Impacts And Implications Of Polytypism On The Evolution Of ...
Transcript of Impacts And Implications Of Polytypism On The Evolution Of ...
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 [email protected].
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