University of Mississippi University of Mississippi

eGrove eGrove

Electronic Theses and Dissertations Graduate School

2018

Impacts And Implications Of Polytypism On The Evolution Of Impacts And Implications Of Polytypism On The Evolution Of

Aposematic Coloration Aposematic Coloration

Justin Philip Lawrence University of Mississippi

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

Part of the Ecology and Evolutionary Biology Commons

Recommended Citation Recommended Citation Lawrence, Justin Philip, "Impacts And Implications Of Polytypism On The Evolution Of Aposematic Coloration" (2018). Electronic Theses and Dissertations. 382. https://egrove.olemiss.edu/etd/382

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 egrove@olemiss.edu.

IMPACTS AND IMPLICATIONS OF POLYTYPISM ON THE EVOLUTION OF

APOSEMATIC COLORATION

A Dissertation

presented in partial fulfillment of requirements

for the degree of Doctor of Philosophy

in the Department of Biology

The University of Mississippi

By

Justin Philip Lawrence

May 2018

Copyright Justin Philip Lawrence 2018

ALL RIGHTS RESERVED

ii

ABSTRACT

Aposematic signaling is a common defensive strategy whereby prey species use

conspicuous signals (i.e., bright coloration) to warn predators of the risks of predation due to a

secondary defense. Theoretical, lab, and field experiments have demonstrated that individuals

who project novel conspicuous signals should experience disproportionately high predation

pressure as predators will not associate novel signals with a secondary defense. Thus, aposematic

signaling should be subject to strong positive frequency-dependent selection (FDS). Numerous

species show considerable intrapopulation variation (polymorphism) in direct conflict with the

expectations of FDS. Interpopulation variation (polytypism) can adhere to expectations of FDS,

but as its origins likely involve polymorphism, FDS likely plays a role in the establishment of

such populations. Both the Dyeing Poison Frog (Dendrobates tinctorius) and the Australian

Brood Frogs (Pseudophryne) exhibit considerable inter- and intrapopulation variation making

them excellent candidates to investigate how aposematic signaling can evolve and under what

circumstances of FDS can be relaxed. Consequently, I approach this question by systematically

investigating how predators perceive conspicuous colors, how secondary defense influences

predator response, how predators respond to known and novel colors, and the role that gene flow

plays in promoting or limiting phenotypic divergence. By using model, naïve predators for

learning experiments, I found that the hue color component influences predators’ abilities to

learn to avoid an aposematic signal. I provide the first among-population characterization of

alkaloid toxins in D. tinctorius and, when using a model avian predator to examine unpalatability

of alkaloid toxins, found that a subset of these alkaloids is driving the predator response. In

iii

examining how predators respond to known and novel phenotypes, I elucidate the function of

conspicuous signals in Pseudophryne and how predators may generalize to novel signals.

Finally, I find that despite field and lab experiments that indicate a selective disadvantage of a

weak aposematic signal, it can persist when isolated with limited gene flow. Together, these

studies provide evidence for the evolution of aposematic signals and propose mechanisms that

can allow the relaxation of FDS and thus allow for phenotypic polymorphism in aposematic

species.

iv

DEDICATION

This dissertation is dedicated to all of those along the way who have helped me prepare

for this advancement in my life. In particular, I would like to thank my parents, Wade and Kim

Lawrence, for their unwavering support through all of the travels abroad. Additionally, my

grandparents, Keith and Marlene Lawrence and Ken and Jean Case, have been integral to making

me the person I am today. And finally, to my siblings Stacey and Austin Lawrence, for pushing

me to be the best person I can be.

v

LIST OF ABBREVIATIONS AND SYMBOLS

θ Theta, mutation-scaled population size

1,4-Q 1,4-disubstituted quinolizidine

3RAD 3-enzyme Restriction-Site Associated Digestion

3,5-I 3,5-disubstituted indolizidine

3,5-P 3,5-disubstituted pyrrolizidine

4,6-Q 4,6-disubstituted quinolizidine

5,6,8-I 5,6,8-trisubstituted indolizidine

5,8-I 5,8-disubstituted indolizidine

ACEC Animal Care and Ethics Committee

ANOVA Analysis of Variance

aPTX Allopumiliotoxin

bp Base Pairs

CI Chemical Ionization

CNRS Centre National de la Recherche Scientifique

ddRAD Double Digest Restriction-Site Associated Digestion

Dehydro-5,8-I Dehyrdo-5,8-disubstituted indolizidine

DHQ Decahydroquinoline

DNA Deoxyribonucleic Acid

FDS Frequency Dependent Selection

GC-MS Gas Chromatography-Mass Spectrometry

vi

GLM Generalized Linear Model

GLMM Generalized Linear Mixed Model

G-PhoCS Generalized Phylogenetic Coalescent Sampler

HTX Histrionicotoxin

IACUC Institutional Animal Care and Use Committee

JND Just Noticeable Difference

LD50 Lethal Dose of 50% of test subjects

m Migration

MCMC Markov Chain Monte Carlo

NEB New England Biolabs

PCA Principle Components Analysis

PCR Polymerase Chain Reaction

Pip Piperidine

SNP Single Nucleotide Polymorphism

Spiro Spiropyrrolizidine

Tri Tricyclic

Unclass Unclassified alkaloid

vii

ACKNOWLEDGEMENTS

Many people deserve thanks and acknowledgement in the making of this dissertation.

Thanks to my adviser, Brice Noonan, for always being enthusiastic about my research and

challenging me to dig deeper into my questions. Thanks to my committee members, Ryan

Garrick, Colin Jackson, John Rimoldi, and Rebecca Symula for helping guide me through the

biggest endeavor of my life as well as handling my constant questioning with grace and patience.

Thanks to Cori Richards-Zawacki for temporarily serving on my committee and for pushing me

to approach questions analytically. Thank you to the staff in the Biology Department for making

my stay here enjoyable and the transition through my dissertation smooth. And to all of the

graduate students in the Biology Department, thank you for making my tenure here enjoyable

and for sharing the trials and tribulations associated with being graduate students. You all are the

ones who truly helped me get through all of this. Finally, I would like to thank the numerous

undergraduate students who aided me through the years in collecting and analyzing data at the

University of Mississippi. Your assistance is the backbone of my research.

This research involved many people in many different countries, and I would be remiss to

mention the people who helped me abroad completing this research. In French Guiana, my

research involved the aid of many people. We were awarded a CEBA research grant from the

Centre National de la Recherche Scientifique (CNRS) which involved the collaboration of

researchers from CNRS (Antoine Fouquet and Elodie A. Courtois) and the University of

Jyväskylä (Bibiana Rojas and Johanna Mappes). Antoine was a particular asset as he helped

coordinate French volunteers (Madeline Segers and Antoine Baglan) as well as collect tissues.

viii

Additionally, Tim Muhich and John Constan were invaluable in their assistance in making and

deploying clay models while in the country.

In Australia, I was awarded a National Science Foundation East Asia and Pacific Summer

Institutes Fellowship that involved collaborating with Michael Mahony of the University of

Newcastle. His assistance in hosting me as well as recommending where to place clay models

ensured a particularly successful season. I would be remiss to mention the aid of Allira Jackman

who helped place models while in the Watagans as well as ensure that my transition to living in

Australia was smooth. The following year, I was awarded an Endeavour Fellowship which

allowed me to collaborate with Kate Umbers at Western Sydney University. Her enthusiasm and

assistance in allowing me to travel all around Australia to conduct research was invaluable to my

experiences there. It was a pleasure to be able to work with Michael Kelly and Griffin Taylor-

Dalton up in the subalpine regions of Australia, and with Stewart MacDonald in Western

Australia. Without all of their help, my successes in Australia would not have been possible.

ix

TABLE OF CONTENTS

ABSTRACT .................................................................................................................................... ii

DEDICATION ............................................................................................................................... iv

LIST OF ABBREVIATIONS AND SYMBOLS ........................................................................... v

ACKNOWLEDGEMENTS .......................................................................................................... vii

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

LIST OF FIGURES OR ILLUSTRATIONS ................................................................................. xi

I. INTRODUCTION ....................................................................................................................... 1

II. AVIAN LEARNING FAVORS COLORFUL, NOT BRIGHT, SIGNALS ............................ 12

III. ALKALOID VARIABILITY IN THE DYEING POISON FROG AND ITS IMPORTANCE

TO PREDATOR RESPONSE ...................................................................................................... 25

IV. DIFFERENTIAL RESPONSES OF AVIAN AND MAMMALIAN PREDATORS TO

PHENOTYPIC VARIATION IN AUSTRALIAN BROOD FROGS .......................................... 43

V. WEAK WARNING SIGNALS CAN PERSIST IN THE ABSENCE OF GENE FLOW ...... 54

VI. CONCLUSIONS .................................................................................................................... 84

LIST OF REFERENCES .............................................................................................................. 87

VITA ........................................................................................................................................... 101

x

LIST OF TABLES

Table 1: Contributions and funding sources for manuscripts from this dissertation .......... 11

Table 2: Alkaloid variation seen in the Matoury and Kaw populations ................................ 36

Table 3: Distribution of attacks for each of the six different models for birds and mammals

....................................................................................................................................................... 48

Table 4: Summary table of the questions, hypotheses, and experiments used for this study

....................................................................................................................................................... 58

Table 5: Distribution of alkaloids in the white (*) and yellow (†) populations ..................... 75

Table 6: Summary table of the questions and results of our experiments ............................ 77

xi

LIST OF FIGURES OR ILLUSTRATIONS

Figure 1: Phenotypic variation seen in the Dyeing Poison Frog (Dendrobates tinctorius) .... 5

Figure 2: Dorsal and ventral phenotypic variation seen in the Australian Brood Frogs

(Pseudophryne) ............................................................................................................................. 7

Figure 3: Aposematic signaling has been documented in a wide variety of taxa .................. 14

Figure 4: Reflectance curves for each color signal. ................................................................. 15

Figure 5: Mean Just Noticeable Differences (JNDs) between different signal comparisons

for both A) chromatic and B) achromatic contrasts. ............................................................... 19

Figure 6: A) Proportion learned and B) mean hesitation time of chicks (Gallus gallus

domesticus) for each of the three signals ................................................................................... 20

Figure 7: Principle Components Analysis (PCA) of the Kaw and Matoury populations of

Dendrobates tinctorius ................................................................................................................. 33

Figure 8: Distribution of alkaloids between the two populations ........................................... 39

Figure 9: Dorsal and ventral phenotypes represented in the models ..................................... 45

Figure 10: Attack proportions for birds and mammals on dorsal and ventral reticulated

phenotype ..................................................................................................................................... 50

Figure 11: Distribution, predation, and spectrometry of the white (A) and yellow (B)

populations................................................................................................................................... 71

Figure 12: Results of the learning experiments for white and yellow models ....................... 72

Figure 13: Proportion of chickens that generalized to a novel signal .................................... 74

Figure 14: Alkaloid quantities of the two populations ............................................................ 76

Figure 15: Results of unpalatability experiments using Blue Tits (Cyanistes caeruleus) ..... 78

Figure 16: Percentage of oats eaten per second in the assays done in A) 2015 and B) 2017 79

1

I. INTRODUCTION

Origins and Mechanisms of Biological Diversity

Inter- and intraspecific communication can employ a number of sensory modalities to

convey signals, and the modality used often depends on the environment through which the

signal is transmitted. Sensory modalities used to investigate this include visual, auditory,

chemosensory, and tactile cues. Regardless of the sensory modality used, natural selection

should ensure that receivers understand messages with little opportunity for confusion. With the

diversity of visual systems that have evolved throughout the animal kingdom, visual

communication is a common methods of conveying signals to conspecifics and heterospecifics.

Visual communication can involve behavioral and/or morphological displays to convey

any number of messages. These signals are employed to communicate everything from sexual

receptivity and mate quality to territoriality and predator deterrence. Sexual selection is

commonly invoked as a driver of the evolution of a broad array of sexual ornaments and modes

of communication. A large number of taxa, including birds (Barraclough et al. 1995), insects

(Thornhill 1976), mammals (Promislow 1992), fishes (Seehausen et al. 1999), amphibians (Ryan

1983), and reptiles (Stuart-Fox and Ord 2004), bear ornaments that are the produce of sexual

selection. For instance, in Anolis lizards, many species display colorful dewlaps as an

energetically-effective ornament that is used to defend territories and to attract mates. In each

species, these dewlaps have evolved to transmit this colorful signal through their specific

2

environment (i.e., understory, canopy; Leal and Fleishman 2004). As in other species, the dewlap

ornament is a colorful and critical component of the signal used to convey messages within and

among species.

A large majority of animals (and plants) employ coloration to communicate to

conspecifics and heterospecifics. Birds most commonly display bright or gaudy colors to attract

mates (and will often lose that coloration in nonbreeding seasons). Big cats use cryptic coloration

to hide themselves from detection by prey (Allen et al. 2011). Many insects (Olofsson et al.

2012, Umbers and Mappes 2015), amphibians (Martins 1989, Lenzi-Mattos et al. 2005), and

cephalopods (Langridge 2009, Staudinger et al. 2011) display coloration that is thought to startle

predators of these organisms, thereby giving them time to escape (Umbers et al. 2015). Though

the uses of color are varied (i.e., sexual display, territoriality), coloration is used in defense in

two different ways: crypsis and aposematism.

Defensive coloration occurs along a spectrum from cryptic to conspicuous, and

interestingly, where a particular color and/or pattern lies on this spectrum can depend on the

observer (Maan and Cummings 2012, Dreher et al. 2015) and circumstances (Rojas et al. 2014b)

allowing prey species to use multiple defensive strategies (i.e., crypsis, aposematism) to increase

survival. Research has largely focused on the ends of this spectrum, from cryptic camouflage to

conspicuous warning coloration. However, a number of examples exist where signals are

somewhere in the middle, displaying both conspicuous and cryptic characteristics (Rojas et al.

2014b).

A particularly interesting portion of the spectrum is the use of conspicuous coloration as a

warning to heterospecifics (aposematism). Aposematism is bright or conspicuous coloration that

3

is coupled with unprofitability of a prey species (Tullberg et al. 2005, Rojas et al. 2014b). This

unprofitability most commonly is in the form of a chemical defense (e.g., alkaloids, cardiac

glycosides, and venoms; (Savage and Slowinski 1992, Nishida 2002, Saporito et al. 2011), but

can also include noxious smells (Stankowich et al. 2011) or physical defenses like spines (Speed

and Ruxton 2005) and hard exoskeletons (Wang et al. 2018). How the defensive strategy

originates and is maintained remains perplexing as aposematic signals should experience strong

positive frequency-dependent selection (FDS; Endler and Greenwood 1988, Alatalo and Mappes

1996, Mappes et al. 2005). Thus, selection should not only make the persistence of a new

conspicuous coloration difficult but also limit variability in signal within prey species after

conspicuous coloration has increased in frequency. Despite this, aposematism has been

implicated in increased diversification rates in poison frogs (Coloma et al. 2014). With the

diversity of aposematic signals seen throughout the animal kingdom, understanding its

proliferation, despite expectations of positive FDS operating against it in the early stages, is an

important concept to investigate.

Lab and theoretical models have shown that experienced predators should limit

intrapopulation variation in aposematic signal (Lindström et al. 1999, 2001). Novel signals are

unknown to experienced predators and are unlikely to be associated with previous negative

experiences. Thus, predators should disproportionately target novel prey. Therefore, when a

signal is aposematic, differentiation of a signal will be constrained by experienced predators, and

polymorphism (intrapopulation variation) should be selected against. As polymorphism is likely

a precursor to polytypism (interpopulation variation)(McLean and Stuart-Fox 2014), reconciling

4

how polymorphism can persist is a pivotal concept in understanding how aposematic signals can

differentiate and result in phenotypically distinct populations.

Interestingly, while aposematism is a defensive strategy, it is thought to be beneficial to

experienced predator and prey. From the predator’s perspective, learning avoidance of

conspicuous coloration reduces the need for expending energy necessary to capture unprofitable

prey. From the prey’s perspective, predator avoidance reduces the risk of predation. While not

universal, other defensive interactions are often antagonistic to at least one species and incur

appropriate costs (e.g., parasitism, predation). Aposematism often operates through optimal

foraging theory (Endler and Greenwood 1988, Mappes et al. 2005) where predators must make

decisions of how much risk is worthwhile given the cost (unprofitability). In some instances,

predators are locked in arms races with defended prey, and locally-adapted predators can cope

with defended prey with minimal consequence while non-local individuals of the same species

cannot (Geffeney et al. 2002).

Despite expectations of positive FDS, aposematic species can be quite variable within

and among populations (Mallet and Joron 1999, Siddiqi et al. 2004, Beukema et al. 2016). Three

hypotheses have been put forth to explain this intraspecific variation. First, predators may

generalize particular components of aposematic signal (Kikuchi and Pfennig 2010, Pfennig and

Kikuchi 2012). When a predator recognizes a range of signals with different components as

equivalent warning signals, it generalizes the signal. Thus, generalization can allow phenotypic

variants to elicit similar avoidance by predators. Second, variable secondary defenses may result

in signals being honest or dishonest predictors of those defenses (Blount et al. 2009, Speed et al.

2010, Zollman et al. 2013). Third, population isolation and genetic drift can counter species-wide

5

expectations of positive frequency-dependent selection homogenizing aposematic signals (Mallet

and Joron 1999, Tazzyman and Iwasa 2010, Chouteau and Angers 2012).

Poison Frogs as Models of Aposematic Signaling Evolution

Poison frogs (Dendrobatidae) are among the most recognizable examples of animals with

warning signaling. As expected with aposematic animals, poison frogs have brightly colored-

bodies, defensive alkaloid skin toxins (Ruxton et al. 2004). Interestingly, a large number of

poison frog species display aposematic polytypism (interpopulation variation) and polymorphism

(intrapopulation variation; Richards-Zawacki et al. 2013). As aposematic signals are expected to

experience strong positive frequency-dependent selection (Endler and Greenwood 1988) novel

are predicted to be rare within a population (Lindström et al. 2001). Among-population variation

could fit within the paradigm of positive FDS if each population has a unique predator

community, but as polymorphism is a likely precursor to polytypism, positive FDS plays an

important role in the evolution of interpopulation variation. The fact that poison frog species

show aposematic polytypism and polymorphism,

despite evolutionary expectations, they offer a

unique opportunity to investigate how

aposematic signals evolve and how novel signals

proliferate. Indeed, much of our understanding of

aposematic polytypism in poison frogs has come

from research conducted on the Strawberry

Poison Frog (Oophaga pumilio). This species’

extraordinary phenotypic radiation in Bocas del

Figure 1: Phenotypic variation seen in the

Dyeing Poison Frog (Dendrobates tinctorius)

6

Toro, Panama has aided our understanding of how sexual selection (Summers et al. 1999, Maan

and Cummings 2008, Richards-Zawacki and Cummings 2011), toxin variation (Maan and

Cummings 2012), and predator learning (Saporito et al. 2007c, Hegna et al. 2012, Preißler and

Pröhl 2017) can constrain or promote phenotypic polytypism.

While the entire spectrum of the rainbow being represented on a frog is an exciting and

interesting phenomenon (Siddiqi et al. 2004), it does pose potential issues when it comes to

identifying trends in the evolution of aposematic signaling, chief among them the lack of

independent replicates of aposematic signals. This is where the Dyeing Poison Frog

(Dendrobates tinctorius) works well and why it is the focus of this dissertation. Dendrobates

tinctorius is a large dendrobatid species found in the Guiana Shield of South America. Across its

range, it shows considerable polytypism (Rojas and Endler 2013). It has four primary base colors

(yellow, white, blue, and black; Figure 1) represented in varying amounts across populations and

a variety of patterns of stripes and spots, which have repeatedly evolved over the course of its

evolutionary history (Noonan and Gaucher 2006). This variety in color and pattern facilitates the

exploration of how aposematism evolves, and how signals diversify. Prior research has shown

that the color and pattern of D. tinctorius is aposematic and can be cryptic to predators from a

distance (Rojas et al. 2014b, 2014a).. Different colored morphotypes even behave differently

(Rojas et al. 2014a). This multifunctional use of color, pattern, and behavior makes D. tinctorius

an ideal system to investigate how conspicuous signals can first evolve and yet do not experience

strong positive FDS selection (Lindström et al. 2001).

The fact that dendrobatid frogs sequester alkaloid toxins from their diet (Hantak et al.

2013) creates some difficulties for investigating phenotypic diversification in aposematic signals.

7

Aposematic polytypism has captured the imagination of researchers as it can help us understand

how this ubiquitous and perplexing defensive strategy can evolve.

Controlling for Toxin Variation

One of the prevailing hypotheses,

particularly for dendrobatid color evolution,

is that of honest signaling (Blount et al.

2009, Speed et al. 2010, Maan and

Cummings 2012). Briefly, honest signaling

occurs when the signal is a reliable predictor

for predators as to the consequence of

predation. Under honest signaling, brighter,

more conspicuous signals are expected to

relay more unprofitable secondary defenses (Sherratt and Beatty 2003). In the case of

dendrobatids, this may explain their polytypism. Dendrobatids obtain alkaloid toxins from their

diet (Daly and Myers 1967, Daly et al. 1987, Saporito et al. 2006), and since invertebrate

community composition likely varies across geographic space, alkaloid profiles of populations

could similarly vary over geographic space (Saporito et al. 2006). This variation in toxin content

would likely result in variation in unprofitability to predators, and selection would be expected to

drive signals to be an honest reflection of this variation (Sherratt and Beatty 2003, Speed et al.

2010). But the underlying problem with this line of thought is that there are other forces that

could explain polytypism (e.g., sexual selection, predator biases), but as we have yet to devise a

way to remove toxin variation as a potential confounding factor, exploring these hypotheses is

Figure 2: Dorsal and ventral phenotypic variation

seen in the Australian Brood Frogs (Pseudophryne)

8

difficult and must always be qualified with toxin variation. Ideally, a similar system that is

variable in phenotype, but not in toxin content would allow testing of these other hypotheses

independent of toxin variability.

Examples of systems in which toxin content remains stable, but phenotypes vary, are few

and far between. However, the Australian Brood Frogs (Pseudophryne) may fit this description

fairly well. There are currently 14 recognized species of Pseudophryne that are distributed

throughout the Australian continent. Broadly, they form two different clades (Donnellan et al.

2012) between central/western species, and eastern species. These frogs generally have a mottled

brown dorsum and a conspicuous black-and-white mottled venter (Figure 2). In the eastern clade,

frogs add to this by incorporating yellow and orange patches into their dorsal coloration, with the

most colorful examples being the Northern and Southern Corroboree Frogs (P. pengilleyi and P.

corroboree, respectively). Like dendrobatids, these frogs have alkaloid skin secretions (Smith et

al. 2002). These alkaloid toxins coupled with conspicuous coloration have led some to

hypothesize that members of Pseudophryne are aposematic (Williams et al. 2000). Unlike the

dendrobatids, however, Pseudophryne species are capable of sequestering alkaloid toxins from

invertebrate prey and synthesizing their own alkaloids (pseudophrynamines) de novo (Smith et

al. 2002). This is extraordinary because alkaloids are primarily plant-derived (Roberts and Wink

1998), and animal creation of alkaloids is rare (Roberts and Wink 1998). Indeed, these are the

only species of frogs currently known to be capable of synthesizing alkaloids (Smith et al. 2002,

Saporito et al. 2011). In contrast, dendrobatids sequester alkaloids from invertebrate prey, and it

is thought that the invertebrates, in turn, obtain those alkaloids from plant sources on which they

feed (Saporito et al. 2011). Furthermore, Pseudophryne species appear to be capable of

9

upregulating production of biosynthesized pseudophrynamines when dietary sources of alkaloids

are lacking (Smith et al. 2002). This suggests that they are capable of controlling the toxin

content in their mucus secretions, and that unprofitability (i.e., unpalatability of alkaloids) may

be less variable within and among populations as compared to dendrobatids. This makes

understanding why conspicuous signal is variable across the genus an interesting and important

question to be answered.

Aims for this Dissertation

Aposematic signaling is an important defensive strategy used throughout the animal

kingdom. Interestingly, a number of taxa display considerable diversity in aposematic signal and

that often corresponds with notable species diversity (e.g., bees and wasps, nudibranchs, poison

frogs). Understanding the underlying mechanisms that promote aposematic signal diversification

will aid in understanding a question that has been driving biologists since Darwin: how does

biological diversity originate?

To address this question, I used the polytypic taxa of Dendrobates tinctorius and

Pseudophryne in a series of ecological and evolutionary experiments to better understand how

polytypism may violate evolutionary expectations for aposematic signaling. In doing so, I

explored the interplay between toxins and predator avoidance, and how these interactions can

explain aposematic polymorphism and polytypism in these and other species, which are likely

precursors to speciation.

10

It is important to note that the questions and experiments in this dissertation are broad-

reaching and could not be simply done by myself. The findings reported in this dissertation are

the result of a large number of collaborators (Table 1) without whom, I would not be able to

accomplish what I have. The following manuscripts have resulted from this research, and are

broken down as follows:

I. Lawrence, J.P. and B.P. Noonan. 2018. Avian learning favors colorful, not bright,

signals. PLoS ONE 13 (3): e0194279.

II. Lawrence, J.P., P. Andes, A. Smith, B. Rojas, R.A. Saporito, A. Fouquet, E.A.

Courtois, J. Mappes, and B.P. Noonan. 2018. Consequences of prey selection on

toxin synthesis and predator avoidance in poison frogs (Dendrobatidae). Manuscript.

III. Lawrence, J.P., M. Mahony, and B.P. Noonan. 2018. Differential responses of avian

and mammalian predators to phenotypic variation in Australian Brood Frogs. PLoS

ONE 13 (4) e0195446.

IV. Lawrence, J.P., B. Rojas, A. Fouquet, J. Mappes, A. Blanchette, R.A. Saporito, R.J.

Bosque, E.A. Courtois, and B.P. Noonan. The origin and evolution of polymorphic

warning signals. Manuscript.

11

Table 1: Contributions and funding sources for manuscripts from this dissertation

I II1 III2,3 IV1,4

Original Idea JPL, BPN JPL, BPN, BR JPL, BPN JPL, BPN, BR

Study Design JPL, BPN JPL, BPN, BR JPL, BPN, MM JPL, BPN, BR

Data Collection JPL JPL, BR, PA,

AS, AF, RAS

JPL JPL, BR, AF,

AB, RAS, RJB,

TM, JC, EAC

Data Analysis JPL JPL, BR, RA,

AS, RAS

JPL JPL, BR, AB,

RAS

Writing JPL, BPN JPL, BR JPL, BPN JPL, BR, BPN,

JM

JPL = J.P. Lawrence, BPN = Brice P. Noonan, BR = Bibiana Rojas, PA = Peter Andes, AS =

André Smith, AF = Antoine Fouquet, RAS = Ralph A. Saporito, MM = Michael Mahony, AB

= Annelise Blanchette, RJB = Renan Janke Bosque, TM = Tim Muhich, JC = John Constan,

EAC = Elodie A. Courtois.

1 Funded by Investissement d’Avenir grant of the Agence Nationale de la Recherche (CEBA:

ANR-10-LABX-25-01)

2 Funded by National Science Foundation East Asia and Pacific Summer Institutes Fellowship

(OISE-1515149).

3 Funded by Australian Endeavour Fellowship.

4 Funded by American Society of Ichthyologists and Herpetologists Gaige Award.

12

II. AVIAN LEARNING FAVORS COLORFUL, NOT BRIGHT, SIGNALS

Abstract

A few colors, such as red and yellow, are commonly found in aposematic (warning)

signaling across taxa, independent of evolutionary relationships. These colors have unique traits

(i.e., hue, brightness) that aid in their differentiation, and perhaps, their effectiveness in

promoting avoidance learning. This repeated use calls into question the influence of selection on

specific warning colors adopted by aposematic prey-predator systems. To disentangle the

influence of color characteristics on this process, we trained week-old chickens (Gallus gallus

domesticus) to learn to avoid distasteful food that was associated with one of three color signals

(yellow, white, red) that varied in both hue and in brightness in order to assess which of these

traits most influenced their ability to learn avoidance. Our results show that while chicks learned

to avoid all three colors, avoidance was based on the hue, not brightness of the different signals.

We found that yellow was the most effective for avoidance learning, followed by red, and finally

white. Our results suggest that while these three colors are commonly used in aposematic

signaling, predators’ ability to learn avoidance differs among them. These results may explain

why yellow is among the most common signals across aposematic taxa.

13

Introduction

Convergent phenotypic evolution, where traits have evolved independently, is

widespread in the natural world, and as such, provides an intriguing opportunity to examine the

evolutionary pressures driving such patterns. Convergence can be found in a wide variety of taxa

from plants (Ashfield et al. 2004) and animals (Emery and Clyaton 2004) to single-celled

organisms (Pupo et al. 2000), making the subject of convergent evolution an important question

in biology. From the evolution of eyes (Goldsmith 1990) to flight (Ostrom 1974), these “natural

experiments” provide perspective into how the natural world functions. Conclusions drawn from

these examples can be quite powerful as they represent a broad, repeated pattern from which

there are apparent evolutionary mechanisms leading to such convergence. Common explanations

for similarity, such as phylogenetic relationships, can be discounted due to the independence of

the phenomena, allowing researchers to get to the underlying evolutionary mechanisms that

could promote such convergent evolution.

Aposematic (warning) signaling is widespread throughout Animalia (Figure 3), and

despite the variety of colors and patterns employed, there are common components to this

signaling (Ruxton et al. 2004). Most common are yellow, red, and white coloration, which

comprise a dominant component to the vast majority of aposematic signals. Aposematic

signaling functions by displaying memorable coloration to would-be predators (Endler and

Greenwood 1988, Mallet and Joron 1999, Pfennig et al. 2001, Endler and Rojas 2009). White,

yellow, and red are also among the most contrasting colors against a wide variety of natural

backgrounds. These commonalities among diverse aposematic taxa beg the question of why such

colors repeatedly appear in independent instances of aposematic signaling. It is important to note

14

that these signals vary not only in color (hue), but also in brightness. It is then possible that the

brightness signal components may be important, perhaps more important than hue, in predator

perception as this directly affects conspicuousness in a variety of lighting environments.

As research into sensory ecology has progressed, our understanding of the constituent

components of signaling has also improved. Aposematic signaling has increasingly been

understood to not be comprised solely of conspicuous colors, but rather a complex combination

of features that aid in predator recognition, learning, and avoidance (Rowe and Guilford 1999,

Ham et al. 2006, Ratcliffe and Nydam 2008). Everything from color, pattern, behavior, and

lighting environment can impact the learning and recognition of an aposematic signal (Rojas et

Figure 3: Aposematic signaling has been documented in a wide variety of taxa

These include, but are not limited to, A) Dendrobatid frogs, B) Salamandrid salamanders, C) Micrurus coral

snakes, D) Nudibranchs, E) Heliconius butterflies, and F) Hymenopterans. Among these, similar colors have

evolved, which are commonly some variation of yellow, red, and/or white. Printed under a CC BY license,

original copyright J.P. Lawrence 2017.

15

al. 2014b). Here, we seek to decouple hue from brightness to determine which is more important

to the learning of an aposematic signal by predators.

While predators of aposematic species are undoubtedly quite varied and possess a variety

of visual systems, birds are commonly used in aposematism studies (Noonan and Comeault

2009, Skelhorn and Rowe 2010, Kikuchi and Pfennig 2010). The primary reasons for this is that

they are common predators of aposematic taxa and have tetrachromatic vision, allowing them to

perceive and discern color very well (Okano et al. 1992). Consequently, we utilized a model

avian predator to examine how well naïve individuals can learn to avoid a variety of aposematic

signals that vary in both hue and brightness.

Methods

This research was approved by the University of Mississippi IACUC and conducted

under University of Mississippi IACUC 14-026.

Color Analysis – We

created three different color

signals to test learning.

Signals were pictures of

drawn frogs with either white,

yellow, or red stripes that

mimic a pattern found in

Dendrobates tinctorius (with

the exception of red). Bodies

were black and legs were

Figure 4: Reflectance curves for each color signal.

Color signals were smoothed and averaged using the pavo package in R.

Lines represent means of all nine measurements for each of the three

colors red, yellow, and white (in black) while the shaded region

represents a 95% confidence interval for each corresponding signal.

16

blue. White and yellow colors were extracted from photographs of D. tinctorius from Grand

Matoury and Kaw in French Guiana, respectively, while the red color was extracted from

photographs of Oophaga pumilio from Bastimentos, Panama. These represent colors common in

aposematic signaling, but they also represent low (red), medium (yellow), and high (white)

brightness (Figure 4). The modeled colors were not meant to exactly match the frogs from which

they were extracted but provide three distinctly different colors varying in both hue and

brightness.

We took three color measurements for each color band (see Figure 4) of each printed frog

using an Ocean Optics Jaz A3305 spectrometer. We measured three different regions on the

color band at the top, middle, and bottom of the band. Measurements were averaged and

smoothed in the pavo package (Maia et al. 2013) in R (R 2016) to create reflectance curves for

each signal (Figure 4).

We analyzed spectra using the pavo package (Maia et al. 2013) to examine the chromatic

(color, hue) and achromatic (luminance, brightness) contrasts between the two colors. The pavo

package calculates contrasts in Just Noticeable Differences (JNDs) where a value of 1 JND

would mean differences could be observed under the modeled vision system when objects are

stationary under bright conditions (Thurman and Seymoure 2016). However, in more natural

conditions, signals with 3 JNDs or less are unlikely to be distinguished from one another under a

modeled visual system (Thurman and Seymoure 2016).

We used the average avian vision system for visual contrasts and chicken double cone

sensitivity to calculate achromatic contrasts under the D65 (standard daylight) illuminant

(Siddiqi et al. 2004). We conducted all pairwise comparisons between the three measurements

for each band. Consequently, each among-band comparison had nine different JND estimates. As

17

JNDs represent a threshold where a modeled visual system can or cannot distinguish signals

from one another, we analyzed whether or not means for each signal differ from 1 JND and, a

more conservative estimate, 3 JNDs using a nonparametric one-tailed sign test (Thurman and

Seymoure 2016).

Learning Experiments – In order to assess how predators learn avoidance, we used one-week-

old chickens (Gallus gallus domesticus; male cornish rock crosses from the commercial supplier, Ideal

Poultry). These were housed in wire brooders (approximately 1m x 1m x 30cm) together while not being

tested and fed a commercial corn meal mash. To examine how naïve predators learn color signals, chicks

were equally divided (N = 15 per treatment; 45 chicks total) into three different treatments: white signal,

yellow signal, and red signal. Trials were done in 30cm x 60cm wooden compartments under full

spectrum lighting. Chicks were placed individually into a compartment and allowed to habituate for 2

hours. Chicks were food-deprived during this acclimation period to ensure motivation to feed during the

trials.

Once in the compartment and after the 2-hour acclimation period, training consisted of

teaching the chicks to eat a dried mealworm (Tenebrio molitor) from a petri dish with a printed

brown frog on a tan background below the transparent dish. The training phase was completed

once the chick had eaten three consecutive times, after which we allowed the birds to rest for a

period of 5 minutes.

The avoidance-learning trials consisted of the consecutive presentation of mealworms on

a petri dish with a printed frog beneath the dish displaying a warning signal matching the dorsal

pattern of D. tinctorius (described above) with either a white, yellow, or red (these colors

hereafter referencing the whole printed frog displayed in Figure 4, not solely the bands) on a tan

background. In order to make them unpalatable, mealworms were soaked in a solution of 10%

18

chloroquine for >1 hour. We recorded the hesitation time (i.e., the time until the chick

approached and ate the mealworm) and registered any behavioral reaction to the distastefulness

of the mealworm after tasted or eaten. Such behaviors most often involved beak wiping and head

shaking. Each trial lasted 5min, followed by 5 resting minutes after which a new petri dish with

an unpalatable mealworm was offered on the same aposematic background. This procedure was

repeated until the chick refused to eat the mealworm over three consecutive trials. The test ended

either when chicks “learned” the signal or proceeded through 10 trials, whichever came first. If

chicks proceeded through 10 trials without three consecutive refusals, they did not learn the

signal. When a chick refused the chloroquine-soaked mealworm on the aposematic pattern, a

palatable mealworm was offered on the training (brown frog) background, which the chick did

not associate with an unpleasant experience. If the palatable mealworm on a brown frog

background was refused (n = 5), these trials were discarded due to the potential for satiation to be

confused with learned avoidance. For this reason, trials were run with additional chicks to ensure

each treatment had 15 replicates. We registered the number of trials (i.e., mealworms consumed)

that it took for the chick to learn to avoid the signal. This research was conducted under

University of Mississippi IACUC animal care protocol 14-026.

Data Analysis – We analyzed the data from the learning experiments using an analysis of

variance (ANOVA) in R (R 2016). We examined the number of attack trials (i.e., how many

trials in which chicks tasted the mealworm) and average hesitation time. If ANOVA yielded

significant results, we conducted post-hoc Tukey’s Honest Significant Difference tests to

examine pairwise comparisons among color signals. We examined whether there were

19

differences in chicks’ ability to learn different signals using a generalized linear hypothesis test

with Tukey contrasts.

Results

Color Differences – As expected, mean Just Noticeable Differences (JNDs) for each

comparison of chromatic contrasts was found to be quite distinct (mean ± SE: red – white: 10.81

± 0.38, yellow vs. red: 10.37 ± 0.11, and white vs. yellow: 11.86 ± 0.16; Figure 5A). However,

achromatic contrasts were distinct between the red – white (mean ± SE: 8.56 ± 0.83) and yellow

– red (mean ± SE: 9.27 ± 0.82), but the white – yellow comparison averaged less than 1 JND

(mean ± SE: 0.8 ± 0.17; Figure 5B). Nonparametric sign tests for chromatic contrasts revealed

that all three comparisons were significantly greater than 1 JND (red – white: p = 0.002; yellow –

red: p = 0.002; white – yellow: p = 0.002) and 3 JNDs (red – white: p = 0.002; yellow – red: p =

0.002; white – yellow: p = 0.002).

Achromatic contrasts were significantly

greater than 1 JND and 3 JNDs for the

red – white (1 JND: p = 0.002; 3 JNDs: p

= 0.002) and yellow – red (1 JND: p =

0.002; 3 JNDs: p = 0.002) comparisons,

but were not significantly greater than 1

JND, and thus 3 JNDs, for the white –

yellow comparison (1 JND: p = 0.91; 3

JNDs: p = 1).

Figure 5: Mean Just Noticeable Differences (JNDs)

between different signal comparisons for both A)

chromatic and B) achromatic contrasts.

Signal comparisons were Red – White (R – W), Yellow –

Red (Y – R), and White – Yellow (W – Y). The dotted line

represents 1 JND and the dashed line represents 3 JNDs.

Error bars represent standard error.

20

Learning Experiments – When examining both the number of trials required to learn avoidance

and the average hesitation time, we find significant differences between color treatments in both

avoidance trials and average hesitation time (F2, 42 = 5.514, p = 0.007 and F2, 42 = 5.295, p =

0.009, respectively). Pairwise comparisons yielded significant differences when comparing

results from the yellow treatment

and white treatment (attack trials: p

= 0.005; hesitation time: p = 0.007).

While none of the other pairwise

comparisons yielded significant

differences, the average hesitation

time between the red and white

treatments did trend towards

significant (p = 0.092). When

comparing whether there was a

difference between signals in how

often chicks learned to avoid signals,

we found that red was not

significantly different from white (p

= 0.498) or yellow (p = 0.157).

However, when comparing yellow

and white, we did find a significant

Figure 6: A) Proportion learned and B) mean hesitation time

of chicks (Gallus gallus domesticus) for each of the three

signals

Error bars on the bar graph represent SE. Box plot shows mean,

25th and 75th percentiles of data while vertical lines indicate data

range. Circles represent extreme values. Different letters above

bars and boxes indicate statistical significance (α = 0.05) from

one another.

21

difference (p = 0.016) with chicks more likely to learn to avoid yellow than white (Figure 6).

Discussion

Many components likely contribute to a predator’s ability to remember and avoid novel

warning signals. These can include, but are not limited to, hue, pattern, and brightness. Here, we

sought to decouple hue and brightness and determine which is more important at eliciting

avoidance behavior. Our results suggest that hue is likely more important than brightness for

predator learning, although not all colors are avoided equally. We saw distinct differences

between yellow and white signals, with red being intermediate for both learning and hesitation

time (Figure 6). Were brightness to be important in determining or limiting predator avoidance,

we would have expected white and yellow to be more similar and for both to have evoked

greater, or more efficient, learning than red.

Indeed, our results concur with findings previously reported in the literature with regard

to hue (color) being the most important factor in determining avian predator learning (Aronsson

and Gamberale-Stille 2008, Kazemi et al. 2014). There may also be an interaction between color

and pattern that enhanced avoidance (Finkbeiner et al. 2014), though as we only examined color

differences, we are unable to support or refute this. To that end, it is important to note that there

are a number of factors that can influence predator learning that are beyond the scope of this

study, including lighting environment (Rojas et al. 2014b), contrasting backgrounds (Preißler and

Pröhl 2017), predator communities (Mappes et al. 2014), etc. Further, our reflectance spectra

show some fluorescence in the white signal. While this is likely an artifact of being printed on

paper, it is important to note that prior literature suggests that this fluorescence should not affect

predator attacks rates between fluorescent and non-fluorescent colors (Finkbeiner et al. 2017).

22

Despite this, we would encourage some caution in broad application of our results as some color

traits are not necessarily broadly applicable to all animals.

While this research was conducted with domestic chickens in a laboratory setting, it is

important to give this research context to wild populations and patterns, particularly because

prior studies have demonstrated that chicks can be neophobic to novel signals (Roper and

Marples 1997). Interestingly, the pattern observed in our study is mimicked in wild poison frog

predator populations where yellow, in general, is better protected than red (Preißler and Pröhl

2017) indicating that patterns observed in the wild are repeatable in laboratory settings.

Likewise, several clay model experiments have shown that yellow is especially protective

against avian predators (Noonan and Comeault 2009, Dreher et al. 2015, Preißler and Pröhl

2017) with red providing moderate protection (Saporito et al. 2007c, Paluh et al. 2015) – though

importantly, these studies did not examine relative differences of red to other aposematic colors).

This research provides interesting insight into the evolution and proliferation of

aposematic signals in the wild. Aposematic signaling is prevalent throughout the animal

kingdom, but a number of colors continually evolve independently as signal components (Figure

3). Our results indicate differential learning abilities of avian predators that may explain why we

see such signal variation in natural systems. While all three colors did lead to learned avoidance

from chickens, the chicks clearly had a more difficult time learning to avoid white, particularly

when compared to yellow. As avian predators likely drive aposematic signal evolution in a

number of taxa (Skelhorn and Rowe 2010, Maan and Cummings 2012, Kraemer and Adams

2013, Valkonen and Mappes 2014), this could explain why warmer colors (yellow, orange, red)

are more commonly used than other, brighter, colors. While birds are capable of seeing into the

ultraviolet spectrum, avian visual systems are more sensitive to longer wavelengths (Prescott and

23

Wathes 1999). This, coupled with innate biases towards warmer colors (Schuler and Hesse 1985,

Lindström et al. 1996), suggests why warmer colors are prevalent in aposematic signaling.

Of course, our research does raise the question as to why suboptimal signals such as red

or white may persist in aposematic species. While we examined the influence of color on

aposematic signaling, these colors likely have a wide variety of functions outside of warning

signaling. Research suggests that, for example, in the Strawberry Poison Frog (Oophaga

pumilio) color may be an important component in mate choice (Maan and Cummings 2008,

2009). Further, aposematic function can be dependent on the distance and behavior of

aposematic species (Rojas et al. 2014b, 2014a). At certain distances, counterintuitively,

aposematic signals can be disruptive or cryptic. These colors likely have a wide variety of

functions that go beyond simple warning signaling, which could explain why suboptimal signals

persist and may even be promoted by selection. And finally, suboptimality of aposematic signal

should be considered in context of the entire predator community. Avian predators are thought to

be important in driving color evolution in poison frogs (Siddiqi et al. 2004, Comeault and

Noonan 2011, Chouteau and Angers 2011), however, recent evidence suggests that alkaloid

toxin defenses may target insects (Weldon et al. 2013). As such, while these signals may be

suboptimal to avian predators, they may be more effective to other predator taxa.

Conclusions

Repeated patterns across independent taxa provide an intriguing opportunity to examine

underlying evolutionary pressures that may promote such patterns. For example, the repeated

banding or longitudinal striping patterns seen in snakes have been shown to have either flicker

effects or cause predators difficult in locating vulnerable parts of the body (Allen et al. 2013). By

24

dissecting color signals into their constituent parts, we can discern what components are most

important for predator decision making. Here, we present evidence that color, not brightness, is

important in predator decision making and learning. While our research is limited to a small

sample of aposematic colors found in natural systems, it provides important insight into why

these colors persist. Further research should focus on why colors, such as yellow, elicit such a

strong response in avoidance learning and if this, perhaps, is due to predator physiology (i.e.,

sensitivity to particular color spectrums) or cognition.

Acknowledgements

We would like to thank R. Bosque and M. Smith for their contributions to conducting the

learning experiments. We would also like to thank J. Yeager for providing spectrometric

measurements for the printed frog models.

25

III. ALKALOID VARIABILITY IN THE DYEING POISON FROG AND ITS IMPORTANCE

TO PREDATOR RESPONSE

Abstract

Aposematic (warningly colored) animals that are chemically defended must evolve

mechanisms that ensure that their signal is accurately interpreted by predators. While some

species have evolved pathways to synthesize their chemical defenses de novo, many species rely

on diet for sequestering the necessary toxins to defend themselves. This dietary acquisition can

lead to variable chemical defenses across geographic space as the community composition of

chemical sources is likely to vary across the range of an aposematic species. We characterized

the alkaloid content of two populations of the Dyeing Poison Frog (Dendrobates tinctorius) in

northeastern French Guiana. In addition, we conducted unpalatability experiments with Blue Tits

(Cyanistes caeruleus) using whole-toxin cocktails to assess how a model predator would respond

to the defense of individuals from each population. While there were some similarities between

the two D. tinctorius populations in terms of alkaloid content, our analysis revealed that these

two populations are largely distinct from one another in terms of overall alkaloid profiles. While

we did not observe any difference in avian predator aversive behavior between the two frog

populations from our unpalatability trials, we did identify candidate alkaloids that may be

responsible for predator response in one frog population. We provide the first between-

26

population comparison of alkaloid profiles for D. tinctorius and describe a novel method for

assessing unpalatability of alkaloids and identifying which elements contribute to the predator

response.

Introduction

Aposematic, or warning, signaling is a ubiquitous defensive strategy in which prey

species use conspicuous signals to warn would-be predators of the risks of a secondary defense,

often in the form of toxins. For a number of aposematic taxa (e.g., dendrobatid frogs, butterflies,

and nudibranchs), toxins are sequestered from food sources rather than being synthesized de

novo (Proksch 1994, Nishida 2002, Saporito et al. 2011). This diet-based toxin sequestration

results in interpopulation variation in toxins because food sources vary over geographic space

(Saporito et al. 2006, 2007a, Daly et al. 2008). Consequences of this diversity vary from

automimicry (Speed et al. 2006) to aposematic polytypism (interpopulation variation in color

pattern phenotypes; Siddiqi et al. 2004).

Honest signaling occurs when a signal is predictive of the consequence of a secondary

defense (Summers et al. 2015). If a signal is poorly defended, predators will have difficulty

associating the signal with a defense. Conversely, a weak signal that is over-defended may result

in predators having difficulty recalling the signal to avoid the defense. Additionally, chemical

defenses are likely to be metabolically costly (Ojala et al. 2005, Sandre et al. 2007), and so

having a well-matched signal should ensure that energy is not wasted. While there are numerous

examples of aposematic species that show honest signaling among populations, this is not a

27

universal trend. Some species of poison frogs, for example, seem to display a tradeoff between

secondary defense and signal (Darst and Cummings 2006, Wang 2011).

Alkaloid variation in dendrobatid poison frogs has been well-characterized in a few

species (Daly et al. 1987, 1992, Bolton et al. 2017). Diet-based alkaloids in poison frogs

primarily come from ants, mites, millipedes (Saporito et al. 2003, 2007b, McGugan et al. 2016),

and some other invertebrates (i.e., beetles) (Saporito et al. 2007a). Over 800 different types of

alkaloids have been described from poison frogs (Daly et al. 2005). As the source of these toxins

are arthropods (Darst et al. 2005, Hantak et al. 2013), toxin profiles among frog populations has

been shown to vary among populations and over time within populations (Saporito et al. 2006,

2007a). Characterizing alkaloid profiles in aposematic species is important for understanding

how defense varies among populations. As the secondary defense can aid in the efficacy of

aposematic signaling, baseline data on alkaloid profiles can provide important insight into why

aposematic signals vary among species and populations.

While characterization of alkaloids is important to better understand the polytypism

displayed by D. tinctorius, these alkaloids may also have important medicinal uses. These

alkaloids affect the sodium and potassium channels of nerve cells (Weldon et al. 2013), making

them a potential painkiller. Epibatidine is an example of such an alkaloid derived from a

dendrobatid frog that has been investigated for its painkilling properties (Badio and Daly 1994,

Fisher et al. 1994, Fitch et al. 2010).

While alkaloid defenses have been the subject of scientific inquiry for decades, their

relationship to aposematic color variation is less understood. When examining the polytypism in

the Strawberry Poison Frog (Oophaga pumilio) in Bocas del Toro, Panama, Daly and Myers

28

(1967) found no relationship between toxicity and color pattern. Conversely, Maan and

Cummings (2012) suggested that the colors are honest, particularly for avian predators, and

therefore there should be a strong relationship between level of toxicity and clarity of aposematic

color signals. This apparent contradiction may be the result of differences in methodological

techniques. Daly and Myers (1967) assessed toxicity by determining LD50 (lethal dose for 50%

of test subjects) of toxin extracts for mice from different populations while Maan and Cummings

(2012) examined the discomfort mice exhibited after injection. Further, mice are often used to

represent avian predators, so not only are injections an unrealistic method of assessing the

biological function of alkaloids, mice are unlikely to prey on frogs. While these techniques are

informative, they may not be biologically relevant as these toxins are experienced by the

predator when capturing and consuming the prey species, rather than entering the bloodstream or

musculature (Holen 2013, Weldon 2017). Accordingly, there is a clear need to develop a more

realistic method for assessing unpalatability (cf. toxicity, per se) of frog toxins.

The Dyeing Poison Frog (Dendrobates tinctorius) is found throughout the Guiana Shield

in northern South America. Throughout its range, it shows considerable polytypism with highly

variable conspicuous components (Noonan and Gaucher 2006, Rojas et al. 2014a). While it is

known that D. tinctorius sequester alkaloids (Summers and Clough 2001, Santos et al. 2003), no

study has yet examined among population variability of alkaloids. With the among-population

variability in aposematic signal in this species, it is important to understand whether 1) alkaloid

defenses vary among populations, and 2) how predators respond to different alkaloid defenses.

Here, we investigated how variable alkaloid toxins are between populations of D.

tinctorius and behavioral responses to these toxins by a model predator. We expected, as in other

29

poison frogs, to find considerable population differences in alkaloid content between

populations. Similarly, we expected that, with this toxin diversity, there will be similarly variable

predator response to the toxins. We also expected that while individuals will have numerous

alkaloid toxins, predator response to these will likely be based on a subset of toxins.

Methods

Field Collection – We collected 12 Dendrobates tinctorius in May-June 2013 from two

populations in French Guiana: Matoury (4.89°N, 52.34 °W; 10 individuals) and Kaw Mountains

(4.57°N, 52.21°W; 2 individuals). An additional six individuals from the Kaw Mountains were

collected in August 2014. For each frog we encountered, we recorded individual variation (sex,

snout-vent length, and spectrometry). Next, frogs were euthanized by cervical transection and

pithing in the field immediately after taking measurements. We skinned frogs and placed whole

skins in 100% methanol in 4mL vials with PTFE caps.

Toxin Extraction – We followed the protocol outlined by Saporito et al. (2010) to conduct

acid-base fractionations of alkaloids from skins. Vials were topped off at 4mL with 100%

methanol. Following this, we took 1mL of methanol and added it to a graduated conical vial

along with 50μL of HCl to acidify the alkaloids. We added 100μL of an internal nicotine

standard (L-Nicotine, 99+%, Acros Organics, 1:10,000 nicotine to methanol). Samples were then

dried down to 100μL using a gentle flow of N2 and then 200μL of DI H2O was added. Following

this, we removed lipids with four washes of 300μL hexanes, after which hexane layers were

removed. The solution was then basified by adding drops of saturated NaHCO3 until the pH was

30

>7.0. We then washed the sample three times with 300μL of ethyl acetate, transferring the ethyl

acetate layer to a new vial with anhydrous NaSO4. The resulting ethyl acetate was then

transferred to a new vial and evaporated to dryness under a gentle flow of N2. The residue was

reconstituted in 100μL of methanol for alkaloid analysis.

GC-MS Analysis - Gas chromatography-mass spectrometry (GC-MS) was performed for

each individual sample on a Varian Saturn 2100T ion trap MS instrument, which was coupled to

a Varian 3900 GC with a 30m x 0.25mm inside diameter Varian Factor Four VF-5ms fused silica

column. GC separation of alkaloids was achieved using a temperature program from 100°C to

280°C at a rate of 10°C per minute with helium as the carrier gas (1 mL/min). Each alkaloid

fraction was analyzed in triplicate with electron impact MS and once with chemical ionization

(CI) MS with methanol as the CI reagent.

Individual alkaloids of D. tinctorius were identified based on comparison of mass

spectrometry properties and GC retention times with those of previously reported alkaloids in

dendrobatid frogs (Daly et al. 2005). Alkaloids in dendrobatid frogs have been assigned a series

of code names that consist of a boldfaced number indicating the alkaloids’ nominal mass, and a

boldfaced letter to distinguish those alkaloids with the same nominal mass (Daly et al. 2005).

Alkaloid quantities for each individual frog were calculated by comparing the average observed

peak area of individual alkaloids to the average peak area of the nicotine standard from the

triplicate EI-MS analyses using a Varian MS Workstation v.6.9 SPI. Only alkaloids that were

present in quantities of ≥ 0.5μg were included in the analyses (Bolton et al. 2017). Similarity and

overlap of toxin profiles between populations was assessed by conducting a principle

components analysis (PCA).

31

Unpalatability Assays - Of the 4mL methanol, we took 1mL of the methanol solution and

evaporated it to dryness under a gentle stream of N2. Since predators experience toxins and non-

toxin skin content (i.e., mucus, peptides, etc.) when attacking a frog, we opted not to purify the

toxins through fractionation and instead used everything present in the methanol solution. We

evaporated off methanol to avoid potential issues with methanol toxicity. These samples were

reconstituted in 0.5mL ethanol to then be used in unpalatability trials with Blue Tits (Cyanistes

caeruleus). While Blue Tits are a paleoarctic species and thus would not encounter Neotropical

frogs, predators of poison frogs are assumed to be birds (Comeault and Noonan 2011, Chouteau

and Angers 2011, Paluh et al. 2014). Bird taste systems are generally conserved across genera

(Wang and Zhao 2015), and as such, use of tits as an analog for an avian taste system in this

study is acceptable.

Prior to experiments, Blue Tits (Cyanistes caeruleus) were trained to eat oats. After

training, birds were given unpalatable oats which lacked any color signal. Aversive behavior

(i.e., beak wiping) and proportion of the oats eaten was recorded. All trials were filmed to be

able to record behaviors. In addition to the two unpalatable treatments, a control treatment

consisted of birds that were given oats that had been treated with only ethanol, which were

allowed to evaporate to dryness in a similar manner as toxin-coated oats. This ensured that any

differences in behavior can be attributed to toxins and not possible residual ethanol.

We tested a total of 25 birds, eight with toxin extracts from the yellow population, 10

with toxin extracts from the white population, and seven controls. Each of two oats were soaked

with 15µl of the extract of one frog skin and left for 24h at room temperature to ensure that all

ethanol had evaporated. Two other oats were soaked each with 15µl of pure ethanol that were

32

used at the beginning and end of the experiment with each bird. Each bird went through four

trials. The first trial consisted of a control oat which needed to be consumed entirely by the bird

before the experiment could be initiated. Following this, two consecutive toxin treatments each

consisted of a single oat with toxin extract. The final trial involved the second control oat which

was offered to ensure that the birds were not refusing to eat the oats coated with toxins out of

satiation or lack of motivation to eat in general. Birds in the control treatment received oats

soaked with pure ethanol four all four treatments in order to compare directly the response of

birds to oats containing frogs’ toxins vs. oats with ethanol only.

Each oat was presented on a hatch that had a visual barrier, which allowed us to identify

the exact moment at which the oat was seen which set the actual beginning of each of the two

trials. We measured the percentage of the oat eaten as an analog for how distasteful the oat was.

Birds were watched for a 2-min period after they finished eating the oats, or for a maximum of 5

min in the cases in which the oat was not fully eaten, to make sure that any delayed response to

the oat taste was not going to be missed.

Data Analysis – While individuals and populations can be quite variable in toxin content,

it is unlikely that the entire suite of toxins is contributing to the behavioral response of predators.

In order to identify the most likely alkaloid candidates driving predator response, we performed

an exploratory factor analysis using the quantities of each type of alkaloid. Factor analyses group

independent variables into loadings that explain population variation in these variables.

Following this, we performed a multiple linear regression to identify which loadings are

explanatory for variation in behavioral response. We performed this analysis for each population

33

to determine what, if any, toxins are important for behavioral response to toxins. All analyses

were conducted using the R platform (R 2016).

Results

We detected 49 different types of alkaloids representing 11 different structural classes

(Table 2; Figure 8) from representatives of the Matoury population of D. tinctorius. Twelve of

the 49 alkaloids were represented in single individuals. Of the 49 alkaloids, 14 (two 3,5-

disubstituted indolizidines [223AB and 275C], four 5,6,8-trisubstituted indolizidines [231B,

251T, 259C, and 267R], two 3,5-disubstituted pyrrolizidines [209Q and 251K], one

histrionicotoxin [235A], two decahydroquinolines [219A and 243A], one 1,4-disubstituted

quinolizidines [231A and

249N], and two new alkaloids

[“247a” and “275”]) were

found in a majority of

individuals in the population.

Factor analysis

revealed five loadings that

explained 76% of the variation

among toxin profiles in the

Matoury population. A

subsequent multiple linear

Figure 7: Principle Components Analysis (PCA) of the Kaw and

Matoury populations of Dendrobates tinctorius

PC1 explains 82.7% of the variance among toxins while PC2

explains 11.6%. Ellipses represent 95% confidence intervals.

34

regression revealed (multiple R2 = 0.97, F5,4 = 34.07, p = 0.002) two loadings that significantly

deviated from the null when examining proportion of oats eaten by blue tits. Loading 1 (t = -

11.47, p = 0.0003), which explains 18% of the population variation, contained the toxins 205B,

207G, 236, 245A, 265L, and 339A. Loading 3 (t = -3.113, p = 0.0350), which explains 17% of

the population variation, contained the toxins a new 245, new 247, 249C, 249N, 251T, 259C,

261B, 263A, and 269B.

Among members of the Kaw population of D. tinctorius, we observed 46 different

alkaloids representing 12 different structural classes (Table 2; Figure 8). Notably, of the eight

Kaw individuals examined one individual showed approximately ten times the quantity of

alkaloids as compared to other individuals in the population indicating that this individual may

have been an outlier or error, thus it was not included in the analysis. Of the 46 different

alkaloids, eighteen are represented in single individuals. Eleven alkaloids are represented in a

majority of individuals in this population including one 3,5-disubstituted indolizidines (223AB

and 275C), one 5,6,8-trisubstituted indolizidine (231B), decahydroquinolines (195J, 219A, and

221D), one 1,4-disubstituted quinolizidine (231A), one spiropyrrolizidine (236), one unclassified

(235BB), one new piperidine (“213”), and one new uncharacterized “247b” alkaloid.

Factor analysis revealed four loadings that explained 70% of the variation observed in

toxin profiles in the Kaw population. After factor analysis, however, we found no differences

among loadings when examining either proportion of oats eaten (multiple R2 = 0.07, F4,3 = 0.05,

p = 0.99) or beak wiping. The inability to identify specific alkaloids responsible for predator

response may be due to a low sample size of individuals for the population.

35

Based on principle components analysis, the two populations partially overlap with one

another. We found that PC1 explained 82.7% of the variance and PC2 explained 11.6% (Figure

7). The partial overlap due to common toxins likely also reflects similarity in diet which may be

explained by the close proximity of the two populations.

36

Table 2: Alkaloid variation seen in the Matoury and Kaw populations

Alkaloids in were found to significantly impact predator behavior in unpalatability experiments. Alkaloids are divided into the following structural classes: 3,5-

disubstituted indolizidines (3,5-I), 5,6,8-trisubstituted indolizidines (5,6,8-I), 3,5-disubstituted pyrrolizidines (3,5-P), histrionicotoxins (HTX),

decahydroquinolines (DHQ), 1,4-disubstituted quinolizidines (1,4-Q), allopumiliotoxins (aPTX), 5,8-disubstituted indolizidines (5,8-I), spiropyrrolizidine (Spiro),

4,6-disubstituted quinolizidines (4,6-Q), dehyrdo-5,8-disubstituted indolizidines (Dehydro-5,8-I), tricyclics (Tri), unclassified alkaloids (Unclass), piperidine

(Pip). The piperidine and fifteen (New) alkaloids have not previously been described but are in quotes as further characterization is needed in order to be

documented as new alkaloids. Major alkaloids (*) are found in quantities greater than 50μg, minor alkaloids (†) are found in quantities between 5μg and 50μg, and

trace alkaloids (o) are found in quantities less than 5μg.

Structural

Class Alkaloid

Matoury Population

Matoury

Average Kaw Population

Kaw

Average

M1 M2 M3 M4 M5 M6 M7 M8 M9 M10 K1 K2 K3 K4 K5 K6 K7

3,5-I 195B

o

o

223AB † † † o † * * † * o † † † † * † † * †

275C

o † o † o †

† o † †

† †

† †

5,6,8-I 193G

o

o

195D

o

o

o

o

205A

o

o

207C

o

o

219N

o

o

223A

o

o

225K

† †

o

o

o o

231B * † * † † * * † † † * o † o † † † † †

233G

o

o

o

o

235E

o o o o

237C

o o o

245G

o

o

249C †

† o

†

†

249BB

† o

† † o †

251T * * * o * * * * * * *

259C * † * † † †

† † † *

261B †

†

o

263A o

o

o

265L o

†

o

265U o

o

o

o

265L

†

† o

267R

o o o o o o

o

37

3,5-P 209Q

†

† † o o o o o o

223B

†

o o o

o

251K † † † o o * * † † * † o o † †

HTX 235A † *

o † † * † † † † * * * *

238A

†

o

245A

†

o

259A

† † † †

o

261A

o

o o o

283A

†

o

285A

†

o

285C

†

o

291A

†

† o

DHQ 195A

o

o

195J

† †

o

† o †

219A o o †

† o

o o † † † † * † * * †

221D

o o o o † o

223F

o

o

243A * * † † † * * † * † * †

o † o

245Q

o

o

269B o o

o

1,4-Q 231A † o o

†

† o

o o o o o o o

o

235U

o o

aPTX 305A

† †

o

339A

†

o

5,8-I 243C

o

o

237D

o

o o

Spiro 236

o †

†

o

† † †

† o †

4,6-Q 195C

† o

o

o

275I

†

† o

Dehydro-5,8-

I 265T

†

o

Tri 205B

o

o

205E

o

o

207G

o

o

Unclass 209G

o

o

38

209M

o

o

227

o

o o

235BB

* †

† † †

247M

o

o

249N † o o o o o o o o

o

Pip “213”

o o o †

o

New “171”

o

o

“193”

o

o

“207”

†

o

“209”

o

o

†

o

“217”

†

o

“223”

o o

o

“229”

o o

o

“233”

†

† o o

“235”

o

† o

“237”

“245” o

o

“247a” o

o

o o

o o o

“247b”

† † † † † †

“253”

o o

o

o

o

“275”

o o o † † † o o o o

Total 17 18 25 13 13 22 19 19 20 17 49 19 22 10 15 16 16 18 46

39

Discussion

Our study revealed the first among-population comparison of alkaloid diversity for D.

tinctorius. The Kaw and Matoury populations show similar diversity of alkaloids, but the

distribution of alkaloids is different between the two populations (Figure 7). Only three alkaloids

are well-represented in both populations (223AB, 231B, and 219A). Mites are known sources for

Figure 8: Distribution of alkaloids between the two populations

Alkaloids common to both populations of D. tinctorius are surrounded by the pink shape. Text in the circles

represent the type of alkaloid. Size of circles represents relative proportions of the population average of the

alkaloid. Color of the circle represents structural class in which the alkaloid is found. Alkaloids are divided into

the following structural classes: 3,5-disubstituted indolizidines (3,5-I), 5,6,8-trisubstituted indolizidines (5,6,8-

I), 3,5-disubstituted pyrrolizidines (3,5-P), histrionicotoxins (HTX), decahydroquinolines (DHQ), 1,4-

disubstituted quinolizidines (1,4-Q), allopumiliotoxins (aPTX), 5,8-disubstituted indolizidines (5,8-I),

spiropyrrolizidine (Spiro), 4,6-disubstituted quinolizidines (4,6-Q), dehyrdo-5,8-disubstituted indolizidines

(Dehydro-5,8-I), tricyclics (Tri), unclassified alkaloids (Unclass), piperidine (Pip), and previously undescribed

(New).

40

both 223AB and 231B (Saporito et al. 2007b, McGugan et al. 2016), and while no source of

219A has been identified, as it is a decahydroquinoline, the source for that alkaloid is likely ants

(Saporito et al. 2007a). These overlapping and common alkaloids suggest prey common to both

populations, which is not surprising as the two populations are less than 50km from one another.

Matoury shows a large diversity of 5,6,8-trisubstituted indolizidines, six of which (249C, 251T,

259C, 261B, 263A, and 265L) are implicated in predator aversive response. Allopumiliotoxins

and tricyclics are unique to Matoury while dehydro-5,8-indolizidines, 4,6-quinolizidines, and

piperidines are unique to Kaw. These classes are known to come from ants, mites, and beetles

(Saporito et al. 2007a), which given the population specificity of these alkaloid classes, suggests

differential availability of these sources between the two populations.

Notably, we present a unique assay for assessing unpalatability of alkaloid toxins in

poison frogs that explicitly accounts for biologically relevant predators and mode of ingestion.

Prior research has focused on LD50 (Daly and Myers 1967) or injection assays (Darst and

Cummings 2006, Maan and Cummings 2012), which does not address the fact that predators

experience alkaloid toxins through ingestion rather than injection into the bloodstream or

musculature (Weldon 2017), and thus, examining toxicity (i.e., response to toxins in the

bloodstream) does not address the how selection favors or disfavors distasteful frogs. Further, it

is possible that toxic compounds are not distasteful and vice versa. Thus, to understand the

evolutionary consequences of alkaloid sequestration, examining unpalatability, rather than

toxicity, is necessary. With over 800 alkaloid toxins known from poison frogs (Daly et al. 2005),

it is likely that palatability varies widely. Whole toxin profile examination provides insight into

how predators select prey, but it lacks the ability to identify the specific alkaloid components of

41

the diverse toxin cocktail driving response. By comparing individual response to toxin profiles,

we provide the first evidence of toxins that may be driving that response. Given the large variety

of alkaloids present in toxin profiles of D. tinctorius and the small sample size of this study,

interpretation of our results is limited but does provide a mechanism for future studies to narrow

down what alkaloids are driving predator response.

As D. tinctorius is highly polytypic throughout the Guiana Shield (Noonan and Gaucher

2006, Rojas et al. 2014a), future research will focus on further characterizing alkaloid diversity

among populations. Given that 16 of the 81 (19.7%) alkaloids described here have not previously

been described in the literature suggests that a large number of undescribed alkaloids will be

present in the other populations of D. tinctorius. Further, our research represents the first study

that seeks to understand the drivers of predator response. Future research should extend upon this

to determine if there is a small suite of toxins that are primarily responsible for predator

response. Doing so will give predictive power to future alkaloid characterization studies in how

predators will respond to toxins.

Effective aposematic signaling is a complicated interplay between a conspicuous signal

and a secondary defense. While a large amount of research has focused on the conspicuous

signal and the psychology of learned avoidance, considerably less research has focused on

understanding secondary defense complexity and its consequences for predator response. By

examining what toxins may be important in driving predator response, we can now begin to

predict how predators will response to highly varied secondary defenses both within and among

populations of aposematic prey. This will allow us to better address concepts such as

42

automimicry and honest signaling that are hypothesized drivers of diversification in aposematic

signals (Speed et al. 2006, 2010, Maan and Cummings 2012).

Acknowledgements

We would like to thank Ralph Saporito for running the GC-MS assays, the use of his

equipment, and expertise in analyzing toxin content among populations. We would also like to

thank Annelise Blanchette for her aid in extracting and identifying toxins. Finally, we would like

to thank Bibiana Rojas for her assistance in predation assays.

43

IV. DIFFERENTIAL RESPONSES OF AVIAN AND MAMMALIAN PREDATORS TO

PHENOTYPIC VARIATION IN AUSTRALIAN BROOD FROGS

Abstract

Anti-predator signaling is highly variable with numerous examples of species employing

cryptic coloration to avoid detection or conspicuous coloration (often coupled with a secondary

defense) to ensure detection and recollection. While the ends of this spectrum are clear in their

purpose, the purpose of intermediate signals is less clear. Australian Brood Frogs

(Pseudophryne) display conspicuous coloration on both their dorsum and venter. Coupled with

the alkaloid toxins these frogs possess, this coloration may be aposematic, providing a protective

warning signal to predators. We assessed predation rates of known and novel color patterns and

found no difference for avian or mammalian predators. However, when Pseudophryne dorsal

phenotypes were collectively compared to the high-contrast ventral phenotype of this genus, we

found birds, but not mammals, attacked dorsal phenotypes significantly less frequently than the

ventral phenotype. This study, importantly, shows a differential predator response to ventral

coloration in this genus which has implications for the evolution of conspicuous signaling in

Pseudophryne.

44

Introduction

Phenotypic coloration and pattern in prey species often serves to evade or deter potential

predators. These signals range from cryptic (i.e., background matching; (Ruxton et al. 2005)) to

conspicuous (i.e., aposematism; (Mappes et al. 2005)) with some signals serving both purposes

(Rojas et al. 2014b, 2014a). In some cases, how prey species utilize a phenotypic signal is easy

to discern (i.e., cryptic coloration of many moths or conspicuous coloration of poison frogs), but

in many cases, how signals are used is not clear (Valkonen et al. 2011). Signal interpretation is

dependent upon the observer and environmental conditions in which the signal is received. For

example, there is experimental evidence that coloration in the Strawberry Poison Frog (Oophaga

pumilio) is aposematic (Saporito et al. 2007c), however, when viewed by conspecifics, these

colors can affect mate choice (Maan and Cummings 2009). Even aposematism may be observer-

specific as modeling has demonstrated that conspicuous signals to avian predators may not be

conspicuous to other predators (i.e., crabs and snakes; Maan and Cummings 2012). The duality

of these signals can result in phenotypes that are complex, and perhaps not readily apparent as to

their purpose.

Pseudophryne are small, terrestrial Australian frogs exhibiting dorsal coloration ranging

from solid brown to brilliant yellow stripes on a black background (e.g., Figure 9A). As these

frogs possess distasteful alkaloid toxins (Smith et al. 2002, Saporito et al. 2011), this coloration

is thought to be aposematic (Williams et al. 2000). However, some Pseudophryne species have

little conspicuous coloration, often confined to inguinal flash marks. Research on Neotropical

poison frogs (Dendrobatidae) suggests there may be an inverse relationship between toxicity and

45

conspicuousness (Darst et al. 2006, Wang 2011), and it is certainly possible that this may be true

of Pseudophryne.

Further complicating matters, however, is that all Pseudophryne have conspicuous black-

and-white ventral patterns (Figure 9G). This coloration has also been hypothesized to function in

an aposematic manner (Williams et al. 2000) as disturbed frogs are likely to freeze position and

Figure 9: Dorsal and ventral phenotypes represented in the models

Dorsal (A) and ventral (G) phenotypes of Pseudophryne coriacea found in Watagans National Park. Dorsal

phenotypes brown (B), orange head (C), yellow head (D), corroboree (E), and flash marks (F) represent dorsal

phenotypes found in P. bibronii, P. australis, P. dendyi, P. corroboree, and P. bibronii/P. dendyi, respectively.

The ventral reticulated (H) phenotype is found throughout the genus. Local and novel (for the study site)

phenotypes are denoted by white circles with black letters and black circles with white letters, respectively.

46

not right themselves if flipped upside down (pers. obs.). Unlike the Neotropics (Saporito et al.

2007c, Noonan and Comeault 2009), mammals (e.g., marsupial predators such as quolls,

possums, and Antechinus) may be important predators of these frogs that contribute to signal

evolution. While dorsal yellows, reds, and oranges would be effective signals for birds which are

sensitive to long wavelength colors (Okano et al. 1992, Bowmaker 1998), many Australian

mammals have limited ability to perceive color (Ebeling et al. 2010), so high-contrast signals

would likely be more effective at deterring predation. Alternatively, this black-and-white

coloration could be disruptive, breaking up the shape of the frog (Allen et al. 2013). We

examined how predators in eastern Australia react to natural dorsal and ventral phenotypes both

known and novel to the region.

Methods

This study was conducted in the Watagans National Park (33°03’S, 151°20’E) in New

South Wales in July 2015. This site was chosen because three Pseudophryne species have been

recorded from the site (P. australis, P. bibronii, and P. coriacea). We constructed 1174 replica

frogs using Monster Clay® (Brodie 1993, Noonan and Comeault 2009) which retains evidence

of predation attempts, a method used in predation experiments of snakes (Pfennig et al. 2001,

Kikuchi and Pfennig 2010), frogs (Noonan and Comeault 2009, Chouteau and Angers 2011), and

even mice (Linnen et al. 2013). We poured melted clay into silicone molds made from plastic

model replicas (Yeager et al. 2011) that were 3D printed from a museum specimen of P.

bibronii. Models were approximately 30mm in length (snout-to-vent). Body shape is largely

47

conserved among Pseudophryne species, thus the use of P. bibronii is appropriate. All models

were brown with one of six phenotypes painted on them with acrylic paint (Chouteau and Angers

2011). Brown Pseudophryne are highly variable in the shade of the dorsum, and thus we could

not include that variability in our clay. For that reason, we used a single brown base color for all

models. Models were constructed to represent naturally occurring dorsal patterns (though some

were novel for the study site): brown, orange head, yellow head, corroboree, flash marks (Figure

9B-F), and the reticulated ventral pattern found in all species (Figure 9H). Notably, the

corroboree pattern was meant to mimic the pattern found in P. corroboree (Osborne 1989), a

species found ~360km away, representing a novel phenotype and not meant to be an exact

representation of P. corroboree, whose base color is normally black.

Models were placed on four transects separated by at least 100m (750m to 2.1km long).

The six model types were randomized with one model every 3m with equal numbers of each

phenotype on each transect (Noonan and Comeault 2009, Hegna et al. 2011, 2012). Models were

left for one week (Noonan and Comeault 2009, Comeault and Noonan 2011), after which models

were collected, and predation attempts recorded by noting bite marks. Avian attacks were

recognized by a U or V shape beak impressions as well as holes indicative of stabbing-type

attacks. Mammals were identified by heterodont tooth impressions. All research was conducted

under University of Newcastle Animal Care and Ethics Committee (ACEC) protocol number A-

2015-513. Research in Watagans National Park was conducted under New South Wales

Scientific Permit SL100190.

48

Data Analysis – Multiple attacks on a single models were scored as a single predation

attempt as we were not able to determine whether one predator attacked multiple times or

multiple predators attacked once (Noonan and Comeault 2009). Consecutive attacks on the same

model type would be counted as one attack to avoid nonindependence of attacks (Brodie 1993).

We observed four instances in which the same type of predator attacked consecutive models, but

in all four instances, consecutive attacks were on different phenotypes. Consequently, attacks

were counted independently as attacks on different phenotypes represented different decisions

made by a predator. Since missing models cannot be determined to be the result of predation or

detectability by the observer, missing models (N = 129) were excluded from the analysis.

Because the frequency of predation attempts in clay model studies are typically low

(Noonan and Comeault 2009, Kikuchi and Pfennig 2010), a G-Test of Independence is most

appropriate for examining frequency of attacks among model types. We used a G-Test of

Independence to compare predation attempts among phenotypes to determine whether there were

differences among phenotypes. This will allow us to examine if there are differential attacks by

phenotype compared to the null expectation of no difference among phenotypes.

Table 3: Distribution of attacks for each of the six different models for birds and mammals

Models labeled with an asterisk (*) are the local phenotypes that can be currently or historically found in

Watagans National Park.

Dorsal Signals Ventral Signals

Brown*

Flash

Marks

Yellow

Head Corroboree

Orange

Head*

Reticulated*

Birds 3 3 0 2 1 9

Mammals 5 4 6 3 5 4

Total

Collected 167 176 176 176 173 177

Total

Missing 29 22 22 18 21 17

49

Results

From 1045 recovered models, we recorded a total of 45 attacks from birds (N = 18) and

mammals (N = 27; Table 3). There was no statistical difference in number of attacks among

naturally occurring dorsal phenotypes (Figure 8B-F) for birds (G4 = 5.6, p = 0.23) or mammals

(G4 = 0.99, p = 0.91).

We further analyzed only the phenotypes that naturally occur in Watagans National Park

(orange head, brown, and reticulated [ventral]) and found that dorsal phenotypes (orange head

and brown) were attacked significantly less than the ventral phenotype (reticulated) by birds (G1

= 6.43, p = 0.01) but not mammals (G1 = 0.2, p = 0.65; Figure 10A). We also found birds, but

not mammals, attacked the reticulated ventral phenotype significantly more when compared to

all dorsal phenotypes (Figure 9B-C; birds: G1 = 9.4, p = 0.002; mammals: G1 = 0.18, p = 0.68;

Figure 10B), the local conspicuous phenotype (Fig. 9C; birds: G1 = 6.12, p = 0.01; mammals: G1

= 0.3, p = 0.59; Figure 10C), and the novel dorsal phenotypes (Figure 9D-F; birds: G1 = 8.5, p =

0.004; mammals: G1 = 0.08, p = 0.77; Figure 10D).

Of the 1174 models placed, 129 (10.9%) were missing. Notably, the majority of these

missing (71) came from a single transect which had a series of six consecutive bird attacks

before most models went missing down the transect, suggesting, perhaps a bird had discovered

the transect and proceeded to attack or obscure the rest of the transect. The disturbed leaf litter

where missing models occurred on this transect suggests an Australian Brushturkey (Alectura

lathami) or Superb Lyrebird (Menura novaehollandiae) discovered the transect and foraged

50

along it. Excluding this transect, and we had 58 missing models of 1050 models placed (5.5%)

which is consistent in clay model studies (e.g., 1.5% to 14%; Saporito et al. 2007c, Chouteau and

Angers 2011, Hegna et al. 2012). Nonetheless, there is no significant difference between missing

models when comparing dorsal to ventral models (G1 = 1.24, p = 0.266; Table 3).

Discussion

Pseudophryne are known

for their toxins (Smith et al. 2002)

and their presumed aposematic

coloration (Williams et al. 2000).

While dorsal coloration can be

variable within and among

species, ventral coloration is far

less variable. Our research

suggests dorsal coloration is an

important signal for avoiding

avian predation. We also provide

insight into the function of ventral

coloration in Pseudophryne.

While caution should be taken in

interpreting these results due to

the low attack rate (4.3%), this is

Figure 10: Attack proportions for birds and mammals on dorsal

and ventral reticulated phenotype

When combining all local dorsal models (A), all dorsal models (B),

the local conspicuous dorsal model (orange head; C), and all novel

dorsal models (D). Proportions are either nonsignificant (NS) or

significant (0.05 > p > 0.01, * and 0.01 > p > 0.001, **).

51

within the range of attack rates reported in other studies. For example, of the 1218 models

recovered by Hegna et al. (2012), a total of 91 were attacked (7.4%), but only 19 (1.5%) were

avian attacks compared to 18 (1.7%) reported in this study.

We found black-and-white ventral reticulation not to be effective at deterring avian

predators. This high-contrast signal had been thought to function as a signal to mammalian

predators as many Australian mammals have limited ability to discern color (Arrese et al. 2002,

Ebeling et al. 2010). We found no difference in the attack rates of dorsal versus black-and-white

signals by mammals suggesting the possibility that both dorsal and ventral signals serve an anti-

predator function for mammals, though it is impossible to differentiate equal avoidance or equal

disregard for signal. For the latter, it may be that traits not captured in plasticine models (i.e.,

smell) contribute to the deterrence of mammalian predators (Roper and Marples 1997).

Plasticine models are an excellent way to assess native predator response to known and

novel phenotypes of conspicuous prey (Noonan and Comeault 2009, Chouteau and Angers 2011,

Hegna et al. 2012, Kikuchi and Pfennig 2013), though use of models does have drawbacks.

Perhaps most notably, many predators, especially avian predators, are particularly attracted to

movement (Schwarzkopf and Shine 2014). And indeed, when incorporating movement into clay

model studies, attack rates increase for models, although conclusions between stationary and

moving models are unchanged (Paluh et al. 2014). While incorporating movement in models is

an important consideration, behavior of the prey species is equally important. Unlike Neotropical

poison frogs (Paluh et al. 2014), Pseudophryne are slow-moving and have the tendency to sit

motionless upon detection (Williams et al. 2000), thus stationary models are an accurate

representation of prey behavior.

52

Clearly, dorsal signals are more effective in deterring avian predators than ventral

coloration, which is consistent with the higher likelihood of avian predators viewing frogs from

above. Dorsal and ventral signals are equally effective against mammalian predators, possibly a

product of foraging mode that likely includes chemosensory cues and/or rooting behavior that

may expose frogs’ dorsal or ventral signals (they often remain immobile on their backs when

exposed). This research provides an important comparison to well-studied Neotropical systems

(Saporito et al. 2007c, Noonan and Comeault 2009, Chouteau et al. 2011), that also demonstrate

avoidance of dorsal signals. While numerous dendrobatid species have elaborate and colorful

ventral signals, to date, no research has examined whether these signals influence predator

behavior.

The reasons for model avoidance are somewhat varied and do deserve further attention.

With their conspicuous coloration and alkaloid toxins, Pseudophryne have been proposed to be

aposematic (Williams et al. 2000). However, other anti-predator strategies, such as disruptive

coloration, could explain the decreased attack rate reported in this study (Allen et al. 2013).

While not explicitly addressed in this study, future research will focus on disentangling why

avian predators attack dorsal phenotypes with reduced frequency when compared to ventral

phenotypes. Here, our results suggest that dorsal coloration in Pseudophryne has antipredator

function, possibly either through aposematism or crypsis (including disruptive coloration).

Conclusions

The research presented here represents the first exploration of signaling in Pseudophryne

in predator deterrence. Further research will focus on whether differential avoidance based on

predator type is widespread across Pseudophryne species or if this is solely confined to the

53

Watagans area in Australia. Additionally, a number of myobatrachid frogs possess a black-and-

white reticulated venter (i.e., Adelotus, Crinia, Arenophryne), indicating broad selection for this

phenotype. Studies of these species could help elucidate the purpose to this recurring phenotype

that our study suggests may play a role in deterring mammalian, but not avian, predators.

Acknowledgements

We would like to thank A. Jackman for her assistance in setting transects. This study was

conducted under University of Newcastle ACEC A-2015-509. Research in Watagans National

Park was conducted under NSW Scientific License SL100190.

54

V. WEAK WARNING SIGNALS CAN PERSIST IN THE ABSENCE OF GENE FLOW

Abstract

Aposematism is a defensive strategy whereby conspicuous coloration coupled with a

secondary defense deters predators from attacking. Novel conspicuous prey are selected against

because predator learning favors common warning signals (positive frequency-dependent

selection [FDS]). How, then, can novel phenotypes persist after they arise? Using a polytypic

poison frog (Dendrobates tinctorius), we explored the directionality and strength of selection on

variable aposematic signals using two phenotypically distinct populations. Surprisingly, the

common phenotype was not locally favored for either population. We found that this was due to

asymmetrical avoidance learning and generalization by predators of the two phenotypes. We

found that behavioral response to alkaloid secondary defenses was not influenced by overall

quantity of toxins. We propose that signals that are easily learned and broadly generalized can

protect rare, novel signals, and that, despite selective disadvantages, weak warning signals can

persist when gene flow among populations is limited. This provides a mechanism for the

persistence of intrapopulation aposematic variation, a likely precursor to polytypism and driver

of speciation.

55

Introduction

Aposematism is a widespread defensive strategy whereby organisms use conspicuous

coloration to warn predators about the possession of secondary defense (Poulton 1890), which

often is chemical in nature (i.e., alkaloid toxins, cardiac glycosides; Ruxton et al. 2005). In

principle, aposematic signaling should be subject to strong positive frequency-dependent

selection (FDS). This is because, in order to function efficiently, predators must recognize and

avoid the aposematic signal, and as such novel signals will then be selected against owing to

their rarity. Indeed, theoretical models (Endler and Mappes 2004), lab (Lindström et al. 2001),

and field experiments (Mallet and Barton 1989, Borer et al. 2010, Chouteau et al. 2016) support

this hypothesis. However, there are numerous examples in nature where aposematic species

display intrapopulation (O’Donald and Majerus 1984, Ueno et al. 1998, Nokelainen 2013, Rojas

and Endler 2013) or interpopulation (Daly and Myers 1967, Hegna et al. 2015, Roland et al.

2017) phenotypic variation (polymorphism or polytypism, respectively). While interpopulation

variation can adhere to expectations of strong positive FDS if local predators’ ranges do not

include multiple phenotypically distinct populations, polymorphism apparently violates the

predictions of FDS.

Understanding how polymorphism can persist is pivotal to understanding how and why

polytypic species occur (Gray and McKinnon 2007). Polytypism is likely to first start as

polymorphism (Hugall and Stuart-Fox 2012, McLean and Stuart-Fox 2014) as in variation likely

evolves within a population first before individuals within that population found new populations

leading to polytypism. A polymorphic population could then be fragmented or individuals within

the existing population could found new ones, for example. Following this, populations may

56

phenotypically diverge (approaching different adaptive peaks) and populations could become

fixed for different phenotypes due to local selective pressures and/or by random chance,

mediated by genetic drift; eventually resulting in a polytypic species or speciation. Thus, the

maintenance of phenotypic variation within an aposematic population is a key concept to

investigate.

Wright (1982) proposed the shifting balance theory using the framework of an adaptive

landscape. Through this landscape, there are peaks and valleys, and selection drives the mean

trait value of a population up a peak that represents a phenotypic optimum for local conditions.

Consequently, if a population’s phenotypic mean is to climb a different adaptive peak than the

one it is currently on, it must travel through a valley, which, generally, is selectively

disadvantageous. Thus, populations can become “stranded” on local peaks across the adaptive

landscape. These peaks can shift due to random genetic drift affecting genes that impact

phenotype, variability in secondary defenses that prey possess, and/or predator community

composition (Mallet and Joron 1999, Arias et al. 2016). As these adaptive peaks shift in space

and over time, individuals within a population will be subjected to differential selection pressures

that may allow for novel signals not only to persist, but also increase in frequency. This process

has been used as a possible explanation for polytypism (among population variation) in

aposematic signals (Mallet and Joron 1999, Chouteau and Angers 2012) as the phenotypic

differences among populations may represent different phenotypic optima for communicating

unprofitability of prey to their potential predators based on local environmental conditions. In

contrast, polymorphic populations may represent a population that has not yet reached an

57

adaptive peak (Mallet and Joron 1999, Mallet 2010), while polytypic populations represent

differing adaptive peaks where each population has reached a phenotypic optimum.

In order to address how novel phenotypes persist within an aposematic population, we

must examine the underlying mechanisms that promote or constrain phenotypic diversity in

aposematic signaling. Species that are both polymorphic and polytypic are ideal because we can

examine within and among population pressures that promote and/or constrain phenotypic

diversity. These include examining how natural populations of predators react to novel

phenotypes and how they learn and react to aposematic signaling. To do so, we examined two

populations of the Dyeing Poison Frog (Dendrobates tinctorius) from northeastern French

Guiana which differ in color, but not pattern (Figure 11A), and examined the ease with which

avian predators learn and avoid signals in order to infer what conditions could sustain the

evolution of novel signals.

We address the question of how phenotypic variation can persist in aposematic species by

examining two phenotypically distinct populations of D. tinctorius (Table 4). We assess whether

similarities between these two populations can be explained by ongoing migration. Following

this, we conducted four experiments. First, we created plasticine clay models of local and novel

phenotypes and presented them to wild, native predators. Second, we examined how model avian

predators learned avoidance and generalization of these aposematic signals. Third, we examined

variability in toxin content among the two populations. Finally, we assessed unpalatability of the

alkaloid toxins for each population using a model avian predator.

58

Methods

Field Sampling

The two focal populations were from Matoury, where frogs have white-stripes, and the

Kaw Mountains, where frogs have yellow-stripes, of French Guiana, hereafter referred to as

white and yellow, respectively. These sites were chosen because frogs have the same pattern

(striped on a black body with blue legs) but differ in the color of their stripes (white and yellow,

respectively; Figure 11C). We collected frogs from white (n = 10) and yellow (n = 8) populations

in 2013 and 2014. Frogs were euthanized by cervical dislocation and pithing in the field. Livers

were preserved in 95% ethanol, and whole skins were preserved in 100% methanol for

subsequent alkaloid analysis. Skins were stored in 4mL glass vials with PTFE caps. Photos,

Table 4: Summary table of the questions, hypotheses, and experiments used for this study

These questions and hypotheses seek to better understand the overall question of how phenotypic

diversity can persist in aposematic species.

Question(s) Hypothesis Experiment

Is there gene flow between

populations?

There is ongoing gene flow

between populations

Gene flow among

populations

How do predators respond to

known and novel signals in the

two populations?

Local phenotypes will be

protected, while novel phenotypes

will receive disproportionately

high predation pressure

Plasticine clay models in

the field

How to predators learn to avoid

different phenotypes, and when

learned, do they extend

experience to novel signals?

Different signals are equally

effective at learned avoidance and

do not evoke generalization to

novel signals

Learning and

generalization assays

using naïve model

predators

Does alkaloid content vary

between populations?

Alkaloid content varies between

populations

Alkaloid characterization

between populations

Do predators respond

differently to alkaloid content

between populations?

Predator response will vary in

relation to alkaloid variation

between populations

Behavioral response to

alkaloids by model

predators

59

demographic (i.e., sex, snout-vent length, etc.), and spectrometric data from for each individual

were collected before euthanasia.

Gene Flow Between Populations

To assess gene flow between populations, we sought to identify single nucleotide

polymorphisms (SNPs) across the D. tinctorius genome that could then be used to estimate

population demography. To identify SNPs, we employed the 3-enzyme restriction-site associated

DNA (3RAD) method of Graham et al. (2015) and Hoffberg et al. (2016). Samples were digested

in a reaction that contained 100ng of DNA, 20 units of each of three restriction enzymes (XbaI,

EcoRI-HF, and NheI-HF; New England Biolabs [NEB]), 1X NEB Cutsmart Buffer, 1µL of both

a forward and reverse double stranded adapter at 5µM and water to 15µL total volume. Enzymes

were chosen based on an in silico digest of the genome of Xenopus laevis which suggested these

enzymes would produce approximately 3,500 unique loci within the desired size range of 300-

350 base pairs (bp). Each adapter contained a 6-9bp barcode and each sample was assigned a

unique forward and reverse barcode combination. Digestions were carried out at 37°C for 1 hour

and followed immediately by the addition of ligation mix. As adapters were already present in

digestion reactions, ligation mix consisted of 1.5µL of 10 mM rATP (NEB), 0.5µL 10X ligase

buffer (NEB), 100 units T4 DNA Ligase (NEB), and water to 5µL. Ligation reactions were

incubated for two cycles of 22°C for 20 minutes and 37°C for 10 minutes followed by 20

minutes at 80°C. Ligated samples were then cleaned by mixing with 1.2 volumes of Sera-mag

Speedbeads (Fisher Scientific; prepared as in Rohland and Reich 2012), two washes of 70%

ethanol, and eluted from beads with 20µL of IDTE pH 8 (Integrated DNA Technologies).

Samples were then amplified in 20 µL reactions with 10µL of post-bead template, 1X HIFI

60

Buffer (Kapa Biosystems), 0.75 µL dNTPs (Kapa Biosystems), 0.5µL HIFI DNA polymerase

(Kapa Biosystems), 0.5 µM iTru5 and iTru7 primers containing sample-specific 8bp barcodes,

and H2O to 25 µL total volume. Library amplification began with 3 minutes at 95°C followed by

16 cycles of 98°C for 20 seconds, 55°C for 15 seconds, and 72°C for 30 seconds, followed by a 5

minute extension of 72°C.

Success of individual PCRs was determined by gel electrophoresis. When digestion,

ligation, and amplification were successful (as evidenced by a smear), 10µL from each PCR

were pooled into sets of 24. Each pool was concentrated down to 30 µL using a QIAGEN PCR

purification column and then the pool was run in a single lane of a 1.5% Pippin Prep gel cassette

(Sage Science) selecting for fragments sized 400-450 bp. Selected sized fragments were removed

from Pippin elution wells and used directly in qPCR library quantification (Kapa Biosystems)

and Illumina Sequencing on a NextSeq 500 to obtain single-ended, dual-indexed reads of 75 bp.

Raw reads were downloaded from Basespace using basespacerundownloader (Illumina)

and demultiplexed using BCL2FASTQ (Illumina) allowing up to 2 bp errors in barcodes as our

barcodes had 3bp degeneracy built in. After demultiplexing, each sample was trimmed from the

beginning of each read to remove remaining internal (adapter) barcode and restriction site

overhang. All samples were then trimmed to 63 bp in length to prevent downstream difficulties

in assignment of homology. Loci were identified within individuals and aligned among

individuals using pyRAD (Eaton 2014).

Population Genetic Analysis – We used the SNP data with the Bayesian coalescent

program G-PhoCS (Generalized Phylogenetic Coalescent Sampler; Gronau et al. 2011) to

61

estimate rates of migration between the two populations. Six independent, simultaneous runs

were conducted, each with a 500,000 MCMC iterations, which were then compiled for a total of

3,000,000 iterations. We used a 10% burn-in, which resulted in 2,700,000 iterations from which

summary statistics were calculated. We defined a constant mutation rate in order to conduct

these analyses. Migration rates were then calculated in G-PhoCS between the two populations

(from yellow to white and from white to yellow).

Avoidance of Novel Phenotypes

Plasticine clay model experiments are commonly used to assess predator responses to

novel signals in aposematic animals (Noonan and Comeault 2009, Comeault and Noonan 2011,

Paluh et al. 2014). These allow observation of predation attempts, which is an otherwise rare

phenomenon. We constructed 1320 replica frogs using Van Aken polymer modeling clay

(Brodie 1993, Noonan and Comeault 2009) which does not harden and retains evidence of

predation attempts. We poured melted clay into silicone molds made from plastic model replicas

(Yeager et al. 2011). Models were 45 mm long (snout to vent). In order to assess how predators

responded to novel colors and patterns, we created models of four different phenotypes: yellow

stripes on a black dorsum, white stripes on a black dorsum, solid yellow dorsum, and solid white

dorsum (Figure 11B). All models had blue legs (as do both populations sampled). Both

dendrobatid frogs and the polymer clay have been shown to lack UV reflectance (Saporito et al.

2007). Both stripes and eyes were created with clay and affixed to models (Comeault and

Noonan 2011).

62

Three model types were randomized with one model every 5 m and equal numbers of

each phenotype on each transect (Noonan and Comeault 2009, Hegna et al. 2011, 2012). The

three model types for each location were local color and pattern, local color and novel pattern,

local pattern and novel color, ensuring that at least one component of the aposematic signal was

familiar to predators. Transects were separated by at least 100m. We placed seven transects from

495m long to 1.5km long in the Matoury (white stripe) location and 13 transects from 495m long

to 1.5km long in Kaw (yellow-striped; Figure 11). Models were left for 72h to allow for

predation attempts (Noonan and Comeault 2009, Comeault and Noonan 2011). Following this

period, we collected models and determined which ones had been preyed upon by scoring bite

marks. Avian attacks were recognized by a characteristic U or V shape mark as well as stabbing

attacks. While mammals and invertebrates are commonly reported as attacking plasticine

models, avian predators are likely most important in driving phenotypic diversity in conspicuous

signals because they have the capability to discern different colored phenotypes (Llaurens et al.

2014).

Clay Model Data Analysis – If models had multiple bite marks, they were scored as a

single predation attempt as we were not able to determine whether one predator attacked multiple

times or multiple predators attacked once (Noonan and Comeault 2009). Missing models were

excluded from the analysis (43 of 1093 and 57 of 1316 for white and yellow, respectively).

We used a G-Test of Independence (Noonan and Comeault 2009, Chouteau and Angers

2011) to compare predation attempts among phenotypes at both sites to determine whether

aposematic signal is a predictor of predation risk. Such chi-square methods are often employed

in clay model studies to examine null expectations of equal attack rates among model types

63

(Brodie 1993, Noonan and Comeault 2009, Comeault and Noonan 2011, Chouteau and Angers

2011). Because predation rates are typically low in clay model studies (Noonan and Comeault

2009, Kikuchi and Pfennig 2010), a G-Test of Independence is most appropriate for examining

distributions of attacks among model types. Additionally, we tested (G-Tests) whether pattern,

irrespective of color (and vice versa), was a predictor of predation risk. If results were found to

be significant, pairwise G-Tests were conducted between the local morph and novel morphs.

Avoidance Learning and Generalization

While avian predators are the likely drivers of aposematic signal evolution in D.

tinctorius (Noonan and Comeault 2009, Comeault and Noonan 2011), we used a model avian

predator to assess how naïve predators learn and extend experience (generalize) with aposematic

signals. To do this, we trained naïve one-week old chickens (Gallus gallus domesticus) to eat

mealworms associated with a picture of either a yellow-striped or white-striped frog. To examine

how naïve predators learn color signals with varying levels of distastefulness, chicks were

equally divided (N = 15 per treatment; 60 chicks total) into four different treatments: 5%

chloroquine with a yellow signal, 5% chloroquine with a white signal, 10% chloroquine with a

yellow signal, and 10% chloroquine with a white signal. We soaked mealworms in a chloroquine

solution (5% or 10%), which is known to be distasteful, but not harmful, to birds (Brodie 1993,

Lindström et al. 1999). Over successive trials, we examined whether chicks would learn to avoid

distasteful mealworms (defined as three consecutive refusals of a distasteful mealworm;

Exnerová et al. 2015), and whether avoidance was related to the associated color and/or

chloroquine concentration. Signals were illustrations of frogs with either white or yellow stripes,

on a black background and blue legs, similar to the clay models. White and yellow colors were

64

taken from photographs of frogs in the white and yellow populations, respectively, using the

color dropper tool in Photoshop. The trials were done in 30cm x 60cm wooden compartments

under full spectrum lights. Chicks were placed individually into a compartment and allowed to

habituate for 2 hours. Chicks were food-deprived during this acclimation period to ensure

motivation to feed during the trials.

Once in the compartment and after the 2h acclimation period, training consisted of

teaching the chicks to eat a dried mealworm (Tenebrio molitor) from a petri dish on top of an

illustration of a brown frog on a tan background. The training phase was completed once the

chick had eaten three consecutive times, after which we allowed the birds to rest for a period of 5

min.

The avoidance-learning trials consisted of the consecutive presentation of mealworms on

a petri dish on top of an illustration of D. tinctorius on a tan background. (i.e., a frog with either

white or yellow dorsal stripes). In order to make them unpalatable, the mealworms were soaked

in a solution of 5% or 10% chloroquine for at least 1 h but no longer than 3 hours. We recorded

the latency (i.e., the time until the chick approached and picked up the mealworm) and noted any

behavioral reaction to the distastefulness of the mealworm after tasted or eaten. Such behaviors

most often involved beak wiping and head shaking. Each trial ran for 5 min., followed by 5 min.

rest, after which a new petri dish with an unpalatable mealworm was offered. This procedure was

repeated until the chick refused to eat the mealworm over three consecutive trials, at which point

the chick had learned the signal. The test ended either when chicks “learned” the signal or

proceeded through 10 trials, whichever came first. If chicks proceeded through 10 trials without

three consecutive refusals, they were not considered to have learned the signal. When a chick did

65

not eat the chloroquine-soaked mealworm on the D. tinctorius pattern, a palatable mealworm

was offered on the neutral (brown frog) background, which the chick did not associate with an

unpleasant experience. In that way, we made sure that the chicks refrained from eating the

treated mealworm because of an association with the warning signal, rather than because of

satiation. In five cases, chicks stopped eating and subsequently refused the palatable mealworms

and as a result, were excluded from analysis. In these cases, we tested additional chicks to ensure

each treatment had 15 replicates. We registered the number of trials (i.e., mealworm consumed)

that it took for the chick to learn to avoid the signal.

The next step was to run a generalization trial, which aimed to test whether once a signal

was learned, the aversion would be extended to other signals. Therefore, after the three trials in

which the chick would refrain from eating the presented mealworm, a new unpalatable

mealworm was presented in association with a different aposematic signal. Chicks that had

learned to avoid yellow-striped frogs were then presented with white-striped frogs, and vice

versa. We recorded the chick’s response both as a binary variable (whether the mealworm was

eaten or not) and the hesitation time. Generalization trials were only conducted on chicks that

learned to avoid a signal in the first set of trials. Chicks were considered to have generalized

avoidance to the novel signal if they avoided the first presentation of the novel signal. This

situation best simulates choices wild predators would make when encountering novel signals. All

trials, both avoidance learning and generalization, were filmed in order to extract details on the

chick’s behavior afterwards.

Learning Experiment Data Analysis – We analyzed the data from the learning

experiments using a generalized linear mixed model (GLMM) in two ways. First, we examined

66

the effect of color and chloroquine content (5% vs. 10%), and the interaction between the two,

on the probability of attacking a mealworm and the latency of attack, using chick ID as a random

factor, and a binomial distribution. Then, for each chick, we counted the number of trials in

which the mealworm was attacked, to a maximum of ten, and whether or not the chicks learned

to avoid the signal they were presented (as defined by three consecutive refusals). In the first

case, we used a Poisson error distribution, whereas in the second we used a binomial distribution.

In both cases, our predicting variables were color, chloroquine concentration, and the interaction

between the two. All analyses were done in R (R Core Team 2016), with the RStudio interface

and using the package lme4 (Bates et al. 2015).

Population Variability of Alkaloids

For whole frog skins from 18 frog specimens, alkaloids were extracted and analyzed from

frog skins using the procedure outlined in Bolton et al. (2017), which is described here briefly.

One mL of each methanol extract was transferred into a 10 mL conical vial and 50 µL of 1N

hydrochloric acid was added. Each sample was then mixed and evaporated with nitrogen gas to a

volume of 100 µL. Subsequently, each sample was diluted with 200 µL of deionized water. The

samples were then extracted with four 300 µL portions of hexane. The resulting hexane (organic)

layer was disposed of and the remaining aqueous layer was basified with saturated sodium

bicarbonate. Basicity was tested with pH paper. Once basic, each sample was extracted with

three 300 µL portions of ethyl acetate. Anhydrous sodium sulfate was added to the newly

extracted mixture to remove any excess water. The remaining samples were carefully evaporated

to dryness with nitrogen gas. Alkaloid fractions were resuspended in 100 µL of methanol and

stored at -20 °C.

67

Gas chromatography-mass spectrometry (GC-MS) was performed for each individual

frog on a Varian Saturn 2100T ion trap MS instrument, which was coupled to a Varian 3900 GC

with a 30 m x 0.25 mm inner diameter Varian Factor Four VF-5ms fused silica column. GC

separation of alkaloids was achieved using a temperature program from 100 to 280°C at a rate of

10°C per minute with helium as the carrier gas (1 mL/min). Each alkaloid fraction was analyzed

in triplicate with electron impact MS and once with chemical ionization (CI) MS with methanol

as the CI reagent.

Individual alkaloids of D. tinctorius were identified based on comparison of mass

spectrometry properties and GC retention times with those of previously reported alkaloids in

dendrobatid frogs (Daly et al. 2005). Alkaloid quantities for each individual frog were calculated

by comparing the average observed peak area of individual alkaloids to the average peak area of

the nicotine standard from the triplicate EI-MS analyses using a Varian MS Workstation v.6.9

SPI. Only alkaloids that were present in quantities of ≥ 0.5 μg were included in the analyses

(Bolton et al. 2017).

Predator Response to Alkaloids

Prior research into alkaloids has largely focused on toxicity of the toxins through either

LD50 (Daly and Myers 1967) or subcutaneous injections (Darst and Cummings 2006, Maan and

Cummings 2012), which, while informative, are limited in their evolutionary significance as

predators experience alkaloids through ingestion, not injection into the bloodstream or

musculature (Weldon 2017). Consequently, we sought to examine unpalatability of alkaloids as

unpalatability is what causes predators to make decisions. We used 1 mL of methanol fraction

68

from the skins abovementioned frogs (10 white, 8 yellow) and prepared toxins for two

unpalatability assays.

The two different assays were identical except for the skin secretion toxin preparation.

The purpose of doing two assays is to attempt to 1) understand how toxins influence predator

response and 2) ensure that non-toxin components of skin secretions do not disproportionately

affect results and mask potential aversion to the alkaloids. For each assay, we tested one frog

sample with one bird (white, N = 10, yellow, N = 7), as well as having a control treatment,

though these differed in number of birds used for the control treatment with the first assay using

six birds and seven birds for the second assay.

For the first assay, we tested equal concentrations of skin content. We evaporated 1 mL

of methanol extracts to dryness and weighed to determine how much toxin was present for each

sample (notably, this included everything in the methanol such as mucus, peptides, and fatty

acids). Samples were then reconstituted in ethanol such that the concentration of toxin for each

sample was the same based on the mass of the dried sample (1:1 toxin mass to ethanol). Toxins

were then transferred to oats and allowed to dry. The amount of toxin extract on each oat was

0.15 mg. This was in an effort to control for the quantity of toxins present in the extracts and not

allow individuals with high or low amounts of alkaloids to skew results. However, if non-toxin

content varies among frogs, this could act as a diluting effect on toxins, thus we performed a

second unpalatability assay to account for variation.

In our second assay, we tested equal proportions of skin content. We took 1mL of

methanol fraction and evaporated to dryness under N2, and then samples were reconstituted with

69

0.5mL ethanol regardless of dry mass of skin secretions. We added 15 µl of the reconstituted

sample to oats and allowed the oats to dry. This would counter potential effects that other

components of skin secretions may have in diluting the effect of the alkaloids. For each assay,

we tested one frog sample with one bird (white, N = 10; yellow, N = 8), as well as having a

control treatment, though these differed in number of birds used with the first assay using six

birds and seven birds for the second assay.

Prior to experiments, Blue Tits (Cyanistes caeruleus) were trained to eat oats. After

training, birds were given unpalatable oats which were not modified in any way other than

addition of toxins (Rojas et al. 2017). Aversive behavior (i.e., beak wiping) and proportion of

oats eaten were recorded. All trials were filmed to be able to accurately score behaviors.

For each assay, each of two oats were soaked with 15 µl of the extract of one frog skin

and left for 24 h at room temperature to ensure that all ethanol had evaporated. Two other oats

were soaked with 15 µl each of pure ethanol and used at the beginning and end of the experiment

with each bird. The first ethanol-only oat needed to be consumed entirely by the bird before the

experiment could be initiated, and the second ethanol-only oat was offered in the final trial to

ensure that the birds were not refusing to eat the oats coated with toxins out of satiation or lack

of motivation to eat in general. Birds in the control treatment received oats soaked with pure

ethanol for all trials in order to compare directly the response of birds to oats containing frogs’

toxins versus oats with ethanol only.

Each oat was presented on a hatch that had a visual barrier, which allowed us to detect

the exact moment at which the oat was seen, which set the actual beginning of the trials. We

70

measured the number of times the bird wiped its beak, which is a known aversive behavior

(Roper and Marples 1997), and the percentage of the oat eaten. Birds were watched for a 2 min.

period after they finished eating the oats, or for a maximum of 5 min. in the instances in which

the oat was not fully eaten, to make sure that any delayed response to the oat taste was not going

to be missed. Data were analyzed using Generalized Linear Mixed Models in which frog

population was entered as the predicting variable, while number of times the beak was wiped and

percentage of oat eaten were entered as the response variables. Because the response of each bird

was measured twice, we included bird ID as a random factor. Given that the duration was not the

same for all trials, we used the function “offset” to account only for the effect of population on

our response variable once the effect of duration was removed. All analyses were done in R (R

Core Team 2016) interface and the package lme4 (Bates et al. 2015).

Results

Gene Flow Between Populations

Our RAD fragment selection resulted in the sequencing of 1505 SNPs across the D.

tinctorius genome. These data were used to estimate migration between the two populations

(Gronau et al. 2011). For both populations, migration rates were found to be < 0.0001 (Figure

11A), effectively indicating complete genetic isolation.

Avoidance of Novel Phenotypes

71

As tetrachromatic predators (i.e., birds) are most likely driving phenotypic diversity in

these frogs, our results only represent avian predators, but it is worth mentioning that our models

were also attacked by mammals and arthropods.

White population – We recovered 364 yellow-striped, 362 white-striped, and 367 solid

white models for a total of 1093 models. Of the 1093 models, 9 yellow-striped, 23 white-striped,

and 16 solid white models were unambiguously attacked by bird predators (Figure 11B).

Interestingly, the local morph, white striped, was attacked more often than expected by chance

(G2 = 6.661, p = 0.036). Pairwise G-Tests between local (white-striped) and novel morphs

indicate that avian predators avoided yellow stripes significantly more than white stripes (G1 =

6.153, p = 0.013). Though the local form was attacked more than the solid white form, this

Figure 11: Distribution, predation, and spectrometry of the white (A) and yellow (B) populations.

A) Map in northeastern French Guiana of the two populations displaying migration rate (m) between the two

populations and estimated mutation-scaled effective population sizes (θ). B) Distribution of attacks on clay

models within the two populations, and C) spectrometric curves of the two populations.

72

difference was not significant (G1 = 1.292, p = 0.256). Birds did avoid yellow more than white

(G1 = 5.474, p = 0.019) but showed no aversion to pattern (G1 = 0.005, p = 0.942).

Yellow population – We recovered 438 yellow-striped, 438 white-striped, and 440 solid

yellow models for a total of 1316 models. Bird predators attacked 20 yellow-striped, 23 white-

striped, and 23 solid yellow models (G2 = 0.353, p = 0.838; Figure 11B). Birds exhibited no

differential avoidance of color (G1 = 0.074, p = 0.785) or pattern (G1 = 0.1, p = 0.752).

Avoidance Learning and Generalization

We found a significant interaction between color and chloroquine concentration affecting

both the probability of attack (estimate ± SE = 0.963 ± 0.382; z = 2.52; p = 0.012), and the

latency to attack

(estimate ± SE = -

0.948 ± 0.411; z = -

2.30; p = 0.021).

Likewise, the

interaction between

color and chloroquine

concentration had a

significant effect on

the number of trials in

which the chicks

attacked the

Figure 12: Results of the learning experiments for white and yellow models

Chickens (Gallus gallus domesticus) were exposed to either the yellow or white

treatment which were each split into a high (10%) and low (5%) chloroquine

treatment. The results are characterized by A) mean latency to attacking distasteful

prey and B) number of trials in which a bird attacked a mealworm. Significant

differences between treatments are denoted by * (p < 0.05).

73

mealworm (estimate ± SE = 0.564 ± 0.239; z = 2.36; p = 0.018). Based on this, mealworms with

a high chloroquine concentration (highly unpalatable) presented on the white signal were more

likely to be attacked, elicited a shorter latency to attack (Figure 12A), and were subject to a

higher number of attacks across trials (Figure 12B) in comparison to highly unpalatable

mealworms offered in association with a yellow signal. However, whether or not the chicks

learned the offered signal depended only on the signal color (estimate ± SE = 2.398 ± 0.870; z =

2.755; p = 0.006), and not on the chloroquine concentration (estimate ± SE = 0.878 ± 0.780; z =

1.125; p = 0.260). Thus, chicks who were offered mealworms in association with a yellow signal

were more likely to learn to avoid them, regardless of distastefulness.

When chicks learned to avoid colors, they were subjected to generalization tests in which

they were exposed to the alternative signal (e.g., learned to avoid yellow, then exposed to white).

Again, chloroquine concentration was found to have no effect (estimate ± SE = -0.911 ± 0.999; z

= -0.911; p = 0.362), but signal color did (estimate ± SE = 1.950 ± 0.974; z = 2.001; p = 0.045),

such that chicks that had learned to avoid yellow were more likely to also avoid white than the

converse (Figure 13).

Population Variability of Alkaloids

We found no difference in toxin content from the two populations (white mean ± SE:

471μg ± 80μg; yellow mean ± SE = 249μg ± 33μg; Mann-Whitney U-Test: W = 54, p = 0.070;

Figure 14), though this likely is due to our small sample size. One yellow individual had eight

times more than the population average (yellow outlier = 2040μg) and this outlier was excluded

from analysis. The white population possessed 49 different alkaloids (with multiple isomers for

74

some) representing 11 different structural

classes. These included eight alkaloids that

had not been isolated from poison frogs. In

contrast, the yellow population possessed

46 different alkaloids (with multiple

isomers for some) representing 12 different

structural classes. This also included nine

alkaloids that had not been isolated from

poison frogs (including one new piperidine;

Table 5).

Predator Response to Alkaloids

Our equal concentrations experiment

revealed that birds exposed to skin extracts

from the yellow population showed aversive behavior (beak wiping) at a significantly higher rate

than those exposed to the ethanol control (Estimate ± SE = 1.250 ± 0.520, z = 2.400, p = 0.016;

Figure 15A) and to the extracts of white frogs (Effect = 1.030 ± 0.380, z = 2.669, p = 0.008;

Figure 15A), suggesting that yellow frogs are more unpalatable. These differences were not

influenced by the quantity of alkaloids found in frogs’ skin (Estimate ± SE. = 0.189 ± 0.226, z =

0.838, p = 0.402). Beak wiping in response to the extracts of white frogs did not differ from that

to oats soaked with ethanol only (Effect = 0.153 ± 0.520, z = 0.293, p = 0.769), and none of the

populations differed from the controls in the proportion of oats eaten by the birds (Figure 16A).

Figure 13: Proportion of chickens that generalized to

a novel signal

Bars in white represent the proportion of birds that

learned avoidance of the white signal and were exposed

to a novel yellow signal while the yellow bars are the

opposite (learned yellow, exposed to novel white).

Differences denoted by * (p < 0.05) or NS

(nonsignificant).

75

Table 5: Distribution of alkaloids in the white (*) and yellow (†) populations

If multiple isomers were found of a particular alkaloid, the number of different isomers is depicted in parentheses. Alkaloid classes are as follows 3,5-

disubstituted indolizidine (3,5-I), 5,6,8-trisubstituted indolizidine (5,6,8-I), 3,5-disubstituted pyrrolizidine (3,5-P), histrionicotoxin (HTX), decahydroquinoline

(DHQ), (1,4-Q), allopumiliotoxin (aPTX), 5,8-disubstituted indolizidine (5,8-I), spiropyrrolizidine (Spiro), 4,6-disubstituted quinolizidine (4,6-Q), dehydro-5,8-

disubstituted indolizidine (Dehydro-5,8-I), Tricyclic (Tri), Unclassified (Unc), piperidine (Pip), and new, undescribed (New) alkaloids. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

3,5-I 5,6,8-I 3,5-P HTX DHQ 1,4-Q aPTX 5,8-I Spiro 4,6-Q Dehydro

– 5,8-I

Tri Unc Pip New

195B† 193G† 209Q(2)*† 235A(2)*† 195A† 231A*† 305A* 243C* 236*† 195C† 265T† 205B* 209G† 213† 171†

223AB(3)*† 195D*† 223B(3)*† 238A† 195J(2)† 235U* 339A* 237D† 275I† 205E* 209M* 193*

275C*† 205A† 251K(2)*† 245A* 219A(5)*† 207G* 227† 207*

207C† 259A* 221D† 235BB(2)† 209(2)*†

219N* 261A† 223F† 247M† 217*

223A† 283A† 243A(7)*† 249N* 223†

225K*† 285A† 245Q† 229†

231B(3)*† 285C† 269B* 233†

233G* 291A* 235†

235E* 245*

236A* 247(2)*†

237C(2)† 253*†

245G* 275*

249C*

249BB(2)*

251T(3)*

259C(2)*

261B*

263A*

265L or

U*

265U*

265L(4)*

267R*

Total

White

2 18 3 4 3 2 2 1 1 0 0 3 2 0 8

Total

Yellow

3 8 3 6 7 1 0 1 1 2 1 0 4 1 8

76

In our equal proportions

experiment, we found both populations to

elicit significantly higher beak wiping

(yellow: Estimate ± SE = 1.627 ± 0.485; z

= 3.35; p < 0.001; white: Estimate ± SE =

1.318 ± 0.461; z = 2.856; p < 0.01; Figure

16B) than the ethanol-soaked oats, but no

differences between the two populations.

Oats coated with the yellow population

extracts were eaten significantly less than

control oats (estimate ± SE = -0.6807 ±

0.2827; z = -2.408; p = 0.016; Figure

16B), while those coated with extracts from the white population did not differ from controls.

These results seem to point at both populations being unpalatable to some extent, but a tendency

for the frogs from Kaw Mountains (yellow) to be less palatable than those from Matoury (white).

In order to facilitate synthesis of the results of our five experiments, we summarized them

in Table 6. We will reference these results with the assay number in our conclusions.

Discussion

Aposematic signals under strong positive FDS should quickly reach an adaptive peak

(Mallet and Joron 1999). Poor or ineffective signals should quickly be selected against as the

population is driven to an optimal signal given the ecological and evolutionary conditions

present. We provide evidence that, while having a selective disadvantage in the field (Assay 2)

Figure 14: Alkaloid quantities of the two populations

77

and the lab (Assay 3), a weak signal can persist in the presence of limited gene flow (Assay 1).

This research supports the idea of shifting balance (Wright 1982) being a possible explanation

for the existence of polytypic populations. The yellow and white populations examined here

appear to represent different adaptive optima based on the local conditions. While the genetically

isolated Matoury population possesses an inferior aposematic signal, this may represent a

suboptimal adaptive peak. Given the asymmetry between the two populations in terms of

protection conferred by novel signals (Assay 2 and 3), our research highlights a plausible

scenario where a novel white signal evolved within a yellow population and subsequently was

isolated through stochastic processes (i.e., genetic drift, founder effects, etc.). Following this, the

white population climbed its adaptive peak, which was limited by gene flow among neighboring

populations, resulting in fixation of the white phenotype.

Table 6: Summary table of the questions and results of our experiments

Experiment Question(s) Species Tested Results and Conclusions

Assay 1

Population

Genetics

Is there gene flow between

yellow and white?

White and Yellow

Dendrobates

tinctorius

Figure 11A; Virtually no migration (m <

0.0001) between populations

Assay 2

Clay Models

How do predators respond

to known and novel

signals in the yellow and

white populations?

Native avian

predator

community

Figure 11B; White stripes attacked more

in white stripe population; no differences

among models in yellow population

Assay 3

Learning

Experiments

How to predators learn to

avoid yellow and white,

and when learned, do they

extend experience to novel

signals?

Naïve Chickens Figure 12 and 13; Naïve predators learnt

to avoid yellow faster, hesitate longer,

and extend experience to novel signals

when compared to white

Assay 4

Alkaloid

Content

Does alkaloid content vary

between populations?

White and Yellow

Dendrobates

tinctorius

Figure 14, Table 5; Wide variation in

toxin content and quantity among

populations

Assay 5

Unpalatability

Do predators respond

differently to alkaloid

content between

populations?

Blue Tits Figure 15 and 16; Alkaloid content

significantly different from controls, but

variably different among populations

78

In examining the white

and yellow populations, we

demonstrated how a weak

aposematic signal can persist

and, further, how novel signals

could persist in a population

despite expectations of strong

positive frequency-dependent

selection operating against

individuals that display such

signals. The two populations

examined here have virtually no gene flow between them (Assay 1), suggesting that the apparent

selective disadvantage of the white signal cannot be recovered from more adaptive alleles from

the yellow population (Assay 3). We found that predators have greater difficulty learning

avoidance of the white signal (Assay 3), and those that do are less likely to generalize novel

signals (Assay 3). Further, we showed that this apparent weakness is not overcome by the

presence of greater quantities of alkaloid toxins (Assays 4 and 5). While we found no differences

between populations means of alkaloid quantity (Assay 4), which is likely due to a low sample

size, we found no effect of alkaloid quantity influencing avian behavior (Assay 5). Toxin

samples ranged from 104 μg to 833 μg (Figure 14), yet we detected no difference in predator

Figure 15: Results of unpalatability experiments using Blue Tits

(Cyanistes caeruleus)

Beak wiping is significantly higher in response to extracts from the

yellow population compared to controls in A) 2015 and B) 2017 and

compared to extracts from the white population in the 2015 assay.

Differences denoted by *** (p < 0.001), ** (0.001<p < 0.01), *

(0.01< p < 0.05), and NS (nonsignificant).

79

behavior. Coupled with the

interaction observed in the

chloroquine learning

experiments (Assay 2), we

can conclude that overall

alkaloid quantity found in

these frogs is not an accurate

predictor of strength of

secondary defense and that

there is likely a subset of the

alkaloids present in these

frogs that is driving predator response. We demonstrate that a weak signal cannot be overcome

simply by a stronger defense, and conversely, a strong signal can further be enhanced by an

increased defense (Assay 3). Our unpalatability assay provided the first practical assessment of

distastefulness of alkaloids in avian predators. Prior research has focused on toxicity of alkaloids

via injections (Daly and Myers 1967, Maan and Cummings 2012), even suggesting unpalatability

cannot be empirically assessed for toxins (Darst and Cummings 2006). While informative,

toxicity assays do not give an evolutionary context to defensive compounds as predators

experience these alkaloids through taste, not via toxicity effects in the bloodstream (Weldon

2017). In contrast, our results show that we can quantitatively assess variation in unpalatability

among skin secretions.

Figure 16: Percentage of oats eaten per second in the assays done in

A) 2015 and B) 2017

Differences denoted by * (0.01< p < 0.05) and NS (nonsignificant).

80

Our results also showed how experienced avian predators response to novel signals when

they learned either yellow or white frog pattern (Assay 3). Predators were more likely to avoid

attacking a novel signal if they have previously seen yellow signals. Importantly, this is

corroborated by our field experiments with clay models (Assay 2). In the yellow population,

counter to expectations of antiapostatic selection (Lindström et al. 2001), we observed no

difference between known and novel phenotypes, indicating that the experienced predators in

these areas were likely to avoid attacking novel signals, thereby allowing novel signals to persist

in the population (if/when they evolve within a population). Interestingly, we observed that the

local form is attacked significantly more than the novel (Assay 2), yellow-striped form, which,

based on our lab experiments, can be explained by the difficulty predators have in learning

avoidance of the white form and, when they do learn to avoid white, they were less likely to

extend that avoidance to novel forms (Assay 3).

The most obvious question from the above results is why does the white population

persist despite possessing a weak aposematic signal. We would expect that if there was exchange

of alleles between the white and yellow populations, there would be strong selection against

white coloration and that the population should quickly become fixed for yellow stripes.

However, we observe virtually no gene flow between the two populations, and as a result, the

individuals in the white population persist in presenting a suboptimal aposematic signal (Ruxton

et al. 2007, McLean et al. 2014). How populations of this species or other polytypic species

become isolated from one another is subject to debate, but it seems plausible that a subset of a

polymorphic population become isolated for some reason (i.e., migration, stochastic processes,

etc.). These population isolates are constrained by the allelic diversity of the original founders,

81

and under this premise, a weak aposematic signal may persist. It is also important to note that our

lab experiments do demonstrate that avian predators can learn to avoid white and their skin

secretions are distasteful, so while it may not be as efficient a signal as yellow, it does apparently

offer some protection. In fact, it has been previously suggested that weak signals could evolve

and be maintained if predators are variable in their response towards defended prey (Endler and

Mappes 2004, Nokelainen et al. 2014), and if accompanied by a behavioral adaptation (Willink

et al. 2013). Individuals in the white population, for example, appeared to be more secretive and

less bold than individuals in the yellow population (JPL, pers. obs.). This behavioral adjustment

coupled with reduced predation pressure (i.e., see clay model experiment) may aid in the

persistence of this population despite their suboptimal signal.

Our research provides important insight into the function and persistence of aposematic

signals. Notably, we provide a reasonable mechanism by which a novel signal persists via broad

generalization of aposematic signals once it appears despite expectations of FDS for aposematic

signaling. While it is important to note that we did not observe white individuals in the yellow

population, this is not unexpected as phenotypic variation could be due to random mutation. Our

results support the idea that if novel signals were to arise in the yellow population, they would be

protected by the strong yellow signal. These results are somewhat supported by other populations

of D. tinctorius. In some yellow populations, we can observe a variety of patterns, indicating that

phenotype is not constrained in these population (Rojas and Endler 2013), and that predators

may indeed generalize among them (Rojas et al. 2014). Interestingly, while few, we do not

observe such polymorphism in populations with white in their signals.

82

Prior research exploring the evolution of polymorphism in aposematic taxa has largely

focused on mimicry complexes. In these instances, Batesian (Bates 1861, Darst and Cummings

2006, Kunte 2009, Katoh et al. 2017) and Müllerian (Müller 1879, Benson 1972, Brakefield

1985, Kapan 2001, Marek et al. 2009, Stuckert et al. 2014) mimics can diversify as selection

matches them to models. While this is an important mechanism to promote phenotypic diversity

in aposematic species, our findings are different in that we provide a mechanism for

intrapopulation polymorphism to persist without the need for model species. As many

aposematic species are not involved in mimicry complexes, this finding, in particular, is

important for our understanding of how aposematic signals evolve and diversify.

The existence of polymorphic aposematic signals is a difficult phenomenon to explain

given theoretical constraints imposed by selection. Here, we document two mechanisms that can

promote phenotypic diversity within and among populations. Polymorphism in aposematic

signals may not be selectively disadvantageous when signals are sufficiently strong to evoke

broad generalization by predators, and weak signals can persist when gene flow among

populations is limited. We provide strong evidence, through both lab and field experiments, for

how novel signals can persist, and why some signals, despite being weak, may endure in a

population. Together, these findings contribute to our understanding of the forces generating the

diverse array of aposematic signals seen across the animal kingdom.

83

Acknowledgements

We acknowledge an Investissement d’Avenir grant of the Agence Nationale de la

Recherche (CEBA: ANR-10-LABX-25-01), the American Society of Ichthyologists and

Herpetologists’ Gaige Award, and the Society for the Study of Amphibians and Reptiles’ Grant

in Herpetology. We thank Tim Muhich, John Constan, Antoine Baglan, and Mathilde Segers for

their invaluable help in creating and placing model frogs in French Guiana. Ombeline Vrignaud

provided valuable information and assistance and secured permission for work within the Grand

Matoury Natural Reserve; J.A. Endler and the members of the Evolution of Predator-Prey

Interactions Group at the U. of Jyväskylä, provided helpful feedback on an earlier version. Chick

and blue tit experiments were conducted under University of Mississippi IACUC protocols 14-

025 and 14-026 and the Central Finland Centre for Economic Development, Transport and

Environment and license from the National Animal Experiment Board

(ESAVI/9114/04.10.07/2014) and the Central Finland Regional Environment Centre

(VARELY/294/2015), respectively. We are grateful to Helinä Nisu at Konnevesi Research

Station for help with the blue tit experiment, and to Sara Calhim and Janne Valkonen for

invaluable statistical advice.

84

VI. CONCLUSIONS

Aposematic signals must be comprised of a conspicuous signal and a secondary defense,

which elicit predator learning and avoidance (Ruxton et al. 2005). In order to understand how

each of these components influence the evolution of aposematic signaling, through this

dissertation, I have sought to break down each individual component in order to better assess

how the strategy works as a whole, and how these components influence the proliferation of

warning signals throughout a population. In doing so, I provide insight into how aposematic

signals function, and what evolutionary scenarios might allow for phenotypic diversity to arise

despite expectations of strong positive frequency-dependent selection.

As a defensive strategy, aposematic signaling is unique in the theoretical limitations to its

proliferation across taxa (Alatalo and Mappes 1996, Speed and Ruxton 2007). Warning signaling

should have considerable difficulty evolving (Mappes et al. 2005), and when it does, should have

considerable difficulty diversifying (Lindström et al. 2001). Despite this, the abundance and

diversity of aposematic signals across the animal kingdom suggests that situations exist that

allow, and even promote, diversification of these signals. While complex on the surface,

conspicuous warning signals can be broken down into constituent parts to better understand how

selection does or does not act, which then allows us to understand how frequency-dependent

selection can be relaxed, thus allowing for diversification of aposematic signals.

85

As a method of communication, it is imperative that aposematic signals are not confused

or misinterpreted by the receiver. There is great risk of negative consequences for prey species if

this happens. Signals must exploit biases (i.e., color sensitivity, taste aversion) in the receiver to

ensure that a message is properly communicated. Throughout this dissertation, I provide

evidence that aposematic signaling exploits these biases to ensure that a signal is clearly

received, which can then allow novel signals to persist despite their initial rarity.

This dissertation highlights several scenarios that can be exploited to allow for

polymorphism to arise within an aposematic population. Whether it is predator selection only

acting on a single color component or wide variation is secondary defense, under certain

circumstances, selection may allow phenotypic diversity within aposematic populations.

Conceptually, Wright’s shifting balance theory (Wright 1982) provides an explanation for the

origin of phenotypic polymorphism and polytypism in the context of aposematic signaling.

Phenotypic diversity can be thought of as a series of peaks separated by valleys. Selection pushes

populations up these peaks, which can be facilitated by genetic drift. Crossing from one adaptive

peak to another requires traveling through a valley which is selectively disadvantageous. Thus,

when a population does manage to cross a valley, it is thought to be able to quickly travel up to

an adaptive optimum on a new peak. When selection is relaxed, however, as I demonstrate

through my dissertation, populations can more easily traverse the valley between peaks to climb

a new adaptive peak. Relaxed selection on traits contributing to aposematic signaling likely

enables populations to overcome the difficulty of crossing an adaptive valley to a new adaptive

optimum.

86

Understanding aposematic signaling, the underlying processes, and evolutionary

exceptions allow us to better understand how aposematic signals can diversify. As one of the

more common defensive strategies among animals, understanding how and why aposematism

evolves is the first step to understanding how speciation can occur within these groups. This

dissertation provides evidence that broadens our understanding of how these signals can

diversify, how selection acts upon known and novel signals, and what conditions would be

necessary for novel signals to persist. This provides integral understanding of the relationship

between biological diversification and aposematic signaling.

87

LIST OF REFERENCES

88

Alatalo, R. V, and J. Mappes. 1996. Tracking the evolution of warning signals. Nature 382:708–

710.

Allen, W. L., R. Baddeley, N. E. Scott-Samuel, and I. C. Cuthill. 2013. The evolution and

function of pattern diversity in snakes. Behavioral Ecology 24:1237–1250.

Allen, W. L., I. C. Cuthill, N. E. Scott-Samuel, and R. Baddeley. 2011. Why the leopard got its

spots: relating pattern development to ecology in felids. Proceedings of the Royal Society

B: Biological Sciences 278:1373–1380.

Arias, M., Y. le Poul, M. Chouteau, R. Boisseau, N. Rosser, M. Théry, and V. Llaurens. 2016.

Crossing fitness valleys: empirical estimation of a fitness landscape associated with

polymorphic mimicry. Proceedings of the Royal Society B: Biological Sciences

283:20160391.

Aronsson, M., and G. Gamberale-Stille. 2008. Domestic chicks primarily attend to colour , not

pattern, when learning an aposematic coloration. Animal Behaviour 75:417–423.

Arrese, C. a, N. S. Hart, N. Thomas, L. D. Beazley, and J. Shand. 2002. Trichromacy in

Australian marsupials. Current Biology 12:657–660.

Ashfield, T., L. E. Ong, K. Nobuta, C. M. Schneider, and R. W. Innes. 2004. Convergent

evolution of disease resistance gene specificity in two flowering plant families. The Plant

Cell 16:309–318.

Badio, B., and J. W. Daly. 1994. Epibatidine, a Potent Analgetic and Nicotinic Agonist.

Molecular Pharmacology 45:563–569.

Barraclough, T. G., P. H. Harvey, and S. Nee. 1995. Sexual selection and taxonomic diversity in

passerine birds. Proceedings of the Royal Society B: Biological Sciences 259:211–215.

Bates, D., M. Mächler, B. M. Bolker, and S. C. Walker. 2015. Fitting linear mixed-effect models

using lme4. Journal of Statistical Software 67:1–48.

Bates, H. W. 1861. Contributions to an insect fauna of the Amazon Valley. Lepidoptera:

Heliconiidae. Transactions of the Linnean Society of London XXIII:495–566.

Benson, W. W. 1972. Natural selection for Mullerian mimicry in Heliconius erato in Costa Rica.

Science 176:936–939.

Beukema, W., A. G. Nicieza, A. Lourenço, and G. Velo-Antón. 2016. Colour polymorphism in

Salamandra salamandra (Amphibia: Urodela), revealed by a lack of genetic and

environmental differentiation between distinct phenotypes. Journal of Zoological

Systematics and Evolutionary Research 54:127–136.

Blount, J. D., M. P. Speed, G. D. Ruxton, and P. A. Stephens. 2009. Warning displays may

function as honest signals of toxicity. Proceedings. Biological sciences / The Royal Society

89

276:871–7.

Bolton, S. K., K. Dickerson, and R. A. Saporito. 2017. Variable Alkaloid Defenses in the

Dendrobatid Poison Frog Oophaga pumilio are Perceived as Differences in Palatability to

Arthropods. Journal of Chemical Ecology 43:273–289.

Borer, M., T. Van Noort, M. Rahier, and R. E. Naisbit. 2010. Positive frequency-dependent

selection on warning color in alpine leaf beetles. Evolution 64:3629–3633.

Bowmaker, J. K. 1998. Evolution of colour vision in vertebrates. Eye 12:541–547.

Brakefield, P. M. 1985. Polymorphic Müllerian mimicry and interactions with thermal melanism

in ladybirds and a soldier beetle: a hypothesis. Biological Journal of the Linnean Society

26:243–267.

Brodie, E. D. 1993. Differential avoidance of Coral Snake banded patterns by free-ranging avian

predators in Costa Rica. Evolution 47:227–235.

Chouteau, M., and B. Angers. 2011. The role of predators in maintaining the geographic

organization of aposematic signals. The American Naturalist 178:810–817.

Chouteau, M., and B. Angers. 2012. Wright’s shifting balance theory and the diversification of

aposematic signals. PloS one 7:e34028.

Chouteau, M., M. Arias, and M. Joron. 2016. Warning signals are under positive frequency-

dependent selection in nature. Proceedings of the National Academy of Sciences 113:2164–

2169.

Chouteau, M., K. Summers, V. Morales, and B. Angers. 2011. Advergence in Müllerian

mimicry: the case of the poison dart frogs of Northern Peru revisited. Biology Letters

7:796–800.

Coloma, L. A., J. C. Santos, M. Baquero, L. K. Erdtmann, A. P. Lima, D. C. Cannatella, I. D. B.

Tropical, and C. Rica. 2014. Aposematism increases acoustic diversification and speciation

in poison frogs. Proceedings of the Royal Society of London Series B-Biological Sciences

281:20141761.

Comeault, A. A., and B. P. Noonan. 2011. Spatial variation in the fitness of divergent aposematic

phenotypes of the poison frog, Dendrobates tinctorius. Journal of Evolutionary Biology

24:1374–1379.

Daly, J. W., H. M. Garraffo, T. F. Spande, L. A. Giddings, R. A. Saporito, D. R. Vietes, and M.

Vences. 2008. Individual and Geographic Variation of Skin Alkaloids in Three Swamp-

Forest Species of Madagascan Poison Frogs (Mantella). Journal of Chemical Ecology

34:252–279.

Daly, J. W., and C. W. Myers. 1967. Toxicity of Panamanian poison frogs (Dendrobates): some

90

biological and chemical aspects. Science 156:970–3.

Daly, J. W., C. W. Myers, and N. Whittaker. 1987. Further classification of skin alkaloids from

Neotropical poison frogs (Dendrobatidae), with a general survey of toxic/noxious

substances in the Amphibia. Toxicon 25:1023–1095.

Daly, J. W., S. I. Secunda, H. M. Garraffo, T. F. Spande, A. Wisnieski, C. Nishihira, and J. F.

Cover Jr. 1992. Variability in alkaloid profiles in neotropical poison frogs (Dendrobatidae):

Genetic versus environmental determinants. Toxicon 30:887–898.

Daly, J. W., T. F. Spande, and H. M. Garraffo. 2005. Alkaloids from Amphibian Skin: A

Tabulation of Over Eight-Hundred Compounds. Journal of Natural Products 68:1556–1575.

Darst, C. R., and M. E. Cummings. 2006. Predator learning favours mimicry of a less-toxic

model in poison frogs. Nature 440:208–211.

Darst, C. R., M. E. Cummings, and D. C. Cannatella. 2006. A mechanism for diversity in

warning signals: Conspicuousness versus toxicity in poison frogs. Proceedings of the

National Academy of Sciences of the United States of America 103:5852–5857.

Darst, C. R., P. A. Menendez-Guerrero, L. A. Coloma, and D. C. Cannatella. 2005. Evolution of

dietary specialization and chemical defense in poison frogs (Dendrobatidae): A comparative

analysis. The American Naturalist 165:56–69.

Donnellan, S. C., M. J. Mahony, and T. Bertozzi. 2012. A new species of Pseudophryne (Anura:

Myobatrachidae) from the central Australian ranges. Zootaxa 3476:69–85.

Dreher, C. E., M. E. Cummings, and H. Pröhl. 2015. An Analysis of Predator Selection to Affect

Aposematic Coloration in a Poison Frog Species. Plos One 10:e0130571.

Eaton, D. A. R. 2014. PyRAD: assembly of de novo RADseq loci for phylogenetic analyses.

Bioinformatics 30:1844–9.

Ebeling, W., R. C. Natoli, and J. M. Hemmi. 2010. Diversity of color vision: Not all Australian

marsupials are trichromatic. PLoS ONE 5:e14231.

Emery, N., and N. Clyaton. 2004. The Mentality of Crows: Convergent Evolution of Intelligence

in Corvids and Apes. Science 306:1903–1907.

Endler, J. A., and J. J. D. Greenwood. 1988. Frequency-Dependent Predation, Crypsis and

Aposematic Coloration [and Discussion]. Philosophical Transactions of the Royal Society

B: Biological Sciences 319:505–523.

Endler, J. A., and J. Mappes. 2004. Predator Mixes and the Conspicuousness of Aposematic

Signals. The American Naturalist 163:532–547.

Endler, J. A., and B. Rojas. 2009. The spatial pattern of natural selection when selection depends

91

on experience. The American naturalist 173:E62-78.

Exnerová, A., D. Ježová, P. Štys, L. Doktorovová, B. Rojas, and J. Mappes. 2015. Different

reactions to aposematic prey in 2 geographically distant populations of great tits. Behavioral

Ecology 26:1361–1370.

Finkbeiner, S. D., A. D. Briscoe, and R. D. Reed. 2014. Warning signals are seductive: Relative

contributions of color and pattern to predator avoidance and mate attraction in Heliconius

butterflies. Evolution; international journal of organic evolution.

Finkbeiner, S. D., D. A. Fishman, D. Osorio, and A. D. Briscoe. 2017. Ultraviolet and yellow

reflectance but not fluorescence is important for visual discrimination of conspecifics by

Heliconius erato:1267–1276.

Fisher, M., D. Huangfu, T. Y. Shen, and P. G. Guyenet. 1994. Epibatidine, an Alkaloid from the

Poison Frog Epipedobates Tricolor, Is a Powerful Ganglionic Depolarizing Agent. Journal

of Pharmacology and Experimental Therapeutics 270:702–707.

Fitch, R. W., T. F. Spande, H. M. Garraffo, H. J. C. Yeh, and J. W. Daly. 2010. Phantasmidine:

an epibatidine congener from the ecuadorian poison frog Epipedobates anthonyi. Journal of

natural products 73:331–7.

Geffeney, S., E. D. Brodie Jr., P. C. Ruben, and E. D. Brodie III. 2002. Mechanisms of

adaptation in a predator-prey arms race: TTX- resistant sodium channels. Science

297:1336–1339.

Goldsmith, T. H. 1990. Optimization, Constraint, and History in the Evolution of Eyes. The

Quarterly review of biology 65:281–322.

Graham, C. F., T. C. Glenn, A. G. McArthur, D. R. Boreham, T. Kieran, S. Lance, R. G.

Manzon, J. A. Martino, T. Pierson, S. M. Rogers, J. Y. Wilson, and C. M. Somers. 2015.

Impacts of degraded DNA on restriction enzyme associated DNA sequencing (RADSeq).

Molecular Ecology Resources 15:1304–1315.

Gray, S. M., and J. S. McKinnon. 2007. Linking color polymorphism maintenance and

speciation. Trends in Ecology and Evolution 22:71–79.

Gronau, I., M. J. Hubisz, B. Gulko, C. G. Danko, and A. Siepel. 2011. Bayesian inference of

ancient human demography from individual genome sequences. Nature Genetics 43:1031–

1034.

Ham, A. D., E. Ihalainen, L. Lindström, and J. Mappes. 2006. Does colour matter? The

importance of colour in avoidance learning, memorability and generalisation. Behavioral

Ecology and Sociobiology 60:482–491.

Hantak, M. M., T. Grant, S. Reinsch, D. McGinnity, M. Loring, N. Toyooka, and R. A. Saporito.

2013. Dietary alkaloid sequestration in a poison frog: an experimental test of alkaloid

92

uptake in Melanophryniscus stelzneri (Bufonidae). Journal of chemical ecology 39:1400–6.

Hegna, R. H., J. A. Galarza, and J. Mappes. 2015. Global phylogeography and geographical

variation in warning coloration of the wood tiger moth (Parasemia plantaginis). Journal of

Biogeography 42:1469–1481.

Hegna, R. H., R. A. Saporito, and M. A. Donnelly. 2012. Not all colors are equal: predation and

color polytypism in the aposematic poison frog Oophaga pumilio. Evolutionary Ecology.

Hegna, R. H., R. A. Saporito, K. G. Gerow, and M. A. Donnelly. 2011. Contrasting colors of an

aposematic poison frog do not affect predation. Annales Zoologici Fennici 48:29–38.

Hoffberg, S. L., T. J. Kieran, J. M. Catchen, A. Devault, B. C. Faircloth, R. Mauricio, and T. C.

Glenn. 2016. RADcap : sequence capture of dual-digest RADseq libraries with identifiable

duplicates and reduced missing data. Molecular Ecology Resources 16:1264–1278.

Holen, O. H. 2013. Disentangling taste and toxicity in aposematic prey. Proceedings. Biological

sciences / The Royal Society 280:20122588.

Hugall, A. F., and D. Stuart-Fox. 2012. Accelerated speciation in colour-polymorphic birds.

Nature 485:631–634.

Kapan, D. D. 2001. Three-buttery system provides a field test of müllerian mimicry. Nature

409:18–20.

Katoh, M., H. Tatsuta, and K. Tsuji. 2017. Rapid evolution of a Batesian mimicry trait in a

butterfly responding to arrival of a new model. Scientific Reports 7:1–7.

Kazemi, B., G. Gamberale-Stille, B. S. Tullberg, and O. Leimar. 2014. Stimulus salience as an

explanation for imperfect mimicry. Current Biology 24:965–9.

Kikuchi, D. W., and D. W. Pfennig. 2010. Predator cognition permits imperfect coral snake

mimicry. The American Naturalist 176:830–4.

Kikuchi, D. W., and D. W. Pfennig. 2013. Imperfect mimicry and the limits of natural selection.

The Quarterly review of biology 88:297–315.

Kraemer, A. C., and D. C. Adams. 2013. Predator Perception of Batesian Mimicry and

Conspicuousness in a Salamander. Evolution; international journal of organic

evolution:1197–1206.

Kunte, K. 2009. The diversity and evolution of batesian mimicry in papilio swallowtail

butterflies. Evolution 63:2707–2716.

Langridge, K. V. 2009. Cuttlefish use startle displays, but not against large predators. Animal

Behaviour 77:847–856.

93

Leal, M., and L. J. Fleishman. 2004. Differences in Visual Signal Design and Detectability

between Allopatric Populations of Anolis Lizards. The American Naturalist 163:26–39.

Lenzi-Mattos, R., M. M. Antoniazzi, C. F. B. Haddad, D. V. Tambourgi, M. T. Rodrigues, and

C. Jared. 2005. The inguinal macroglands of the frog Physalaemus nattereri

(Leptodactylidae): Structure, toxic secretion and relationship with deimatic behaviour.

Journal of Zoology 266:385–394.

Lindström, L., R. V Alatalo, A. Lyytinen, and J. Mappes. 2001. Strong antiapostatic selection

against novel rare aposematic prey. Proceedings of the National Academy of Sciences of the

United States of America 98:9181–9184.

Lindström, L., R. V Alatalo, and J. Mappes. 1996. Reactions of hand-reared and wild-caught

predators toward warningly colored, gregarious, and conspicuous prey. Behavioral Ecology

10:317–322.

Lindström, L., R. V Alatalo, J. Mappes, M. Riipi, and L. Vertainen. 1999. Can aposematic

signals evolve by gradual change? Nature 397:249–251.

Linnen, C. R., Y.-P. Poh, B. K. Peterson, R. D. H. Barrett, J. G. Larson, J. D. Jensen, and H. E.

Hoekstra. 2013. Adaptive evolution of multiple traits through multiple mutations at a single

gene. Science 339:1312–6.

Llaurens, V., M. Joron, and M. Théry. 2014. Cryptic differences in colour among Müllerian

mimics: how can the visual capacities of predators and prey shape the evolution of wing

colours? Journal of Evolutionary Biology 27:531–540.

Maan, M. E., and M. E. Cummings. 2008. Female preferences for aposematic signal components

in a polymorphic poison frog. Evolution 62:2334–2345.

Maan, M. E., and M. E. Cummings. 2009. Sexual dimorphism and directional sexual selection on

aposematic signals in a poison frog. Proceedings of the National Academy of Sciences of

the United States of America 106:19072–7.

Maan, M. E., and M. E. Cummings. 2012. Poison frog colors are honest signals of toxicity,

particularly for bird predators. The American Naturalist 179:E1–E14.

Maia, R., C. M. Eliason, P. P. Bitton, S. M. Doucet, and M. D. Shawkey. 2013. pavo: An R

package for the analysis, visualization and organization of spectral data. Methods in

Ecology and Evolution 4:906–913.

Mallet, J. 2010. Shift happens! Shifting balance and the evolution of diversity in warning colour

and mimicry. Ecological Entomology 35:90–104.

Mallet, J., and N. H. Barton. 1989. Strong natural selectio in a warning-color hybrid zone.

Mallet, J., and M. Joron. 1999. Evolution of diversity in warning color and mimicry:

94

Polymorphisms, shifting balance, and speciation. Annual Review of Ecology and

Systematics 30:201–233.

Mappes, J., H. Kokko, K. Ojala, and L. Lindström. 2014. Seasonal changes in predator

community switch the direction of selection for prey defences. Nature Communications

5:1–7.

Mappes, J., N. Marples, and J. A. Endler. 2005. The complex business of survival by

aposematism. Trends in Ecology & Evolution 20:598–603.

Marek, P. E., J. E. Bond, and M. R. Berenbaum. 2009. A Müllerian mimicry ring in Appalachian

millipedes. Proceedings of the National Academy of Sciences of the United States of

America 106:9755–9760.

Martins, M. 1989. Deimatic behavior in Pleurodema brachyops. Journal of Herpetology1 3:305–

307.

McGugan, J. R., G. D. Byrd, A. B. Roland, S. N. Caty, N. Kabir, E. E. Tapia, S. A. Trauger, L.

A. Coloma, and L. O’Connell. 2016. Ant And Mite Diversity Drives Toxin Variation In The

Little Devil Poison Frog. Journal of Chemical Ecology 42:537–551.

McLean, C. A., and D. Stuart-Fox. 2014. Geographic variation in animal colour polymorphisms

and its role in speciation. Biological Reviews 89:860–873.

McLean, C. a, a Moussalli, and D. Stuart-Fox. 2014. Local adaptation and divergence in colour

signal conspicuousness between monomorphic and polymorphic lineages in a lizard.

Journal of evolutionary biology 27:2654–64.

Müller, F. 1879. Ituna and Thyridia: A remarkable case of mimicry in butterflies. Transactions of

the Entomological Society of London 1879:20–29.

Nishida, R. 2002. Sequestration of defensive substances from plants by Lepidoptera. Annual

Review of Entomology 47:57–92.

Nokelainen, O. 2013. Many forms of the wood tiger moth (Parasemia plantaginis): Selective

heterogeneity favours polymorphic warning signals. University of Jyväskylä.

Nokelainen, O., J. Valkonen, C. Lindstedt, and J. Mappes. 2014. Changes in predator community

structure shifts the efficacy of two warning signals in Arctiid moths. Journal of Animal

Ecology 83:598–605.

Noonan, B. P., and A. A. Comeault. 2009. The role of predator selection on polymorphic

aposematic poison frogs. Biology Letters 5:51–54.

Noonan, B. P., and P. Gaucher. 2006. Refugial isolation and secondary contact in the dyeing

poison frog Dendrobates tinctorius. Molecular Ecology 15:4425–35.

95

O’Donald, P., and M. E. N. Majerus. 1984. Polymorphism of melanic ladybirds maintained by

frequency‐dependent sexual selection. Biological Journal of the Linnean Society 23:101–

111.

Ojala, K., R. Julkunen-Tiitto, L. Lindström, and J. Mappes. 2005. Diet affects the immune

defence and life-history traits of an Arctiid moth Parasemia plantaginis. Evolutionary

Ecology Research 7:1153–1170.

Okano, T., D. Kojima, Y. Fukada, Y. Shichida, and T. Yoshizawa. 1992. Primary structures of

chicken cone visual pigments: vertebrate rhodopsins have evolved out of cone visual

pigments. Proc Natl Acad Sci U S A 89:5932–5936.

Olofsson, M., S. Eriksson, S. Jakobsson, and C. Wiklund. 2012. Deimatic Display in the

European Swallowtail Butterfly as a Secondary Defence against Attacks from Great Tits.

PLoS ONE 7.

Osborne, W. S. 1989. Distribution, relative abundance and conservation status of Corroboree

Frogs, Pseudophryne-corroboree Moore (Anura, Myobatrachidae). Australian Wildlife

Research 16:537–547.

Ostrom, J. H. 1974. Archaeopteryx and the Origin of Flight. The Quarterly review of biology

49:27–47.

Paluh, D. J., M. M. Hantak, and R. A. Saporito. 2014. A test of aposematism in the dendrobatid

poison frog Oophaga pumilio: The importance of movement in clay model experiments.

Journal of Herpetology 48:249–254.

Paluh, D. J., E. K. Kenison, and R. A. Saporito. 2015. Frog or Fruit? The Importance of Color

and Shape to Bird Predators in Clay Model Experiments. Copeia 2015:58–63.

Pfennig, D. W., W. R. Harcombe, and K. S. Pfennig. 2001. Frequency-dependent Batesian

mimicry. Nature 410:323.

Pfennig, D. W., and D. W. Kikuchi. 2012. Competition and the evolution of imperfect mimicry.

Current Zoology 58:608–619.

Poulton, E. B. 1890. The Colour of Animals: Their meaning and use, especially considered in the

case of insects. D. Appleton.

Preißler, K., and H. Pröhl. 2017. The effects of background coloration and dark spots on the risk

of predation in poison frog models. Evolutionary Ecology 31:683–694.

Prescott, N. B., and C. M. Wathes. 1999. Spectral sensitivity of the domestic fowl (Gallus g.

domesticus). British Poultry Science 40:332–339.

Proksch, P. 1994. Defensive roles for secondary metabolites from marine sponges and sponge-

feeding nudibranchs. Toxicon 32:639–655.

96

Promislow, D. E. L. 1992. Costs of Sexual Selection in Natural Populations of Mammals.

Proceedings of the Royal Society B: Biological Sciences 247:203–210.

Pupo, G. M., R. Lan, and P. R. Reeves. 2000. Multiple independent origins of Shigella clones of

Escherichia coli and convergent evolution of many of their characteristics. Proceedings of

the National Academy of Sciences of the United States of America 97:10567–72.

R, C. T. 2016. R: A language and environment for statistical computing. R Foundation for

Statistical Computing. Vienna, Austria.

Ratcliffe, J. M., and M. L. Nydam. 2008. Multimodal warning signals for a multiple predator

world. Nature 455:96–9.

Richards-Zawacki, C. L., and M. E. Cummings. 2011. Intraspecific reproductive character

displacement in a polymorphic poison dart frog, Dendrobates pumilio. Evolution 65:259–

267.

Richards-Zawacki, C. L., J. Yeager, and H. P. S. Bart. 2013. No evidence for differential survival

or predation between sympatric color morphs of an aposematic poison frog. Evolutionary

Ecology 27:783–795.

Roberts, M. F., and M. Wink. 1998. Alkaloids Biochemistry, Ecology, and Medical

Applications. Page (M. F. Roberts, Ed.). First edition. Springer US.

Rohland, N., and D. Reich. 2012. Cost-effective , high-throughput DNA sequencing libraries for

multiplexed target capture. Genome Research 22:939–946.

Rojas, B., E. Burdfield-steel, H. Pakkanen, K. Suisto, M. Maczka, S. Schulz, and J. Mappes.

2017. How to fight multiple enemies: target-specific chemical defences in an aposematic

moth. Proceedings Of The Royal Society B-Biological Sciences 284:20171424.

Rojas, B., J. Devillechabrolle, and J. A. Endler. 2014a. Paradox lost: variable colour-pattern

geometry is associated with differences in movement in aposematic frogs. Biology Letters

10:20140193–20140193.

Rojas, B., and J. A. Endler. 2013. Sexual dimorphism and intra-populational colour pattern

variation in the aposematic frog Dendrobates tinctorius. Evolutionary Ecology 27:739–753.

Rojas, B., P. Rantiala, and J. Mappes. 2014b. Differential detectability under varying light

environments: an alternative explanation for the maintenance of polymorphic warning

signals? Behavioural Processes 109:164–172.

Roland, A. B., J. C. Santos, B. C. Carriker, S. N. Caty, E. E. Tapia, L. A. Coloma, and L. A.

O’Connell. 2017. Radiation of the polymorphic Little Devil poison frog (Oophaga

sylvatica) in Ecuador. Ecology and Evolution 7:9750–9762.

Roper, T. J., and N. Marples. 1997. Odour and colour as cues for taste-avoidance learning in

97

domestic chicks. Animal Behaviour 53:1241–1250.

Rowe, C., and T. Guilford. 1999. The evolution of multimodal warning displays. Evolutionary

Ecology 13:655–671.

Ruxton, G. D., T. N. Sherratt, and M. P. Speed. 2004. Avoiding Attack.

Ruxton, G. D., T. N. Sherratt, and M. P. Speed. 2005. Avoiding Attack: The Evolutionary

Ecology of Crypsis, Warning Signals, and Mimicry. Oxford University Press, USA, Oxford.

Ruxton, G. D., M. P. Speed, and M. Broom. 2007. The importance of intial protection of

conspicuous mutants for the coevolution of defense and aposematic signaling of the

defense: a modeling study. Evolution 61:2165–2174.

Ryan, M. J. 1983. Sexual selection and communication in a neotropical frog, Physalaemus

pustulosus. Evolution 37:261–272.

Sandre, S. L., T. Tammaru, T. Esperk, R. Julkunen-Tiitto, and J. Mappes. 2007. Carotenoid-

based colour polyphenism in a moth species: Search for fitness correlates. Entomologia

Experimentalis et Applicata 124:269–277.

Santos, J. C., L. A. Coloma, and D. C. Cannatella. 2003. Multiple, recurring origins of

aposematism and diet specialization in poison frogs. Proceedings of the National Academy

of Sciences of the United States of America 100:12792–12797.

Saporito, R. A., M. A. Donnelly, M. A. Garraffo, H. M. Spande, and J. W. Daly. 2006.

Geographic and seasonal variation in alkaloid-based chemical defenses of Dendrobates

pumilio from Bocas del Toro, Panama. Journal of Chemical Ecology 32:795–814.

Saporito, R. A., M. A. Donnelly, R. L. Hoffman, H. M. Garraffo, and J. W. Daly. 2003. A

siphonotid millipede (Rhinotus) as the source of spiropyrrolizidine oximes of dendrobatid

frogs. Journal of Chemical Ecology 29:2781–2786.

Saporito, R. A., M. A. Donnelly, P. Jain, H. M. Garraffo, T. F. Spande, and J. W. Daly. 2007a.

Spatial and temporal patterns of alkaloid variation in the poison frog Oophaga pumilio in

Costa Rica and Panama over 30 years. Toxicon 50:757–778.

Saporito, R. A., M. A. Donnelly, R. A. Norton, H. M. Garraffo, T. F. Spande, and J. W. Daly.

2007b. Oribatid mites as a major dietary source for alkaloids in poison frogs. Proceedings of

the National Academy of Sciences of the United States of America 104:8885–90.

Saporito, R. A., M. A. Donnelly, T. F. Spande, and H. M. Garraffo. 2011. A review of chemical

ecology in poison frogs. Chemoecology 22:159–168.

Saporito, R. A., M. a Donnelly, A. a Madden, H. M. Garraffo, and T. F. Spande. 2010. Sex-

related differences in alkaloid chemical defenses of the dendrobatid frog Oophaga pumilio

from Cayo Nancy, Bocas del Toro, Panama. Journal of natural products 73:317–21.

98

Saporito, R. A., R. Zuercher, M. Roberts, G. Kenneth, and M. A. Donnelly. 2007c. Experimental

evidence for aposematism in the dendrobatid poison frog Oophaga pumilio. Copeia

2007:1006–1011.

Savage, J. M., and J. B. Slowinski. 1992. The colouration of the venomous coral snakes (family

Elapidae) and their mimics (families Aniliidae and Colubridae). Biological Journal of the

Linnean Society 45:235–254.

Schuler, W., and E. Hesse. 1985. On the function of warning coloration: a black and yellow

pattern inhibits prey-attack by naive domestic chicks. Behavioral Ecology and Sociobiology

16:249–255.

Schwarzkopf, L., and R. Shine. 2014. Costs of reproduction in lizards: escape tactics and

susceptibility to predation. Behavioral Ecology and Sociobiology 31:17–25.

Seehausen, O., J. J. M. Van Alphen, and R. Lande. 1999. Color polymorphism and sex ratio

distortion in a cichlid fish as an incipient stage in sympatric speciation by sexual selection.

Ecology Letters 2:367–378.

Sherratt, T. N., and C. D. Beatty. 2003. The Evolution of Warning Signals as Reliable Indicators

of Prey Defense. The American Naturalist 162:377–389.

Siddiqi, A., T. W. Cronin, E. R. Loew, M. Vorobyev, and K. Summers. 2004. Interspecific and

intraspecific views of color signals in the strawberry poison frog Dendrobates pumilio.

Journal of Experimental Biology 207:2471–2485.

Skelhorn, J., and C. Rowe. 2010. Birds learn to use distastefulness as a signal of toxicity.

Proceedings. Biological sciences / The Royal Society 277:1729–34.

Smith, B. P., M. J. Tyler, T. Kaneko, H. M. Garraffo, T. F. Spande, and J. W. Daly. 2002.

Evidence for biosynthesis of pseudophrynamine alkaloids by an Australian myobatrachid

frog (Pseudophryne) and for sequestration of dietary pumiliotoxins. Journal of Natural

Products 65:439–447.

Speed, M. P., and G. D. Ruxton. 2005. Warning displays in spiny animals: One (more)

evolutionary route to aposematism. Evo 59:2499–2508.

Speed, M. P., G. D. Ruxton, J. D. Blount, and P. A. Stephens. 2010. Diversification of honest

signals in a predator-prey system. Ecology Letters 13:744–753.

Speed, M. P., G. D. Ruxton, and M. Broom. 2006. Automimicry and the evolution of discrete

prey defences. Biological Journal of the Linnean Society 87:393–402.

Stankowich, T., T. Caro, and M. Cox. 2011. Bold coloration and the evolution of aposematism in

terrestrial carnivores. Evolution 65:no-no.

Staudinger, M. D., R. T. Hanlon, and F. Juanes. 2011. Primary and secondary defences of squid

99

to cruising and ambush fish predators: Variable tactics and their survival value. Animal

Behaviour 81:585–594.

Stuart-Fox, D. M., and T. J. Ord. 2004. Sexual selection, natural selection and the evolution of

dimorphic coloration and ornamentation in agamid lizards. Proceedings of the Royal

Society B: Biological Sciences 271:2249–2255.

Stuckert, A. M. M., P. J. Venegas, and K. Summers. 2014. Experimental evidence for predator

learning and Müllerian mimicry in Peruvian poison frogs (Ranitomeya, Dendrobatidae).

Evolutionary Ecology 28:413–426.

Summers, K., and M. E. Clough. 2001. The evolution of coloration and toxicity in the poison

frog family (Dendrobatidae). Proceedings of the National Academy of Sciences of the

United States of America 98:6227–6232.

Summers, K., M. P. Speed, J. D. Blount, and A. M. M. Stuckert. 2015. Are aposematic signals

honest? A review. Journal of Evolutionary Biology 28:1583–1599.

Summers, K., R. Symula, M. E. Clough, and T. W. Cronin. 1999. Visual mate choice in poison

frogs. Proceedings of the Royal Society of London Series B-Biological Sciences 266:2141–

2145.

Tazzyman, S. J., and Y. Iwasa. 2010. Sexual selection can increase the effect of random genetic

drift-a quantitative genetic model of polymorphism in Oophaga pumilio, the strawberry

poison-dart frog. Evolution 64:1719–1728.

Thornhill, R. 1976. Sexual selection and parental investment in insects. The American Naturalist

110:153–163.

Thurman, T. J., and B. M. Seymoure. 2016. A bird’s eye view of two mimetic tropical

butterflies: Coloration matches predator’s sensitivity. Journal of Zoology 298:159–168.

Tullberg, B. S., S. Merilaita, and C. Wiklund. 2005. Aposematism and crypsis combined as a

result of distance dependence: Functional versatility of the colour pattern in the swallowtail

butterfly larva. Proceedings Of The Royal Society B-Biological Sciences 272:1315–1321.

Ueno, H., Y. Sato, and K. Tsuchida. 1998. Colour-associated mating success in a polymorphic

Ladybird Beetle , Harmonia axyridis. Functional Ecology 12:757–761.

Umbers, K. D. L., J. Lehtonen, and J. Mappes. 2015. Deimatic displays. Current Biology

25:R58–R59.

Umbers, K. D. L., and J. Mappes. 2015. Postattack deimatic display in the mountain katydid,

Acripeza reticulata. Animal Behaviour 100:68–73.

Valkonen, J. K., and J. Mappes. 2014. Resembling a viper: implications of mimicry for

conservation of the endangered smooth snake. Conservation Biologyogy 28:1568–74.

100

Valkonen, J. K., M. Niskanen, M. Björklund, and J. Mappes. 2011. Disruption or aposematism?

Significance of dorsal zigzag pattern of European vipers. Evolutionary Ecology 25:1047–

1063.

Wang, I. J. 2011. Inversely related aposematic traits: reduced conspicuousness evolves with

increased toxicity in a polymorphic poison-dart frog. Evolution 65:1637–1649.

Wang, K., and H. Zhao. 2015. Birds generally carry a small repertoire of bitter taste receptor

genes. Genome Biology and Evolution 7:2705–2715.

Wang, L.-Y., W.-S. Huang, H.-C. Tang, L.-C. Huang, and C.-P. Lin. 2018. Too hard to swallow:

A secret secondary defence of an aposematic insect. Journal of Experimental Biology

221:jeb172486.

Weldon, P. J. 2017. Poison frogs, defensive alkaloids, and sleepless mice: critique of a toxicity

bioassay. Chemoecology 27:123–126.

Weldon, P. J., Y. J. Cardoza, R. K. Vander Meer, C. W. Hoffman, J. W. Daly, and T. F. Spande.

2013. Contact toxicities of anuran skin alkaloids against the fire ant (Solenopsis invicta).

Naturwissenschaften 100:185–192.

Williams, C. R., E. D. Brodie, M. J. Tyler, S. J. Walker, E. B. Jr, E. D. Brodie, M. J. Tyler, S. J.

Walker, and E. B. Jr. 2000. Antipredator mechanisms of Australian frogs. Journal of

Herpetology 34:431–443.

Willink, B., E. Brenes-Mora, F. Bolaños, and H. Pröhl. 2013. Not everything is black and white:

color and behavioral variation reveal a continuum between cryptic and aposematic

strategies in a polymorphic poison frog. Evolution; international journal of organic

evolution 67:2783–94.

Wright, S. 1982. The shifting balance theory and macroevolution. Annual Review of Genetics

16:1–19.

Yeager, J., C. Wooten, and K. Summers. 2011. A new technique for the production of large

numbers of clay models for field studies of predation 42:357–359.

Zollman, K. J. S., C. T. Bergstrom, and S. M. Huttegger. 2013. Between cheap and costly

signals: the evolution of partially honest communication. Proceedings. Biological sciences /

The Royal Society 280:20121878.

101

VITA

Education

2011

2006

Work

Experience

2011-

2018

2005-

2012

Awards

2009

2010

2011

2012

2013

2015

Michigan State University, East Lansing, Michigan

Masters of Science in Fisheries and Wildlife

Michigan State University, East Lansing, Michigan

Bachelors of Science in Zoology

University of Mississippi

Teaching Assistant, Biological Science I Laboratory (BISC 161) – Fall ‘11

Teaching Assistant, Genetics (BISC 336) – Spring ’12 – Spring ‘18

Lab Coordinator, Genetics (BISC 336) – Fall ’16, Spring ‘17

Michigan State University

Teaching Assistant, Introductory Biology (LB 144) – Fall ’07-Fall ‘10

Teaching Assistant, Biology of Amphibians and Reptiles (ZOL 384) – Fall ‘05,

Fall ‘06, Fall ’07, and Fall ‘10

Teaching Assistant, Tropical Biodiversity and Conservation in Panama (LB 493)

– Summer ’07, Summer ‘09

Teaching Assistant, Environmental Studies in Costa Rica (NSC 490, NSC491) –

Winter ‘05-’06; Winter ‘09-’10

Teaching Assistant, Wild Borneo! Biodiversity in Southeast Asia (NSC 490,

NSC 491) – Summer '10

Course Instructor, Wild Borneo! Biodiversity in Southeast Asia (NSC 490, NSC

491) – Summer ‘11

Course Instructor, Environmental Studies in Costa Rica (NSC 475, NSC 491,

ZOL 490, ZOL 494) – Winter ‘10-’11, Winter ’11-‘12

Total Support Received ($45,025)

Michigan State University College of Agriculture and Natural Resources

Summer Research Fellowship, $6000

Michigan State University Graduate School Research Extension Fellowship,

$1000

Michigan State University Graduate School Travel Funding Fellowship, $350

Michigan State University Fisheries and Wildlife Graduate Student Organization

Travel Fellowship, $100

American Society of Ichthyologists and Herpetologists Gaige Award, $500

University of Mississippi Summer Research Assistantship, $2600

Smithsonian Tropical Research Institute A. Stanley Rand Fellowship, $3100

Society for the Study of Amphibians and Reptiles Grant in Herpetology, $500

University of Mississippi Graduate Students Council Research Grant, $1000

102

2016

Peer

Reviewed

Publications

North American Nature Photography Association College Scholarship, $1000

National Science Foundation East Asia and Pacific Summer Institutes

Fellowship, $10965

Endeavour Research Fellowship, $17910

Lawrence, J.P., M. Mahony, and B. Noonan. 2018. Avian predators avoid dorsal,

but not ventral, signals in Australian Brood Frogs (Pseudophryne). PLoS ONE

13 (4) e0195446.

Lawrence, J.P. and B.P. Noonan. 2018 Avian learning favors colorful, not

luminous, signals. PLoS ONE 13 (3): e0194279.

Lawrence, J.P. 2017. Ineffectiveness of call surveys for estimating population

size in a widespread Neotropical frog, Oophaga pumilio. Journal of Herpetology

51:52-57.

Lawrence, J.P. 2017. Hypsiboas cinerascens predation. Herpetological Review

48:409.

Lawrence, J.P. 2011. Oophaga pumilio (Strawberry Dart Frog) habitat use.

Herpetological Review 42: 90.

Lawrence, J.P. 2018. Variable conservation of morphological features in a

polytypic frog (Oophaga pumilio). In review.

Bosque, R.J., J.P. Lawrence, R. Buchholz, G.R. Colli, J. Heppard, amd B.P.

Noonan. 2018. Diversity of warning signal and social interaction influences the

evolution of imperfect mimicry. In review.

Kelly, M.B.J., G. Taylor-Dalton, J.P. Lawrence, P. Byrne, and K.D.L. Umbers.

2018. Warning colors, camouflage and decoys as tactics to enhance the survival

of reintroduced southern corroboree frogs in the Australian Alps. In review

Lawrence, J.P., B. Rojas, A. Fouquet, J. Mappes, A. Blanchette, R.A. Saporito,

R.J. Bosque, E.A. Courtois, and B. Noonan. 2018. Weak warning signals can

persist in absence of gene flow. In prep.

Lawrence, J.P. 2018. Population estimates and spatial distributions of Oophaga

pumilio in a Forest Edge Landscape and Implications for Conservation. In prep.

Lawrence, J.P. and G.R. Urquhart. 2018. Demographic responses in a population

of the Strawberry Poison Dart Frog (Oophaga pumilio) after the addition of

artificial rearing sites. In prep.

103

Conference

Presentations

Conference

Presentations

Evolution and origin of polymorphism in aposematic species.

Evolution 2017, Portland, Oregon. 2017

Aposematic Signal Optimization in a Polytypic Species 2016

Australian Society of Herpetologists Meeting 2016, Launceston, Australia

Directional Selection on Color in the Dyeing Poison Frog Dendrobates

tinctorius

Behaviour 2015, Cairns, Australia 2015

Directional Selection on Color in the Dyeing Poison Frog, Dendrobates

tinctorius

Evolution 2014, Raleigh, North Carolina 2014

The Diversity, Distribution, and Conservation of a Polymorphic Frog (Oophaga

pumilio) in Western Panama 2010

Michigan Amphibian and Reptile Research Symposium, Central Michigan

University

Conservation of a Polymorphic Frog 2010

Michigan Amphibian and Reptile Research Symposium, University of Michigan

- Flint

Conservation of a Polymorphic Frog in Western Panama 2010

Michigan State University Fisheries and Wildlife GSO Symposium

Conservation of a Polymorphic Frog 2009

Michigan Amphibian and Reptile Research Symposium, Michigan State

University