by increments of 5 individuals): 3, 8, 13, 18, 23, 28, 33, and 38. These densities, which ranged from 2.5 to 31.67 individuals/m3, cover the range of densities seen in natural ponds (Semlitsch 1987). Each treatment was replicated five times (n = 5) and each treatment was represented once in each of the five rows (= blocks) (8 treatments × 5 blocks = 40 mesocosms).

Egg masses of A. talpoideum were collected from ponds at UMFS in the winter (2016-

2017). They were subsequently housed in small outdoor wading pools for hatching and kept separated by date of oviposition. A sufficient number of eggs could not be collected from one night of oviposition for all mesocosms, so the addition of hatchlings to mesocosms occurred in three phases. On 20 Jan, mesocosms in block 1 received randomly selected A. talpoideum hatchlings from the earliest cohort of egg masses, and the same was done on 21 Jan for blocks 2 and 3 and 24 Jan for blocks 4 and 5.

On 11 Feb, screen lids were depressed into the water to allow insect oviposition, which adds more invertebrate prey to the mesocosms’ food supplies for larval salamanders. I began night checks for emerging metamorphs on 12 May, and they occurred every three days until late

October when metamorphs stopped emerging. Metamorphs were collected by hand, weighed, photographed and then released into the terrestrial environment at UMFS. During 14–15 Dec, the experiment was terminated and all remaining individuals (larvae and paedomorphs) were collected, weighed and photographed. Paedomorphs were distinguished from larvae by enlarged testes on males and a swollen cloaca with gravid body shape on females.

Statistical analyses

Using the lme4 package v1.1.17 (Bates et al. 2015) in R v3.5.1 (R Core Team 2018), I used linear mixed models (LMMs) to analyze larval response variables: snout-vent length, mass, larval period, and body condition. Snout-vent length and mass were analyzed with initial density and phenotype as fixed effects (to examine differences in body size across the different phenotypes), and survival rate as a fixed covariate (to control for changes in density). Phenotype was included as a fixed factor to examine differences in body size among phenotypes. Larval period was analyzed with initial density as a fixed factor and survival rate as a fixed covariate.

Larval body condition (size-independent mass) was assessed in an LMM by performing a second analysis of mass with the addition of SVL as a fixed covariate (Garcia-Berthou 2001). I modeled all individuals as data points (as opposed to averaging per mesocosm), so I included mesocosm nested within block as a random effect in our models to account for this non-independence among individuals sharing a mesocosm. Significance for linear mixed effects models was tested with Approximate F Tests (Type III Satterthwaite) from the lmerTest package v3.0.1

(Kuznetsova et al. 2015).

Survival and phenotype proportions were analyzed with logistic GLMMs made with the lme4 package (Bates et al. 2015). Survival analysis data were aggregated by mesocosm and, thus, excluded the nested random term mesocosm, but still included the random term block.

Phenotype proportions were analyzed using three separate logistic GLMMs to compare metamorphs vs. paedomorphs, metamorphs vs. larvae, and paedormorphs vs. larvae. Three separate regressions were conducted instead of multinomial regression since multinomial regression requires repeated analyses with multiple reference levels to perform a complete set of comparisons. Additionally, GLMMs readily accommodate the nested data structure (mesocosm nested within block) of our design. Logistic GLMMs followed the structure of the previously mentioned LMMs by including initial density and survival rate as fixed effects, along with mesocosm nested within block as a random effect. The significance of logistic GLMMs was tested with likelihood ratio tests (Bolker et al. 2009, Warton and Hui 2011, Bates et al. 2015).

Since this experiment was explicitly concerned with density effects, fourteen mesocosms with exceptionally low survival (< 30%) were excluded from analyses. All analyses set α = 0.05.

Plots were made using ggplot2 v3.0.0 (Wickham 2009) and sciplot v1.1.1 (Morales and R Core

Team 2012).

RESULTS

Survival rate across all initial larval densities for analyzed mesocosms averaged 75.0%

and there was no difference in average survival across different densities ( = 0.026, p =

0.872). Of all individuals, 71.9% remained as larvae, 17.9% became paedomorphic, and 10.2% metamorphosed. Initial density had no effect on the relative proportion of paedomorphosis to

metamorphosis ( = 0.080, p = 0.777), and there was also no effect of survival ( = 0.476, p =

0.490). However, when comparing the proportion of metamorphosis to those remaining as

larvae, fewer individuals metamorphosed and more remained as larvae as initial density ( =

29.899, p < 0.001) and survival increased ( = 10.790, p = 0.001). The same pattern occurred with paedomorphosis —fewer individuals became paedomorphic and more remained as larvae as

initial density, = 31.187, p < 0.001) and survival increased ( = 9.268, p = 0.002) (Fig.

2.1).

As expected, there was a logarithmic relationship between mass and snout-vent length

(Fig. 2.2). Size variables reflected density-dependent growth patterns as they all decreased with increasing initial larval density and survival rate. Mass declined with increasing initial larval density (F1, 22.84 = 65.65, p < 0.001) and survival rate (F1, 22.86 = 14.73, p < 0.001) (Fig. 2.3), as did body condition (density: F1, 31.00 = 8.49, p = 0.007; survival: F1, 29.26 = 9.15, p = 0.005).

Paedomorphs had the largest body sizes, followed by metamorphs and then larvae (F2, 364.50 =

110.88, p < 0.001) (Fig. 2.4). Mean SVL decreased with initial larval density (F1, 25.29 = 126.41, p

< 0.001) and survival rate (F1, 26.44 = 17.78, p < 0.001). SVL also differed across phenotypes (F1,

373.38 = 78.13, p < 0.001) with paedomorphs reaching larger sizes than metamorphs or larvae

(Fig. 2.5). Larval period of metamorphs showed no differences across initial larval densities (F1,

15.71 = 1.38, p = 0.258) and there was no effect of survival (F1, 11.10 = 0.044, p = 0.838).

Figure 2.2 Phenotype proportions of all individuals across all initial larval densities. The proportion of paedomorphs was relatively constant until reaching 23.33 individuals/m3, where proportion of individuals remaining as larvae increased. Metamorph proportions decreased tended to decrease after reaching 6.67 individuals/m3.

Figure 3.2 The relationship between body mass and snout-vent length is logarithmic, reflecting a decelerating larval growth pattern. The shaded region surrounding the curve represents the 95% confidence interval.

Figure 2.4 Mean body mass (± SE) decreased with increasing larval density.

Figure 2.5 Mean body mass (± SE) of each phenotype.

Figure 2.6 Mean snout-vent length (± SE) of A. talpoideum over different initial larval densities and separated by different phenotypes expressed at the end of the experiment.

DISCUSSION

Larval growth and body size was largely determined by density, with increasing density having predictably negative effects on body sizes. More interesting were the effects of larval density on phenotype frequency. The proportion of paedomorphs remained relatively constant until densities of 23.33 individuals/m3, when the proportion of larvae dramatically increased. The rate of metamorphosis was expected to increase with increasing density since high population density is considered “unfavorable” habitat conditions. However, the likelihood of metamorphosis was clearly negatively density-dependent with a potential threshold near 6.67 individuals/m3. Instead of increased rates of metamorphosis, there were increasing rates of individuals remaining as larvae, which made up the difference as rates of metamorphosis decreased (Fig. 2.1). These results only partially support predictions made by the PA and DP hypotheses, and I found no support for the BOBL hypothesis. Under the predictions of the DP hypothesis, the proportion of paedomorphs is expected to be greatest at both low and high larval densities (i.e large and small body sizes), while the expected maximum proportion of metamorphs is expected at intermediate densities. While I did see high rates of paedomorphosis at low larval densities, I did not see a similar pattern at high larval densities and I also did not see increased rates of metamorphosis at intermediate densities. Thus, our results with this population of A. talpoideum mostly support the PA hypothesis.

Following the PA hypothesis, the paedomorphic individuals reached the largest body sizes in our study, while metamorphs were consistently smaller across all densities (Fig. 2.5).

Unsurprisingly, individuals remaining as larvae were the smallest among all phenotypes. There

was no obvious differentiation between large-bodied (PA) and small-bodied (BOBL) paedomorphs as they showed a rather continuous distribution (Fig. 2.5). Since many individuals overall were small and many remained as larvae, the data suggests that many of these individuals did not reach the minimum size required for metamorphosis (~30mm; M*; Fig. 1.8) (Semlitsch

1987, Whiteman 1994). Our experiment used natural density ranges found for A. talpoideum

(Semlitsch 1987), yet I still found patterns that suggest most individuals need to overwinter as larvae for additional growth. This is a common strategy for many amphibians, but it can be risky if habitats are temporally variable. However, there is little decision making involved since the minimum size required for metamorphosis appears to represent some physiological body size limit that overrides any polyphenic responses.

Our results have narrowed the range of densities in which the polyphenic thresholds occur for this population (23.33 individuals/m3 for paedomorphs, 6.67 individuals/m3 for metamorphs). Densities above 23.33 individuals/m3 appear to have the strongest effect on phenotype since nearly all individuals remained larval at those densities, and the few that did not emerged from mesocosms with relatively lower survival (i.e., effectively lower density). Other thresholds were not as sharply demarcated since all phenotypes emerged at nearly all initial larval densities. This variability in phenotype frequencies can arise from multiple sources. For example, we know that variation in initial body size can influence phenotype frequency (Doyle and Whiteman 2008) and many factors can generate this variation in mesocosms such as stochastic spatial distribution of prey, genetic growth factors, and aggressive behavior from conspecific.. Small variations in body size can potentially expand into larger variations in body size (Walls 1998) because larger individuals can monopolize resources, harass smaller conspecifics (Walls and Jaeger 1987), and even cannibalize them, which is common in

Ambystoma (Pfennig and Collins 1993, Hoffman and Pfennig 1999). Cannibalism, especially, can cause variation in body size among individuals because it simultaneously decreases larval density and increases the cannibal’s energy intake. Another potential factor is individual variation (genetic) in sensitivities to and production of hormones that regulate metamorphosis

(Denver 2017). Since metamorphosis is inherently controlled through environmentally-induced hormone production (Boorse and Denver 2002), some individuals may be more likely than others to metamorphose under certain circumstances. Furthermore, many of these factors can also interact, e.g., variation in body size may increase agonistic interactions, which thereby increase hormone production in smaller individuals.

What is the mechanism of density-dependence? It is classically assumed that the main mechanism of density-dependence is decreased per capita resources (e.g., prey, light, nutrients, refuge). However, more factors than just per capita resources are affected when populations increase. For example, larval injury rates scale with increasing density, suggesting that agonistic interactions can play an important role in density-dependent growth (Semlitsch 1987, Petranka

1989, Walls 1998, Wildy et al. 2001). Agonistic interactions are also stressful interactions and environmental stress is known to induce the production of hormones that induce metamorphosis

(CORT, thyroxine) (Boorse and Denver 2002, Denver 2017). In fact, density-dependence in amphibians has been directly related to CORT production (Glennemeier and Denver 2002), but there is some conflicting evidence concerning this relationship (Belden et al. 2007). In aquatic organisms, increased conspecific density can also lower water quality through excessive toxic ammonia waste, which is a major issue in the development of aquacultures (Huchette et al.

2003). Lastly, many of these alternative factors are non-consumptive effects (the sum total costs of avoiding the predation) and they have been implicated as a widespread, but unrecognized,

mechanism regulating density-dependent processes such as the famous Lynx-Hare predator-prey oscillations (Peckarsky et al. 2008).

These results illustrate the density-dependent nature of facultative paedomorphosis, which showed a relatively constant frequency of paedomorphosis until densities of 23.33 individuals/m3. Our results did not fit precisely with current theory, suggesting that theory may not fully explain patterns for all populations. However, using final body size may not be entirely sufficient to explain why individuals expressed a given phenotype. Instead, it may be useful to investigate growth trajectories, which will naturally respond to changes in density resulting from individuals metamorphosing. In other words, it is important to consider how density changes over time and how this may affect growth patterns and, in turn, phenotype. This type of study is difficult since handling itself is likely to increase the probability of metamorphosis due to the relationship between stress and the induction of metamorphosis (Denver 2017). Regardless, polyphenisms are important responses to environmental variation, and they can only be fully understood when considering that organisms must respond to some threshold that initiates a switch to an alternative phenotype.

CHAPTER 3:

ARE DIRECT DENSITY CUES, NOT RESOURCE COMPETITION, DRIVING LIFE

HISTORY TRAJECTORIES IN A POLYPHENIC SALAMANDER?

ABSTRACT1

Polyphenisms, where multiple, discrete, environmentally-cued phenotypes can arise

from a single genotype, are extreme forms of phenotypic plasticity. Cue acquisition and

interpretation are vital for matching phenotypes to varying environments, but can be difficult

if cues are unreliable indicators or if multiple cues are present simultaneously. Facultative

paedomorphosis, where juvenile traits are retained at sexual maturity, is a density-dependent

polyphenism exhibited by many salamanders. Favorable conditions such as low larval

densities and stable hydroperiod delay metamorphosis and promote a paedomorphic strategy.

I investigated proximate cues affecting facultative paedomorphosis in order to understand

how larval newts (Notophthalmus viridescens louisianensis) assess conspecific density. To

isolate the effects of density cues from the effects of resources and agonistic behavior, I

caged larval newts in mesocosms in a 2×2 factorial design that manipulated both background

larval newt densities (high or low) and food levels (ambient or supplemented). I found strong

effects of both food and density on caged individuals. Under high densities, caged larvae

were more likely to become efts, a long-lasting juvenile terrestrial stage, across both food

levels, while paedomorphs were more common under low densities. Though food levels

increased growth rates, density had strong independent effects on metamorphic timing and

phenotype. Competition for food and space are classical density-dependent processes, but

1 This chapter was published in this article: Bohenek, J. R., and W. J. Resetarits. 2018. Are direct density cues, not resource competition, driving life history trajectories in a polyphenic salamander? Evolutionary Ecology 32:335–357.

density cues themselves may be a mediator of density-dependent effects on polyphenisms and life history responses.

INTRODUCTION

In variable environments, a singular, inflexible phenotype may not be optimal compared to polyphenisms, which are a type of phenotypic plasticity where multiple, environmentally-cued phenotypes arise from a single genotype (West-Eberhard 1989). Polyphenisms are potentially adaptive in that each alternative phenotype maximizes survival and reproductive output under particular environmental conditions (Moran 1992). The environment bombards organisms with cues and parsing informative, reliable cues from uninformative, unreliable cues is necessary to optimize responses. Phenotype mismatching, due to either poor cue acquisition or integration, can result in potentially severe fitness consequences, thus, we should expect strong selection on acquisition and evaluation of reliable environmental cues (Getty 1996).

A variety of biotic and abiotic cues influence polyphenisms across a range of environments and taxa (Grunt and Bayly 1981; Pfennig 1990; McCollum and Van Buskirk 1996;

Moczek 1998; Michimae and Wakahara 2002; Nijhout 2003; Laforsch et al. 2009; Maher et al.

2013). Many polyphenisms are density-dependent, where crowding elicits alternative phenotypes such as dispersal phenotypes in insects (Uvarov 1921; Nijhout 2003; Pener and Simpson 2009), and terrestrial (Harris 1987b; Grayson and Wilbur 2009) and cannibalistic phenotypes in salamanders (Collins and Cheek 1983). Density-dependence is an important driver of competition, population size and community structure (Verhulst 1838; Pearl and Reed 1920;

Volterra 1926; Lotka 1932; Gause 1934; Brook and Bradshaw 2006). Through a variety of mechanisms, density-dependence affects fitness components such as survival, growth rate, fecundity, and parasite load (Brockelman 1969; Wilbur and Collins 1973; Hassell 1975;

Anderson and Gordon 1982; Petranka 1989b; Turchin 1999). Organisms can respond to negative effects of high population density via phenotypic plasticity, initiating changes in morphology

(Hoffman and Pfennig 1999), developing dispersal phenotypes (Harrison 1980; Applebaum and

Heifetz 1999; Cisse et al. 2015), or initiating ontogenetic niche shifts (Collins and Cheek 1983;

Harris 1987b; Semlitsch 1987; Pfennig 1992; Newman 1994; Hoffman and Pfennig 1999).

Many amphibians are explosive breeders, resulting in the potential for extreme crowding and density-dependence in the larval stage (Wilbur and Collins 1973; Petranka 1989a; Van

Buskirk and Smith 1991; Wildy et al. 2001). Thus, larval amphibians possess phenotypic plasticity in development rate through metamorphosis that can be modulated to either escape deteriorating conditions or exploit favorable conditions (Wilbur and Collins 1973; Werner and

Gilliam 1984; Newman 1992; Denver et al. 1998). High conspecific density can restrict growth via exploitative competition to the point where organisms cannot reach minimum size required for metamorphosis (Newman 1987; Scott 1990). However, density-dependent effects may also arise from alternative mechanisms (Richter et al. 2009). Stress, as a result of agonistic behavior

(Walls and Jaeger 1987; Petranka 1989a; Semlitsch and Reichling 1989; Wildy et al. 2001;

Glennemeier and Denver 2002), tactile and visual cues (Rot-Nikcevic et al. 2005, 2006) or chemical cues, has been implicated in increasing amphibian development rates through metamorphosis (Glennemeier and Denver 2002). Thus, larval growth and survival patterns typically attributed to exploitative competition may actually be a result of stress from a variety of cue sources, which can scale with conspecific density and influence development rates (Wildy et al. 2001; Rot-Nikcevic et al. 2005, 2006; Richter et al. 2009). Understanding the effects of density cues (tactile, visual, and chemical), independent of the effects of competition and injury

from agonistic behavior, is problematic, as they are confounded in nature (Petranka 1989a;

Kuzmin 1995; Richter et al. 2009).

Some salamanders are polyphenic and can delay or prevent metamorphosis in favor of paedomorphosis (Fig. 1.9), which is broadly defined as the retention of aquatic juvenile characteristics at sexual maturity (Gould 1977). Paedomorphosis in salamanders is either obligate, where the ability to metamorphose has been lost, or facultative, which is a polyphenism where either metamorphic or paedomorphic adult phenotypes are possible (Harris 1987b;

Semlitsch 1987; Whiteman 1994; Denoël et al. 2005; Denoël and Ficetola 2014). Delaying or preventing metamorphosis can be a viable strategy if aquatic conditions are favorable (Wilbur and Collins 1973; Werner and Gilliam 1984; Whiteman 1994), and such conditions have been shown to increase the frequency of paedomorphosis (“paedomorph advantage” hypothesis)

(Wilbur and Collins 1973; Harris 1987b; Semlitsch 1987; Whiteman 1994; Denoël and Ficetola

2014), but not always (see "best of a bad lot" scenario, Whiteman 1994; Whiteman et al. 2012).

Relevant environmental factors that affect the expression of paedomorphosis include pond drying (Semlitsch 1987; Semlitsch et al. 1990), conspecific density (Harris 1987b; Semlitsch

1987), temperature (Sprules 1974), and food availability (Sprules 1974; Ryan and Semlitsch

2003). Under poor aquatic conditions, metamorphosis may be a better alternative; however, initiating metamorphosis in a productive aquatic habitat precludes substantial growth opportunities that can lead to greater fecundity and offspring opportunities (Denoël et al. 2005).

Maximizing growth in the aquatic stage is optimal because size at metamorphosis is a strong correlate of fitness (Semlitsch et al. 1988). Additionally, favorable conditions may allow larvae to skip juvenile stages and develop directly into paedomorphs or metamorphosed adults, thus decreasing age at maturity, which has strong fitness effects (Cole 1954; Stearns and Koella

1986). Environmental cues affecting timing of metamorphosis, feeding rates and/or onset of sexual maturity have a large impact because opportunities in the larval stage set the trajectory of future fitness.

Plentiful food should be essential for “favorable conditions” and should have obvious effects on growth. However, the interaction between food and facultative paedomorphosis are complex, conflicting and unresolved. For example, Semlitsch (1987) found no effect of food levels on the expression of paedomorphosis in Ambystoma talpoideum Holbrook. In contrast,

Denoël and Poncin (2001) found that captive paedomorphic newts (Ichthyosaura alpestris

Laurenti) metamorphosed later with ad libitum food and metamorphosed earlier with low food levels. The effect of food on the expression of paedomorphosis may interact with the various stages of larval development. Ryan and Semlitsch (2003) found that high food levels later in development promote metamorphosis, while low food levels late in development promote paedomorphosis. Food levels should theoretically interact with larval density (but see Petranka

(1989a)), but Semlitsch (1987) crossed these factors and found only density effects, suggesting that larvae directly assess density. However, food and density are still confounded in these studies and it is unknown whether larvae assess density via growth effects arising from competition and stress from agonistic behavior or via alternative density cues like tactile, visual and chemical cues.

An experiment was conducted using central newts (Notophthalmus viridescens louisianensis Rafinesque) (Fig. 1.9) to determine the cues used by larval newts to assess density.

I utilized a fully factorial design with two levels of conspecific density (high and low) crossed with two levels of food (supplemented and ambient). Our response individuals were individually caged in each mesocosm to prevent physical interaction, agonistic behavior and environmental

exploration, but still allowed access to water-borne cues, which I hypothesized as the most informative and reliable indicator of conspecific density (Dettner and Liepert 1994) and, thus, habitat quality. I expected that cues indicating high density (i.e., waterborne cues) can independently promote terrestrial life history strategies, increase development rate and decrease length of larval periods and body size at metamorphosis.

METHODS

Breeding Mesocosms

On 5 April 2013, metamorphosed adult newts were collected from ponds at the Tyson

Research Center. Twelve cattle tanks (breeding mesocosms) were filled with 1kg leaf litter and

1200L of well water. For oviposition substrate, four sprigs of either Egeria densa (Planch) and/or

Elodea Canadensis (Michx.) were planted in plastic pots and added to each mesocosm. On 7

April, one male and one female newt were added to each of nine mesocosms and on 7 May, a pair of newts was added to two more breeding mesocosms. On 26 May, seven male and three female newts were added to the final breeding mesocosm — all offspring from this mesocosm were used only as background density (see below) due to multiple females. Eggs were collected every few days starting 6 May by searching through the water plants. The eggs were transferred to the lab and placed into individually marked hatching containers filled with aged tap water

(5.68L, 34.3 × 21.0 × 12.1 cm), separating the eggs by consanguinity. The larvae were fed a mixture of bloodworms (San Francisco Bay Brand, Inc., Newark, CA & Hikari BIO-PURE

Blood Worms, Hikari Sales USA Inc., Hayward, CA) ad libitum until they reached a minimum total length of 10mm, which was large enough to prevent them from passing through the 1.3 x

1.13mm cage mesh.

Experimental Mesocosms

Sixteen cylindrical, 1200L plastic mesocosms (1.8 m diameter, 50 cm depth; ARM-

10138, Ace Roto-Mold, Hospers, Iowa, USA) were constructed at Tyson from 9–10 May, filled with well-water and allowed to age for ~50 days. Each mesocosm had 0.5 kg of dry leaf litter added on 13 June, and were covered with fiberglass screen lids (1.3 x 1.13 mm mesh) to prevent colonization and oviposition by other organisms. The experiment was a randomized complete block factorial design: two levels of density (low density [LD] = 8 and high density [HD] = 40 larval newts), similar to those found in natural ponds (Harris 1987b; Harris et al. 1988), crossed with two levels of food (ambient [AF] and supplemented food [SF]). Mesocosms were assigned into blocks based on date of larval addition (see below). Each treatment was represented once in each of the four blocks. Overall, the design consisted of four distinct treatment combinations (k =

4) replicated across four blocks (n = 4) for a total of 16 mesocosms (N = 16).

Each tank held eight cages consisting of a large, black plastic plant pot (28 cm height ×

32 cm diameter) with an open cylindrical mesh top (1.3 x 1.13mm mesh, 35.5 cm h × 32 cm d) extending halfway through the water column and above the water level, supported and propped open by four wooden dowels. Cages were tall enough to rest on the bottom (stabilized with a randomly selected rock) and extend out of the water so individuals that developed lungs could gulp air and metamorphosed efts could crawl out of the water. The bottom half of the cages were opaque except for 1.3 x 1.13 mm screen covering five ~1cm diameter holes. LD treatments contained 8 total individuals (6 larval newts/m3) that were all separately caged and HD treatments contained 40 individuals (32 larval newts/m3), 8 caged and 32 not caged. The caged individuals were “response” individuals, since they were subject only to density cues. LD individuals are only exposed to cues from other caged individuals while HD individuals are

exposed to cues from both caged and background individuals. Thus, HD response individuals should be exposed to greater density cues than LD response individuals.

Larvae were assigned one at a time to each mesocosm by first randomly selecting a source breeding mesocosm, and then randomly selecting an individual hatchling. This randomized process was performed for all individuals in all mesocosms, ensuring that each mesocosm received randomly selected individuals from a randomly selected breeding mesocosm. Since newts oviposit single eggs over multiple weeks, larvae were introduced one block at a time, between the dates of 27 June and 13 July, once sufficient numbers were accumulated. Each cage was marked with the caged individual’s source breeding mesocosm so that genetic differences could be accounted for in statistical analyses. Only caged response individuals could be tracked as background individuals were randomized, but not individually marked. Of the eleven total females that contributed to mesocosm cages, each experimental mesocosm received input from 4–6 different breeding mesocosms (mean = 5.3), providing a relatively balanced contribution across the experimental array.

To parse the effects of resource competition from density effects, supplemental food was added to SF treatments once per week beginning 7 July until 6 October. Supplemental food consisted of 0.5g/individual/week of frozen bloodworms (Omega One Whole Frozen

Bloodworms, Omega Sea Ltd., Sitka, AK, 6.3% min. crude protein, 0.8% min. crude fat, 0.3% max. crude fiber, 91.2% max. crude moisture), with amount based on the ad libitum quantities consumed in hatching containers. Individual rations were greater than individual larval body mass (>100%) for nearly the entire experiment across all treatments. HDSF tanks received

20g/wk of bloodworms (4g evenly distributed among the 8 cages and 16g outside the cages) and

LDSF received 4g/wk (distributed evenly among the cages). All caged individuals were briefly

removed (< 3min) and photographed for measurements of snout-vent length (SVL) on 27 July

(Day 30) and 17 August (Day 51) in order to track growth rates. The open cage tops were closed from 31 August – 3 September to prevent climbing efts from escaping and mixing with background individuals.

The larval period ended with larvae either becoming paedomorphic or metamorphosing into efts, aquatic juveniles or semi-aquatic adults. At the end of the larval period, newts were removed, massed, and photographed. Newts began to metamorphose in August and continued into mid-October. Mesocosms were checked every other night for emerging metamorphs beginning 15 August. Caged individuals were checked weekly for signs of paedomorphosis or metamorphosis. The experiment ended on 26 October and all remaining background and caged individuals were removed, massed and photographed. All caged individuals were euthanized with Tricaine-S (MS-222; Ferndale, WA), fixed with 10% neutral buffered formalin solution

(Sigma-Aldrich Corporation, LLC., St. Louis, MO) and then preserved in 70% ethanol. To properly evaluate phenotypes, a total of ten representative individuals (5 efts, 3 aquatic juveniles and 2 paedomorphs) representing all blocks and treatments were selected from the preserved specimens and had their hyobranchial apparati cleared and double-stained (Hanken and

Wassersug 1981). Staining was done to ensure gilled individuals were actually paedomorphs and not simply large larvae, as the two phenotypes can be ambiguous before secondary sexual characteristics develop during the breeding season. Stained individuals were evaluated for the presence of larval ceratobranchials, which are a key skeletal trait in the head that distinguishes larval newts from post-larval phenotypes. Any gilled individuals lacking larval ceratobranchials were considered paedomorphs (Reilly 1987).

Data Analysis

Final SVL was determined from photographs using a standard 1×1mm grid background and Image-J v1.49 (Schneider et al. 2012). I used linear mixed effects models (LMMs) for variables that followed a normal distribution and generalized linear mixed effects models

(GLMMs) for binomial variables. Models were developed using the lme4 v1.1.13 package

(Bates et al. 2015) in R v3.4.0 (R Core Team 2017). Significance for LMMs was tested with lmerTest v2.0.33 Approximate F Tests (Type III Satterthwaite denominator degrees of freedom approximation) (Kuznetsova et al. 2015), while fixed effects of GLMMs were analyzed using z- test statistics from lme4 summary output (Bolker et al. 2009). AICc was calculated with MuMIn v1.40.4 (Barton 2018). All analyses used α = 0.05, and all figures were made using raw data with ggplot2 v2.2.1 (Wickham 2009) and sciplot v1.1.1 (Morales and R Core Team 2012).

Survival and phenotype proportion were modeled as logistic GLMMs (Warton and Hui

2011). Survival data was data aggregated by mesocosm (counts per mesocosm). Phenotype proportions were also modeled as a logistic GLMM, but with non-aggregated data. Since newt post-larval phenotypes are not binomial, two phenotype analyses were conducted: efts vs. non- efts and paedomorphs vs. non-paedomorphs. These analyses were chosen to contrast the terrestrial life history choice (eft) with aquatic life history choices (aquatic juveniles, paedomorphs) and to contrast metamorphosed individuals (efts, aquatic juveniles) with paedomorphs. However, due to the statistical issue of “complete separation,” the analysis of paedomorphs vs. non-paedomorphs was modeled with a LMM using proportional data aggregated by mesocosm.

Length of the larval period (days), SVL, growth rate (SVL / day) and body condition were analyzed using LMMs. Body condition (size independent mass) was analyzed in two ways.

First by analyzing mass with SVL as a fixed covariate in a LMM (Garcia-Berthou 2001), then by mean-scaling masses of caged individuals to decouple variance from the measurement scale and means, regressing against SVL, and modeling the residuals (Berner 2011). The base statistical model for all analyses included density, food and their interaction as fixed effects and mesocosm nested with block as a random effect. AICc was used to compare the base model with those that also included any combination of overall larval survival per mesocosm as a fixed covariate and source breeding mesocosm (to account for genetic effects) as a random effect (Table A.1).

Using the same approach to alternative models as above, repeated measures was conducted on

SVL by including time as a fixed effect crossed with density and food to explore patterns over time and individual as a random effect to control for pseudoreplication (Table A.1).

RESULTS

One LDSF mesocosm was excluded from analyses because of failed cages. Of the 376 individuals in the study, 74.7% survived until the end of the experiment, with 83.3% of the 120 focal, caged individuals surviving and 66.4% of the 256 background individuals surviving.

Survival was not different across densities, food levels, or the density × food interaction (Table

A.2).

Phenotypic Proportions

Individuals started metamorphosing in late August, which eventually slowed in early

October and then stopped in mid October when temperatures began to drop, which directly mirrored patterns seen by Harris (1987b). Newts were categorized as efts, aquatic juveniles or paedomorphs. All ten individuals (5 efts, 3 aquatic juveniles and 2 paedomorphs) that were cleared and stained were clearly non-larval in structure (Reilly 1987). All paedomorphs were partially metamorphosed with closed gill slits and partially reduced gills, as is typical for

Notophthalmus (Reilly 1987). Due to logistical constraints, the experiment was not carried out long enough into their breeding season to properly assess sexual maturity. All metamorphosed newts that remained in the water until the end of the experiment were categorized as aquatic juveniles because they retained smooth skin, remained in the water for extended periods while metamorphosed and showed no attempts at dispersal. Besides differences in size, aquatic juveniles are anatomically indistinguishable from metamorphosed adults until the following breeding season when any adults will show visible secondary sexual characteristics. Efts were

easily identified as they had dry, rough, hydrophobic skin and attempted to disperse from the mesocosms.

Efts were the most common phenotype among caged individuals at 79%, while aquatic juveniles and paedomorphs represented 15% and 6%, respectively (Fig. 3.1). The proportion of efts vs. non-efts varied across density, but not food levels or the density × food interaction (Fig.

3.1) (Table A.2). Paedomorphs were uncommon, but were marginally more common in low density with no difference across food levels or the density × food interaction (Fig. 3.1) (Table

A.2).

Figure 3.1. Phenotype proportions of response (caged) individuals across the four treatments. There was a significant density effect on proportion of efts (p = 0.001) and a marginal effect on proportion of paedomorphs (p = 0.095). HD = high density, LD = low density, SF = supplemented food and AF = ambient food.

Life History Traits

Larvae had fast development times in high density as evidenced by their shorter larval periods. Larval period was also shorter with supplemented food compared to ambient food, but

there was no significant density × food interaction or any effect of survival (Fig. 3.2a) (Table

A.2). Shorter larval periods resulted in reduced SVL (Fig. 3.2b) (Table A.2), which incurs a fitness cost to individuals because SVL has a positive relationship with fitness (Semlitsch et al.

1988). Unlike with larval period, food levels did not have the same strong effect on SVL, and similarly there was no density × food interaction or survival effect (Fig. 3.2b) (Table A.2).

Supplemental food increased growth rates but there was no difference across densities and no density × food interaction, or survival effect (Fig. 3.2c) (Table A.2). These patterns suggest that supplemental food shortened larval period via accelerated growth rates that allowed larvae to reach minimum size required for metamorphosis earlier (Fig. 3.2a). However, the timing of metamorphosis and likelihood of its onset was more strongly predicted by density than food levels (Fig. 3.2a). The two body condition analyses produced nearly identical results: there was no difference in body condition across density, food level or their interaction, but SVL strongly predicted mass in the covariate approach (Table A.2). I present the residual index for visualization (Fig. 3.2d).

Body size differences between treatments were gradually accumulated over time.

Repeated measures revealed a significant density × time interaction as SVL was initially indistinguishable among treatments, but started diverging by the second sampling period (Day

51) and showed the greatest differences in SVL at the end of the larval period (Fig. 3.3).

Repeated measures also showed main effects of time and a food × time interaction and marginal density main effect. There was no effect of food level, the density × food interaction, the density

× food × time interaction or survival (Table A.2).

Figure 3.2 (a) The larval period of response (caged) individuals measured as days to metamorphosis (mean ± SE). There were significant density (p < 0.001) and food level (p = 0.010) effects, but no density × food interaction (p = 0.702) or survival effect (p = 0.324). (b) Final snout-vent length (mm) (mean ± SE) of response (caged) individuals. There was a significant density effect (p = 0.009), but no food level effect (P = 0.191) or density × food interaction (p = 0.257) or survival effect (p = 0.501). (c) Growth rate (SVL/day) (mean ± SE) of response (caged) individuals. There was a significant food (p = 0.001) effect, but no density effect (p = 0.284) or density × food interaction (p = 0.260) or survival effect (p = 0.236). (d) Body condition (mean ± SE) of response (caged) individuals was not different between density (p = 0.542) or food levels (p = 0.973) and there was no density × food interaction (p = 0.303) or effect of survival (p = 0.967). Ambient food is represented by closed circles and supplemented food by open circles. Raw data is used in all graphs except body condition, which is a derived index.

Figure 3.3. Mean snout-vent length (mm) (mean ± SE) of response individuals of each treatment over time. SVL measurements were taken 30 (27 July) and 51 days (17 August) after larvae were introduced into the first block and the third measurement was taken at the end of the larval period, which varied between individuals. There was a significant density × time interaction (p < 0.001) and food × time interaction (p = 0.014) (Table A.2). LDSF = open squares, LDAF = closed squares, HDSF = open circles, HDAF = closed circles.

DISCUSSION

Cues indicating density come in a variety of forms and via different sensory modalities, from internal cues like stress, hunger or growth rate, to external cues like visual or tactile cues while encountering conspecifics. Here, the effects of food and density appear to be mostly additive due to their non-interactive, parallel effects (Fig. 3.2a, b, c). Food levels play an obvious role in promoting growth (Fig. 3.2c), but our results suggest that cues indicating high density, unrelated to resource competition or agonistics behavior, are an independent driver of metamorphosis timing (Fig. 3.2a) and thus phenotype (Fig. 3.1) in developing larval newts. It has long been assumed that increased competition for food was the main consequence of high conspecific density (Wilbur and Collins 1973), but the mechanism of density-dependent effects is confounded with a multitude of factors. Past studies have suggested that these plastic growth and developmental responses are unrelated to resource competition (Semlitsch 1987; Petranka

1989a; Semlitsch and Reichling 1989) and others have related density effects to an integrated stress response (Glennemeier and Denver 2002).

In the Wilbur-Collins model (1973), metamorphosis optimally occurs once growth in the aquatic environment diminishes below some threshold, while the Werner-Gilliam model (1984). predicts that ontogenetic niche shifts (metamorphosis, in this case) should occur when the ratio of mortality (µ) to growth (g) of the occupied habitat (e.g., aquatic environment) outweighs that of another habitat (e.g., terrestrial environment). In a natural setting, high larval density combined with an unproductive environment (e.g., low food) should exhaust growth opportunities quickly due to competition and the associated stress of crowding. These

“unfavorable” environments should also have high perceived µ/g ratio throughout the experiment relative to other treatments due to the outcomes of competition and agonistic behavior, as well as cannibalism, which is common in larval newts (Harris 1987a). However, despite the HDAF treatment having perceived conditions that are theoretically “unfavorable,” larvae from HDAF metamorphosed later than those from the “favorable” HDSF treatment. These results mirror previous studies that found exceptions to the Wilbur-Collins model (reviewed in Morey and

Reznick 2000) and likely occurred because larvae in HDAF grew more slowly than larvae in

HDSF thereby taking longer to reach the minimum body size required for metamorphosis

(Wilbur and Collins 1973; Semlitsch 1987). Our results suggest that the optimal body size for metamorphosis is inversely related to conspecific density, as larvae from HDSF larvae grew rapidly (since growth was unrestricted due to supplemented food) (Fig. 3.2c), but traded additional SVL growth (Fig. 3b) for an earlier onset of metamorphosis (Fig. 3.2a).

Our primary goal was to investigate the effects of density cues on post-larval phenotype.

The general pattern across densities was as predicted — high densities resulted in more efts and fewer aquatic phenotypes (Fig. 3.1), while “favorable” conditions in the form of supplemented food had no effect on phenotype. All individuals from HDSF became efts, despite relatively high body condition and growth rates that would predict extended larval periods, suggesting that density cues are the primary indicator of habitat quality (Fig. 3.1). High density seemed to truncate SVL growth via an early onset of metamorphosis and thus canalize the ontogenetic response resulting in efts. Due to earlier metamorphosis, individuals from HDSF only attained body sizes typical of efts, suggesting that larval density can affect population dynamics, since efts take multiple years (2 years in North Carolina, USA (Harris 1987b) and 3–8 years in

Massachusetts, USA populations (Healy 1974)) to reach sexual maturity (Cole 1954; Stearns and

Koella 1986). Additionally, the potential for mortality during the lengthy eft stage would reduce fitness to zero, thereby increasing the risk associated with that phenotype. Nevertheless, the eft stage has its obvious benefits as an alternative strategy. Efts are chemically protected and mobile, which allows individuals to disperse and colonize other, potentially more suitable, ponds (Gill

1978). Since efts are terrestrial and adults and paedomorphs are aquatic, habitat (and therefore resource) partitioning may be occurring (Lejeune et al. 2018), where smaller, less competitive individuals (efts) are avoiding habitats with larger, competitively superior individuals (adults or paedomorphs) (Denoël et al. 2005).

Maintenance of polyphenisms, especially those involving more than two phenotypes, is a fascinating question in evolutionary biology (West-Eberhard 1989; Moran 1992; Whiteman

1994; Getty 1996). Imposing discrete phenotypic traits onto an environmental gradient (e.g., conspecific density, food level) is problematic because developmental thresholds have no clear environmental correlates. Under continuous environmental conditions, a generalist phenotypic strategy or continuous phenotypic plasticity should be favored. However, if phenotype- environment matching is accurate and the fitness advantages are large, then polyphenisms can be maintained (West-Eberhard 1989; Moran 1992). Polyphenisms have naturally selected thresholds creating reaction norms, but individuals must still accurately assess and respond to environmental conditions. Therefore, reliable cues are of utmost importance because they can be the difference between a fitness-increasing phenotype match or a fitness-reducing mismatch.

Since density-dependence is of considerable importance in a variety of systems, we expect strong selection on an organism’s ability to assess density. Encounter rates can be a useful indicator of conspecific density, but only if organisms can distinguish between repeated encounters with the same individuals and encounters with different individuals. Visual cues may provide this

information for a variety of organisms, but they are less reliable in aquatic environments (Dettner and Liepert 1994).

Adult newts are capable of utilizing pheromonal chemical cues to assess conspecific density, at least while breeding (Park and Propper 2001; Rohr et al. 2005). Our results suggest larvae may also be able to assess conspecific density via chemical cues, which ultimately translates as a proxy of long-term environmental quality. The nature of these cues is unknown, but possible origins are conspecific diet cues, secondary metabolites, prey alarm cues, exogenous hormones, cannibalism or conspecific death. Since background individuals occupied the same mesocosm, the possibility of visual or tactile cues affecting response individuals is not entirely eliminated. However, visual or tactile cues seem unlikely to transmit across the cage mesh, because (1) background larvae take refuge in leaf litter except at night, when visual cues are limited or absent, (2) caged larvae spent most of their time in the cage bottoms, which were opaque, (3) visual cues are less reliable in aquatic environments relative to chemical cues

(Dettner and Liepert 1994; Wisenden 2000), (4) larval salamanders have sluggish behavior that would not likely transmit tactile cues well, and (5) visual and tactile cues transmitted from predators, competitors, and prey are likely indistinguishable and thus uninformative.

Polyphenism reaction norms vary between species and populations (Semlitsch and

Gibbons 1985; Semlitsch et al. 1990; Takahashi and Parris 2008; Takahashi et al. 2011). Though a weaker density-dependent response was observed here compared to some other studies (Harris

1987b; Semlitsch 1987), I found results for this subspecies similar to those of Takahashi and

Parris (2008). The design of Takahashi and Parris (2008) allowed full interactions between individuals (competition, agonistic behavior, visual, tactile and chemical cues) and our design only permitted chemical cues, yet there were similar phenotype proportions for both studies. This

comparison suggests that physical interactions may not be necessary to elicit effects on polyphenisms in this system.

The importance of chemical cues in aquatic systems (Dettner and Liepert 1994;

Wisenden 2000) is widely recognized (Chivers and Smith 1998; Kats and Dill 1998; Brönmark and Hansson 2000; Wisenden 2000; Ferrari et al. 2010). They are known to elicit many alternative phenotypes, principally in the form of inducible defenses (Grunt and Bayly 1981;

Harvell 1990; McCollum and Van Buskirk 1996; Gilbert 1999; Kuhlmann et al. 1999; Laforsch et al. 2009; van Donk et al. 2011). Chemical cues have advantages over other cues because of their specificity, which can reveal not only the presence of competitors and predators, but also their identity through species-specific signatures, density through cue concentration, and preferred prey through dietary cues, all of which can be integrated into a threat level. In the case of newts, cannibalism is a real predation threat that simulatenouly scales and is confounded with intraspecific competition, thus chemical cues can provide information about both potential predation and competition. I minimized the potential for cannibalism in our background populations by introducing similar-sized individuals, however percieved threat of cannibalism may be inherent in cues indicating high density.

I provide evidence that larval salamanders utilize chemical density cues to modulate developmental trajectories and assess habitat quality. Conspecific density itself may be a more reliable, and more comprehensive, indicator of long term habitat quality than food resources, or at least be predictive of future food resources. Understanding the modality and relative importance of different cues will allow us to decipher complex life history decisions in polyphenic organisms, and understand the contributions of those decisions to population dynamics and evolutionary maintenance of alternative phenotypes.

CHAPTER 4:

AMBYSTOMA HAVE CONVERGENT PLASTIC RESPONSES AMONG ALTERNATIVE

HABITAT TYPES SEPARATED BY SMALL SPATIAL SCALES

ABSTRACT

Polyphenisms are an important component of species traits that facilitate macroevolutionary change. Since polyphenisms are defined as the expression of discrete, alternative phenotypes in a single population, they are ripe for reproductive isolation if alternative phenotypes are spatially or temporally isolated. Facultatively paedomorphic salamanders are polyphenic in that their aquatic larvae can either metamorphose into terrestrial phenotypes or they can reach adulthood in an aquatic juvenile state. Past studies have shown this polyphenism’s reaction norms to differ between relatively distant populations that are historically associated with ponds of different hydroperiods, where more permanent ponds favor paedomorphosis and vice versa. Utilizing a common garden mesocosm design in which I manipulated water levels, I investigated this relationship by comparing developing salamanders sourced from ponds separated by small geographic distances (1200 m) that also vary in hydroperiod. I found that water level treatments (constant vs. drying), and not pond source, were the main determinant of phenotype with more metamorphosis occurring in drying treatments.

Size and growth variables did not show any notable differences among treatments with the exception of body condition, which showed a water level × pond source interaction. Individuals from temporary ponds performed better (heavier per unit length) in drying treatments than in constant water level treatments and, likewise, individuals from permanent ponds performed better in constant water level treatments. Our results suggest that gene flow over our small spatial scale prevented divergence of most traits, especially phenotype thresholds.

INTRODUCTION

Speciation is the fundamental process that creates biodiversity. One potential pathway to the formation of new species is via polyphenisms, where a single population of organisms can express two or more environmentally-cued, discrete phenotypes (Pfennig et al. 2010).

Polyphenisms enable organisms to tolerate temporal and spatial variability by utilizing plastic traits to excel under different circumstances. While many plastic traits, such as size, are often continuously distributed, they can also be discrete, resulting in a polyphenism, where two or more alternative phenotypes exist in a population. Polyphenisms can be important for evolutionary change (West-Eberhard 2005, Pfennig et al. 2010). For example, if each alternative phenotype was more likely to mate with the same phenotype rather than the alternative, either due to spatiotemporal, morphological, or behavioral reproductive isolation, then polyphenisms can increase genetic divergence and facilitate sympatric speciation (West-Eberhard 2005,

Whiteman and Semlitsch 2005, Pfennig et al. 2010).

Local adaptation is often contrasted with phenotypic plasticity (Sultan and Spencer

2002), but the two are not mutually exclusive (Torres-Dowdall et al. 2012). For example, it is widely documented that populations can show differences in their polyphenic reaction norms

(Semlitsch and Gibbons 1985, Semlitsch et al. 1990, Torres-Dowdall et al. 2012), which are the relationships between environmental variation and phenotype (Woltereck 1913). Different selection pressures acting on these populations can shape reaction norms by selecting against one or more alternative phenotypes in a given population and thereby increasing the fitness of another phenotype (Semlitsch and Gibbons 1985, Semlitsch et al. 1990). Alternatively, the

phenotypes may exist as a locally adjusted evolutionarily stable strategy where the specific localities can support specific ratios of the different phenotypes (Pfennig 1992). If selection were to be extreme for one phenotype over another, then the phenotype can become canalized and alternative phenotypes can be lost in the population (Levis and Pfennig 2018). Phenotypic plasticity therefore represents important intraspecific biodiversity that allows organisms to adapt to changing landscapes (Miner et al. 2005, Pfennig et al. 2010).

Pond permanence is a defining characteristic of aquatic habitats that determines species composition, community structure, and functional traits (Wellborn et al. 1996). Temporary and permanent aquatic habitats are especially distinctive because of regular drying in temporary habitats preventing long-term presence of top level predators (fish), but they also widely differ in abiotic characteristics, with permanent ponds providing more stable conditions (Wellborn et al.

1996). While many species exclusively utilize either temporary or permanent aquatic habitats, others may be found in both habitat types. Many larval amphibians have evolved plastic traits to improve performance in ponds of variable hydroperiod (Wilbur and Collins 1973, Newman

1992, Székely et al. 2017). Those that colonize temporary ponds typically have fast and plastic growth rates to accommodate the ephemerality of the habitat, while others that colonize permanent ponds typically show less plasticity and greater investment in predator defenses

(Wellborn et al. 1996). Ambystoma talpoideum is a facultatively paedomorphic mole salamander that breeds in both temporary and permanent ponds. Facultative paedmorphosis is a polyphenism where sexually mature individuals can either remain in an aquatic, juvenile phenotype

(paedomorphosis) or metamorphose into a terrestrial phenotype (Whiteman 1994). Since paedomorphs are fully aquatic, the expression of this phenotype is restricted to permanent aquatic habitats, whereas metamorphosed phenotypes can occur in either habitat. This

polyphenism is proximally regulated by environmental factors such as intraspecific density and pond drying (Semlitsch 1987, Doyle and Whiteman 2008, Whiteman et al. 2012). However, the likelihood of either phenotype has shown to be divergent between populations that experience different hydroperiod regimes. For example, Ambystoma talpoideum sourced from temporary ponds have a higher likelihood of metamorphosis, while salamanders sourced from permanent ponds have shown a higher likelihood of paedomorphosis (Semlitsch and Gibbons 1985).

While it is expected that two isolated populations exposed to two different hydroperiod regimes would have different reaction norms, it also expected that gene flow between these populations could homogenize reaction norms, which occurs with facultatively paedomorphic newts (Oromi et al. 2016). Thus, if gene flow disrupts localized adaptation, then any difference in reaction norms over such short distances may be due to habitat selection, where adults with a higher likelihood of either phenotype (potentially due to their own life history trajectory) preferentially oviposit in habitats that optimize their offspring’s performance. While this process may be self-canalizing due to limitations of paedomorphs to disperse (since they are restricted to their permanent, natal aquatic habitats), many individuals in Ambystoma talpoideum do eventually metamorphose and, thus, can disperse to different oviposition sites in later breeding seasons. However, in order to investigate these questions further, I must first know whether salamanders colonizing temporary and permanent ponds separated by small distances show any differences in reaction norms.

Utilizing the unique habitats of UMFS, which contains over 200 ponds of various degrees of permanence, I sought to compare the performance of larval salamanders from both temporary and permanent habitats in a common garden experiment. While past studies have shown population divergence across distances as short as 13.1 km (Semlitsch and Gibbons 1985), I was

interested in comparing individuals across small geographic scales (i.e., individuals sourced from ponds separated by only 1200m of forest habitat). Our goals were 1) to compare the reaction norms of individuals sourced from divergent habitats separated by small spatial scales in a common garden experiment and 2) to test whether natal habitat relates to offspring performance by measuring size and growth variables.

METHODS

Experimental Design

To test the effects of oviposition habitat on performance in habitats with different hydroperiods, I performed a 2 × 2 factorial, common garden experiment where larval individuals were exposed to both a simulation of their original oviposition habitat hydroperiod and the alternative. I constructed an experimental array of twenty-four 1.8 m diameter, 1200L cattle tanks (mesocosms; N = 24) in a mowed field at UMFS. Mesocosms were filled with well water and randomly received assigned aliquots of both zooplankton inocula (1.5 L) from a fishless pond (UMFS Pond #5) and dry, hardwood leaf litter (2 kg). Mesocosms were fitted with window screen lids to prevent colonization of unwanted organisms, but they were also sunk into the water on 11 Feb to add insect colonists to the food web structure of the mesocosms. Mesocosms were assigned one level of each of the following factors: larval source pond (temporary or permanent) and water level (drying or constant). Each treatment was replicated six times (n = 6) and each treatment was represented once in each of the five columns (= blocks) (4 treatments × 6 blocks =

24 mesocosms).

On 8–9 April, I haphazardly collected larvae via dipnet from UMFS Ponds #7 and #137.

UMFS Pond #7 is a representative temporary pond that dries yearly in July–August, while

UMFS Pond #137 is a representative permanent pond, but may potentially dry in years of severe drought given its absence of predatory fish. UMFS Pond #7 and #137 are separated by a distance of ~1200m that crosses two streams and upland, mixed forests. Each mesocosm received twelve randomized larvae. Larvae from UMFS Pond #7 were assigned to the twelve mesocosms

designated as “temporary larval source pond,” while larvae from UMFS Pond #137 were assigned to the twelve mesocosms designated as “permanent larval source pond.” Since temperatures were warm, individuals were very sensitive to handling stress and mortality was high during the first transfer attempt, especially when measurements were taken. In order to reduce mortality during transfer, individuals were sized visually and only those individuals that minimized size variation within and across mesocosms were introduced.

Every two weeks, starting on 8 May, I measured physicochemical parameters (DO, specific conductance (µS/cm), TDS, pH, and temperature) of each mesocosm between 10–11 a.m. On 19 May, the standpipes of all drying treatments mesocosms were lowered by 7 cm.

Standpipes were lowered 7cm every two weeks until water depth was 14cm, which was maintained until the end of the experiment. Constant water level mesocosms were maintained at

49cm depth for the duration of the experiment. On 6 June, temperature loggers (Onset HOBO pendant UA-001-08) were introduced into eight select tanks distributed across the mesocosm array (two in each of the four treatment combinations; four in constant water level, four in short drying treatments) to measure temperature fluctuations over time in 120 minute intervals.

I began night checks for emerging metamorphs on 12 May, and they occurred every three days until late October when metamorphs stopped emerging. Metamorphs were collected by hand, weighed, photographed and then released into the terrestrial environment at UMFS. During

11–12 Dec, the experiment was terminated and all remaining individuals (larvae and paedomorphs) were collected, weighed and photographed. Paedomorphs were distinguished from larvae by enlarged testes on males and a swollen cloaca with gravid body shape on females.

Statistical analyses

Using the lme4 package v1.1.21 (Bates et al. 2015) in R v3.5.1 (R Core Team 2018), I used linear mixed models (LMMs) to analyze size and growth response variables: snout-vent length (SVL), mass, larval period, growth rate (SVL/day), and body condition. Analyses of SVL, mass, larval period, and growth rate used identical statistical models that included pond source

(permanent or temporary), treatment (drying or constant), and their interaction as fixed factors, as well as survival rate as a covariate (to control for changes in density). Body condition (size- independent mass) was assessed by performing a second analysis of mass with SVL as an added covariate (Garcia-Berthou 2001). Body condition was visualized by mean-scaling masses to decouple variance from the measurement scale and means, regressing against SVL, and modeling the residuals (Berner 2011). I modeled all individuals as data points (as opposed to averaging per mesocosm), so I included mesocosm nested within block as a random effect in our models to account for this non-independence among individuals sharing a mesocosm.

Physicochemical parameters (DO, specific conductance (µS/cm), TDS, pH) were modeled in a LMM using pond source, water level treatment, their interaction, and sampling date

(time component) as fixed factors along with mesocosm nested within block as a random effect in our models to account for non-independence among repeated measures. Temperature was analyzed in a LMM with only water level treatment and sampling time (time component) as predictors since only four of each water level treatments were logged. Temperature analysis also included mesocosm nested within block as a random effect in our models to account for non- independence among repeated measures. Mean temperature variation between constant and drying treatments was tested in a general linear model using water level treatment as the only fixed factor. The random term block was removed from temperature variance analysis because there were not enough data points for mixed effects models. Significance for linear mixed effects

models was tested with ANOVA approximate F Tests (Type III Satterthwaite) from the lmerTest package v3.0.1 (Kuznetsova et al. 2015), while the general linear model was analyzed with an

ANOVA.

Phenotype proportions were analyzed using three separate logistic GLMMs to compare metamorphs vs. paedomorphs, metamorphs vs. larvae, and paedormorphs vs. larvae. Three separate regressions were conducted instead of multinomial regression since multinomial regression requires repeated analyses with multiple reference levels to perform a complete set of comparisons. Additionally, GLMMs readily accommodate the nested data structure (mesocosm nested within block) of our design. Logistic GLMMs followed the structure of the previously mentioned LMMs by including pond source, water level treatment, their interaction, and survival rate as fixed factors, along with mesocosm nested within block as a random effect. Five mesocosms with exceptionally low survival (< 40%) were excluded from analyses. The significance of the logistic GLMMs was tested with Wald tests (Bolker et al. 2009, Warton and

Hui 2011, Bates et al. 2015). All analyses set α = 0.05. Plots were made using ggplot2 v3.0.0

(Wickham 2009) and sciplot v1.1.1 (Morales and R Core Team 2012).

RESULTS

Survival rate across all initial larval densities of analyzed mesocosms averaged 73.7% and there was no difference in average survival among pond sources (z = –1.47, p = 0.143), water level treatment (z = 1.18, p = 0.239), or pond source × water level treatment (z = –0.23, p

= 0.816). For all individuals, 32.7% remained as larvae, 52.4% metamorphosed, and 14.9% became paedomorphic. The frequency of metamorphosis increased relative to paedomorphosis in drying treatments (z = –2.48, p = 0.013), but there was no difference between pond sources (z =

–1.03, p = 0.303) or the pond source × water level treatment interaction (z = –1.22, p = 0.221)

(Fig. 4.1). As survival rate increased, the frequency of paedomorphosis relative to metamorphosis decreased (z = –2.38, p = 0.017). Pond source (z = 0.48, p = 0.631), water level treatment (z = 0.26, p = 0.792), their interaction (z = 0.83, p = 0.409), and survival rate (z = –

1.72, p = 0.086) had no effects on the relative frequency of paedomorphosis vs larvae (Fig. 4.1).

When comparing the frequency of metamorphosis to larvae, drying treatments increase the frequency of metamorphosis (z = 4.55, p < 0.001), but pond source (marginally non-significant; z

= 1.84, p = 0.065), their interaction (z = 0.08, p = 0.937), and survival (z = 0.13, p = 0.898) had no effect (Fig. 4.1).

Figure 4.7 Phenotype proportions across all treatment combinations (left), along with the same results aggregated by pond sources (center) and water level treatments (right) to display the overriding effect of water level treatment on phenotype expression.

Mass negatively scaled with survival rate (F1, 14.01 = 4.86, p = 0.045), and there was a marginal effect of water level treatment (F1, 14.07 = 4.26, p = 0.058), and no effect of pond source

(F1, 13.49 = 0.08, p = 0.778) or the pond source × water level treatment interaction (F1, 13.23 = 0.02, p = 0.869) (Fig. 4.2). As expected, SVL followed the same pattern, with a significant negative effect of survival (F1, 15.50 = 12.33, p = 0.003), but no effect of pond source (F1, 14.51 = 1.01, p =

0.332), water level treatment (F1, 13.95 = 0.366, p < 0.555), or their interaction (F1, 13.99 = 0.04, p =

0.842). There was a pond source × water level treatment effect on body condition (F1, 12.01 =

6.63, p = 0.024) (Fig. 4.3), where individuals from temporary ponds did better in drying treatments and vice versa, and there was no effect of survival (F1, 20.86 = 0.063, p = 0.804).

Growth rate was higher in drying treatments (F1, 159.73 = 11.42, p < 0.001) and in individuals sourced from temporary ponds (F1, 162.84 = 4.38, p = 0.038), with no effect of survival (F1, 144.01 =

0.05, p = 0.819) or the pond source × water level treatment interaction (F1, 138.84 = 0.05, p =

0.828). Larval period of metamorphs did not differ among pond sources (F1, 5.98 = 0.23, p =

0.647), water level treatment (F1, 6.10 = 0.34, p = 0.581), their interaction (F1, 7.39 = 0.00, p =

0.975), nor did it trend with survival (F1, 5.74 = 1.19, p = 0.318).

Most physicochemical parameters (DO, TDS, specific conductance) showed no differences across treatments (all p > 0.05). However, there were a few exceptions. The mesocosm pH levels differed across pond sources (F1, 379 = 7.98, p = 0.005), but there was no effect of water level treatment (F1, 379 = 0.07, p = 0.794), or their interaction (F1, 379 = 0.29, p =

0.588). Analyzing the temperature time series (Fig. 4.4) revealed that constant water level treatments had warmer temperatures than drying treatments (F1, 4.40 = 29.45, p = 0.004).

Temperature variance (s2) also differed between water level treatments, with more variation in drying treatments than in constant water level treatments (F1, 6 = 15.49, p = 0.007) (Fig. 4.5).

Figure 4.8 Mean final mass (g) ± SE of all individuals from temporary source ponds (open circles) and permanent source ponds (closed circles) across drying and constant water level treatments.

Figure 4.9 Mean final body condition index of all individuals from temporary source ponds (open circles) and permanent source ponds (closed circles) across the two water level treatments.

Figure 4.10 Temperature over time across constant (blur circles, upper trend line) and drying (multicolor triangles, lower trendline) water level treatments. Each point represents a temperature measurement taken at 120 min intervals from a temperature logger. Constant water level treatments were consistently warmer than drying treatments. As the drying treatment had its water level reduced (indicated here by changing colors), the daily temperature variance increased.

Figure 4.11 Temperature variance (s2) was greater in drying treatments when compared to constant water level treatments.

DISCUSSION

Our results show that there is no difference in the reaction norm of facultative paedomorphosis from the two different pond sources, suggesting that there is no local adaptation in reaction norms of facultative paedomorphosis across small geographic scales. Water level treatments were the main determinant of phenotype, with drying treatments producing 79.2% metamorphs compared to only 30.0% in constant water level treatments (Fig. 4.1). These results reaffirm the primacy of proximate environmental cues during the drying process that facilitate or induce metamorphosis (Denver et al. 1998). While paedomorphosis was more common in constant water level treatments, it was the only phenotype that was had a negative relationship with survival rate, providing more evidence for the density-dependent patterns of facultative paedomorphosis (Harris 1987, Semlitsch 1987). Likewise, more individuals remained as larvae in constant water level treatments and, had the experiment not concluded, they would have overwintered as such until the next season. Since metamorphosis was uncommon in constant water level treatments, the average density was relatively high throughout the experiment.

Through density-dependent processes, this effect kept body sizes small and physiologically constrained ontogenetic pathways to increase the number of individuals overwintering as larvae.

Body size and growth variables were generally undifferentiated among treatments.

Higher growth rates in drying treatments were mostly an artifact of the large number of metamorphs, which have a shorter larval period (smaller growth rate denominator) that biases growth rate to the fast growing early stages of larval development. However, body condition showed a clear interaction between pond source and water level treatments — individuals from

temporary ponds performed better in drying treatments and individuals from permanent ponds performed better in constant water level treatments. While this result suggests some form of local adaptation, I cannot rule out the possibility of environmental effects affecting growth and developmental trajectories early in ontogeny before individuals were introduced into our experiment. However, tail clips were taken from all individuals to potentially investigate any genetic relationship to body condition in the future. Salamanders may be able to quickly adapt to local conditions or they are making informed oviposition choices to increase offspring success.

While we expect drying ponds will endure fluctuations in many biotic (e.g., intra- and interspecific density, prey density) and abiotic (temperature, dissolved oxygen, TDS) variables, it is difficult to dissociate these factors. Our results show that drying and constant water level treatments did not differ in most physicochemical parameters. Temperature, however, did show substantial differences between water level treatments. Constant water level treatments had average higher temperatures throughout the experiment. Amphibian growth rates are known to be temperature-dependent, with higher growth rates occurring in warmer temperatures, but this effect was absent in our study as density took primacy. Drying treatments had considerable variability in temperature, with higher maximums and lower minimums than constant water level treatments. This variability increased over time as the water level was decreased in drying treatments (Fig. 4.4). While these daily fluctuations in temperature may generate a stress response and increase the likelihood of metamorphosis, past laboratory work with anurans has established that adaptive plasticity in response to drying is independent of temperature or water quality effects (Denver et al. 1998). Thus, temperature variability combined with the physical constraints of drying (i.e., density-dependence) may compound.

I investigated patterns of phenotypic plasticity across microgeographic spatial scales to see if either local adaptation is shaping reaction norms of facultatively paedomorphic salamanders. Most metrics showed no differences between pond sources except for body condition index. Whether or not this specific difference is due to genetic divergence is unknown, but I failed to detect any differences in those traits that have been shown to differ over large geographic scales (Semlitsch and Gibbons 1985, Harris et al. 1990). Populations of Ambystoma maculatum in New York were shown to integrated by gene flow over distances of 4.8 km

(Zamudio and Wieczorek 2007), while another study from Ohio showed no isolation by distance with populations separated by distances up to 55 km (Purrenhage et al. 2009). A study examining swimming performance in Ambystoma barbouri across gradients of predator exposure revealed differences in populations separated by 14 km, but not by 1 km (Storfer et al. 2006). Distances used by Storfer et al. (2006) are directly analogous to the results of our study (1.2 km) when compared with that of Semlitsch and Gibbons (1985), which found population divergence in paedomorphosis expression across distances of 13.2 km. While I cannot overlook the differences in body condition, I found no other differences between pond sources, suggesting that salamanders studied at our geographic scale likely have significant levels of gene flow or there was no selection on reaction norms (perhaps only in their ability to affect their body condition).

Despite our results, there has been growing evidence that population divergence can happen over small geographic scales (Skelly 2004, Storfer and Sih 2006). A study by Urban

(2010) showed population divergence of Ambystoma maculatum in response to predation risk across distances of only ~1.2 km. The degree to which populations can diverge over such small geographic scales will depend on the balance between gene flow and selection pressures driving local adaptation (Wright 1931, Slatkin 1985). Ambystoma are rather mobile salamanders given

their seasonal migrations to and from breeding ponds. Adult males have been found to migrate as far as ~280 m from breeding sites (Semlitsch 2009). Thus, the distances between ponds in our study are not unreasonable to facilitate high gene flow. Polyphenisms are an important type of phenotypic plasticity that can facilitate macroevolutionary change. While I did not see much evidence over short distances in our system, past studies have noted trait divergence over larger geographic scales, supporting the notion that polyphenic species may be intermediary steps in the process of speciation.

CHAPTER 5:

FUNCTIONAL DIVERSITY OF PREDATORS DRIVE VARIABLE TRAIT- AND DENSITY-

MEDIATED EFFECTS ON PREY LIFE HISTORY

ABSTRACT

The consumptive effects of predators have been central to ecological theory, but non- consumptive predator effects have emerged as major components of predator-prey interactions.

Consumptive effects typically translate into density-mediated effects on prey, where a change in the number of individuals affects per-capita resource availability. Non-consumptive effects translate into trait-mediated effects, where an organism’s behavioral or morphological traits change in response to predator presence. Consumptive and non-consumptive predator effects are expected to vary with predator functional traits (hunting mode, gape size). In a field mesocosm experiment, I exposed developing larval salamanders (Ambystoma talpoideum) to the consumptive (free-roaming) and non-consumptive (caged) effects of three predators with different hunting modes and gape sizes (Lepomis cyanellus, Aphredoderus sayanus, Fundulus chrysotus). I measured life history variables (body size, growth rates) along with plastic phenotypic traits known to respond to predator presence: tail color and the expression of paedomorphic life history strategies. Lepomis cyanellus and Aphredoderus sayanus had equivalent consumptive effects on prey, but induced opposite effects on body size. Consumptive

(but not non-consumptive) L. cyanellus evoked trait-mediated effects that suppressed growth, induced achromatic tails, and overrode density-mediated effects. Surviving larvae exposed to consumptive A. sayanus reached the largest body sizes, showing only density-mediated, and not trait-mediated, effects. Fundulus chrystous imposed intermediate effects on prey survival and body size. Predator effects varied with functional traits with larger gaped predators having stronger consumptive effects, but active predators induced stronger trait-mediated effects

compared to sit-and-pursue predators. Non-consumptive predator cues had minimal effects on prey suggesting prey may have habituated to predator cues and emphasizing the importance of immediate threat over time. Our results emphasize the role that species identity and functional traits play in determining density- and trait-mediated effects on prey. Instead of investigating the most intense predators, researchers should investigate a diversity of predators with varying functional traits to determine natural impacts on prey populations.

INTRODUCTION

Predator-prey dynamics hold a central place in classical ecological theory, which has historically emphasized the consumptive effects of predators on prey (Lotka 1925, Volterra

1926, Elton and Nicholson 1942, Brooks and Dodson 1965, Gotelli 2008). Consumptive effects reduce prey population densities resulting in density-mediated effects, which can affect intraspecific competition dynamics (Wilbur et al. 1983) and demography of prey (Brooks and

Dodson 1965, Rood and Reznick 1997). Modern predator-prey theory has recognized that predators can affect prey through non-consumptive effects (Peckarsky et al. 2008, Wissinger et al. 2010), where the threat of predation induces trait-mediated responses such as changes in prey behavior or morphology (Abrams 1984, Lima and Dill 1990, Abrams 1995, Lima 1998).

Predicting the strength and outcomes of both consumptive and non-consumptive effects in predator-prey interactions is vital to generalizing community ecology. While much attention has been given to variation in prey responses to predators, predators are too often treated as static sources of risk in predator-prey interactions, which can limit our understanding of how predators can vary in their functional consumptive and non-consumptive effects on prey (Lima 2002,

Chalcraft and Resetarits 2003).

Predator-prey dynamics often rely upon the species identity and corresponding traits of the interacting predators and prey, thereby producing functionally diverse predator-prey combinations that differ in interaction strength (Chalcraft and Resetarits 2003, Schmitz 2012).

Predator functional diversity can manifest through variation in a variety of functional traits, such as body size (Rudolf 2012), hunting mode (Preisser et al. 2007, Schmitz 2012), and metabolic

rates (Brown et al. 2004, Chalcraft and Resetarits 2004). Predator hunting mode, which can be classified as either active, sit-and-pursue, or sit-and-wait, has been shown to strongly predict prey responses to non-consumptive predator presence (Preisser et al. 2007). Sit-and-pursue predators tend to elicit the strongest non-consumptive effects, which is thought to occur because cues from sit-and-pursue predators are more reliable and localized indicators of immediate predation risk than cues from active predators, which may saturate an environment (Preisser et al. 2007). However, analyses of the effects of different hunting modes on prey with complex life histories showed that active predators reduced size at metamorphosis, which is critical to lifetime expected fitness (Semlitsch et al. 1988), compared to sit-and-pursue predators (Davenport et al.

2014). Predators with ambush hunting modes (sit-and-wait/pursue) should face selection pressure to be cryptic to avoid detection by prey and improve predation success. In this scenario, predators with ambush hunting modes will elicit density-mediated effects on prey through consumptive effects, but would have reduced or absent non-consumptive effects thereby limiting trait-mediated effects. Active predators will induce both density and trait-mediated effects through consumptive and non-consumptive effects, respectively. Thus, consideration of predator hunting mode is important for predicting the net consumptive and non-consumptive effect in predator-prey interactions (Preisser et al. 2007, Davenport et al. 2014).

Consumptive effects are mainly manifested through reductions in prey density, which can decrease intraspecific competition of prey and, thereby, increase prey growth rates (Peacor and

Werner 2001), which can translate into earlier age at first reproduction or higher fecundity

(Semlitsch et al. 1988). Density-mediated effects can therefore have large effects on fitness trajectories for individual organisms. Likewise, trait-mediated predator effects can alter growth rates and fitness trajectories, but not via changes in competition, but by trading foraging, energy

investment, and growth for behavioral or morphological predator defense and survival (Werner and Gilliam 1984, Werner and Hall 1988, Relyea and Werner 1999, Benard 2004, McPeek

2004). Because consumptive and non-consumptive effects of predators co-occur (Peacor and

Werner 2001, Peacor et al. 2013), density-mediated and trait-mediated responses of prey also co- occur and can interact in nature. These interactions can produce unique contexts in natural systems that can explain patterns seen in complex species interactions (Werner and Peacor

2003).

I conducted a predation experiment using a replicated mesocosm design with both free- roaming (consumptive) and caged (non-consumptive) predator treatments crossed with three species with different hunting modes and ecology: green sunfish (Lepomis cyanellus; active, generalist hunter), golden topminnows (Fundulus chrysotus; active, surface-feeding hunter) and pirate perch (Aphredoderus sayanus; sit-and-pursue hunter). For prey, I used a model amphibian species (Ambystoma talpoideum), which shows density- (Semlitsch 1987a) and trait-mediated

(Semlitsch 1987b, Jackson and Semlitsch 1993) responses to predators. Our goal was twofold: 1) to determine how variation in predator functional traits drives both trait- and density-mediated effects in prey and 2) how the trait- and density-mediated effects may interact to uniquely affect prey.

METHODS

Study species

The model predators consisted of three different species with varied predatory capabilities and trophic niches: the voracious, opportunistic green sunfish (Lepomis cyanellus) the smaller gaped, surface-feeding golden topminnow (Fundulus chrysotus), and the cryptic, ambushing pirate perch (Aphredoderus sayanus). Green sunfish are one of the most widespread fishes in North America (Lee et al. 1980a). They are very effective, generalist predators that actively forage for prey. They have strong consumptive (Sexton and Phillips 1986) and non- consumptive (Petranka et al. 1987) effects on amphibians. Golden topminnows are small (71 mm maximum standard length) (Foster 1967), gape-limited fish that occur in ponds and slow moving streams (Shute 1980) across the southeastern United States (Hubbs et al. 2008). They are active, surface-feeding fish whose diet consists primarily of small invertebrates (Hunt 1953). Pirate perch are a unique, but common, freshwater fish that is the only member of the monotypic family

Aphredoderidae and only one of nine extant described species in the order Percopsiformes

(Dillman et al. 2011). Adults are intermediately sized (144 mm maximum total length) (Moore and Burris 1956) and moderately-gaped (Poly 2004). They are nocturnal, ambush predators that occupy streams, swamps, lakes and rivers of the eastern and central United States (Gunning and

Lewis 1955, Parker and Simco 1975, Shepherd and Huish 1978, Lee et al. 1980b, Monzyk et al.

1997). They feed on a wide variety of prey, including larval anurans, small fishes, insects and crustaceans (Forbes 1888, Forbes and Richardson 1908, Flemer and Woolcott 1966, Shepherd

and Huish 1978, Benke et al. 1985, Albecker and Vance-Chalcraft 2015). All three of these species have extensive range overlap with A. talpoideum and co-occur naturally at UMFS.

Experimental Design

To test the effects of predator diversity on plastic life history responses of prey, I conducted an experiment using three predatory fishes — green sunfish, golden topminnows, and pirate perch — and examined their effect on growth and development of larval A. talpoideum. In

Dec 2016, I constructed an experimental array of forty-two 1.8 m diameter, 1200 L cattle tanks

(mesocosms; N = 42) in a mowed field at UMFS. Each mesocosm was filled with well water and randomly received assigned aliquots of both dry, hardwood leaf litter (1.5 kg) and zooplankton inocula (1.5 L) from a fishless pond (#61) at UMFS. One cylindrical mesh cage (height = 0.61m, diameter = 0.58m, volume = 0.16 m3, mesh = 1.3 × 1.13 mm openings) was added to each mesocosm. Mesocosms were fitted with a window screen lid that closed it to colonization by other organisms and prevented the escape of metamorphs. Mesocosms were assigned one of the following treatments: (a) free-ranging, consumptive pirate perch, (b) free-ranging, consumptive golden topminnow, (c) free-ranging, consumptive green sunfish, (d) caged, non-consumptive consumptive pirate perch, (e) caged, non-consumptive golden topminnow, (f) caged, non- consumptive green sunfish, and (g) fishless controls. Each treatment was replicated six times

(n=6) and each treatment was represented once in each of the seven rows (= blocks) (6 treatments × 7 blocks = 42 mesocosms). Fish were collected from local ponds and streams at

UMFS (UMFS Pond #144, Bay Springs Branch, Bramlett Pond), weighed, and assigned to mesocosms on 12 Jan so that all fish within a block were of similar relative size and mass (2.65 g

± 0.27). All predators were initially placed in cages, along with a handful of leaf litter for refuge,

until the addition of salamander hatchlings. Due to the long duration of the experiment (~1 year), not all fish remained in cages or survived until the end of the experiment. Ten mesocosms with missing fish were removed from analyses.

Egg masses of A. talpoideum were collected from ponds at UMFS during December and were reared in small outdoor wading pools until hatching. Eggs were separated by date of oviposition. I could not collect a sufficient number of eggs from one night of oviposition for all mesocosms, thus addition of hatchlings to mesocosms occurred in two phases. On 25 Jan, each mesocosm in blocks 1–4 received 12 randomly selected A. talpoideum hatchlings from the earliest cohort of egg masses and the same was done on 31 Jan for blocks 5 and 6 with a later cohort of egg masses. Hatchling larvae were allowed to acclimate for 24 hours before predators in consumptive treatments were released from their cages into the surrounding mesocosms.

Predators in non-consumptive treatments were retained in their cages. Screen lids were permanently depressed into the water in early February to allow for spring insect colonists to increase the food supply of invertebrates for larval salamanders.

Night checks for emerging metamorphs began on 12 May and occurred every three days until late October when metamorphs stopped emerging. Metamorphs were collected by hand, weighed, photographed and then released into the terrestrial environment at UMFS. During 12–

14 Dec, the experiment was terminated and all remaining individuals (larvae, paedomorphs and fish) were collected, weighed and photographed. Paedomorphs were distinguished from larvae by enlarged testes on males and a swollen cloaca with gravid body shape on females.

Tail color analyses

All individuals collected at the end of the experiment were analyzed for tail color.

Individuals were photographed in standardized positions on a white and black balance card for analysis in ImageJ (Schneider et al. 2012). Color analysis procedure followed the methods of

Touchon and Warkentin (2008). For each individual, their image color was corrected using the balance card, their tails were outlined, and values were determined for levels of red, green, and blue (RGB) and Hue, Saturation, and Brightness (HSB).

Statistical analyses

Using the lme4 package v1.1.17 (Bates et al. 2015) in R v3.5.1 (R Core Team 2018), I used linear mixed effects models (LMMs) to analyze larval response variables: snout-vent length

(SVL), larval period, growth rate, and body condition. Snout-vent length (SVL), larval period, and growth rate were analyzed as LMMs with “predator species,” “predator exposure”

(consumptive or non-consumptive), and their interaction as fixed effects, along with survival rate as a fixed covariate (to control for changes in density). Body condition (size-independent mass) was assessed in a LMM by performing a second analysis of mass with the addition of SVL as a fixed covariate (Garcia-Berthou 2001). Body condition was visualized by mean-scaling masses to decouple variance from the measurement scale and means, regressing against SVL, and modeling the residuals (Berner 2011). I modeled all individuals as data points (as opposed to averaging per mesocosm), so I included mesocosm nested within block as a random effect in our models to account for this non-independence among individuals sharing a mesocosm. All mesocosms with dead or missing predators (10) were removed from analyses. Significance for

LMMs was tested with Approximate F Tests (Type III Satterthwaite) from the lmerTest package

v3.0.1 (Kuznetsova et al. 2015).

Survival and phenotype proportions were analyzed with logistic GLMMs (Warton and

Hui 2011) made with the lme4 package (Bates et al. 2015). Survival analysis data were aggregated by mesocosm and, thus, excluded the nested random term mesocosm, but still included block as a random effect. Phenotype proportions were analyzed using three separate logistic GLMMs to compare metamorphs vs. paedomorphs, metamorphs vs. larvae, and paedormorphs vs. larvae. Three separate regressions were conducted instead of multinomial regression since multinomial regression requires repeated analyses with multiple reference levels to perform a complete set of comparisons. Additionally, GLMMs readily accommodate the nested data structure (mesocosm nested within block) of our design. Logistic GLMMs followed the structure of the previously mentioned LMMs by including predator species, predator exposure, their interaction, and mesocosm survival rate as fixed factors, along with mesocosm nested within block as a random effect. When the interaction was non-significant, that term was dropped from the GLMM to investigate effects of predator species and predator exposure individually. The consumptive Lepomis cyanellus treatment produced singular logistic models, so it was dropped from phenotype analyses. The significance of the logistic models was tested with likelihood ratio tests (Bolker et al. 2009, Warton and Hui 2011, Bates et al. 2015). All analyses set α = 0.05. Plots were made using ggplot2 v3.0.0 (Wickham 2009) and sciplot v1.1.1

(Morales and R Core Team 2012).

Tail color metrics (R, G, B, Hue, Saturation, and Brightness) were averaged per mescosm and analyzed using the Bray-Curtis Permutational multivariate analysis of variance

(PERMANOVA) method in the adonis function available in the vegan package v2.5.2 (Oksanen et al. 2015). As before, predator species, predator exposure, and their interaction were fixed

factors. These metrics were visualized with a principle component analysis, which was plotted using ggbiplot v0.55 (Vu 2011).

RESULTS

Across all treatments, 53.5% of the salamanders survived to the end of the experiment;

there was a significant treatment effect of predator species ( = 23.81, p < 0.001) and predator

exposure on salamander survival ( = 121.34, p < 0.001) (Fig. 5.1). Salamander survival in consumptive pirate perch (9.72% ± 0.06; mean ± SE) and consumptive green sunfish (12.5% ±

0.06) treatments was reduced compared to consumptive golden topminnows and all non- consumptive treatments. Consumptive golden topminnows (40.0% ± 0.10) had an intermediate effect, which showed reduced salamander survival compared to nearly all non-consumptive treatments (Fig. 5.1). Differences in SVL of salamanders among treatments were a result of density-mediated effects created by differential survival (F1, 23.47 = 22.22, p < 0.001) (Fig. 5.2) — the growth of surviving larvae increased through density-mediated effects. However, there was also a significant predator species × predator exposure interaction for SVL (F2, 26.25 = 17.20, p <

0.001) as these growth-enhancing density-mediated effects were absent with consumptive green sunfish, but very strong with consumptive pirate perch (Fig. 5.2). Body condition and growth rate followed the same patterns as SVL, with a significant predator species × predator exposure interaction (F2, 30.17 = 7.77, p = 0.002, F2, 29.01 = 3.92, p = 0.031, respectively) and negative effect of survival (F1, 30.68 = 8.82, p < 0.001; F1, 25.52 = 23.43, p < 0.001, respectively).

Figure 5.1. Proportion of individuals that survived until the end of the experiment in consumptive (closed circles) and non-consumptive treatments (open circles). Different letters indicate significant differences (Tukey HSD; p < 0.05).

Figure 5.2. Snout-vent length (mm) of individuals across consumptive (closed circles) and non-consumptive treatments (open circles). Different letters indicate significant differences (Tukey HSD; p < 0.05).

Figure 5.3. Body condition index of individuals across consumptive (closed circles) and non-consumptive treatments (open circles). Different letters indicate significant differences (Tukey HSD; p < 0.05).

When comparing metamorphs vs. paedomorphs, there was no effect of the predator

species × predator exposure interaction ( = 1.29, p = 0.257) or survival rate ( = 1.01, p =

0.314). While there was no effect of predator exposure ( = 0.184, p = 0.668), predator species

did affect the likelihood of metamorphosis and paedomorphosis ( = 8.52, p = 0.036), driven by a negative effect of F.chrysotus on the expression of paedomorphosis. When comparing

larvae and paedomorphs, there was a predator species × predator exposure interaction ( =

7.87, p = 0.005) and survival rate effect ( = 21.22, p < 0.001). Survival rate negatively affected the expression of paedomorphosis and likewise increased the number of individuals remaining as larvae, while the significant interaction was driven by consumptive, but not non- consumptive, F. chrysotus, which decreased the expression of paedomorphosis relative to larvae.

When comparing metamorphs vs. larvae, there was no effect of the predator species × predator

exposure interaction ( = 1.49, p = 0.222), but there was a negative effect of survival rate (

= 13.32, p < 0.001). After dropping the interaction term, predator exposure ( = 9.15, p =

0.002), but not predator species ( = 1.92, p = 0.590), had a significant effect driven by an increase in metamorphosis, relative to remaining as larvae, in non-consumptive treatments compared to consumptive treatments. While not included in analyses, consumptive green sunfish appeared to have the strongest effects on phenotype with few paedomorphs and no metamorphs emerging, despite large predator-induced density-mediated effects (Fig. 5.1).

Tail color metrics showed a predator species × predator exposure interaction (F2, 25 =

3.10, p = 0.0497). The variables R, G, B, Hue, and Brightness were explained by principal component 1 and explained 74.2% of the variation, while principal component 2 was saturation and explained 18.0% of the vartion (Fig. 5.5). Most treatments had significant overlap, but consumptive L. cyanellus showed marked dissimilarity with lower values for all metrics (i.e., less colorful).

Figure 5.4. Proportions of each phenotype at the end of the experiment in consumptive (+) and non-consumptive treatments (-).

Figure 5.5. Principle component analysis of color traits (R, G, B, Hue, Saturation, and Brightness). Principle component axis 1 (PC1; R, G ,B, Hue, Brightness) explained 74.2% of the variation, while principle component axis 2 (PC2; Saturation) explained 18.0% of the variation. Most treatments had high degrees of overlap except consumptive Lepomis cyanellus, which had lower values for all variables indicating larvae exposed to consumptive Lepomis cyanellus were less conspicuous.

DISCUSSION

The predators in this experiment varied in their effects on prey. Both larger gaped predators, A. sayanus and L. cyanellus, had equivalent consumptive effects, whereas the smaller gaped F. chrysotus had less intense consumptive effects. The consumptive effects imposed by A. sayanus translated directly into density-mediated effects, which released constraints on larval growth of the surviving salamanders and allowed them to reach the largest body sizes among all treatments. These patterns suggest that A. sayanus only has density-mediated effects on prey, especially when considering that this pattern did not occur with F. chrysotus and L. cyanellus.

While L. cyanellus consumption dramatically reduced salamander survival, there was no density- mediated growth release. These results suggest two things: 1) that consumptive L. cyanellus also imposes a significant trait-mediated effect on salamander growth rate and 2) that chemical cues of L. cyanellus alone are insufficient to elicit this trait-mediated effect because non-consumptive

L. cyanellus did not show any effects.

Both consumptive and non-consumptive Fundulus chrysotus had indistinguishable, intermediate effects on salamander body sizes despite the reduced density in consumptive treatments. Like with L. cyanellus, this pattern of both reduced density and reduced growth rates suggests that consumptive Fundulus chrysotus imposed trait-mediated effects on salamanders that were not seen with non-consumptive Fundulus chrysotus. While there were stark differences in the patterns seen among consumptive treatments and between consumptive and non- consumptive treatments, the four non-consumptive treatments showed no differences, indicating that salamanders either did not respond to predator chemical cues or habituated to them over

time due to lack of immediate threat. This lack of effect of chemical cues was opposite to what we have seen with gray tree frogs (Hyla chrysoscelis), which showed variable survival and growth patterns in response to chemical cues alone, albeit over a shorter time period that may not have allowed habituation (Bohenek et al., in review).

There were strong density-mediated effects on phenotype, evidenced by survival strongly reducing the frequency of both paedomorphosis and metamorphosis. While this effect of survival was expected for paedomorphosis, metamorphosis is expected to increase with increasing survival (i.e., increased survival acts as a proxy for higher conspecific density) (Harris 1987,

Semlitsch 1987c). However, if densities are too high, then developing larvae may not grow large enough to reach a threshold body size required for metamorphosis (Whiteman 1994, Whiteman et al. 2012). Fundulus treatments reduced paedomorph frequency relative to controls, despite the reductions in density, which should increase the frequency of paedomorphosis. Survival and consumptive L. cyanellus had the largest effects on metamorphosis, but F. chryostus also reduced paedomorph frequency. Similarly, salamander metamorphosis frequency was reduced when exposed to consumptive treatments, but non-consumptive treatments had no such effects suggesting that non-consumptive predator cues differentially affect paedomorph and metamorph frequencies (i.e., cues only strongly affect paedomorph frequency). Consumptive L. cyanellus had dramatic effects on metamorphosis as no salamanders metamorphosed from this treatment, likely due to the intense trait-mediated effects on body size. This pattern was not entirely unexpected with L. cyanellus, which has been shown to suppress paedomorphosis and metamorphosis in A. talpoideum (Jackson and Semlitsch 1993).

The trait-mediated effects on tail color were primarily driven by the effects of consumptive L. cyanellus, which differed markedly from all other treatments. Consumptive L.

cyanellus elicited desaturated tails with low values for R, G, B, Hue, and Brightness, but non- consumptive L. cyanellus showed overlap with all other treatments. While consumptive L. cyanellus invoked trait-mediated responses in the expected direction, it is surprising that non- consumptive L. cyanellus cues had no effect (Touchon and Warkentin 2008). No other predator treatments, including non-consumptive L. cyanellus, appeared to have effects on tail coloration.

The tail color and body size effects of consumptive L. cyanellus may be confounded as both occurred simultaneously. It could be that since salamanders exposed to consumptive L. cyanellus remained at small body sizes (due to trait-mediate behavioral effects), they were constantly below the gape limit of L. cyanellus and especially vulnerable to predation. Therefore, inconspicuous tail coloration was necessary for survival. Meanwhile, salamanders in other treatments were able to outgrow gape limitations of their predators and their phenotypes converged with controls. However, I do not have data to explicitly support this hypothesis and additional studies would be needed.

Salamanders and fish predators may compete over the same food resources, which can potentially explain the low growth rates seen with consumptive L. cyanellus. However, a similar experiment was conducted with gray tree frog larvae (Hyla chrysoscelis) and the effect of consumptive L. cyanellus on larval growth was the same, despite no trophic overlap between the two (Bohenek et al. in review). I conclude that this growth effect is likely trait-mediated (either due to stress, suppressed foraging, or both). That being said, our design can still be improved because trait- and density-mediated effects are confounded in our study due to consumptive predators eliciting both, while non-consumptive predators can only elicit trait-mediated effects.

Separating these effects is possible if a separate treatment group was maintained that removed individuals (Peacor and Werner 2001). However, the density-mediated response of A. talpoideum

100

has been previously studied and has known effects (Semlitsch 1987c). I have also conducted separate density experiments without predators (Bohenek, unpublished data) that confirm density-mediated effects are predictable and follow the patterns seen with A. sayanus and differ markedly from patterns seen with L. cyanellus.

While a recent meta-analysis concluded that sit-and-pursue predator cues had stronger non-consumptive effects on prey than active hunters (Preisser et al. 2007), our results show that predators had no non-consumptive effects. In fact, the physical presence and immediate threat of an active hunting predator had the greatest effect, suggesting that predator cues of active predators may need to be paired with an immediate threat stimulus (attacks) to be effective long term. Similarly, sit-and-pursue predators may not be as easily detected because selection should favor cryptic behavior and cue camouflage (Resetarits and Binckley 2013), which may explain high growth seen with A. sayanus. This same pattern was also noted in a meta-analysis that showed prey metamorphose at larger sizes in the presence of sit-and-pursue predators compared to active predators (Davenport et al. 2014).

Our results show that predator functional traits can have varying trait- and density- mediated effects on prey. Though consumptive A. sayanus and L. cyanellus had equivalent lethal effects on prey, they induced dramatically opposite shifts in larval body size suggesting that L. cyanellus, but not A. sayanus, had significant trait-mediated effects that completely overridden density-mediated traits. This pattern was also seen with the suppression of metamorphosis and achromatic tail coloration in the presence of consumptive L. cyanellus. Fundulus chrysotus appeared to have intermediate effects. Thus, active hunting predators may have long lasting effects on prey that impact fitness, but these effects may scale with gape size as the larger of the two active hunters had greater effects. Many aquatic ecology studies use L. cyanellus as a model

101

predator, but researchers should recognize that this species may be extreme in terms of effects on prey; however, its study is encouraged given its widespread distribution and introduction. While we continue to understand the importance of predator functional diversity (Chalcraft and

Resetarits 2003, Schmitz 2012), it is also important to understand that many predators co-occur and we still do not understand how prey respond to conflicting demands of predator functional diversity.

102

CHAPTER 6:

GLUCOCORTICOIDS REGULATE LIFE HISTORY STRATEGIES IN A FACULTATIVELY PAEDOMORPHIC SALAMANDER

103

ABSTRACT

The role of glucocorticoids in regulating a wide array of vertebrate ontogenetic transitions is becoming increasingly recognized. The timing and rate of these ontogenetic transitions are often plastic with glucocorticoids being key regulators of this plasticity. Extreme forms of phenotypic plasticity can result in polyphenisms where two or more alternative, environmentally-cued phenotypes are produced from the same genotype. Facultatively paedomorphic salamanders are polyphenic in that they can delay metamorphosis, the typical amphibian ontogenetic transition into terrestrial adults, and instead reproduce as aquatic paedomorphic adults. Paedomorphosis often occurs when aquatic conditions remain favorable, while metamorphosis typically occurs in response to deteriorating aquatic conditions. Since glucocorticoids are central to the vertebrate stress response and are known to play a central role in regulating obligate metamorphosis in amphibians, they may be key regulators of understudied paedomorphic life history strategies. Here, I manipulated corticosterone (CORT, the primary glucocorticoid in amphibians) to examine the effects on development in the facultatively paedomorphic mole salamander, Ambystoma talpoideum. I compared developmental trajectory of larvae in outdoor mesocosms exposed to low, medium, and high exogenous concentrations of

CORT. Results revealed that the proportion of paedomorphs and body size were both inversely related to exogenous CORT concentrations, which corresponded to whole-body CORT levels.

Consistent with known effects of CORT on obligate metamorphosis in amphibians, our results suggest that the suite of environmental cues known to influence facultative paedmorphosis may

104

be linked to the endocrine stress response, providing a proximate basis for theoretical models concerning ontogenetic transitions in facultatively paedomorphic salamanders.

105

INTRODUCTION

Phenotypic plasticity is the ability of one genotype to express multiple phenotypes under different environmental conditions (Nijhout 2003, West-Eberhard 2003, Pfennig et al. 2010).

While phenotypic plasticity can involve continuous trait variation, it may also manifest as non- continuous variation (two or more discrete, alternative phenotypes), referred to as polyphenisms.

Polyphenisms can provide fitness advantages over fixed phenotypes because they allow organisms to respond to spatial and temporal environmental heterogeneity (West-Eberhard

2003).

Facultative paedomorphosis is a polyphenism found in caudate amphibians represented across five families including Ambystomatidae, Salamandridae, Dicamptodontidae, Hynobiidae, and Plethodontidae (Denoël et al. 2005). In these salamanders, adults may have either a fully aquatic larval-like phenotype complete with gills (paedomorphosis) (Gould 1977) or a terrestrial phenotype after losing larval traits (metamorphosis) (Whiteman 1994). Facultative paedomorphosis differs markedly from the fixed life history strategies of most amphibians, which involves either obligate metamorphosis, wherein metamorphosis occurs prior to sexual maturity (e.g., all anurans and most Ambystomatidae, Hynobiidae, Salamandridae), or obligate paedomorphosis, wherein larval characteristics are retained throughout life (e.g., Amphiumidae,

Cryptobranchidae, Proteidae, Sirenidae) (Denoël et al. 2005).

Multiple hypotheses have been proposed to explain variation in phenotype frequency within and across populations (Wilbur and Collins 1973, Whiteman 1994). For example, the

Paedomorphic Advantage hypothesis predicts that paedomorphosis occurs among the largest

106

individuals when aquatic conditions are favorable and metamorphosis occurs when aquatic conditions are poor or deteriorate (Harris 1987, Semlitsch 1987, Jackson and Semlitsch 1993,

Denoël and Poncin 2001, Denoël and Ficetola 2014). The Best of a Bad Lot hypothesis predicts that paedomorphosis occurs when smaller larvae with slower growth rates do not reach the minimum size necessary for metamorphosis and thus maximize fitness by shortening time to first reproduction (Whiteman et al. 2012). Finally, the Dimorphic Paedomorph hypothesis predicts that both the smaller and larger individuals become paedomorphs (for the same reasons of

Paedomorph Advantage and Best of a Bad Lot), but also that the intermediate size classes metamorphose to escape competition with larger, dominant paedomorphs (Whiteman 1994). Of these hypotheses, the Paedomorph Advantage hypothesis is most well supported as numerous studies have provided evidence that deteriorating environmental conditions such as short hydroperiod (Semlitsch 1987), increased conspecific density (Harris 1987, Semlitsch 1987), low food availability (Denoël and Poncin 2001, Ryan and Semlitsch 2003) and high predation risk

(Jackson and Semlitsch 1993) have all been shown to induce metamorphosis in facultatively paedomorphic salamanders. Interestingly, these environmental conditions are linked by a common thread – that is, they represent environmental stressors known to stimulate the production of stress hormones (i.e., glucocorticoids) (Crespi and Denver 2005, Smith and Vale

2006, Denver 2017), implicating endocrine stress as a potential common proximate factor regulating polyphenic expression.

Adrenal glucocorticoids are known to play a pivotal role in amphibian metamorphosis

(Niki et al. 1981, Bonett et al. 2010, reviewed in Denver 2017). In amphibians with obligate metamorphosis, activation of the hypothalamic-pituitary-interrenal (HPI) axis stimulates thyroid stimulating hormone (TSH) via actions of corticotrophin-releasing hormone (CRH) on pituitary

107

thyrotropes (Denver 2009). Corticosterone (CORT, the primary glucocorticoid in amphibians) upregulates 5’-deiodinase activity that converts thyroxine (T4) to triiodothyronine (T3), which is the more potent form of the hormone (Galton 1990, Hayes and Wu 1995, Darras et al. 2002,

Kühn et al. 2005, Bonett et al. 2010). CORT also increases maximal T3 nuclear binding capacity

(Niki et al. 1981, Suzuki and Kikuyama 1983, Kikuyama et al. 1993) and increases tissue sensitivity to thyroid hormones by initiating transcription of thyroid hormone receptors in target tissues (Bonett et al. 2010).

CORT can also differentially impact growth and development depending on the developmental stage (Kulkarni and Buchholz 2014). In anuran larvae, for example, elevated

CORT inhibits growth in early stages of development (premetamorphosis), but increases development rate in later stages (prometamorphosis)(Frieden and Naile 1955, Kobayashi 1958,

Kikuyama et al. 1983, Gray and Janssens 1990, Hayes et al. 1993, Wright et al. 1994, Darras et al. 2002, Glennemeier and Denver 2002a, Belden et al. 2005, Hu et al. 2008), when thyroid hormone actions are more prevalent (Leloup and Buscaglia 1977, Polis 1981, Suzuki and Suzuki

1981, Carr and Norris 1988, Denver 1998, Glennemeier and Denver 2002b, Krain and Denver

2004).

Much of what is known regarding the role of endocrine stress in metamorphosis is based on work in anurans (frogs and toads), all of which are obligate metamorphs. Hence, it is unclear whether glucocorticoids also play a role in regulating facultative paedomorphosis (Denver 2017).

However, stress hormones also appear to play a central role in regulating the phenotype in other amphibians (i.e., salamanders), including obligate paedomorphs (Boorse and Denver 2002,

Bonett 2016). For example, metamorphosis in the obligate paedomorphic axolotl (Ambystoma mexicanum) can be induced by treatment with a combination of CORT and thyroid hormones

108

(Darras et al. 2002, Kühn et al. 2004, 2005) suggesting that these hormones also play a role in regulating facultative paedomorphosis.

Given that a variety of environmental stressors can affect the expression of facultative paedomorphosis, and the clear links between stress hormones and obligate metamorphosis in anurans (Denver 1997a, 1997b, Denver et al. 1998, Marino et al. 2014), I examined how stress hormones regulate facultative paedomorphosis in the mole salamander Ambystoma talpoideum

(Ambystomatidae) by manipulating CORT concentrations in the rearing water of developing larvae in outdoor mesocosms. Guided by the Paedomorph Advantage hypothesis, I predicted that

CORT treatment would negatively affect growth and body size, but increase the frequency of metamorphosis by increasing development rate.

109

METHODS

Mesocosm Experiment

Thirty-two 1200 L mesocosms (1.83 m diameter × 0.61 m depth) were constructed on 12

Dec 2016 in an 8 row × 4 column arrangement in an open field at UMFS. Initially, each mesocosm received 2 kg of dry leaf litter (mixed hardwoods) and 50 mL inoculum of concentrated zooplankton from a fishless pond (designated pond #61 by the UMFS). Mesocosms were filled with well-water, covered with fiberglass screening (1.3 mm × 1.13 mm opening) and allowed to age until 14–Jan when the screens were sunk below the water surface to allow chironomid and mosquito oviposition. An additional 2.5 L of fishless pond water and filtrate obtained by running 175 L of fishless pond water through 80 µm mesh was added to each mesocosm on 30–Jan.

Ambystoma talpoideum egg masses were collected from ponds and placed in small outdoor wading pools from 4–11 Jan. Upon hatching and before yolk sacs were absorbed, larvae were haphazardly separated into groups of four. Each group of four was then randomly assigned to a mesocosm so that each mesocosm received a total of 12 larvae derived from three different egg masses, well within the natural range of density (10/m3) for this species (Semlitsch 1987).

This procedure was carried out for twelve mesocosms on 3 Feb and 20 mesocosms on 6 Feb.

Larvae failed to establish in two blocks and those mesocosms were removed from the experiment. All mesocosms were treated identically at this stage of the experiment and salamanders were allowed to grow and develop freely.

Visual monitoring for metamorphs began on 6 Apr and was conducted every three nights.

110

The first metamorph was collected on 27 May. At this time, the pools were assigned CORT treatments in a randomized complete block design. Crystalline CORT was obtained from

Steraloids, Newport, RI (Cat #Q1550-000) and dissolved in ethanol before being added to mesocosms. Pools within each of the eight rows (row = block) of four mesocosms (8 × 4 = 32 mesocosms) were randomly assigned one of four treatments: Control (ethanol only), Low CORT

(0.006 g CORT in 20 mL ethanol, for a 14.43 nM mesocosm water concentration ), Medium

CORT (0.03 g, 72.16 nM) and High CORT (0.06 g, 144.31 nM). CORT dissolved in water is passively absorbed by salamanders (Glennemeier and Denver 2002a, Krain and Denver 2004,

Maher et al. 2013). A concentration of 125 nM has been shown to raise whole-body CORT content of Rana pipiens tadpoles by approximately 35% (Glennemeier and Denver 2002c), which corresponds to larval salamander CORT level that falls within the natural range of circulating CORT for Ambystoma (Houck et al. 1996, Homan et al. 2003a, 2003b, Cooperman et al. 2004).

Beginning on 31 May, mesocosms received their assigned treatment once every two weeks until 18 Sept, when the dosing interval was increased to every three days. The interval was increased mid-September because most metamorphosis occurs during this period. CORT degrades in water after 48 hours, suggesting that treatments were not compounding (Krain and

Denver 2004). The final treatment was administered on 25 Oct as occurrences of metamorphosis became infrequent, which is a seasonal pattern observed in past studies conducted at UMFS

(Pintar and Resetarits 2018; unpublished data).

The experiment concluded 13–14 Dec 2016 when all remaining individuals were collected from mesocosms. Upon collection, all individuals were weighed, photographed on a standard 1×1 mm grid background. Snout-vent length (SVL), total length (TL), head width (HW)

111

and head length (HL) were measured for each individual using photographs in Image-J v1.49

(Schneider et al. 2012). To minimize stress, individuals were only measured at the end of the experiment. Growth rates were determined by dividing SVL by the number of days in the experiment.

Radioimmunoassay

To evaluate the effect of treatment on CORT levels, a subset of 19 individuals were sampled twenty-four hours after the first treatment and were used for whole-body CORT concentration (n = 5 for Control, Medium and High CORT treatments, n = 4 for Low CORT treatment). Individuals were quickly captured from mesocosm tanks with a net and immediately frozen in liquid nitrogen (< 2 min). Since established densities used in our salamander mesocosm study were relatively low, I restricted sampling to only 4-5 individuals per treatment.

Individuals were stored at -20°C until assayed for whole-body CORT. At that time, frozen individuals were weighed, rinsed, measured (SVL), and homogenized while on ice. Each homogenized sample was diluted with purified water (5 ml water/per gram of salamander), vortexed vigorously for 5 min, and centrifuged at 2200 rpm for 10 min at 4°C and immediately placed back on ice. I then removed 1 ml of the liquid homogenate fraction for radioimmunoassay. I also created “pooled” homogenate samples by combining aliquots from all individuals, which were used as controls in the assay. Controls consisted of water blanks, whole- body pooled samples that were stripped with dextran-coated charcoal, and unstripped samples.

The pooled stripped and unstripped samples received 4 different volumes of standard CORT (0

µl, 10 µl, 50 µl and 100 µl), each in duplicate.

Samples were then incubated overnight with radiolabeled CORT (PerkinElmer, Inc.

Hebron, KY, USA) for determination of recoveries. CORT was then extracted from samples

112

using dichloromethane, dried under nitrogen gas at 40°C, resuspended in 10% ethyl acetate in iso-octane, and loaded onto diatomaceous earth columns containing a diatomaceous earth:distilled water “glycol trap” and a propanediol:ethylene glycol mixture for hormone separation (see Leary and Crocker-Buta 2018). The CORT fraction was collected using stepwise elution with increasingly polar mixtures of ethyl acetate in iso-octane. The CORT fraction was dried under nitrogen and resuspended in phosphate-buffered saline. The CORT antibody was purchased from MP Biomedicals, LLC (Solon, OH, USA, Cat #07120016). All samples were assayed in duplicate. Samples were analyzed in a single assay. The mean intra-assay coefficient of variation for CORT based on 4 standards was 4.48% (10µl of standard CORT), 3.59% (50 µl standard CORT) and 10.61% (100 µl CORT). The assay was validated by examining parallelism of the pooled stripped and unstripped samples. There was no evidence of heterogeneity in slopes across concentrations of standard CORT (F2, 18 = 2.26, p = 0.133).

Statistical analyses

MANOVA was used on morphometrics and life history variables and, if significant, was followed with univariate linear mixed effects models (LMMs). Larval period (i.e., number of days since mesocosm introduction until metamorphosis), mass, SVL, head length, head width, total length, and growth rate were analyzed using LMMs. CORT treatment and survival were included as fixed effects with the random term of mesocosm nested within block to assure statistical independence. Body condition (size corrected mass) analysis was conducted as a second analysis of mass with the inclusion of SVL as an additional covariate (Garcia-Berthou

2001). Significance for MANOVA and LMMs was tested with Approximate F Tests (Pillai’s trace and Type III Satterthwaite).

113

Survival analyses were performed on data aggregated by mesocosm and they were modeled as a logistic GLMM (Warton and Hui 2011) with CORT treatment as a fixed effect and block as a random effect. Phenotype proportions were analyzed using GLMM; however, due to binomial model singularity, phenotype proportions (aggregated by mesocosm) were analyzed with LMMs with treatment as a fixed effect and block as a random effect. I explored the effect of body mass on phenotype by fitting a logistic GLMM and using body mass as a fixed effect and mesocosm nested within block as a random effect. Significance for binomial GLMMS was tested with likelihood ratio tests of nested models (Bolker et al. 2009, Warton and Hui 2011).

Tukey HSD post hoc tests were used to comparemeans following significant treatment effects. All analyses used α = 0.05 and were performed in R v3.4.0 (R Core Team 2018) using the lme4 v1.1.13 package for mixed models (Bates et al. 2015), multcomp v1.4.6 for post hoc tests (Hothorn et al. 2008) and lmerTest v2.0.33 (Kuznetsova et al. 2015) for Approximate F

Tests. Figures were made with ggplot2 v2.2.1 (Wickham 2009) and sciplot v1.1.1 (Morales and

R Core Team 2012).

114

RESULTS

Mean whole-body CORT levels in larval A. talpoideum increased proportionately with increasing CORT concentrations across treatments (Fig. 6.1). Average survival did not differ

among treatments ( = 0.811, p = 0.847), but the proportions of paedomorphs (F3, 19 = 4.33, p =

0.017) and metamorphs (F3, 19 = 6.07, p = 0.004) differed significantly among treatments.

Controls had the highest proportion of paedomorphs (39%) while the proportion of paedomorphs decreased with increasing CORT concentrations; treatments with higher CORT concentrations produced no paedomorphs (Fig. 6.2).

MANOVA results revealed that treatment had a significant effect on morphological and life history variables (Pillai’s trace = 1.51, F21,42 = 2.02, p = 0.026) but not survival (Pillai’s trace

= 0.60, F7,12 = 2.57, p = 0.073). There was also an effect of CORT on body size; higher CORT concentrations resulted in lower body mass (F3, 77 = 7.04, p < 0.001) and shorter SVL (F3, 77 =

5.09, p = 0.003), total length (F3, 77 = 5.88, p = 0.001), head width (F3, 77 = 8.37, p < 0.001), and head length (F3, 77 = 6.66, p < 0.001) compared to controls (Fig. 6.3A-F). There was no difference in growth rate among treatments (F3, 74.83 = 2.37, p = 0.077) (Fig. 6.3F), while body condition was marginally lower in higher CORT concentrations (F3, 9.46 = 3.69, p = 0.057) after controlling for SVL (F1, 72.89 = 683.70, p < 0.001). Metamorphs did not differ in larval period among treatments (F3, 53.27 = 0.52, p = 0.672). Survival had a significant effect only on SVL, mass, TL, and HW (p < 0.05), consistent with classic density-dependent growth patterns seen in mesocosm studies (Wilbur and Collins 1973). Smaller individuals were more likely to metamorphose while the largest individuals were more likely to be paedomorphic across all

115

treatments ( = 69.78, p < 0.001) (Fig. 6.4).

Figure 6.1. Mean whole-body CORT concentrations (n = 19) of larval salamanders across the four treatments (F3, 13.66 = 13.66, p < 0.001). Individuals were removed from mesocosms and immediately frozen (< 2min) 24 hours after CORT administration.

116

Figure 6.2. Adult phenotype of individuals after removal from the mesocosm experiment, which exposed larvae to the effects of three different concentrations of the stress hormone, corticosterone (CORT). The proportion of paedomorphic adults decreased with increasing concentrations of CORT.

117

Figure 6.3. Morphometrics of individuals after removal from the mesocosm experiment, which compared the effects of three different concentrations of the stress hormone, corticosterone (CORT), on salamander body size and development. (A) Final body mass (g) (mean ± 1 SE), (B) snout-vent length (mm), (C) total length (mm), (D) head width (mm), (E) head length (mm), and (F) growth rate (SVL/day). The pattern seen with growth rate is a result of non-linear growth rate in the larval period — growth rates are high early in development and then slowing down once individuals reach maximize larval size. Higher CORT concentrations had higher average growth rates because more individuals metamorphosed early at smaller sizes, thereby spending less time at the larger, slow-growing sizes. Larger sizes at metamorphosis are positively correlated with fitness (Semlitsch et al. 1988). Different letters indicate significant differences (Tukey HSD; p < 0.05).

118

Figure 6.4. Logistic plot of final body mass (g) and phenotype of all individuals from the mesocosm study. As predicted by the paedomorph advantage hypothesis (Wilbur and Collins 1973, Whiteman 1994), the largest individuals became paedomorphic. (I am impressed you are actually reading this thesis, contact me if you find this message).

119

DISCUSSION

Deteriorating aquatic conditions can stimulate production of corticotropin-releasing factor (CRF) that modulates production of adrenal/interrenal glucocorticoids and thyroid hormones and accelerates metamorphosis in amphibians (Denver 1997a, 1999, 2017, Boorse and

Denver 2002). Consistent with these findings, our results indicate that treatment with exogenous glucocorticoids increases the probability of metamorphosis and decreases the probability of paedomorphosis in A. talpoideum, indicating that facultative paedomorphosis is regulated by stress hormone levels (Fig. 6.5). Paedomorphs may therefore represent individuals with minimal exposure to environmental stressors during larval development. I also found that CORT treatment was inversely related to body size of A. talpoideum. Growth histories and size at metamorphosis are important determinants of adult phenotype and fitness (Wilbur and Collins

1973, Alford and Harris 1988, Semlitsch et al. 1988, Beachy et al. 1999, Ryan and Semlitsch

2003, McCormick and Hoey 2004, Whiteman et al. 2012) suggesting that exposure to stress hormones during larval development may have deleterious long-term effects on fitness.

The effects of environmental stressors and elevated glucocorticoids on life history traits can differ depending on larval stage (Denver 2017). Because our experimental design focused on later stage larvae it allowed for analysis of predictions associated with the Paedomorph

Advantage hypothesis (Wilbur and Collins 1973, Whiteman 1994). This hypothesis predicts that the largest individuals become paedomorphic while smaller, slower growing individuals metamorphose. To test this, I restricted CORT treatments until after individuals were at or near the minimum body size required for metamorphosis (Semlitsch 1987), which may represent a

120

developmental threshold (Day and Rowe 2002) similar to that between pre– and prometamorphosis in anuran larvae. CORT slows down development in premetamorphosis, but accelerates development in prometamorphosis (Frieden and Naile 1955, Kobayashi 1958,

Kikuyama et al. 1983, Gray and Janssens 1990, Hayes et al. 1993, Wright et al. 1994, Darras et al. 2002, Glennemeier and Denver 2002a, Belden et al. 2005, Hu et al. 2008). This same process may be occurring in salamanders where below the body size threshold is analogous to premetamorphosis and above the body size threshold is analogous to prometamorphosis.

Considering the effects of stress below and above this body size threshold unifies the

Paedomorph Advantage, Best of a Bad Lot, and Dimorphic Paedomorph hypotheses proposed by

Whiteman (1994). Environmental stressors acting on a population of larval salamanders whose size distribution span this body size threshold may explain all possible scenarios described in the

Dimorphic Paedomorph hypothesis. For example, stress below the body size threshold will decrease somatic development and, if body sizes never crosses the body size threshold during the growing season, individuals will develop into Best of a Bad Lot (i.e., small) paedomorphs during the next reproductive season. Sufficient stress experienced above the body size threshold during the growing season will elicit metamorphosis, while individuals experiencing low or limited stress while above the body size threshold will develop into Paedomorph Advantage (i.e., large) paedomorphs during the next reproductive season. More work is required to firmly establish the growth and developmental effects of CORT above and below this threshold.

121

Figure 6.5 Proposed model incorporating endocrine mechanisms regulating facultative paedomorphosis. Under the paedomorph advantage hypothesis, and above a minimum body size, larvae that are exposed to environmental stressors (e.g., crowding, pond drying, predator presence, starvation) and have at least reached the minimum body size required for metamorphosis (35 mm; (Semlitsch and Wilbur 1988)) are more likely to metamorphose than become paedomorphic. Stressors activate the HPI axis (blue arrow) releasing CRF from the hypothalamus and ACTH and TSH from the anterior pituitary. ACTH stimulates the adrenal gland to release CORT, while TSH stimulates the thyroid gland to release THs (T3/T4). CORT and THs act synergistically to affect gene expression in target tissues resulting in metamorphosis (left) (reviewed in Denver 2017). In the absence of stressors, the HPI axis is not activated and metamorphosis is delayed resulting in aquatic paedomorphs (right). Both metamorphs and paedomorphs attain sexual maturity, but paedomorphs reach sexual maturity earlier (Ryan and Semlitsch 1998), increasing lifetime expected fitness (Cole 1954, Roff 1992, Stearns 1992).

122

Facultative paedomorphosis has been viewed as a specific plastic response to spatially and temporally variable hydroperiod (Wilbur and Collins 1973, Semlitsch and Gibbons 1985,

Semlitsch 1987), conspecific larval densities (Harris 1987, Semlitsch 1987), resource availability

(Semlitsch 1987, Denoël and Poncin 2001, Ryan and Semlitsch 2003), temperature (Sprules

1974) and other relevant environmental factors (Denoël and Ficetola 2014). However, instead of responding directly to variation in specific environmental factors (cues), larval salamanders may be responding to the cumulative effects of one or more interchangeable environmental stressors that affect CORT levels. Under a stress framework, polyphenic salamanders have a general response that is useful for unpredictable, heterogeneous habitats (Wilbur and Collins 1973,

Moran 1992). However, a generic response to all environmental stressors can be problematic if those stressors have disparate effects on fitness, e.g., pond drying or predator presence can be directly fatal whereas competition with conspecifics may only affect growth rates. Selection can tune sensitivity to different stressors, thereby creating optimally calibrated responses to specific environments. Additionally, since metamorphosis in facultatively paedomorphic salamanders is potentially risky and irreversible, there must be strong pressures to elicit abandonment of the aquatic environment.

Salamander species are present in a large diversity of ecosystems that vary in the relative quality of available habitat types (aquatic vs. terrestrial), which can select for either fixation of a phenotype or polyphenism maintenance, thereby producing a wide range of life history strategies across diverse lineages. For example, evolutionary processes may be different for populations with harsh, surrounding terrestrial environments, like alpine lakes, where selection favors paedomorphs (Wilbur and Collins 1973, Sprules 1974, Whiteman 1994) and metamorphosis may

123

be infrequent (e.g., montane Ambystoma tigrinum) or lost (e.g., Ambystoma mexicanum – axolotls). However, the evolution of facultative paedomorphosis does not require harsh terrestrial environments to exert strong selection against metamorphs, because facultative paedomorphosis is common in areas with suitable terrestrial habitats (Denoël and Ficetola 2014). Paedomorphosis can be maintained with available suitable terrestrial habitat due to greater expected fitness of paedomorphs in certain contexts (Denoël et al. 2005). For example, paedomorphs can reach large body sizes (Rose and Armentrout 1976, Whiteman 1994), have reduced age at first reproduction

(Semlitsch 1985, Semlitsch et al. 1988, Ryan and Semlitsch 1998) and earlier seasonal oviposition (Ryan and Plague 2004), which can both have large fitness benefits for paedomorphs over metamorphs (Cole 1954, Roff 1992, Stearns 1992).

The dynamics between stress and facultative paedomorphosis can have significant impacts on population dynamics because paedomorphs typically mature years earlier than metamorphs and at larger body sizes (Healy 1974, Harris 1987, Ryan and Semlitsch 1998), thereby increasing expected lifetime fitness (Cole 1954, Roff 1992, Stearns 1992). The maintenance of this polyphenism may therefore depend not only on evolutionary fitness tradeoffs between the alternative phenotypes, but also on spatial and temporal heterogeneity of the landscape of stress. Our experiment shows that the dynamics of life histories and phenotypic expression of polyphenisms can be affected by stress hormones. Rather than requiring a specific link between each individual environmental stressor and the expression of facultative paedomorphosis, our data suggest a more parsimonious, comprehensive mechanism based on the common currency of stress hormones. However, characterizing the stress response of individuals in natural ponds and the differential endocrine impacts of environmental stressors is worthy of further investigation. This mechanism provides a framework for the integration of multiple

124

environmental stressors that simultaneously (or sequentially) impact life history trajectories and may be broadly applicable to other life history polyphenisms and polymorphisms in a variety of taxa.

125

LIST OF REFERENCES

126

Abrams, P. A. 1984. Foraging Time Optimization and Interactions in Food Webs. The American Naturalist 124:80–96. Abrams, P. A. 1995. Implications of Dynamically Variable Traits for Identifying, Classifying, and Measuring Direct and Indirect Effects in Ecological Communities. The American Naturalist 146:112–134. Albecker, M., and H. D. Vance-Chalcraft. 2015. Mismatched anti-predator behavioral responses in predator-naïve larval anurans. PeerJ 3:e1472. Alford, R. A., and R. N. Harris. 1988. Effects of larval growth history on anuran metamorphosis 131:91–106. Altwegg, R. 2003. Multistage density dependence in an amphibian. Oecologia 136:46–50. Anderson, R. M., and D. M. Gordon. 1982. Processes influencing the distribution of parasite numbers within host populations with special emphasis on parasite-induced host mortalities. Parasitology 85:373–398. Anderson, T. L., C. L. Mott, T. D. Levine, and H. H. Whiteman. 2013. Life Cycle Complexity Influences Intraguild Predation and Cannibalism in Pond Communities. Copeia 2013:284– 291. Anderson, T. L., F. E. Rowland, and R. D. Semlitsch. 2017. Variation in phenology and density differentially affects predator–prey interactions between salamanders. Oecologia 185:475– 486. Anderson, T. L., and H. H. Whiteman. 2015. Asymmetric effects of intra- and interspecific competition on a pond-breeding salamander. Ecology 96:1681–1690. Applebaum, S. W., and Y. Heifetz. 1999. Density-dependent physiological phase in insects. Annual Review of Entomology 44:317–341. Aristotle. (n.d.). IX, Part 1. Page (D. W. Thompson, Ed.) Historia Animalium. Barton, K. 2018. MuMIn: Multi-Model Inference. Bates, D., M. Mächler, B. Bolker, and S. Walker. 2015. Fitting Linear Mixed-Effects Models Using {lme4}. Journal of Statistical Software 67:1–48. Beachy, C. K., T. H. Surges, and M. Reyes. 1999. Effects of developmental and growth history on metamorphosis in the gray treefrog, Hyla versicolor (Amphibia, Anura). Journal of Experimental Zoology 283:522–530. Belden, L. K., I. T. Moore, J. C. Wingfield, and A. R. Blaustein. 2005. Corticosterone and Growth in Pacific Treefrog (Hyla regilla) Tadpoles. Copeia 2005:424–430. Belden, L. K., M. J. Rubbo, J. C. Wingfield, and J. M. Kiesecker. 2007. Searching for the physiological mechanism of density dependence: does corticosterone regulate tadpole responses to density? Physiological and biochemical zoology : PBZ 80:444–51. Benard, M. F. 2004. Predator-induced phenotypic plasticty in organisms with complex life histories. Annual Review of Ecology and Systematics 35:651–673. Benke, A. C., R. L. Henry, D. M. Gillespie, and R. J. Hunter. 1985. Importance of Snag Habitat for Animal Production in Southeastern Streams. Fisheries 10:8–13. Berenbaum, M. R., and A. R. Zangerl. 1999. Coping with life as a menu option: inducible defences of the wild parsnip. Pages 10–32 The Ecology and Evolution of Inducible Defenses. Berner, D. 2011. Size correction in biology: How reliable are approaches based on (common) principal component analysis? Oecologia 166:961–971. Berven, K. A. 1990. Factors affecting population fluctuations in larval and adult stages of the wood frog (Rana sylvatica).

127

Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. H. Stevens, and J. S. S. White. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in Ecology and Evolution 24:127–135. Bonett, R. M. 2016. An Integrative Endocrine Model for the Evolution of Developmental Timing and Life History of Plethodontids and Other Salamanders. Copeia 104:209–221. Bonett, R. M., E. D. Hoopfer, and R. J. Denver. 2010. Molecular mechanisms of corticosteroid synergy with thyroid hormone during tadpole metamorphosis. General and Comparative Endocrinology 168:209–219. Boorse, G. C., and R. J. Denver. 2002. Acceleration of Ambystoma tigrinum metamorphosis by corticotropin-releasing hormone. Journal of Experimental Zoology 293:94–98. Brandon, R. A., and D. J. Bremer. 1966. Neotenic newts, Notophthalmus viridescens louisianensis, in southern Illinois. Herpetologica 22:213–217. Brockelman, W. Y. 1969. An analysis of density effects and predation in Bufo americanus tadpoles. Ecology 50:632–644. Brönmark, C., and L. Hansson. 2000. Chemical communication in aquatic systems: an introduction. Oikos 88:1–7. Brönmark, C., L. B. Pettersson, and P. A. Nilsson. 1999. Predator-induced defense in Crucian carp. Pages 203–217 Ecology and Evolution of Inducible Defenses. Brook, B. W., and C. J. A. Bradshaw. 2006. Strength of evidence for density dependence in abundance time series of 1198 species. Ecology 87:1445–1451. Brooks, J., and S. I. Dodson. 1965. Predation, Body Size, and Composition of Plankton. Science 150:28–35. Brown, J. H., J. F. Gillooly, A. P. Allen, V. M. Savage, and G. B. West. 2004. Toward a metabolic theory of ecology. Ecology 85:1771–1789. Van Buskirk, J., and D. C. Smith. 1991. Density-dependent population regulation in a salamander. Ecology 72:1747–1756. Cappuccino, N., and P. W. Price. 1995. Population Dynamics: New Approaches and Synthesis. Carr, J. A., and D. O. Norris. 1988. Interrenal activity during metamorphosis of the tiger salamander, Ambystoma tigrinum. General and Comparative Endocrinology 71:63–69. Chalcraft, D., and W. J. Resetarits. 2003. Predator identity and ecological impacts: functional redundancy or functional diversity? Ecology 84:2407–2418. Chalcraft, D., and W. J. Resetarits. 2004. Metabolic rate models and the substitutability of predator populations. Journal of Animal Ecology 73:323–332. Chesson, P. 1996. Matters of scale in the dynamics of populations and communities. Pages 353– 368 Frontiers of population ecology. Chivers, D. P., and R. J. F. Smith. 1998. Chemical alarm signalling in aquatic predator-prey systems: A review and prospectus. Ecoscience 5:338–352. Cisse, S., S. Ghaout, A. Mazih, M. A. Ould Babah Ebbe, and C. Piou. 2015. Estimation of density threshold of gregarization of desert locust hoppers from field sampling in Mauritania. Entomologia Experimentalis et Applicata 156:136–148. Cole, L. C. 1954. The population consequences of life history phenomena. The Quarterly review of biology 29:103–137. Collins, J. P., and J. E. Cheek. 1983. Effect of food and density on development of typical and cannibalistic salamander larvae in Ambystoma tigrinum nebulosum. Integrative and Comparative Biology 23:77–84. Cooperman, M. D., J. M. Reed, and L. M. Romero. 2004. The effects of terrestrial and breeding

128

densities on corticosterone and testosterone levels in spotted salamanders, Ambystoma maculatum. Canadian Journal of Zoology 82:1795–1803. Crespi, E. J., and R. J. Denver. 2005. Roles of stress hormones in food intake regulation in anuran amphibians throughout the life cycle. Comparative Biochemistry and Physiology - A Molecular and Integrative Physiology 141:381–390. Croissant, Y. 2018. mlogit: multinomial logit model. Darras, V. M., S. Van Der Geyten, C. Cox, I. B. Segers, B. De Groef, and E. R. Kühn. 2002. Effects of dexamethasone treatment on iodothyronine deiodinase activities and on metamorphosis-related morphological changes in the axolotl (Ambystoma mexicanum). General and Comparative Endocrinology 127:157–164. Davenport, J. M., and D. R. Chalcraft. 2014. Increasing conspecific density weakens the ability of intermediate predators to develop induced morphological defences to top predators. Freshwater Biology 59:87–99. Davenport, J. M., B. R. Hossack, and W. H. Lowe. 2014. Partitioning the non-consumptive effects of predators on prey with complex life histories. Oecologia 176:149–155. Day, T., and L. Rowe. 2002. Developmental thresholds and the evolution of reaction norms for age and size at life-history transitions. The American naturalist 159:338–350. Denoël, M., and G. F. Ficetola. 2014. Heterochrony in a complex world: Disentangling environmental processes of facultative paedomorphosis in an amphibian. Journal of Animal Ecology 83:606–615. Denoël, M., P. Joly, and H. H. Whiteman. 2005. Evolutionary ecology of facultative paedomorphosis in newts and salamanders. Biological reviews of the Cambridge Philosophical Society 80:663–671. Denoël, M., and P. Poncin. 2001. The effect of food on growth and metamorphosis of paedomorphs in Triturus alpestris apuanus. Archiv fur Hydrobiologie 152:661–670. Denver, R. J. 1997a. Environmental stress as a developmental cue: corticotropin-releasing hormone is a proximate mediator of adaptive phenotypic plasticity in amphibian metamorphosis. Hormones and behavior 31:169–179. Denver, R. J. 1997b. Proximate Mechanisms of Phenotypic Plasticity in Amphibian Metamorphosis. American Zoologist 37:172–184. Denver, R. J. 1998. Hormonal correlates of environmentally induced metamorphosis in the Western spadefoot toad, Scaphiopus hammondii. General and Comparative Endocrinology 110:326–336. Denver, R. J. 1999. Evolution of the corticotropin-releasing hormone signaling system and its role in stress-induced phenotypic plasticity. Annals of the New York Academy of Sciences 897:46–53. Denver, R. J. 2009. Structural and functional evolution of vertebrate neuroendocrine stress systems. Annals of the New York Academy of Sciences 1163:1–16. Denver, R. J. 2017. Endocrinology of Complex Life Cycles: Amphibians. Pages 145–168 in D. W. Pfaff and M. Joels, editors. Hormones, Brain and Behavior Online. 3rd edition. Academic Press, Oxford. Denver, R. J., N. Mirhadi, and M. Phillips. 1998. Adaptive plasticity in amphibian metamorphosis: Response of Scaphiopus hammondii tadpoles to habitat desiccation. Ecology 79:1859–1872. Dettner, K., and C. Liepert. 1994. Chemical Mimicry and Camouflage. Annual Review of Entomology 39:129–154.

129

Dillman, C. B., D. E. Bergstrom, D. B. Noltie, T. P. Holtsford, and R. L. Mayden. 2011. Regressive progression, progressive regression or neither? Phylogeny and evolution of the Percopsiformes (Teleostei, Paracanthopterygii). Zoologica Scripta 40:45–60. van Donk, E., A. Ianora, and M. Vos. 2011. Induced defences in marine and freshwater phytoplankton: A review. Hydrobiologia 668:3–19. Doyle, J. M., and H. H. Whiteman. 2008. Paedomorphosis in Ambystoma talpoideum: effects of initial body size variation and density. Oecologia 156:87–94. Dudley, S. A., and J. Schmitt. 1996. Testing the Adaptive Plasticity Hypothesis: Density- Dependent Selection on Manipulated Stem Length in Impatiens capensis. The American Naturalist 147:445. Duellman, W. E., and L. Trueb. 1986. The Biology of Amphibians. McGraw-Hill, new York. Elton, C., and M. Nicholson. 1942. The ten-year cycle in numbers of the lynx in Canada. Journal of Animal Ecology 11:215–244. Ferrari, M. C. O., B. D. Wisenden, and D. P. Chivers. 2010. Chemical ecology of predator–prey interactions in aquatic ecosystems: a review and prospectus. Canadian Journal of Zoology 88:698–724. Flemer, D. A., and W. S. Woolcott. 1966. Food Habits and Distribution of the Fishes of Tuckahoe Creek, Virginia, with Special Emphasis on the Bluegill, Lepomis m. macrochirus Rafinesque. Chesapeake Science 7:75–89. Forbes, S. A. 1888. On the food relations of freshwater fishes: a summary and discussion. Bulletin of the Illinois State Laboratory of Natural History 2:475–538. Forbes, S. A., and R. E. Richardson. 1908. The Fishes of Illinois. Illinois State Laboratory of Natural History. Foster, N. 1967. Comparative studies on the biology of killifishes (Pisces: Cyprinodontidae). Cornell, Ithaca, NY. Frieden, E., and B. Naile. 1955. Biochemistry of Amphibian Metamorphosis : I . Enhancement of Induced Metamorphosis. Science 121:37–39. Galton, V. A. 1990. Mechanisms underlying the acceleration of thyroid hormone-induced tadpole metamorphosis by corticosterone. Endocrinology 127:2997–3002. Garcia-Berthou, E. 2001. On the misuse of residuals in ecology: testing regression residual vs the analysis of covariance. Journal of Animal Ecology 70:708–711. Gause, G. 1934. The Struggle for Existence. Williams and Wilkins, Baltimore. Getty, T. 1996. The maintenance of phenotypic plasticity as a signal detection problem. The American Naturalist 148:378–385. Gilbert, J. J. 1999. Kairomone-induced morphological defenses in rotifers. Pages 127–141 in R. Tollrian and C. D. Harvell, editors. Ecology and Evolution of Inducible Defenses. Princeton University Press, Princeton, NJ. Gill, D. E. 1978. The metapopulation ecology of the red-spotted newt, Notophthalmus viridescens (Rafinesque). Ecological Monographs 48:145–166. Glennemeier, K. A., and R. J. Denver. 2002a. Small changes in whole-body corticosterone content affect larval Rana pipiens fitness components. General and Comparative Endocrinology 127:16–25. Glennemeier, K. A., and R. J. Denver. 2002b. Role for corticoids in mediating the response of Rana pipiens tadpoles to intraspecific competition. Journal of Experimental Zoology 292:32–40. Glennemeier, K. A., and R. J. Denver. 2002c. Developmental changes in interrenal

130

responsiveness in anuran amphibians. Integrative and comparative biology 42:565–573. Gotelli, N. J. 2008. A Primer of Ecology. 4th edition. Sinauer Associates, Sunderland, MA. Gould, S. J. 1977. Ontogeny and phylogeny. Harvard University Press, Cambridge. Gray, K. M., and P. A. Janssens. 1990. Gonadal hormones inhibit the induction of metamorphosis by thyroid hormones in Xenopus laevis tadpoles in vivo, but not in vitro. General and Comparative Endocrinology 77:202–211. Grayson, K. L., and H. M. Wilbur. 2009. Sex- and context-dependent migration in a pond- breeding amphibian. Ecology 90:306–312. Grunt, J. W. G., and I. Bayly. 1981. Predator induction of crests in morphs of the Daphnia carinata King complex. Limnology and Oceanography 26:201–218. Gunning, G., and W. Lewis. 1955. The fish population of a spring-fed swamp in the Mississippi bottoms of southern Illinois. Ecology 36:552–558. Hanken, J., and R. J. Wassersug. 1981. The visible skeleton. Functional Photography 16:22- 26,44. Harris, R. N. 1987a. Density-dependent paedomorphosis in the salamander Notophthalmus viridescens dorsalis. Ecology 68:705–712. Harris, R. N. 1987b. An experimental study of population regulation in the salamander, Notophthalmus viridescens dorsalis (Urodela: Salamandridae). Oecologia 71:280–285. Harris, R. N., R. A. Alford, and H. M. Wilbur. 1988. Density and phenology of Notophthalmus viridescens dorsalis in a natural pond. Herpetologica 44:234–242. Harris, R. N., R. D. Semlitsch, H. M. Wilbur, and J. E. Fauth. 1990. Local variation in the genetic basis of paedomorphosis in the salamander Ambystoma talpoideum. Evolution 44:1588–1603. Harrison, R. G. 1980. Dispersal Polymorphisms in Insects. Annual Review of Ecology and Systematics 11:95–118. Harvell, C. D. 1990. The ecology and evolution of inducible defenses. The Quarterly Review of Biology 65:323–340. Hassell, M. P. 1975. Density-Dependence in Single-Species Populations. The Journal of Animal Ecology 44:283. Hayes, T. B., and T. H. Wu. 1995. Interdependence of corticosterone and thyroid hormones in toad larvae (Bufo boreas). II. Regulation of corticosterone and thyroid hormones. Journal of Experimental Zoology 271:103–111. Hayes, T., R. Chan, and P. Licht. 1993. Interactions of temperature and steroids on larval growth, development, and metamorphosis in a toad (Bufo boreas). Journal of Experimental Zoology 266:206–215. Healy, W. 1974. Population consequences of alternative life histories in Notophthalmus v. viridescens. Copeia 1974:221–229. Healy, W. R. 1973. Life history variation and the growth of juvenile Notophthalmus viridescens from Massachusetts. Copeia 1973:641–647. Hellriegel, B. 2000. Single- or Multistage Regulation in Complex Life Cycles: Does It Make a Difference? Oikos 88:239–249. Hoffman, E. A., and D. W. Pfennig. 1999. Proximate causes of cannibalistic polyphenism in larval tiger salamanders. Ecology 80:1076–1080. Homan, R. N., J. M. Reed, and L. M. Romero. 2003a. Corticosterone concentrations in free- living spotted salamanders (Ambystoma maculatum). General and Comparative Endocrinology 130:165–171.

131

Homan, R. N., J. V. Regosin, D. M. Rodrigues, J. M. Reed, B. S. Windmiller, and L. M. Romero. 2003b. Impacts of varying habitat quality on the physiological stress of spotted salamanders (Ambystoma maculatum). Animal Conservation 6:11–18. Hothorn, T., F. Bretz, and P. Westfall. 2008. Simultaneous Inference in General Parametric Models. Biometrical Journal 50:346–363. Houck, L. D., M. T. Mendonça, T. K. Lynch, and D. E. Scott. 1996. Courtship behavior and plasma levels of androgens and corticosterone in male marbled salamanders, Ambystoma opacum (Ambystomatidae). General and Comparative Endocrinology 104:243–252. Hu, F., E. J. Crespi, and R. J. Denver. 2008. Programming neuroendocrine stress axis activity by exposure to glucocorticoids during postembryonic development of the frog, Xenopus laevis. Endocrinology 149:5470–5481. Hubbs, C., R. J. Edwards, and G. P. Garrett. 2008. An annotated checklist of the freshwater fishes of Texas, with keys to identification of species. Texas Journal of Science 43:1–56. Huchette, S. M. H., C. S. Koh, and R. W. Day. 2003. Growth of juvenile blacklip abalone (Haliotis rubra) in aquaculture tanks: Effects of density and ammonia. Aquaculture 219:457–470. Hunt, B. P. 1953. Food Relationships Between Florida Spotted Gar and Other Organisms in the Tamiami Canal, Dade County, Florida. Transactions of the American Fisheries Society 82:13–33. Jackson, M., and R. Semlitsch. 1993. Paedomorphosis in the salamander Ambystoma talpoideum: effects of a fish predator. Ecology 74:342–350. Johnson, D. J., W. T. Beaulieu, J. D. Bever, and K. Clay. 2012. Conspecific negative density dependence and forest diversity. Science 336:904–7. Kats, L. B., and L. M. Dill. 1998. The scent of death: Chemosensory assessment of predation risk by prey animals. Ecoscience 5:361–394. Keiser, E. 1999. Salamanders of the University of Mississippi Field Station. University of Mississippi, Oxford, MS. Keiser, E. 2001. Turtles of the University of Mississippi Field Station. University of Mississippi, Oxford, MS. Keiser, E. 2008. Frogs of the University of Mississippi Field Station. University of Mississippi, Oxford, MS. Keiser, E. 2010. Snakes of the University of Mississippi Field Station. University of Mississippi, Oxford, MS. Keiser, E. 2014. Lizards of the University of Mississippi Field Station. University of Mississippi, Oxford, MS. Kikuyama, S., K. Kawamura, S. Tanaka, and K. Yamamoto. 1993. Amphibian Metamorphosis: Hormonal Control. International Review of Cytology 145:105–148. Kikuyama, S., K. Niki, M. Mayumi, R. Shibayama, M. Nishikawa, and N. Shintake. 1983. Studies on corticoid action on the toad tadpole tail in vitro. General and Comparative Endocrinology 52:395–399. Kobayashi, H. 1958. Effect of desoxycorticosterone acetate on metamorphosis induced by throxine in anuran tadpoles. Endocrinology 62:371–377. Krain, L. P., and R. J. Denver. 2004. Developmental expression and hormonal regulation of glucocorticoid and thyroid hormone receptors during metamorphosis in Xenopus laevis. Journal of Endocrinology 181:91–104. Kuhlmann, H.-W., J. Kusch, and K. Heckman. 1999. Predator-induced Defenses in Ciliated

132

Protozoa. Pages 142–159 in R. Tollrian and C. D. Harvell, editors. Ecology and Evolution of Inducible Defenses. Princeton University Press, Princeton, NJ. Kühn, E. R., B. De Groef, S. Van Der Geyten, and V. M. Darras. 2005. Corticotropin-releasing hormone-mediated metamorphosis in the neotenic axolotl Ambystoma mexicanum: Synergistic involvement of thyroxine and corticoids on brain type II deiodinase. General and Comparative Endocrinology 143:75–81. Kühn, E. R., B. De Groef, S. V. H. Grommen, S. Van Der Geyten, and V. M. Darras. 2004. Low submetamorphic doses of dexamethasone and thyroxine induce complete metamorphosis in the axolotl (Ambystoma mexicanum) when injected together. General and Comparative Endocrinology 137:141–147. Kulkarni, S. S., and D. R. Buchholz. 2014. Corticosteroid signaling in frog metamorphosis. General and Comparative Endocrinology 203:225–231. Kuzmin, S. L. 1995. The problem of food competition in amphibians. Herpetological Journal 5:252–256. Kuznetsova, A., P. B. Brockhoff, and R. H. B. Christensen. 2015. Package ‘lmerTest.’ Laforsch, C., R. Tollrian, and E. Gene. 2009. Cyclomorphosis and phenotypic changes. Pages 1159–1166 in G. E. Likens, editor. Encyclopedia of Inland Waters. Academic Press, Oxford. Leary, C. J., and S. Crocker-Buta. 2018. Rapid effects of elevated stress hormones on male courtship signals suggest a major role for the acute stress response in intra- and intersexual selection. Functional Ecology 32:1214–1226. Lee, D. S., C. R. Gilbert, C. H. Hocutt, R. E. Jenkins, D. E. McAllister, and J. R. Stauffer. 1980. Aphredoderus sayanus (Gilliams), pirate perch. Page 484 in D. S. Lee, McAllister, C. Jenkins, C. Gilbert, and C. Hocutt, editors. Atlas of North American Freshwater Fishes. North Carolina State Museum of Natural History, Raleigh. Lejeune, B., N. Sturaro, G. Lepoint, and M. Denoël. 2018. Facultative paedomorphosis as a mechanism promoting intraspecific niche differentiation. Oikos 127:427–439. Leloup, J., and M. Buscaglia. 1977. Triiodothyronine, hormone of amphibian metamorphosis. Comptes Rendus Hebdomadaires Des Seances De L Academie Des Sciences Serie D 284:2261–2263. Levis, N. A., and D. W. Pfennig. 2018. Phenotypic plasticity, canalization, and the origins of novelty: Evidence and mechanisms from amphibians. Seminars in Cell and Developmental Biology:12–20. Elsevier Ltd. Lima, S. L. 1998. Nonlethal Effects in the Ecology of Predator-Prey Interactions. BioScience 48:25–34. Lima, S. L. 2002. Putting predators back into behavioral predator–prey interactions. Trends in Ecology & Evolution 17:70–75. Lima, S. L., and L. M. Dill. 1990. Behavioral decisions made under the risk of predation: a review and prospectus. Canadian Journal of Zoology 68:619–640. Lively, C. M. 1986. Predator-induced shell dimorphism in the acorn barnacle Chthamalus anisopoma. Evolution 40:232–242. Loman, J. 2004. Density regulation in tadpoles of Rana temporaria: a full pond field experiment. Ecology 85:1611–1618. Lotka, A. J. 1925. Elements of Physical Biology. Williams and Wilkins Company, Baltimore. Lotka, A. J. 1932. The growth of mixed populations: two species competing for a food supply. Journal of the Washington Academy of Sciences 22:461–469.

133

Maher, J. J. M., E. E. Werner, and R. J. Denver. 2013. Stress hormones mediate predator- induced phenotypic plasticity in amphibian tadpoles. Proceedings of the Royal Society B: Biological Sciences 280:1–9. Malthus, T. R. 1798. An essay on the principle of population, as it affects the future improvement of society. J. Johnson, London. Marino, J. A., M. P. Holland, and J. Middlemis Maher. 2014. Predators and trematode parasites jointly affect larval anuran functional traits and corticosterone levels. Oikos 123:451–460. McCollum, S. A., and J. Van Buskirk. 1996. Costs and benefits of a predator-induced polyphenism in the gray treefrog Hyla chrysoscelis. Evolution 50:583–593. McPeek, M. a. 2004. The growth/predation risk trade-off: so what is the mechanism? The American naturalist 163:E88–E111. Menon, R., and M. M. Holland. 2012. Study of Understory Vegetation at The University of Mississippi Field Station in North Mississippi. Castanea 77:28–36. Michimae, H., K. Nishimura, and M. Wakahara. 2005. Mechanical vibrations from tadpoles’ flapping tails transform salamander’s carnivorous morphology. Biology letters 1:75–7. Michimae, H., and M. Wakahara. 2002. A tadpole-induced polyphenism in the salamander Hynobius retardatus. Evolution 56:2029–38. Miner, B. G., S. E. Sultan, S. G. Morgan, D. K. Padilla, and R. A. Relyea. 2005. Ecological consequences of phenotypic plasticity. Trends in Ecology and Evolution 20:685–692. Moczek, A. P. 1998. Horn polyphenism in the beetle Onthophagus taurus: larval diet quality and plasticity in parental investment determine adult body size and male horn morphology. Behavioral Ecology 9:636–641. Monzyk, F. R., W. E. Kelso, and D. A. Rutherford. 1997. Characteristics of Woody Cover Used by Brown Madtom. Transactions of the American Fisheries Society 126:665–675. Moore, G. A., and W. E. Burris. 1956. Description of the lateral line system of the pirate perch, Aphredoderus sayanus. Copeia 1956:18–20. Morales, M., and R Core Team. 2012. Sciplot: Scientific Graphing Functions for Factorial Designs. R Foundation for Statistical Computing, Vienna. Moran, N. A. 1992. The Evolutionary Maintenance of Alternative Phenotypes. The American Naturalist 139:971–989. Morey, S., and D. Reznick. 2000. A Comparative Analysis of Plasticity in Larval Development in Three Species of Spadefoot Toads. Ecology 81:1736. Morin, P. J. 1981. Predatory salamanders reverse the outcome of competition among three species of anuran tadpoles. Science 212:1284–1286. Newman, R. A. 1987. Effects of density and predation on Scaphiopus couchi tadpoles in desert ponds. Oecologia 71:301–307. Newman, R. A. 1992. Adaptive Plasticity in Amphibian Metamorphosis. BioScience 42:671– 678. Newman, R. A. 1994. Effects of changing density and food level on metamorphosis of a desert amphibian, Scaphiopus couchii. Ecology 75:1085–1096. Nicholson, A. 1954. An outline of the dynamics of animal populations. Australian Journal of Zoology 2:9–65. Nijhout, H. F. 2003. Development and evolution of adaptive polyphenisms. Evolution & Development 5:9–18. Niki, K., K. Yoshizato, and S. Kikuyama. 1981. Augmentation of Nuclear-Binding Capacity for Triiodothyronine by Aldosterone in Tadpole Tail. Proceedings of the Japan Academy Series

134

B-Physical and Biological Sciences 57:271–275. Noble, G. 1926. The Long Island newt: a contribution to the life history of Triturus viridescens. American Museum Novitates 228:1–11. Noble, G. K. 1929. Further observations on the life-history of the newt, Triturus viridescens. American Museum Novitates 348:1–22. Oksanen, J., F. G. Blanchet, M. Friendly, R. Kindt, P. Legendre, D. McGlinn, P. R. Minchin, R. B. O’Hara, G. L. Simpson, P. Solymos, M. H. H. Stevens, E. Szoecs, and H. Wagner. 2018. vegan: Community Ecology Package. Oromi, N., J. Michaux, and M. Denoël. 2016. High gene flow between alternative morphs and the evolutionary persistence of facultative paedomorphosis. Scientific Reports 6:32046. Park, D., and C. R. Propper. 2001. Repellent function of male pheromones in the red-spotted newt. Journal of Experimental Zoology 289:404–408. Parker, N. C., and B. A. Simco. 1975. Activity Patterns, Feeding and Behavior of the Pirateperch, Aphredoderus sayanus. Copeia 1975:572. Peacor, S. D., B. L. Peckarsky, G. C. Trussell, and J. R. Vonesh. 2013. Costs of predator-induced phenotypic plasticity: A graphical model for predicting the contribution of nonconsumptive and consumptive effects of predators on prey. Oecologia 171:1–10. Peacor, S. D., and E. E. Werner. 2001. The contribution of trait-mediated indirect effects to the net effects of a predator. Proceedings of the National Academy of Sciences 98:3904–3908. Pechmann, J. 1994. Population regulation in complex life cycles: aquatic and terrestrial density- dependence in pond-breeding amphibians. Duke University. Peckarsky, B. L., P. A. Abrams, D. I. Bolnick, and L. M. Dill. 2008. Revisiting the Classics : Considering Nonconsumptive Effects in Textbook Examples of Predator–Prey Interactions. Ecology 89:2416–2425. Pener, M. P., and S. J. Simpson. 2009. Locust Phase Polyphenism: An Update. Advances in Insect Physiology, Vol 36 36:1–272. Petranka, J., L. Kats, and A. Sih. 1987. Predator-prey interactions among fish and larval amphibians: use of chemical cues to detect predatory fish. Animal Behaviour 35:420–425. Petranka, J. W. 1989a. Chemical interference competition in tadpoles: Does it occur outside laboratory aquaria? Copeia 1989:921–930. Petranka, J. W. 1989b. Density-Dependent Growth and Survival of Larval Ambystoma: Evidence from Whole-Pond Manipulations. Ecology 70:1752–1767. Petranka, J. W. 1998. Ambystoma talpoideum. Pages 96–102 Salamanders of the United States and Canada. Smithsonian Institution Press, Washington, D.C. Pfennig, D. 1990. The adaptive significance of an environmentally-cued developmental switch in an anuran tadpole. Oecologia 85:101–107. Pfennig, D. W. 1992a. Polyphenism in Spadefoot Toad Tadpoles as a Locally Adjusted Evolutionarily Stable Strategy. Evolution 46:1408–1420. Pfennig, D. W. 1992b. Proximate and functional causes of polyphenism in an anuran tadpole. Functional Ecology 6:167–174. Pfennig, D. W., and J. P. Collins. 1993. Kinship affects morphogenesis in cannibalistic salamanders. Nature 362:836–838. Pfennig, D. W., M. A. Wund, E. C. Snell-Rood, T. Cruickshank, C. D. Schlichting, and A. P. Moczek. 2010. Phenotypic plasticity’s impacts on diversification and speciation. Trends in Ecology and Evolution 25:459–467. Pierce, B. A., and H. M. Smith. 1979. Neoteny or Paedogenesis? Journal of Herpetology 13:119–

135

121. Pintar, M. R., and W. J. Resetarits. 2018. Filling ephemeral ponds affects development and phenotypic expression in Ambystoma talpoideum. Freshwater Biology 63:1173–1183. Polis, G. A. 1981. The evolution and dynamics of intraspecific predation. Annual Review of Ecology and Systematics 12:225–251. Poly, W. J. 2004. Family Aphredoderidae Bonaparte 1846: pirate perches. Pages 1–7 California Academy of Sciences Annotated Checklists of Fishes No. 24. California Academy of Sciences Annotated Checklists of Fishes no. 24, San Francisco. Poly, W. J., and G. S. Proudlove. 2004. Family Amblyopsidae Bonaparte 1846: cavefishes. Page 7 California Academy of Sciences Annotated Checklists of Fishes No. 25. California Academy of Sciences Annotated Checklists of Fishes no. 25, San Francisco. Pope, P. H. 1921. Some Doubtful Points in the Life-history of Notophthalmus viridescens. Copeia 91:14–15. Pope, P. H. 1924. The life-history of the common water-newt (Notophthalmus viridescens), together with observations on the sense of smell. Annals of Carnegie Museum 15:305–368. Pope, P. H. 1928. The life-history of Triturus viridescens - some further notes. Copeia 168:61– 73. Preisser, E. L., J. L. Orrock, and O. J. Schmitz. 2007. Predator Hunting Mode and Habitat Domain Alter Nonconsumptive Effects in Predator – Prey Interactions 88:2744–2751. Purrenhage, J. L., P. H. Niewiarowski, and F. B. G. Moore. 2009. Population structure of spotted salamanders (Ambystoma maculatum) in a fragmented landscape. Molecular Ecology 18:235–247. R Core Team. 2018. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria, Austria. Reilly, S. M. 1987. Ontogeny of the Hyobranchial apparatus in the salamanders Ambystoma talpoideum (Ambystomatidae) and Notophthalmus viridescens (Salamandridae): the ecological morphology. Journal of Morphology 214:205–214. Relyea, R. A., and E. E. Werner. 1999. Quantifying the relation between predator-induced behavior and growth performance in larval anurans. Ecology 80:2117–2124. Resetarits, W. J., and C. A. Binckley. 2013. Is the pirate really a ghost? Evidence for generalized chemical camouflage in an aquatic predator, pirate perch Aphredoderus sayanus. The American Naturalist 181:690–9. Resetarits, W. J., and A. Silberbush. 2016. Local contagion and regional compression: Habitat selection drives spatially explicit, multiscale dynamics of colonisation in experimental metacommunities. Ecology Letters 19:191–200. Richter, J. A. R., L. Martin, and C. K. Beachy. 2009. Increased Larval Density Induces Accelerated Metamorphosis Independently of Growth Rate in the Frog Rana sphenocephala. Journal of Herpetology 43:551–554. Roff, D. 1992. Evolution Of Life Histories: Theory and Analysis. Chapman & Hall, London. Rohr, J. R., D. Park, A. M. Sullivan, M. McKenna, C. R. Propper, and D. M. Madison. 2005. Operational sex ratio in newts: Field responses and characterization of a constituent chemical cue. Behavioral Ecology 16:286–293. Rood, F. H., and D. N. Reznick. 1997. Variation in the demography of guppy populations: The importance of predation and life histories. Ecology 78:405–418. Rose, F. L., and D. Armentrout. 1976. Adaptive strategies of Ambystoma tigrinum green inhabiting the Llano Estacado of West Texas. Journal of Animal Ecology 45:713–729.

136

Rot-Nikcevic, I., R. J. Denver, and R. J. Wassersug. 2005. The influence of visual and tactile stimulation on growth and metamorphosis in anuran larvae. Functional Ecology 19:1008– 1016. Rot-Nikcevic, I., C. N. Taylor, and R. J. Wassersug. 2006. The role of images of conspecifics as visual cues in the development and behavior of larval anurans. Behavioral Ecology and Sociobiology 60:19–25. Rudolf, V. H. W. 2012. Seasonal shifts in predator body size diversity and trophic interactions in size-structured predator-prey systems. Journal of Animal Ecology 81:524–532. Ryan, T. J., and G. R. Plague. 2004. Hatching asynchrony, survival, and the fitness of alternative adult morphs in Ambystoma talpoideum. Oecologia 140:46–51. Ryan, T. J., and R. D. Semlitsch. 1998. Intraspecific heterochrony and life history evolution : Decoupling somatic and sexual development in a facultatively paedomorphic salamander. Proceedings of the National Academy of Science of the United States of America 95:5643– 5648. Ryan, T. J., and R. D. Semlitsch. 2003. Growth and the expression of alternative life cycles in the salamander Ambystoma talpoideum (Caudata: Ambystomatidae). Biological Journal of the Linnean Society 80:639–646. Schmitz, O. J. 2012. Effects of Predator Functional Diversity on Grassland Ecosystem Function Effects of predator functional diversity on function grassland ecosystem. Ecology 90:2339– 2345. Schneider, C. A., W. S. Rasband, and K. W. Eliceiri. 2012. NIH Image to ImageJ: 25 years of image analysis. Nature Methods 9:671–675. Scott, D. E. 1990. Effects of larval density in Ambystoma opacum: an experiment in large-scale field enclosures. Ecology 71:296–306. Scott, D. E. 1993. Timing of reproduction of paedomorphic and metamorphic Ambystoma talpoideum. American Midland Naturalist 129:397–402. Semlitsch, R. 1985. Reproductive strategy of a facultatively paedomorphic salamander Ambystoma talpoideum. Oecologia 65:305–313. Semlitsch, R. D. 1987a. Paedomorphosis in Ambystoma talpoideum: effects of density, food, and pond drying. Ecology 68:994–1002. Semlitsch, R. D. 1987b. Density-dependent growth and fecundity in the paedomorphic salamander Ambystoma talpoideum. Ecology 68:1003–1008. Semlitsch, R. D. 1987c. Interactions between fish and salamander larvae - Costs of predator avoidance or competition? Oecologia 72:481–486. Semlitsch, R. D., and J. W. Gibbons. 1985. Phenotypic variation in metamorphosis and paedomorphosis in the salamander Ambystoma talpoideum. Ecology 66:1123–1130. Semlitsch, R. D., R. N. Harris, and H. M. Wilbur. 1990. Paedomorphosis in Ambystoma talpoideum: maintenance of population variation and alternative life-history pathways. Evolution 44:1604–1613. Semlitsch, R. D., and S. B. Reichling. 1989. Density-dependent injury in larval salamanders. Oecologia 81:100–103. Semlitsch, R. D., D. E. Scott, and J. H. K. Pechmann. 1988. Time and size at metamorphosis related to adult fitness in Ambystoma talpoideum. Ecology 69:184–192. Semlitsch, R., and H. Wilbur. 1989. Artificial selection for paedomorphosis in the salamander Ambystoma talpoideum. Evolution 43:105–112. Sexton, O. J., and C. Phillips. 1986. A qualitative study of fish-amphibian interactions in 3

137

Missouri ponds. Transactions of the Missouri Academy of Science 20:25–35. Shepherd, M. E., and M. T. Huish. 1978. Age, growth, and diet of the pirate perch in a coastal plain stream of North Carolina. Transactions of the American Fisheries Society 107:457– 459. Shoop, R. C. 1964. Ambystoma talpoideum (Holbrook) Mole Salamander. Page 8 in W. Riemer and H. Dowling, editors. Catalogue of American Amphibians and Reptiles. American Society of Ichthyologists and Herpetologists, Kensington, Maryland. Shute, J. 1980. Fundulus chrystotus (Günther), Golden topminnow. Page 510 in D. Lee, editor. Atlas of North American Freshwater Fishes. North Carolina State Musuem of Natural History, Raleigh. Skelly, D. K. 2004. Microgeographic countergradient variation in the wood frog, Rana sylvatica. Evolution 58:160–165. Smith, S. M., and W. W. Vale. 2006. The role of the hypothalamic-pituitary-adrenal axis in neuroendocrine responses to stress. Sprules, W. G. 1974. The adaptive significance of paedogenesis in North American species of Ambystoma (Amphibia: Caudata): an hypothesis. Canadian Journal of Zoology 52:393–400. Stearns, S. C. 1992. The Evolution of Life Histories. Oxford University Press, New York. Stearns, S. C., and J. C. Koella. 1986. The Evolution of Phenotypic Plasticity in Life-History Traits: Predictions of Reaction Norms for Age and Size at Maturity. Evolution 40:893. Storfer, A., J. Cross, V. Rush, and J. Caruso. 2006. Adaptive Coloration and Gene Flow as a Constraint to Local Adaptation in the Streamside Salamander, Ambystoma barbouri. Evolution 53:889. Sultan, S. E., and H. G. Spencer. 2002. Metapopulation structure favors plasticity over local adaptation. The American Naturalist 160:271–283. Suzuki, M. R., and S. Kikuyama. 1983. Corticoids Augment Nuclear Binding Capacity for Triiodothyronine in Bullfrog Tadpole Tail Fins. General and Comparative Endocrinology 52:272–278. Suzuki, S., and M. Suzuki. 1981. Changes in thyroidal and plasma lodine compounds during and after metamorphosis of the bullfrog, Rana catesbeiana. General and Comparative Endocrinology 45:74–81. Székely, D., M. Denoël, P. Székely, and D. Cogălniceanu. 2017. Pond drying cues and their effects on growth and metamorphosis in a fast developing amphibian. Journal of Zoology 303:129–135. Takahashi, M. K., and M. J. Parris. 2008. Life cycle polyphenism as a factor affecting ecological divergence within Notophthalmus viridescens. Oecologia 158:23–34. Takahashi, M. K., Y. Y. Takahashi, and M. J. Parris. 2011. Rapid Change in Life-Cycle Polyphenism across a Subspecies Boundary of the Eastern Newt, Notophthalmus viridescens. Journal of Herpetology 45:379–384. Thom, R. H., and J. H. Wilson. 1980. The Natural Divisions of Missouri: An Introuction to the Natural History of the State. Transactions of the Missouri Academy of Science 14:9–23. Torres-Dowdall, J., C. A. Handelsman, D. N. Reznick, and C. K. Ghalambor. 2012. Local adaptation and the evolution of phenotypic plasticity in trinidadian guppies (Poecilia reticulata). Evolution 66:3432–3443. Touchon, J. C., and K. M. Warkentin. 2008. Fish and dragonfly nymph predators induce opposite shifts in color and morphology of tadpoles. Oikos 117:634–640. Turchin, P. 1999. Population Regulation: A Synthetic View. Oikos 84:153–159.

138

Urban, M. C. 2010. Microgeographic adaptations of spotted salamander morphological defenses in response to a predaceous salamander and beetle. Oikos 119:646–658. Uvarov, B. P. 1921. A Revision of the Genus Locusta, L. (= Pachytylus, Fieb.), with a New Theory as to the Periodicity and Migrations of Locusts. Bulletin of Entomological Research 12:135–163. Verhulst, P. F. 1838. Notice sur la loi que la population suit dans son accroissement. Correspondance Mathematique et Physique 10:113–121. Volterra, V. 1926. Fluctuations in the abundance of a species considered mathematically. Nature 118:558–560. Voss, S. 1995. Genetic basis of paedomorphosis in the axolotl, Ambystoma mexicanum: a test of the single-gene hypothesis. Journal of Heredity 86:441–447. Vu, V. Q. 2011. ggbiplot: A ggplot2 based biplot. Walls, S. C. 1998. Density dependence in a larval salamander: The effects of interference and food limitation. Copeia 1998:926–935. Walls, S. C., and R. G. Jaeger. 1987. Aggression and exploitation as mechanisms of competition in larval salamanders. Canadian Journal of Zoology 65:2938–2944. Warton, D. I., and F. K. C. Hui. 2011. The arcsine is asinie: the analysis of porportions in ecology. Ecology 92:3–10. Wellborn, G. A., D. K. Skelly, and E. E. Werner. 1996. Mechanisms creating community structure across a freshwater habitat gradient. Annual Review of Ecological Systematics 27:337–363. Werner, E. E. 1986. Amphibian Metamorphosis: Growth Rate, Predation Risk, and the Optimal Size at Transformation. The American Naturalist 128:319–341. Werner, E. E., and J. F. Gilliam. 1984. The Ontogenetic Niche and Species Interactions in Size- Structured Populations. Annual Review of Ecology and Systematics 15:393–425. Werner, E. E., and S. D. Peacor. 2003. A review of trait- mediated indirect interactions in ecological communities. Ecology 84:1083–1100. Werner, E., and D. Hall. 1988. Ontogenetic Habitat Shifts in Bluegill: The Foraging Rate- Predation Risk Trade-off. Ecology 69:1352–1366. West-Eberhard, M. J. 1989. Phenotypic plasticity and the origins of diversity. Annual Review of Ecology and Systematics 20:249–278. West-Eberhard, M. J. 2003. Developmental plasticity and evolution. Oxford University Press, Oxford. West-Eberhard, M. J. 2005. Developmental plasticity and the origin of species differences. Proceedings of the National Academy of Sciences of the United States of America 102. Whiteman, H. H. 1994. Evolution of Facultative Paedomorphosis in Salamanders. The Quarterly Review of Biology 69:205–221. Whiteman, H. H., and R. D. Semlitsch. 2005. Asymmetric reproductive isolation among polymorphic salamanders. Biological Journal of the Linnean Society 86. Whiteman, H. H., S. A. Wissinger, M. Denoël, C. J. Mecklin, N. M. Gerlanc, and J. J. Gutrich. 2012. Larval growth in polyphenic salamanders: Making the best of a bad lot. Oecologia 168:109–118. Wickham, H. 2009. ggplot2: Elegant Graphics for Data Analysis. Springer-Verlag New York. Wilbur, H. M. 1977. Density-dependent aspects of growth and metamorphosis in Bufo americanus. Ecology 58:196–200. Wilbur, H. M. 1987. Regulation of structure in complex systems: experimental temporary pond

139

communities. Ecology 68:1437–1452. Wilbur, H. M. 1997. Experimental Ecology of Food Webs : Complex Systems in Temporary Ponds. Ecology 78:2279–2302. Wilbur, H. M., and J. P. Collins. 1973. Ecological Aspects of Amphibian Metamorphosis. Science. 182:1305–1314. Wilbur, H., P. Morin, and R. Harris. 1983. Salamander predation and the structure of experimental communities: anuran responses. Ecology 64:1423–1429. Wildy, E. L., D. P. Chivers, J. M. Kiesecker, and A. R. Blaustein. 2001. The effects of food level and conspecific density on biting and cannibalism in larval long-toed salamanders, Ambystoma macrodactylum. Oecologia 128:202–209. Wisenden, B. D. 2000. Olfactory assessment of predation risk in the aquatic environment. Philosophical Transactions of the Royal Society B: Biological Sciences 355:1205–1208. Wissinger, S. A., H. H. Whiteman, M. Denoel, M. L. Mumford, and C. B. Aubee. 2010. Consumptive and nonconsumptive effects of cannibalism in fluctuating age-structured populations. Ecology 91. Woltereck, R. 1913. Weitere experimentelle untersuchungen über Artänderung, speziell über das Wesen quantitativer Artunterschiede bei Daphniden. Zeitschrift für induktive Abstammungs- und Vererbungslehre 9:146. Wright, M. L., L. J. Cykowski, L. Lundrigan, K. L. Hemond, D. M. Kochan, E. E. Faszewski, and C. M. Anuszewski. 1994. Anterior pituitary and adrenal cortical hormones accelerate or inhibit tadpole hindlimb growth and development depending on stage of spontaneous development or thyroxine concentration in induced metamorphosis. Journal of Experimental Zoology 270:175–188. Wright, S. 1931. Evolution in Mendelian populations. Genetics 16:97–159. Zamudio, K. R., and A. M. Wieczorek. 2007. Fine-scale spatial genetic structure and dispersal among spotted salamander (Ambystoma maculatum) breeding populations. Molecular Ecology 16:257–274.

140

APPENDIX A

141

Table A.1. Summary of alternative models in lme4 notation. Model ID Model Efts vs. Non-Efts (binomial GLMM)† 1 Y ~ Density * Food + Survival + (1|Block/Mesocosm)+(1|BreedingMesocosm) 2 Y ~ Density * Food + Survival + (1|Block/Mesocosm) 3 Y ~ Density * Food + (1|Block/Mesocosm) Paedomorphs vs. Non-Paedomorphs (LMM)‡ 1 Y ~ Density * Food + Survival + (1|Block) 2 Y ~ Density * Food + (1|Block) Larval period, growth rate, SVL and body condition 1 Y ~ Density * Food + Survival + (1|Block/Mesocosm)+(1|BreedingMesocosm) 2 Y ~ Density * Food + Survival + (1|Block/Mesocosm) 3 Y ~ Density * Food + (1|Block/Mesocosm)+(1| BreedingMesocosm) 4 Y ~ Density * Food + (1|Block/Mesocosm) SVL repeated measures 1 Y ~ Density * Food * Time + Survival + (1|Block/Mesocosm)+(1|ID)+(1|BreedingMesocosm) 2 Y ~ Density * Food * Time + Survival + (1|Block/Mesocosm)+(1|ID) 3 Y ~ Density * Food * Time + (1|Block/Mesocosm)+(1|ID)+(1|BreedingMesocosm) 4 Y ~ Density * Food * Time + (1|Block/Mesocosm)+(1|ID) Only the factors in parentheses are random effects † Failure to converge of model “Y ~ Density * Food + (1|Block/Mesocosm) + (1| BreedingMesocosm),” which was excluded. ‡ This analysis was conducted on aggregated data (proportions) and necessarily excluded Mesocosm and Breeding Mesocosm terms.

142

Table A.2. Summary of complete model parameters † Survival (binomial GLMM) Model: Y ~ Density * Food + (1|Block/Mesocosm) Source Estimate SE z p (>|z|) Fixed effects Intercept 1.686 0.487 3.46 <0.001 Density 0.582 0.778 0.75 0.454 Food −0.351 0.700 −0.50 0.616 Density × Food −0.644 1.020 −0.63 0.528 Random effects Variance SE Mesocosm × Block 0.000 0.000 Block 0.000 0.000

Efts vs. Non-Efts (binomial GLMM) Model: 3 AICc: 101.1 Δi AICc: 0.92 Source Estimate SE z p (>|z|) Fixed effects Intercept −0.355 0.454 −0.78 0.434 Density −2.321 0.850 −2.73 0.006 Food −0.444 0.657 −0.68 0.499 Density × Food 0.579 1.230 0.47 0.638 Random effects Variance SE Mesocosm × Block ~0.000 ~0.000 Block 0.183 0.427

Paedomorphs vs. Non-Paedomorphs (LMM) Model: 2 AICc: 12.6 Δi AICc: 8.41 Source Estimate SE df t p (>|t|) Fixed effects Intercept 0.107 0.047 11 2.27 0.044 Density −0.107 0.067 11 −1.60 0.137 Food −0.007 0.072 11 0.10 0.923 Density × Food 0.035 0.098 11 0.35 0.732 Random effects Variance SE Block 0.000 0.000 Residuals 0.009 0.094

Larval period (days) Model: 1 AICc: 852.0 Δi AICc: 3.38 Source Estimate SE df t p (>|t|) Fixed effects Intercept 107.138 15.007 86.56 7.14 <0.001 Density −18.019 4.791 88.44 −3.76 <0.001 Food −11.344 5.303 91.82 −2.14 0.035 Survival −16.404 16.557 90.05 −0.99 0.324 Density × Food 2.797 7.296 94.20 0.38 0.702 Random effects Variance SE Mesocosm × Block ~0.000 ~0.000 Breeding Mesocosm 48.010 6.929 Block 22.370 4.730 Residuals 268.890 16.400

143

SVL (mm) Model: 1 AICc: 442.8 Δi AICc: 2.27 Source Estimate SE df t p (>|t|) Fixed effects Intercept 23.027 3.050 8.81 7.55 <0.001 Density −3.926 1.038 6.90 −3.78 0.007 Food 0.229 1.125 7.51 0.20 0.844 Survival 2.430 3.448 8.12 0.71 0.501 Density × Food 1.886 1.537 7.43 1.23 0.257 Random effects Variance SE Mesocosm × Block 1.482 1.217 Breeding Mesocosm 0.612 0.782 Block 0.632 0.795 Residuals 3.316 1.821

‡ Growth rate (mm SVL/day) Model: 1 AICc: 615.3 Δi AICc: 0.44 Source Estimate SE df t p (>|t|) Fixed effects Intercept 23.250 4.384 14.26 5.30 <0.001 Density 0.081 1.474 10.49 0.06 0.957 Food 3.719 1.622 12.14 2.29 0.041 Survival 6.082 4.898 12.93 1.24 0.236 Density × Food 2.629 2.226 12.10 1.18 0.260 Random effects Variance SE Mesocosm × Block 0.165 0.406 Breeding Mesocosm 3.371 1.836 Block 0.000 0.000 Residuals 24.817 4.982

‡ Residuals body condition approach Model: 2 AICc: 877.2 Δi AICc: 2.42 Source Estimate SE df t p (>|t|) Fixed effects Intercept −13.596 21.183 11.85 −0.64 0.533 Density −0.292 7.270 9.55 −0.04 0.969 Food −4.610 7.886 10.34 −0.59 0.571 Survival 16.211 23.978 11.22 0.68 0.513 Density × Food 12.499 10.612 9.66 1.18 0.267 Random effects Variance SE Mesocosm × Block 34.440 5.869 Block 0.000 0.000 Residuals 430.910 20.758

‡ Covariate body condition approach Model: 2 AICc: 676.0 Δi AICc: 2.48 Source Estimate SE df t p (>|t|) Fixed effects Intercept −77.742 10.450 26.53 −7.44 <0.001 Density −0.313 2.826 12.33 −0.11 0.914 Food −1.551 2.735 9.47 −0.57 0.584

144

SVL 4.587 0.328 64.25 13.97 <0.001 Survival 5.659 8.350 10.22 0.68 0.513 Density × Food 4.349 3.727 9.16 1.17 0.273 Random effects Variance SE Mesocosm × Block 4.500 2.120 Block 0.000 0.000 Residuals 49.440 7.031

SVL (mm) repeated measures Model: 2 AICc: 1305.1 Δi AICc: 0.86 Source Estimate SE df t p (>|t|) Fixed effects Intercept 16.418 2.945 8.25 5.58 <0.001 Density −0.699 1.098 11.16 −0.64 0.537 Food −1.478 1.175 10.91 −1.26 0.235 Time−Day 51 5.415 0.668 171.90 8.10 <0.001 Time−Final 0.155 0.653 183.16 14.02 <0.001 Survival −0.501 3.192 6.59 −0.16 0.880 Density × Food 1.256 1.572 10.27 0.80 0.442 Density × Time−Day 51 −1.306 0.871 172.14 −1.50 0.135 Density × Time−Final −3.476 0.867 180.74 −4.01 <0.001 Food × Time−Day 51 1.014 0.967 174.27 1.05 0.296 Food × Time−Final 1.843 0.970 187.96 1.90 0.059 Density × Food × Time−Day 51 0.131 1.253 173.87 0.11 0.917 Density × Food × Time−Final 0.014 1.264 183.06 0.01 0.991 Random effects Variance SE ID 0.914 0.956 Mesocosm × Block 1.131 1.145 Block 3.226 1.796 Residuals 4.330 2.081 Models used default treatment contrasts, which sets the first level of factors (Low Density, Ambient Food and Day 30) as the reference level and then compares with the additional factor levels (High Density, Supplemented Food, Time−Day 51 and Time−Final) of the listed fixed effects. Δi = difference in AICc from second lowest model SE = standard error Bold = Significant † Survival was modeled on aggregated data using the methods of Warton and Hui (2011). ‡ × 100 to reduce variance decimal places

145

VITA

JASON R. BOHENEK

EDUCATION 2019 Ph.D. Biology, The University of Mississippi, Oxford, MS Advisor: William J. Resetarits Jr. 2012 B.Sc. Biology, Environmental Science, and Philosophy The University of Scranton, Scranton PA Advisor: Robert J. Smith

PUBLICATIONS Bohenek, J. R. and W. J. Resetarits Jr. (in prep for Ecology). Predator functional diversity drives patterns of life history responses of prey (Ambystoma talpoideum). Bohenek, J. R., C. J. Leary, & W. J. Resetarits Jr. (in review at Proceedings of the Royal Society B). Stress hormone regulates life history strategies in a facultatively paedomorphic salamander. Bohenek, J. R., M. R. Pintar, T. M. Breech, & W. J. Resetarits Jr. (in review at Oecologia). Predation risk vs. risk perception: larval tree frogs have divergent responses to convergent risk from fish predators Resetarits, W. J. Jr., M. R. Pintar J. R. Bohenek, & T. M. Breech. (accepted at The American Naturalist). Patch size and predation risk generate behavioral species sorting in colonizing aquatic insects. Bohenek, J. R. & W. J. Resetarits Jr. 2018. Are direct density cues, not resource competition, driving life history trajectories in a polyphenic salamander? Evolutionary Ecology 32: 335–357. Pintar, M. R., J. R. Bohenek, L. L. Eveland, & W. J. Resetarits Jr. 2018. Colonization across gradients of risk and reward: nutrients and predators generate species-specific responses among aquatic insects. Functional Ecology 32: 1589–1598. Resetarits, W. J. Jr., J. R. Bohenek, T. M. Breech, & M. R. Pintar. 2018. Predation risk and patch size jointly determine perceived patch quality in ovipositing treefrogs, Hyla chrysoscelis. Ecology 99: 661–669. Bohenek, J. R. & W. J. Resetarits Jr. 2017. An optimized method to quantify large numbers of amphibian eggs. Herpetology Notes 10:573–578. Bohenek, J. R., M. R. Pintar, T. M. Breech, & W. J. Resetarits Jr. 2017. Patch size influences perceived patch quality for colonising Culex mosquitoes. Freshwater Biology 62:1614– 1622.

146

Eveland, L. L., J. R. Bohenek (equal co-author), A. Silberbush, & W. J. Resetarits Jr. 2016. Detection of fish and newt kairomones by ovipositing mosquitoes. Pages 247–259 in B. A. Schulte, T. E. Goodwin, and M. H. Ferkin, eds. Chemical Signals in Vertebrates 13. Springer, Cham, Switzerland. Bohenek J. R. 2013. A Comparative Analysis of Nietzsche’s Will to Power and Darwin’s Natural Selection Theory. Annual Undergraduate Philosophy Journal Discourse at the University of Scranton

ORAL PRESENTATIONS Bohenek, J. R. 2019. “Extended Adolescence: the Ecophysiology of Facultative Paedomorphosis.” The Department of Biology Dissertation Defense. April 24, Oxford, MS. Bohenek, J. R., C. J. Leary, & W. J. Resetarits Jr. 2018. “Corticosterone prevents a paedomorphic life history strategy in the salamander Ambystoma talpoideum” The University of Mississippi Field Station Science Conference, April 21, Lafayette County, MS. Bohenek, J. R., C. J. Leary, & W. J. Resetarits Jr. 2017. “Corticosterone prevents a paedomorphic life history strategy in the salamander Ambystoma talpoideum.” Ecological Society of America, August 6–11, Portland, OR Bohenek, J. R. & W. J. Resetarits Jr. 2016. “Density-dependent polyphenisms are mediated by chemical cues in Notophthalmus viridescens louisianensisis.” Joint Meeting of Ichthyologists and Herpetologists, July 7–10, New Orleans, LA. Bohenek, J. R. & W. J. Resetarits Jr. 2016. “A density-dependent polyphenism in Notophthalmus viridescens louisianensisis is mediated by chemical cues.” Evolution, June 17–21, Austin, TX. Bohenek, J. R. 2015. “Cue Modality and Ecological Significance in Larval Anurans and Polyphenic Salamanders.” University of Mississippi Department of Biology Prospectus Defense, Nov. 12, Oxford, MS. Bohenek, J. R., Eveland LL & W. J. Resetarits Jr. 2014. “The influence of newts (Notophthalmus viridescens louisianensis) on oviposition site choice of mosquitoes.” Joint Meeting of Ichthyologists and Herpetologists, July 30–Aug 3, Chattanooga, TN. Bohenek, J. R. 2012. “A Comparative Analysis of Nietzsche’s Will to Power and Darwin’s Natural Selection Theory.” Phi Sigma Tau, Penn-Tau Chapter Forum. The University of Scranton, Scranton, PA

POSTER PRESENTATIONS Bohenek, J. R. & W. J. Resetarits Jr. 2018. “The ecology of polyphenisms: density-dependence, predation threat and hydroperiod as drivers of alternative morphs in a facultatively paedomorphic salamander.” Ecological Society of America, August 6–11, New Orleans, LA. Bohenek, J. R., C. J. Leary, W. J. Resetarits Jr. 2017. “Corticosterone prevents paedomorphic life history strategy in salamanders.” GSC 7th Annual Research Symposium, March 2, Oxford, MS. Eveland, L. L., J. R. Bohenek (equal co-author), A. Silberbush & W. J. Resetarits Jr. 2014. The effects of different aquatic predators on mosquito oviposition site choice. International

147

Society for Chemical Ecology - Chemical Signals in Vertebrates meeting (ISCE-CSiV), July 8-12, Urbana-Champaign, IL. Bohenek, J. R. & W. J. Resetarits Jr. 2013. “Search for Proximate Environmental Cues of Polyphenism in Facultatively Paedomorphic Newts.” Texas Tech Annual Biological Sciences Symposium. Texas Tech University, Lubbock, TX Bohenek, J. R., K. M. Druther, R. J. Smith. 2012. “The Effects of Invasive Shrubs in Early Successional Habitats on Migratory Birds.” Twelfth Annual Celebration of Student Scholars. The University of Scranton, Scranton, PA

148