Access to this full-text is provided by PLOS.
Content available from PLOS One
This content is subject to copyright.
RESEARCH ARTICLE
Poison frog dietary preference depends on
prey type and alkaloid load
Nora A. Moskowitz, Rachel D’Agui, Aurora Alvarez-Buylla, Katherine FioccaID, Lauren
A. O’ConnellID*
Department of Biology, Stanford University, Stanford, CA, United States of America
*loconnel@stanford.edu
Abstract
The ability to acquire chemical defenses through the diet has evolved across several major
taxa. Chemically defended organisms may need to balance chemical defense acquisition
and nutritional quality of prey items. However, these dietary preferences and potential trade-
offs are rarely considered in the framework of diet-derived defenses. Poison frogs (Family
Dendrobatidae) acquire defensive alkaloids from their arthropod diet of ants and mites,
although their dietary preferences have never been investigated. We conducted prey prefer-
ence assays with the Dyeing Poison frog (Dendrobates tinctorius) to test the hypothesis that
alkaloid load and prey traits influence frog dietary preferences. We tested size preferences
(big versus small) within each of four prey groups (ants, beetles, flies, and fly larvae) and
found that frogs preferred interacting with smaller prey items of the fly and beetle groups.
Frog taxonomic prey preferences were also tested as we experimentally increased their
chemical defense load by feeding frogs decahydroquinoline, an alkaloid compound similar
to those naturally found in their diet. Contrary to our expectations, overall preferences did
not change during alkaloid consumption, as frogs across groups preferred fly larvae over
other prey. Finally, we assessed the protein and lipid content of prey items and found that
small ants have the highest lipid content while large fly larvae have the highest protein con-
tent. Our results suggest that consideration of toxicity and prey nutritional value are impor-
tant factors in understanding the evolution of acquired chemical defenses and niche
partitioning.
Introduction
Animals must make efficient use of available food to satisfy their energetic and nutritional
demands that are necessary for day-to-day function, reproduction, and survival [1]. Optimal
foraging theory predicts that organisms forage to maximize fitness by reducing the energetic
costs associated with consuming low-quality prey [1]. Chemically defended animals across
many taxa acquire toxins through their diet, and therefore, their foraging decisions should be
based on both chemical defense acquisition and nutritional quality of prey items. Most studies
on organisms with acquired chemical defense focus on cataloging prey items from stomach
contents, such as diet studies in bufonid (Melanophyriniscus) toads, Pitohui birds and
PLOS ONE
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 1 / 18
a1111111111
a1111111111
a1111111111
a1111111111
a1111111111
OPEN ACCESS
Citation: Moskowitz NA, D’Agui R, Alvarez-Buylla
A, Fiocca K, O’Connell LA (2022) Poison frog
dietary preference depends on prey type and
alkaloid load. PLoS ONE 17(12): e0276331. https://
doi.org/10.1371/journal.pone.0276331
Editor: Mainul Haque, University of Nottingham,
CHINA
Received: January 24, 2022
Accepted: October 4, 2022
Published: December 1, 2022
Peer Review History: PLOS recognizes the
benefits of transparency in the peer review
process; therefore, we enable the publication of
all of the content of peer review and author
responses alongside final, published articles. The
editorial history of this article is available here:
https://doi.org/10.1371/journal.pone.0276331
Copyright: ©2022 Moskowitz et al. This is an open
access article distributed under the terms of the
Creative Commons Attribution License, which
permits unrestricted use, distribution, and
reproduction in any medium, provided the original
author and source are credited.
Data Availability Statement: All relevant data are
available on Dryad: https://doi.org/10.5061/dryad.
m0cfxpp6q.
Funding: This work was supported by the Pew
Charitable Trusts (www.pewtrusts.org; award
Natricine snakes [2–4]. However, quantification of prey availability and nutritional content
are rarely considered, leaving a gap in our current understanding of how dietary preferences
evolve within the context of acquired chemical defenses. Examining how prey availability and
nutritional value may influence selection of prey items contributes to our general understand-
ing of community ecology and the evolution of acquired chemical defense.
Generalist predators must strike a balance between the energetic costs of foraging and prey
nutritional value and availability [5,6]. For example, Schizocosa wolf spiders make choices
based on their nutritional and energetic demands by selecting for certain prey, even if they are
low in abundance in their environments [7]. Some predators must further consider the physio-
logical burden of prey ingestion, especially in species that regularly consume chemically
defended prey items. For example, the European Starling (Sturnus vulgaris) can learn to avoid
toxic prey, but still choose to ingest them in the presence of undefended prey [8]. This suggests
that defended prey are worth consuming despite their physiological burden, as they may be
nutritionally valuable [8]. Such a nutritional tradeoff has also been proposed in possums,
which regularly consume chemically defended plants that are metabolically costly but nutri-
tionally valuable [9]. Choice of prey is arguably even more complex in organisms that acquire
chemical defenses from their prey items. For example, Acalymma vittatum beetles have diet-
derived chemical defenses and there is a tradeoff between protection from predators and the
physiological costs of detoxification [6]. These choices are not only important for short term
energy needs, but they can also impact development and have long lasting consequences on fit-
ness. In Battus philenor butterflies, sequestering defensive chemicals from plants increases sur-
vivorship against predation as caterpillars, but at the cost of reduced fat content as adults [5].
Together, these examples suggest that environmental availability, chemical defense, and nutri-
tional content of prey items are important considerations for predators with diet-derived
defenses [2,3,10]. However, the interplay between these factors in foraging decisions made by
animals that acquire chemical defenses from their diet is poorly understood.
Diet-derived chemical defenses have evolved multiple times in Central and South American
poison frogs (Dendrobatidae), which acquire alkaloid-based chemical defenses from their
diets [11,12]. Chemical defenses have evolved at least four times within Dendrobatidae, which
co-evolved with a dietary specialization on ants and mites in some species [13,14]. Stomach
content analyses have established that alkaloid-containing ants and mites constitute the major-
ity of diet within all chemically-defended dendrobatids and some non-defended species,
although there is high intra- and inter-specific variation [13,15,16]. Consuming an alkaloid-
rich diet while also acquiring enough lipids and protein for metabolism and reproduction can
be challenging. In particular, lipids are important for gamete production and metabolic main-
tenance in amphibians [10,17], and yet are a limiting resource among arthropod food webs
[18]. Chitinous arthropods, such as ants and mites have high levels of insoluble carbohydrates
[19], and therefore frequent ingestion is expected to incur a nutritional tradeoff [20]. Although
poison frog stomach content analyses have increased our understanding of diet in the dendro-
batid clade, how dietary preferences based on prey phenotype or frog alkaloid load have never
been tested. Thus, nutritional analyses paired with prey selectivity assays are necessary to assess
the possible tradeoffs between acquiring chemical defenses and necessary nutrients.
Here, we examined prey preference of Dyeing poison frogs (Dendrobates tinctorius) in a
laboratory setting. We tested the hypothesis that chemical defense acquisition, prey size and
prey type influence dietary preferences in poison frogs. We first tested preference based on
size within prey categories, as a prior study in the Malagasy Mantellidae poison frogs suggests
frogs have a dietary preference towards smaller prey items [21]. Further, it is known that prey
size is a crucial factor for dietary selectivity in frogs generally [22,23]. We predicted the frogs
would show a greater preference towards smaller chitinous prey categories (ants, beetles) [18,
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 2 / 18
#00034087) and the New York Stem Cell
Foundation (www.nyscf.org; award #NYSCF-R-
NI58) to LAO. This work was also supported by a
Student Research Grant from the Society for
Animal Behavior (www.animalbehaviorsociety.org;
no award number) to NAM, a postdoctoral
research fellowship awarded by the National
Science Foundation (www.nsf.gov; DBI-2109400)
to KF, and a graduate research fellowship awarded
by the National Science Foundation (www.nsf.gov;
DGE-1656518) to NAM and AAB, and an HHMI
Gilliam fellowship (www.hhmi.org, GT13330)
awarded to AAB. LAO is a New York Stem Cell
Foundation – Robertson Investigator. The funders
had no role in study design, data collection and
analysis, decision to publish, or preparation of the
manuscript.
Competing interests: The authors have declared
that no competing interests exist.
19], but larger versions of non-chitinous prey (flies, fly larvae), as fly larvae tend to have higher
protein and lipid content compared to chitinous adult arthropods [24]. We next tested prefer-
ence for prey taxonomic identity and nutritional value and predicted D.tinctorius would prefer
ants, which make up the majority of poison frog stomach contents in the wild [13,16,25].
Finally, we tested whether chemical defense changes prey preference by experimentally manip-
ulating alkaloid load and measuring prey selection. We expected frogs to prefer more nutrient-
rich prey as they acquire alkaloids. Together, our work helps to fill the gap in our understand-
ing of how prey size, type and nutritional quality affect the dietary choices of chemically
defended predators.
Methods
Animals
Dendrobates tinctorius (N = 20) were purchased from Josh’s Frogs (Owosso, MI, USA) and
housed in pairs in glass terraria with sphagnum moss, live philodendrons, shelters, egg laying
sites, and a water pool. Frogs were fed three times per week with fruit flies dusted with vitamin
supplements. Terraria were misted with water ten times per day and maintained in a tempera-
ture-controlled room with a 12h:12h light-dark cycle immediately preceded and followed by a
10 min period of dim lighting to simulate dawn and dusk. The Institutional Animal Care and
Use Committee at Stanford University approved all frog experiments (Protocol #33839). Frogs
were not euthanized for this study and were housed in our frog colony for future research.
The arthropods used in this study are comparably sized to wild prey items [26], although
they are not the same species, which are not commercially available in the United States. Four
groups of arthropod prey of two sizes each were used in frog feeding assays as follows (large,
small): Ants: Liometopum apiculatum [7.0–10.0 mm, ~5 mg], Linepithema humile [2.2–2.6
mm, ~1.5 mg], Flies: Drosophila hydei [3.0 mm, ~2.5 mg], Drosophila melanogaster [1.6 mm,
~1.3 mg], Beetles: Stethorus punctillum [1.5 mm, 400–500 μg], Dalotia coriaria [3.0–4.0 mm,
78–80 μg], fly larvae: Callifora vomitoria [10 mm, ~66 mg], Musca domestica [3–5 mm, ~12
mg] [27,28]. Ants were collected the morning of frog feeding trials on the Stanford University
campus by sweeping ants off their nesting trees with a bristle brush into a clean plastic bag. All
beetles and fly larvae were purchased from online vendors (Arbico Organics: Tucson, Arizona,
USA; Evergreen Growers Supply: Clackamas, Oregon, USA; Nature’s Good Guys: Ventura
County, California, USA; Speedyworm Bait Suppliers, Alexandria, Minnesota, USA). Live fly
larvae were kept refrigerated at 4˚C and neither fed nor hydrated to prevent them from pupat-
ing into adult flies. Flies were maintained at room temperature in 32 oz. plastic deli cups with
aerated lids using fruit fly media (Josh’s Frogs). Beetles were stored in 32 oz. plastic deli cups
with aerated lids with packaging bedding and cotton balls soaked with a 1:1 table sugar:water
solution. The beetles were misted with water three times per week and occasionally were fed
grain mites (Acarus siro) sourced from the lids of mite-infested fly cultures. To our knowledge,
none of the eight prey species chosen are chemically defended. We examined chemical profiles
of L.apiculatum and L.humile ants using gas chromatography / mass spectrometry (GC/MS,
described below) and found no known frog alkaloids.
Behavior
All dietary preference assays were conducted in the same room where frogs are housed, in
which temperature and humidity replicate conditions of their native habitats. Before feeding
trials began, all frogs were exposed to arthropod prey items six times over three weeks to
reduce neophobia. Trials were conducted in a small transparent plastic bin (33 cm x 20 cm x
12 cm) stacked within an identical secondary bin with a layer of dried magnolia leaves to
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 3 / 18
simulate a forest-floor environment while keeping the texture even throughout the arena.
Inner sides of the arena were coated with Fluon™liquid teflon (Bioquip Products Inc, Rancho
Dominguez, CA, USA) to prevent the arthropod prey from escaping. While piloting arena
configurations, we confirmed frogs eat their regular diet of fruit flies within this setup. Of the
88 videos, 14 could not be scored due to camera malfunction. Both ingestion and an attempt
to pursue or ingest prey of each type were scored using BORIS software. The scorer was not
aware of the treatment group at the time of scoring.
Size-mediated dietary preference assay
Adult frogs (11 female, 9 male; age unknown) were chosen at random and placed in the arena
with 10 individual insects (five large, five small) of one of the given prey groups (ants, beetles,
flies, or fly larvae). Each trial was conducted for 5 min and recorded with a GoPro camera
(GoPro Inc, San Mateo, CA, USA) mounted above the arena. Each frog participated in one
trial with each prey group and the order in which frogs and prey were tested was randomized.
Intertrial intervals were roughly 10 min and the arena was emptied and wiped down with a
moist paper towel during this time. Videos of all behavior trials were scored using BORIS
(Behavior Observation Research Interaction Software) [29]. The number of “eat” events and
“attempt to eat” events were scored for each size prey item within a trial.
Alkaloid feeding, collection and taxon dietary preference assay
Dietary preference trials were conducted to determine frog preference for different insect
taxa and how this preference may change with chemical defense (Fig 1A). Both male and
female frogs (N = 9 male, N = 10 female) were orally administered 15 μl of a vehicle control
solution (0.85% sodium chloride, 1% ethanol, 98.15% deionized water). One week later,
frogs were chosen at random and placed in the arena with 20 individual insects based on
which item within prey categories was most chosen by frogs in the size trials described
above. The chosen candidates were: five large fly larvae (Callifora vomitoria), five small ants
(Linepithema humile), five small beetles (Dalotia coriaria) and five small flies (Drosophila
melanogaster). Only one prey type per category was chosen to avoid confounding taxo-
nomic and size preferences. Each trial was conducted for 5 minutes and recorded with a
GoPro camera mounted above the arena. The arena was emptied and wiped down with a
moist paper towel between trials. Only female frogs (N = 10, age unknown) were used in
this behavioral assay, as male frogs consistently exited the arena before trial completion.
Why males exited the arena in this assay is unknown, although we hypothesize it may have
been due to stress from many prey items.
To test how preferences may change as frogs acquire chemical defenses, both male and
female frogs were orally administered 15 μl of 0.01% decahydroquinoline (DHQ) in vehicle
solution three times over the course of one week. The following week, these “low-DHQ”
female frogs were re-tested in the taxon dietary preference trials described above. One week
later, both male and female frogs were orally administered 15 μl 0.01% DHQ three times per
week over two consecutive weeks, representing a “high-DHQ” treatment. The following week,
female frogs were retested in a final preference trial. Previous work has established this alka-
loid-feeding paradigm as appropriate in creating low and high DHQ groups [30,31]. Each
female frog was assayed once for dietary preference between ants, beetles, flies and larvae at
each toxicity level for a total of three trials. Although males continued to receive DHQ treat-
ments, they were not used in behavior assays due to the consistent escape from the behavior
arena during the control period.
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 4 / 18
Detection of alkaloids in frog secretions and ant prey using mass
spectrometry
Ant alkaloids were quantified via Gas Chromatography / Mass Spectrometry (GC/MS) to
determine whether ants may carry alkaloids that could influence frog dietary preference dur-
ing the feeding assays. Alkaloids were extracted by soaking a volume of 1 mL of each ant spe-
cies in 1 mL of methanol placed in a glass vial (Wheaton, PTFE caps, 60940A-2). After a
1-hour incubation, 1 mL of the methanol from each ant species was syringe filtered through a
0.45 um PTFE syringe filter (Thermo Scientific, 44504-NP) into a new glass vial. These metha-
nol extractions were incubated for 24 hours at -80˚C to precipitate any lipids and proteins
within the sample. A 100 uL aliquot of each sample was added to an individual GC/MS auto-
sampler vial and the remaining samples were stored at -80˚C. GC/MS analysis was performed
on a Shimadzu GCMS-QP2020 instrument with a Shimadzu 30m x 0.25 mmID SH-Rxi-5Sil
MS column as described previously [16]. In brief, separation of alkaloids was achieved with
helium as the carrier gas (flow rate: 1 mL/min) using a temperature program from 100 to
Fig 1. Experimental workflows of diet preference and nutritional assays. (A) The workflow of the alkaloid-mediated diet preference assay is depicted on a
7-week timeline. (B) Workflow of ether-soluble lipid quantification of prey items are shown. After the 6–8 day ether step, the arrow points to the drying and
weighing steps, as they must be repeated to obtain lipid-free weight. (C) Workflow of the Bradford Protein assay of prey items are shown. Although our
experiment used 5 replicates per prey group in lipid and protein assays, only single replicates are depicted here.
https://doi.org/10.1371/journal.pone.0276331.g001
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 5 / 18
280˚C at a rate of 10˚C/minute. This was followed by a 2-minute hold and additional ramp to
320˚C at a rate of 10˚C/minute for column protection reasons, and no alkaloids appeared dur-
ing this part of the method. Compounds were analyzed with electron impact-mass spectrome-
try (EI-MS), and the spectra were searched against the AMDIS (NIST) mass spectral library to
identify peaks and then were visually inspected for poison frog alkaloids. All of the included
compound spectra had a similarity score of above 50% in the NIST library searches.
Skin secretions from the nine male D.tinctorius frogs that followed an identical vehicle
alkaloid-feeding regime alongside the female frogs used in the behavior trials were obtained to
confirm the presence of DHQ uptake. We used males that did not participate in the taxonomic
diet trials to avoid stressing the experimental female frogs, which may have influenced behav-
ior during feeding trials. We collected alkaloids from 2–3 frogs one week after each set of feed-
ing trials. No individual male frog had its alkaloids collected more than once. Alkaloids were
collected using a small pulse stimulator (Tens1, Largo, Florida, USA) at 1 volt for 30 seconds.
After stimulation, the frog’s dorsal skin was wiped with a Kimwipe™(Kimberly-Clark, Irving,
Texas, USA) and then stored in 5 ml methanol in glass vials. From the methanol in which the
Kimwipe™was stored, 1 ml was syringe filtered through a 0.45 um PTFE syringe filter (Thermo
Scientific, 44504-NP) into the new glass vial (Wheaton, PTFE caps, 60940A-2) supplemented
with 25 ug (-)-nicotine (Sigma Aldrich, N3876). All tubes were then capped, vortexed, and
stored at -80˚C to precipitate lipids and proteins. After precipitating for 24 hours, the superna-
tant was filtered through a 0.45 um PTFE syringe filter again into a new glass vial. We first ana-
lyzed frog secretions using GC/MS as described above and did not detect DHQ. We then used
liquid chromatography / mass spectrometry (LC/MS) as previously described [30,31] given
this method is more sensitive. The preliminary LC/MS data is consistent with an increase in
DHQ with longer feeding times.
Nutritional content assays
Crude lipid content was quantified across the eight prey groups used in the dietary preference
behavior experiments using an ether-soluble lipid extraction method (Fig 1B) [24]. All arthro-
pods were fresh-frozen at -80˚C and stored for 1–2 weeks. After thawing, specimens were sep-
arated into five 1–2 mL biological replicates per prey group. The number of specimens used
per prey type varied due to density and size. Each sample was placed in an incubator at 60˚C
until they were dried to a constant mass, defined as less than a 0.10 percent change after an
additional 20 minutes of drying [24]. One replicate per arthropod sample group was weighed
once after 2 hours, and again after three hours, when a constant mass was reached for all prey
groups except for fly larvae, which reached a constant mass after ~12 hours of drying. Once a
constant mass was reached for all prey groups, the final mass was recorded and samples were
placed in a glass culture tube filled with 5 ml of petroleum ether (Thermofisher Scientific, Wal-
tham, MA, USA) and ground with a Dounce Homogenizer. Specimens were then incubated in
ether for 8 days at room temperature. Every two days during this period, the ether was care-
fully decanted from each tube to prevent loss of non-lipid mass and then replaced with fresh
ether. Specimens were then left in their tubes uncovered for �24 h in a fume hood to allow for
evaporative removal of the remaining ether. Samples were then placed into an incubator at
60˚C for �3 h for additional drying. Once specimens were completely dried, they were
weighed and lipid-free mass was subtracted from total mass, indicating the biological portion
made up of ether-soluble lipids.
Crude proteins were extracted from each of the eight prey groups. All arthropods were
fresh-frozen live at -80˚C and stored for 1–2 weeks. Five replicates of 10–150 mg of each
arthropod type were put into 2 ml tubes, each containing 2.4 mm stainless steel beads (Omni
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 6 / 18
Life Science Inc, Raynham, Massachusetts, USA, SKU: 19–640) and 0.8 ml of cold 1x phos-
phate-buffered saline (PBS, pH 7.4). Samples were homogenized at 4˚C using a Beadmill 24
Homogenizer (Thermofisher Scientific) with the following settings: Speed in m/s (S): 4.5,
Number of cycles (C): 10, Cycle time (T): 30s, Interval paused between cycles (D): 10s. Samples
were then centrifuged at 2,000 rpm at 4˚C for five minutes. The supernatant containing the
protein lysate was moved into a new 1.5 mL microfuge tube and stored for 1 week at -80˚C.
Proteins were then quantified using a Bradford Protein assay according to manufacturer’s
instructions (Pierce Coomassie Plus Bradford Protein Assay Reagent, Thermofisher Scientific).
Samples and albumin standards were run in triplicate on a 96-well optical plate and quantified
on a BioTek Synergy H1 96-well plate reader (Winooski, Vermont, USA). Absorbance was
measured at 595 nm and an endpoint absorbance value was calculated for each standard and
sample. Crude protein concentrations of arthropod samples were calculated based on the
absorbance values of samples compared to the standard curve.
Data analysis and visualization
Statistical analyses were run in R version 4.1.1 using the glmmTMB package [32]. To assess dif-
ferences in large and small prey consumption, we fit zero-truncated generalized linear mixed
models (GLMM) for each of the four prey assays (large vs. small ants, beetles, flies, fly larvae)
based on a truncated Poisson distribution of the number of “eats” or “attempts to eat” across
all 19 frogs. We use “attempts” because our goal is to assess preference from a time and ener-
getic investment from the frog’s perspective. Since the frogs were willing to spend time and
energy pursuing prey items (often, the same individual item) over others, we felt it was impor-
tant to include their attempts in our preference quantification. To control for individual varia-
tion in the subjects, frog ID was used as a random effect, and since each frog differed in its
total number of prey interactions across the four size preference assay categories, “total inter-
actions with prey” was used as an offset in order to account for these differences in consump-
tion. Sex was also used as a covariate because of previous knowledge of differences in diet
between male and female dendrobatids [33].
To assess differences of prey interactions across control, low-DHQ, and high-DHQ groups,
we used our taxonomic prey preference assay data to fit a zero-truncated generalized linear
mixed models (GLMM) to compare the number of prey interactions for each prey type: ants,
beetles, flies, fly larvae based on a truncated Poisson distribution of the number of “eats” or
“attempts” across the 10 female frogs that were assayed. To account for individual variation in
the subjects, frog ID was used as a random effect. As each frog differed in its total number of
prey interactions across the four prey choices, we used “total interactions with prey” as an off-
set in order to account for these differences in consumption. Sex was not considered as a vari-
able in the model, since all frogs used in the taxonomic assay were female. To check for
differences in the general structure of frog diet categories across the three groups, we per-
formed a permutational multivariate analysis of variance (PERMANOVA) on Bray-Curtis dis-
similarities. We used a zero-truncated GLMM (glmmTMB package, R version 4.1.1) to
compare total frog interactions (eats or attempts of all prey items) across the three groups: con-
trol, low and high DHQ. The model was based on a truncated Poisson distribution as
described above. To control for repeated measures in the subjects, frog ID was again used as a
random effect.
P-values were adjusted using the Benjamini-Hochberg procedure to correct for multiple
hypothesis testing across all dietary preference and nutritional analyses. Data were visualized
with boxplots and barplots created using the package ‘ggplot2’ in R version 4.1.1 and info-
graphics were created with images from BioRender.com.
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 7 / 18
Results
Frogs prefer smaller flies and beetles within prey groups
Overall, frogs interacted (ate or attempted) more with smaller prey items (Fig 2). Frogs interact
more with small versus large flies (GLMM, Z: 16.02, BH-adjusted p <0.0001) and small versus
large beetles (GLMM, Z: 2.558, BH-adjusted p = 0.021). There was no preference for small ver-
sus large ants nor fly larvae within this behavioral assay (Generalized linear mixed model
(GLMM), Ants: Z: 0.844, BH-adjusted p = 0.477; fly larvae: Z: 0.712, BH-adjusted p = 0.477).
The relative frequency of prey interactions did not differ between prey groups (Kruskal-Wallis
chi-squared = 2.6556, df = 3, p-value = 0.448).
Alkaloid ingestion slightly changes dietary preference
The total number of interactions across all prey types did not differ between the three alkaloid
groups (PERMANOVA; F: 0.8932, p = 0.470). When frogs without DHQ were given the choice
between small beetles, small flies, small ants and large fly larvae, they interacted most with
large fly larvae (GLMM, Z: 4.95, p<0.00001). Additionally, high DHQ frogs interacted less
with fly larvae when compared to control frogs (GLMM, Z: -2.598, BH-adjusted p = 0.036),
although interactions with fly larvae were extremely variable across groups. Within all three
groups, larvae were interacted with most frequently. There were no differences found among
ant (GLMM, Z: 0.326, BH-adjusted p = 0.744), fly (GLMM, Z: 0.879, p = 0.379, BH-adjusted
p = 0.744), or beetle interactions (z: -0.409, BH-adjusted p: 0.744) among groups.
Prey items differ in lipid and protein content
All eight prey species were evaluated for two important dietary components: lipids and pro-
teins (Fig 3). Overall, lipid content differed across all prey types (Kruskal-Wallis χ
2
= 29.564,
Fig 2. Frogs preferred interacting with smaller flies and beetles. Dendrobates tinctorius frogs mainly interacted with (ate or attempted) small prey items in
prey size assays. Each frog was presented with one group at a time and given the choice of small and large versions of ants, beetles, flies and fly larvae. (A)
Percent breakdown of all interactions across prey groups are plotted in a stacked bar chart and (B) The frequencies of individual frog interactions among large
and small prey are visualized as box plots, where each dot represents a frog.
https://doi.org/10.1371/journal.pone.0276331.g002
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 8 / 18
df = 7, p-value = 0.0001). Small ants (Linepithema humile), beetles (Dalotia coriaria) and fly
larvae (Musca domestica) had a higher proportion of lipids than their larger counterparts (ant:
Liometopum apiculatum, beetle: Stethorus punctillum, and fly larvae: Callifora vomitoria),
while large flies (Drosophila hydei) had a higher proportion of lipids than small flies (D.mela-
nogaster) (Fig 3,Table 1). Protein content also differed across species (Kruskal-Wallis χ
2
=
29.206, df = 7, p-value = 0.0001), where large fly larvae have more protein than most prey
items, followed by small beetles, which have more protein than small or large ants and flies
(Fig 3,Table 2). Protein content did not differ within taxa for large compared to small prey. In
summary, small ants have the most lipids, while large fly larvae have the most protein, of all
eight species tested.
Discussion
We tested the hypothesis that frog chemical defenses and prey traits like size and nutritional
value would influence diet choices in poison frogs. We expected laboratory D.tinctorius to pre-
fer small prey items, reflecting their wild diet [13–16], which was generally true. However, we
Fig 3. Nutritional content of prey from dietary preference assays. (A) Average crude lipid contents across replicates of the eight arthropod species are
depicted in a bar chart, and (B) Variation of these values across replicates is shown in a box plot, where each dot represents a replicate. (C) Average crude
protein content across replicates of the eight arthropod species are depicted in a bar chart, and (D) variation of these values across replicates is shown in a
box plot, where each dot represents a replicate.
https://doi.org/10.1371/journal.pone.0276331.g003
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 9 / 18
were surprised to find frogs did not prefer ants over other prey items, as these frogs are thought
to have a dietary specialization on ants and mites in the wild. Rather, D.tinctorius generally
prefers prey items with high protein content, and frog alkaloid load can influence prey prefer-
ences. We discuss these laboratory results in the context of field studies in amphibians and the
importance of replicating these studies under natural conditions in the future.
Prey selectivity of arthropods
Predators’ ability to consume and obtain energy from prey is primarily limited by the prey
size, making size an important factor in dietary selection [1]. In our study, D.tinctorius frog
dietary preference was assessed using feeding assays where prey varied in size. Frogs preferred
small flies and beetles over large ones, with no significant preference for size in ants or fly lar-
vae. Our findings are similar to a study conducted with Mantella aurantiaca poison frogs
(Mantellidae), which have a slight preference for smaller prey items [21]. We did not find a
size preference for ants or fly larvae, which has been documented in Chiasmocleis mehelyi
Table 1. Differences in crude lipid content (% dry weight, mg) between arthropods.
Comparison Z Raw P BH-adjusted P
large fly larvae—small ants -4.0307 0.0001 0.0008
large beetles—small ants -4.1659 0.0000 0.0009
small ants—small flies 3.1109 0.0019 0.0131
large flies—large fly larvae 3.0298 0.0025 0.0137
large beetles—large flies -3.1650 0.0016 0.0145
small ants—small beetles 2.6240 0.0087 0.0348
large beetles—small fly larvae -2.6240 0.0087 0.0406
large ants—small ants -2.4617 0.0138 0.0430
large fly larvae—small fly larvae -2.4887 0.0128 0.0449
large flies—small flies 2.1100 0.0349 0.0976
Statistics of arthropod taxa differences in crude lipid content using Kruskal-Wallis test with a Dunn Multiple
Comparison Posthoc test. P-values were adjusted using the Benjamini-Hochberg procedure to correct for multiple
hypothesis testing.
https://doi.org/10.1371/journal.pone.0276331.t001
Table 2. Differences in crude protein content (conc [μg/ml]) between arthropod prey.
Comparison Z Raw P BH-adjusted P
large fly larvae—small ants 4.1932 0.0000 0.0008
large ants—large fly larvae -3.5709 0.0004 0.0050
small ants—small fly larvae -3.4086 0.0007 0.0061
large beetles—large fly larvae -3.1246 0.0018 0.0125
small ants—small beetles -2.9217 0.0035 0.0195
large ants—small fly larvae -2.7864 0.0053 0.0249
large flies—small ants 2.5159 0.0119 0.0475
large ants—small beetles -2.2995 0.0215 0.0668
large beetles—small fly larvae -2.3401 0.0193 0.0675
large fly larvae—small flies 2.1507 0.0315 0.0882
Statistics of arthropod taxa differences in crude protein content using Kruskal-Wallis test with a Dunn Multiple
Comparison Posthoc test. P-values were adjusted using the Benjamini-Hochberg procedure to correct for multiple
hypothesis testing.
https://doi.org/10.1371/journal.pone.0276331.t002
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 10 / 18
(Microhylidae) frogs that show a preference for small ants and avoidance of large ants relative
to environmental abundance [23]. However, we did not test ants from the natural range of D.
tinctorius and testing such species may reveal other preferences based on size or other prey
traits like alkaloid content. Additionally, these captive frogs have been exclusively fed Drosoph-
ila melanogaster (small flies) until the start of this study, which may have influenced their die-
tary preferences. Despite these limitations, our study and previous literature show that prey
size is a crucial factor for dietary choices of predators [21].
Taxonomic dietary preference was assayed using the most frequently chosen species (either
large or small) from each prey group (small ants, small beetles, small flies, large fly larvae), and
frogs interacted more large fly larvae over other prey items, though the difference was not sta-
tistically significant. Although poison frogs consume a diverse array of non-ant prey items [13,
14,34], the stomach contents of D.tinctorius and many other chemically defended poison
frogs are composed of over 50% ants [13,26,34]. Unexpectedly, we found that D.tinctorius
frogs preferred fly larvae over ants, beetles, and flies. Although surprising, we must note that
our lack of native prey items in these experiments could have influenced these results, as frogs
may have preferences for native alkaloid-defended ants, which could not be uncovered by our
study. It would be beneficial to test whether preference differs when using native prey items, as
it is possible that frogs’ foraging behavior would have been different when presented with ants
from their natural habitat. For example, a field study with Chiasmocleis mehelyi (Microhylidae)
frogs, which are hypothesized to sequester chemical defenses from their diet, showed preferen-
tial selection for ants over other arthropod categories [23,35]. Frogs in general can be selective
predators, as the stomach contents of bufonid toads (Bufo cognatus and Bufo woodsii) differ
from local arthropod abundance [36]. However, prey availability may strongly contribute to
ant abundance in poison frog diets, as frogs may occupy microhabitats of the forest with high
ant abundance relative to other dietary arthropods. For example, a study that compared stom-
ach contents of alkaloid-defended Malagasy poison frogs (Mantellidae) to environmental
availability of arthropods found no significant differences between them, suggesting that their
diet may reflect environmental availability rather than prey selectivity [37]. Given the known
importance of prey availability, we encourage future studies on organisms with diet-derived
defenses to consider comparing environmental availability of prey items’ effect on dietary pat-
terns and preferences.
Chemical defenses and prey preference
While the acquisition of chemical defenses from diet can be physiologically burdensome,
defended prey may also be nutritionally valuable. We tested the hypothesis that acquiring
chemical defenses would change prey preferences and predicted that alkaloid-associated prey,
such as ants, would become less desirable due to low nutritional value reported in the literature
[19,24]. In other words, we expected that overall metabolic and nutritional demand would
increase with alkaloid sequestration. Contrary to our predictions, frogs decreased their prefer-
ence for fly larvae, with high DHQ frogs showing a lower preference for fly larvae over control
groups. It was surprising that larvae are the most frequently chosen category across groups
given that the stomach contents of wild poison frogs tend to be composed of over 50% ants
[14,16,25]. As preference changes slightly with DHQ consumption, it is possible that alkaloids
can change foraging behavior as described in other animals. For example, in Red Knot birds
(Calidris canutus), bodily levels of ingested toxins slow their feeding rates [38,39]. Another
example comes from Mytilus mussels, where previous ingestion of toxins slows their toxin
uptake rate and feeding rate [40,41]. Further, although not chemically defended themselves,
possums consistently eat toxic, metabolically costly plants for their nutritional value [42].
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 11 / 18
Based on these studies, a positive feedback relationship could be expected between frog alka-
loid levels and ingestion of alkaloid-associated prey up to a physiological threshold of chemical
burden. In the future, using ant species with alkaloids that poison frogs sequesters may give
more ecologically-relevant insights into how behavior changes with alkaloid uptake.
Our study has several limitations that should be accounted for in future studies. First, frog
behavior may change over time with repeated exposure to prey items [43–45]. Our experimen-
tal design utilizes within-subject testing as frogs went from non-toxic to various loads of DHQ
levels and thus repeated exposure and testing may reflect learning through experience rather
than changes associated with alkaloid levels. We cannot verify that learning does not influence
the prey preference patterns observed in our study. Second, the alkaloid load of frogs and prey
items in this study do not reflect natural conditions. All eight prey species in the present study
are not known to possess small-molecule alkaloids [46,47], and we confirmed that the ants in
our study do not have frog alkaloids. Moreover, while DHQ was detected in experimental
frogs, it was at very low levels, being detectable by the more sensitive LC/MS but not by GC/
MS. GC/MS is used by our lab and others to quantify frog alkaloids [48], suggesting that the
DHQ levels in the present study are not ecologically relevant. However, it is currently
unknown how alkaloid load influences frog foraging behavior. Future studies on distinct
groups of chemically defended and undefended frogs being presented with native prey items
that contain alkaloids the frogs sequester are needed to thoroughly understand dietary prefer-
ence. Ideally, future studies on taxonomic preference would include both male and female
sexes, as our taxonomic trials included females only. The results from our captive study suggest
that dietary preference research across sexes would be a worthwhile line of research to pursue
in the future.
Prey selectivity and nutritional content
Dietary lipids and proteins are important for amphibian reproduction and survival. For exam-
ple, the survival rate of Epidalea calamita toad tadpoles is positively correlated with protein
consumption [49], and leopard frog tadpoles given a protein-rich diet are more likely to over-
come Batrachochytrium dendrobatidis infection [50], a disease which has ravaged frog popula-
tions globally [51]. Lipids are also of known importance to frogs, specifically for metabolic
demand and gamete production [17]. A study that monitored whole-body lipid content in
domestically-reared H.marmoratus taeniatus frogs showed a significant decrease in whole-
body lipid content following their breeding period, suggesting that fat stores are quickly
depleted through reproduction [52]. Another study conducted in Rana tigrina found that
frogs with small abdominal fat bodies showed significantly lower fecundity and decreased egg
size than those with large abdominal fat bodies, and these differences were caused by diet qual-
ity [53]. Given the importance of lipids and proteins to gamete production, metabolic demand,
immunity, and survival, nutrient content of prey items should be quantified when investigat-
ing factors that shape prey preference in amphibians.
In the current study, we assessed crude protein and lipid content of eight prey species, and
unexpectedly found small ants were high in lipid content. Much of the literature describes ants
as being composed primarily of insoluble fibers due to their chitinous exoskeletons [18,19,
54]. On this basis, we expected eating ants to impose a trade-off in poison frog foraging deci-
sions between acquiring alkaloids versus nutrition. Yet, there are very few studies examining
ant nutritional value, although our ant lipid results are similar to one study assessing nutrient
quality of small ant species, which report lipid contents of 40–60% [55]. However, our small
ant lipid results are substantially higher than the only study which examined lipid percentage
of the same small ant species as used in our study, Linepithema humile, as they report lower
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 12 / 18
lipid contents of 18–20% [56]. However, the ants in that study were given strict feeding regi-
mens for the purposes of studying behavioral changes associated with manipulated protein :
carbohydrate ratios in the diet. Therefore, those ants are expected to have a different nutri-
tional makeup than the ones in our experiment, which are sourced from the wild. Lipid and
protein content of large fly larvae (Calliphora vomitora) from our current study matched pre-
viously reported values [57]. Overall, the relative lipid values among orders from our study
(Diptera, Coleoptera, Hymenoptera) are quite similar to the only large-scale study which
quantified and compared lipid content across major arthropod orders [24]. However, a recent
review of the nutritional quality of insects for animal feed shows that both lipid and protein
contents are remarkably variable across studies of all insect taxa, even within single species
[58]. More studies with larger sample sizes across a wider range of arthropods are necessary to
understand their nutritional value as prey items. We note that a fruitful future direction would
be to measure lipid, protein, and alkaloid content from prey item species within the D.tinctor-
ius natural range.
Although arthropods are an important dietary component of many taxa [57,58], there are a
few studies which quantify lipid or protein content of individual arthropod species. The vast
majority of studies which examine arthropod crude protein and lipid content are focused on
the benefits of arthropod consumption for humans and domestic animals [59–61], with little
focus on wild animal diets. Within our study, small ants and large fly larvae seem to be the
most nutrient-rich of all prey items. Large fly larvae have the highest protein content but are
lipid-poor, while small ants have the highest lipid content but are protein-poor (Fig 4). Both
sizes of beetles and flies have similar protein levels to one another, but flies had a higher lipid
content than beetles. Therefore, a nutritional trade-off arises between the decision of D.tinc-
torius individuals to consume either lipid and protein-rich prey items in the presence of both.
Assuming the small ants in wild poison frog diet possess similar lipid and protein levels to the
small ants in the present study (Linepithema humilis), chemical defense acquisition may not
signify a total nutritional tradeoff, given the lipid richness of small ants. However, we encour-
age longer-term studies to be conducted using a higher sample size, prey items which more
accurately reflect poison frog wild diet, and a wider variety of poison frog species and alkaloids,
since diets between defended and undefended species vary [13].
Fig 4. Frog prey preference changes with increasing chemical defense. (A) Percent breakdown of all prey interactions (eats or attempts) across DHQ-fed and
vehicle control groups are plotted in a stacked bar chart and (B) the raw frequencies of prey interactions among frog toxicity groups are visualized as box plots
to show the variation of prey interactions across individuals, where each dot represents a frog.
https://doi.org/10.1371/journal.pone.0276331.g004
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 13 / 18
Prey choice becomes more complex when considering prey with high protein content, like
fly larvae, in the presence of lipid-rich small ants, which frogs might innately associate with
chemical defenses. Although this tradeoff has not been demonstrated, we have surmised this
tradeoff from our data since we found that prey species used in this experiment differ in nutri-
tional quality. We expected a clearer-cut tradeoff between consuming ants, which have high
levels of indigestible carbohydrates [19,57] while missing out on the nutritional value of unde-
fended prey. However, given the lipid richness of small ants and the ant-dependent nature
dendrobatid defenses, we propose that poison frogs consume alkaloid-laden prey and higher
protein prey opportunistically, considering that protein is a limiting resource in arthropod
food webs [18] and required for amphibian health [49,50]. Similarly, a particularly high
demand for lipids should be expected in neotropical frogs that tend to breed for the majority
of the year [62,63] and should have a high lipid demand. Poison frogs likely encounter and eat
many more ants than larvae in their natural habitats, as previous literature shows >50% ants
making up most poison frog stomach contents [13,14], implying that availability may be even
more important than preference in shaping poison frog diet. Since the small ants in our experi-
ment are the most lipid-rich prey item, we propose the possibility that those in their native
diets may also be lipid-rich, which could create a tradeoff for frogs given the choice between a
high-lipid ant or a high-protein larvae. In our controlled lab setting, the frogs chose to interact
the most with large larvae, the prey item with a substantially higher protein content than all
prey categories. Taken together, these wild diet studies paired with our results suggest that typ-
ical ant-rich wild poison frog stomach contents likely reflect a combination of their ant-rich
habitats and innate dietary preferences as shown in our study, and as proposed previously [13,
26,64].
Conclusions
We found that poison frogs prefer interacting with large larvae and that alkaloid uptake influ-
ences dietary choice, in that preference for larvae decreases with alkaloid acquisition (between
control and high DHQ categories only). The known importance of lipids to amphibian repro-
duction and survival, taken together with our prey nutrient and preference assay results, show
that poison frogs may have nutritionally benefitted from a dietary specialization on ants before
they evolved an ability to acquire chemical defenses from them. We suggest that innate prey
preferences, the nutritional value of prey, and prey availability are all important for our under-
standing of how dietary alkaloid sequestration evolved multiple times within the Dendrobati-
dae clade. The influence of nutritional value on dietary choices is essential for understanding
the evolution of acquired chemical defenses and niche partitioning across heterotrophs. Future
studies should evaluate the ecological significance of our findings by investigating preference
and nutritional content of prey items available to wild frogs across a larger geographical and
taxonomic scale that includes defended and undefended species.
Supporting information
S1 File. This excel file contains all data spreadsheets, including frog behavior, arthropod
lipid and protein content, and ant alkaloid data.
(XLSX)
Acknowledgments
We thank all the frog caretakers that maintain our poison frog and tadpole laboratory colony
and Dr. Marie-Therese Fischer for feedback on an early version of this manuscript. We thank
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 14 / 18
the Stanford University Mass Spectrometry (SUMS) core for their expert guidance and sup-
port on measuring DHQ in frogs. We acknowledge that our work takes place on the ancestral
and unceded land of the Muwekma Ohlone tribe.
Author Contributions
Conceptualization: Nora A. Moskowitz, Lauren A. O’Connell.
Data curation: Nora A. Moskowitz.
Formal analysis: Nora A. Moskowitz.
Funding acquisition: Nora A. Moskowitz, Lauren A. O’Connell.
Investigation: Nora A. Moskowitz, Rachel D’Agui, Aurora Alvarez-Buylla, Katherine Fiocca.
Methodology: Nora A. Moskowitz, Aurora Alvarez-Buylla, Lauren A. O’Connell.
Project administration: Lauren A. O’Connell.
Resources: Nora A. Moskowitz, Lauren A. O’Connell.
Supervision: Lauren A. O’Connell.
Visualization: Nora A. Moskowitz.
Writing – original draft: Nora A. Moskowitz.
Writing – review & editing: Rachel D’Agui, Aurora Alvarez-Buylla, Katherine Fiocca, Lauren
A. O’Connell.
References
1. Emlen JM. The Role of Time and Energy in Food Preference. Am Nat. 1966 Nov; 100(916):611–7.
2. Hutchinson DA, Savitzky AH, Burghardt GM, Nguyen C, Meinwald J, Schroeder FC,et al. Chemical
defense of an A sian snake reflects local availability of toxic prey and hatchling diet. J Zool. 2013 Apr;
289(4):270–8.
3. Weldon PJ. Avian chemical defense: Toxic birds not of a feather. Proc Natl Acad Sci. 2000 Nov 21; 97
(24):12948–9. https://doi.org/10.1073/pnas.97.24.12948 PMID: 11087849
4. Gregory PT. Responses of Natricine Snakes to Predatory Threat: A Mini-Review and Research Pro-
spectus. J Herpetol. 2016 Jun; 50(2):183–95.
5. Fordyce JA, Nice CC. ANTAGONISTIC, STAGE-SPECIFIC SELECTION ON DEFENSIVE CHEMICAL
SEQUESTRATION IN A TOXIC BUTTERFLY. Evolution. 2008 Jul; 62(7):1610–7. https://doi.org/10.
1111/j.1558-5646.2008.00388.x PMID: 18373625
6. Smyth RR, Tallamy DW, Renwick JAA, Hoffmann MP. Effects of age, sex, and dietary history on
response to cucurbitacin in Acalymma vittatum. Entomol Exp Appl. 2002 Jul; 104(1):69–78.
7. Whitney TD, Sitvarin MI, Roualdes EA, Bonner SJ, Harwood JD. Selectivity underlies the dissociation
between seasonal prey availability and prey consumption in a generalist predator. Mol Ecol. 2018 Apr;
27(7):1739–48. https://doi.org/10.1111/mec.14554 PMID: 29543392
8. Skelhorn J, Rowe C. Predators’ Toxin Burdens Influence Their Strategic Decisions to Eat Toxic Prey.
Curr Biol. 2007 Sep; 17(17):1479–83. https://doi.org/10.1016/j.cub.2007.07.064 PMID: 17716896
9. Wiggins NL, McArthur C, Davies NW. Diet switching in a generalist mammalian folivore: fundamental to
maximising intake. Oecologia. 2006 Apr; 147(4):650–7. https://doi.org/10.1007/s00442-005-0305-z
PMID: 16328546
10. Brenes-Soto A, Dierenfeld ES, Muñoz-Saravia A, Janssens GPJ. No longer a leap in the dark: the
importance of protein as an energy source in amphibians. Wildl Biol [Internet]. 2019 Sep 18 [cited 2021
Dec 8]; 2019(1). Available from: https://bioone.org/journals/wildlife-biology/volume-2019/issue-1/wlb.
00551/No-longer-a-leap-in-the-dark—the-importance/10.2981/wlb.00551.full
11. Bolton SK, Dickerson K, Saporito RA. Variable Alkaloid Defenses in the Dendrobatid Poison Frog
Oophaga pumilio are Perceived as Differences in Palatability to Arthropods. J Chem Ecol. 2017 Mar; 43
(3):273–89. https://doi.org/10.1007/s10886-017-0827-y PMID: 28289966
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 15 / 18
12. Hovey KJ, Seiter EM, Johnson EE, Saporito RA. Sequestered Alkaloid Defenses in the Dendrobatid
Poison Frog Oophaga pumilio Provide Variable Protection from Microbial Pathogens. J Chem Ecol.
2018 Mar; 44(3):312–25. https://doi.org/10.1007/s10886-018-0930-8 PMID: 29427191
13. Darst CR, Menendez-Guerrero PA, Coloma LA, Cannatella DC. Evolution of Dietary Specialization and
Chemical Defense in Poison Frogs (Dendrobatidae): A Comparative Analysis.: 14.
14. Santos JC, Coloma LA, Cannatella DC. Multiple, recurring origins of aposematism and diet specializa-
tion in poison frogs. Proc Natl Acad Sci. 2003 Oct 28; 100(22):12792–7. https://doi.org/10.1073/pnas.
2133521100 PMID: 14555763
15. Valderrama-Vernaza M, RamI
´rez-Pinilla MP, Serrano-Cardozo Vı
´H. Diet of the Andean Frog Ranito-
meya virolinensis (Athesphatanura: Dendrobatidae). J Herpetol. 2009 Mar; 43(1):114–23.
16. McGugan JR, Byrd GD, Roland AB, Caty SN, Kabir N, Tapia EE, et al. Ant and Mite Diversity Drives
Toxin Variation in the Little Devil Poison Frog. J Chem Ecol. 2016 Jun; 42(6):537–51. https://doi.org/10.
1007/s10886-016-0715-x PMID: 27318689
17. Fitzpatrick LC. Life History Patterns of Storage and Utilization of Lipids for Energy in Amphibians. Am
Zool. 1976 Nov; 16(4):725–32.
18. Wilder SM, Norris M, Lee RW, Raubenheimer D, Simpson SJ. Arthropod food webs become increas-
ingly lipid-limited at higher trophic levels. Jordan F, editor. Lett Ecol. 2013 Jul; 16(7):895–902.
19. James B. McClintock. On estimating energetic values of prey: implications in optimal diet models. Oeco-
logia. 1986; 70(1):161–2.
20. Simon MP, Toft CA. Diet Specialization in Small Vertebrates: Mite-Eating in Frogs. Oikos. 1991 Jun; 61
(2):263.
21. Vences M, Glaw F, Zapp C. Mageninhaltsanalysen bei madagassischen Fro
¨schen der Gattungen
Tomopterna, Aglyptodactylus, Boophis und Mantidactylus (Anura: Ranidae).: 8.
22. Ingle D, Cook J. The effect of viewing distance upon size preference of frogs for prey. Vision Res. 1977
Jan; 17(9):1009–13. https://doi.org/10.1016/0042-6989(77)90003-7 PMID: 304272
23. Meurer W, Gonc¸alves FG, Bovendorp RS, Percequillo AR, Bertoluci J. Diet Electivity and Preferences
for Food Resources in Chiasmocleis leucosticta (Anura: Microhylidae). J Herpetol [Internet]. 2021 Sep
27 [cited 2022 Jan 3]; 55(4).
24. Lease HM, Wolf BO. Lipid content of terrestrial arthropods in relation to body size, phylogeny, ontogeny
and sex. Physiol Entomol. 2011 Mar; 36(1):29–38.
25. Santos JC, Tarvin RD, O’Connell LA. A Review of Chemical Defense in Poison Frogs (Dendrobatidae):
Ecology, Pharmacokinetics, and Autoresistance. In: Schulte BA, Goodwin TE, Ferkin MH, editors.
Chemical Signals in Vertebrates 13 [Internet]. Cham: Springer International Publishing; 2016 [cited
2020 Jan 10]. p. 305–37. Available from: http://link.springer.com/10.1007/978-3-319-22026-0_21
26. Moskowitz NA, Dorritie B, Fay T, Nieves OC, Vidoudez C, 2017 Biology Class CRL, et al. Land use
impacts poison frog chemical defenses through changes in leaf litter ant communities. Neotropical Bio-
divers. 2020 Jan 1; 6(1):75–87.
27. Tourtois J, Grieshop M. Susceptibility of Dalotia coriaria (Kraatz) (Coleoptera: Staphylinidae) to Ento-
mopathogenic Nematodes (Rhabditida: Heterorhabditidae and Steinernematidae). Insects. 2015 Mar
18; 6(1):224–35. https://doi.org/10.3390/insects6010224 PMID: 26463077
28. Kalra B, Parkash R. Sex-specific divergence for body size and desiccation-related traits in Drosophila
hydei from the western Himalayas. Comp Biochem Physiol A Mol Integr Physiol. 2014 Nov; 177:1–10.
https://doi.org/10.1016/j.cbpa.2014.07.011 PMID: 25045840
29. Friard O, Gamba M. BORIS: a free, versatile open-source event-logging software for video/audio cod-
ing and live observations. Fitzjohn R, editor. Methods Ecol Evol. 2016 Nov; 7(11):1325–30.
30. Alvarez-Buylla A, Payne CY, Vidoudez C, Trauger SA, O’Connell LA. Molecular physiology of pumilio-
toxin sequestration in a poison frog [Internet]. Physiology; 2020 Nov [cited 2022 Jan 3]. Available from:
http://biorxiv.org/lookup/doi/10.1101/2020.11.03.367524
31. O’Connell LA, LS50: Integrated Science Laboratory Course, O’Connell JD, Paulo JA, Trauger SA, Gygi
SP, et al. Rapid toxin sequestration modifies poison frog physiology. J Exp Biol. 2021 Feb 1; 224(3):
jeb230342. https://doi.org/10.1242/jeb.230342 PMID: 33408255
32. Brooks M E, Kristensen K, Benthem K J, van, Magnusson A, Berg C W, Nielsen A, et al. glmmTMB Bal-
ances Speed and Flexibility Among Packages for Zero-inflated Generalized Linear Mixed Modeling. R
J. 2017; 9(2):378.
33. Donnelly MA. Feeding Patterns of the Strawberry Poison Frog, Dendrobates pumilio (Anura: Dendroba-
tidae). Copeia. 1991 Aug 1; 1991(3):723.
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 16 / 18
34. Born M, Bongers F, Poelman EH, Sterck FJ. Abrigos durante a estac¸ão seca e mudanc¸as na dieta do
sapo-veneno-de-flecha Dendrobates tinctorius (Anura: Dendrobatidae). Phyllomedusa J Herpetol.
2010 Jul 1; 9(1):37.
35. Dı
´az JA. Dietary selectivity and sexual size dimorphism of Chiasmocleis mehelyi(Anura: Microhylidae)
in a Cerrado area of southwest Brazil.: 7.
36. Flowers MA, Graves BM. Prey Selectivity and Size-Specific Diet Changes in Bufo cognatus and B.
woodhousii during Early Postmetamorphic Ontogeny. J Herpetol. 1995 Dec; 29(4):608.
37. Woodhead C, Vences M, Vieites DR, Gamboni I, Griffiths RA. Specialist or generalist? Feeding ecology
of the Malagasy poison frog Mantella aurantiaca.: 13.
38. van Gils JA, van der Geest M, Leyrer J, Oudman T, Lok T, Onrust J, et al. Toxin constraint explains diet
choice, survival and population dynamics in a molluscivore shorebird. Proc R Soc B Biol Sci. 2013 Jul
22; 280(1763):20130861.
39. Oudman T, Onrust J, de Fouw J, Spaans B, Piersma T, van Gils JA. Digestive Capacity and Toxicity
Cause Mixed Diets in Red Knots That Maximize Energy Intake Rate. Am Nat. 2014 May; 183(5):650–9.
https://doi.org/10.1086/675759 PMID: 24739197
40. Moroño A, Franco J, Miranda M, Reyero MI, Blanco J. The effect of mussel size, temperature, seston
volume, food quality and volume-specific toxin concentration on the uptake rate of PSP toxins by mus-
sels (Mytilus galloprovincialis Lmk). J Exp Mar Biol Ecol. 2001 Feb; 257(1):117–32. https://doi.org/10.
1016/s0022-0981(00)00336-1 PMID: 11165303
41. Navarro JM, Contreras AM, Chaparro O
´R. Short-term feeding response of the mussel Mytilus chilensis
exposed to diets containing the toxic dinoflagellate Alexandrium catenella. Rev Chil Hist Nat [Internet].
2008 Mar [cited 2020 Jan 22]; 81(1). Available from: http://www.scielo.cl/scielo.php?script=sci_
arttext&pid=S0716-078X2008000100004&lng=en&nrm=iso&tlng=en
42. Wiggins NL, McArthur C, Davies NW, McLean S. Behavioral Responses of a Generalist Mammalian
Folivore to the Physiological Constraints of a Chemically Defended Diet. J Chem Ecol. 2006 Jun; 32
(6):1133–47. https://doi.org/10.1007/s10886-006-9076-1 PMID: 16770709
43. Eide NE, Eid PM, Prestrud P, Swenson JE. Dietary responses of arctic foxes Alopex lagopus to chang-
ing prey availability across an Arctic landscape. Wildl Biol. 2005 Jun; 11(2):109–21.
44. Greenlees MJ, Phillips BL, Shine R. Adjusting to a toxic invader: native Australian frogs learn not to prey
on cane toads. Behav Ecol. 2010; 21(5):966–71.
45. Caton SJ, Ahern SM, Remy E, Nicklaus S, Blundell P, Hetherington MM. Repetition counts: repeated
exposure increases intake of a novel vegetable in UK pre-school children compared to flavour–flavour
and flavour–nutrient learning. Br J Nutr. 2013 Jun 14; 109(11):2089–97. https://doi.org/10.1017/
S0007114512004126 PMID: 23110783
46. Attygalle AB, Morgan ED. Chemicals from the glands of ants. Chem Soc Rev. 1984; 13(3):245.
47. Kishi Y, Parker J. Cell type innovation at the tips of the animal tree. Curr Opin Genet Dev. 2021 Aug;
69:112–21. https://doi.org/10.1016/j.gde.2021.01.009 PMID: 33784538
48. Daly JW. The chemistry of poisons in amphibian skin. Proc Natl Acad Sci. 1995 Jan 3; 92(1):9–13.
https://doi.org/10.1073/pnas.92.1.9 PMID: 7816854
49. Martins FMS, Oom M do M, Rebelo R, Rosa GM. Differential effects of dietary protein on early life-his-
tory and morphological traits in natterjack toad (Epidalea calamita) tadpoles reared in captivity: Differen-
tial Effects of Diet on Tadpoles. Zoo Biol. 2013 Jul; 32(4):457–62.
50. Venesky MD, Wilcoxen TE, Rensel MA, Rollins-Smith L, Kerby JL, Parris MJ. Dietary protein restriction
impairs growth, immunity, and disease resistance in southern leopard frog tadpoles. Oecologia. 2012
May; 169(1):23–31. https://doi.org/10.1007/s00442-011-2171-1 PMID: 22038058
51. Lips KR. Overview of chytrid emergence and impacts on amphibians. Philos Trans R Soc B Biol Sci.
2016 Dec 5; 371(1709):20150465. https://doi.org/10.1098/rstb.2015.0465 PMID: 28080989
52. Grafe TU, Schmuck R, Linsenmair KE. Reproductive Energetics of the African Reed Frogs, Hyperolius
viridiflavus and Hyperolius marmoratus. Physiol Zool. 1992 Jan; 65(1):153–71.
53. Girish S, Saidapur SK. Interrelationship between food availability, fat body, and ovarian cycles in the
frog,Rana tigrina, with a discussion on the role of fat body in anuran reproduction. J Exp Zool. 2000 Apr
1; 286(5):487–93. https://doi.org/10.1002/(sici)1097-010x(20000401)286:5<487::aid-jez6>3.0.co;2-z
PMID: 10684572
54. Chapman RF. The Insects: Structure and Function. Cambridge University Press; 2012. 963 p.
55. Bernadou A, Hoffacker E, Pable J, Heinze J. Lipid content influences division of labour in a clonal ant. J
Exp Biol. 2020 Jan 1;jeb.219238.
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 17 / 18
56. Kay AD, Zumbusch T, Heinen JL, Marsh TC, Holway DA. Nutrition and interference competition have
interactive effects on the behavior and performance of Argentine ants. Ecology. 2010 Jan; 91(1):57–64.
https://doi.org/10.1890/09-0908.1 PMID: 20380196
57. Redford KH, Dorea JG. The nutritional value of invertebrates with emphasis on ants and termites as
food for mammals. J Zool. 1984 Jul; 203(3):385–95.
58. Miller MJ, Bermingham E, Turner BL, Touchon JC, Johnson AB, Winker K. Demographic conse-
quences of foraging ecology explain genetic diversification in Neotropical bird species. Ruegg K, editor.
Ecology Letters. 2021 Mar; 24(3):563–71. https://doi.org/10.1111/ele.13674 PMID: 33389805
59. Bbosa T, Tamale Ndagire C, Muzira Mukisa I, Fiaboe KKM, Nakimbugwe D. Nutritional Characteristics
of Selected Insects in Uganda for Use as Alternative Protein Sources in Food and Feed. Weintraub P,
editor. J Insect Sci. 2019 Nov 1; 19(6):23. https://doi.org/10.1093/jisesa/iez124 PMID: 31853552
60. Sa
´nchez-Muros M-J, Barroso FG, Manzano-Agugliaro F. Insect meal as renewable source of food for
animal feeding: a review. J Clean Prod. 2014 Feb; 65:16–27.
61. Hussein M, Pillai VV, Goddard JM, Park HG, Kothapalli KS, Ross DA, et al. Sustainable production of
housefly (Musca domestica) fly larvae as a protein-rich feed ingredient by utilizing cattle manure.Qiu X,
editor. PLOS ONE. 2017 Feb 7; 12(2):e0171708.
62. Morrison C, Hero J-M. Geographic variation in life-history characteristics of amphibians: a review. J
Anim Ecol. 2003 Mar; 72(2):270–9.
63. Sheridan JA. Reproductive variation corresponding to breeding season length in three tropical frog spe-
cies. J Trop Ecol. 2009 Nov; 25(6):583–92.
64. Saporito Ralph A, Spande Thomas F, Garraffo H Martin, Donnelly Maureen A. Arthropod Alkaloids in
Poison Frogs: A Review of the ‘Dietary Hypothesis.’ HETEROCYCLES. 2009; 79(1):2.
PLOS ONE
Poison frog prey preference
PLOS ONE | https://doi.org/10.1371/journal.pone.0276331 December 1, 2022 18 / 18
Available via license: CC BY 4.0
Content may be subject to copyright.
