© The Author 2017. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved.For permissions, please e-mail: journals.permissions@oup.com

The official journal of the

ISBEInternational Society for Behavioral Ecology

Behavioral Ecology (2017), 28(4), 1122–1130. doi:10.1093/beheco/arx076

Behavioral Ecology

Original Article

Differential predation drives the geographical divergence in multiple traits in aposematic frogsChangku Kang,a Thomas N. Sherratt,a Ye Eun Kim,b Yujin Shin,b Jongyeol Moon,b Uhram Song,c Jae Yeon Kang,d Kyungmin Kim,b and Yikweon Jangb aDepartment of Biology, Carleton University, 1125 Colonel by Drive, Ottawa, K1S 5B6, Canada, bDivision of EcoScience, Ewha Womans University, 52, Ewhayeodae-gil, Seodaemun-gu, Seoul, 120–570, Republic of Korea, cDepartment of Biology and Research Institute for Basic Sciences, Jeju National University, 102, Jejudaehak-ro, Jeju-si, Jeju, 690–756, Republic of Korea, and dDivision of Basic Research, National Institute of Ecology, 1210, Geumgang-ro, Songnae-ri, Maseo-myeon, Seocheon, 325–813, Republic of KoreaReceived 29 December 2016; revised 4 May 2017; editorial decision 7 May 2017; accepted 10 May 2017; Advance Access publication 3 June 2017.

Prey have evolved a range of traits to enhance their survival against predators. These traits often show geographical variation due to the differences in local predation pressure. To date, there exists ample evidence of the geographical variation in single anti-predator traits such as coloration induced by differential predation. However, predation pressure often induces the shift in a suite of (correlated) prey traits, such as coloration and behavior, rather than a single trait in nature. In this study, we investigated the nature and extent of geographical variation in both color and behavioral traits in the oriental fire-bellied toad, Bombina orientalis, testing whether the frogs experience different predation pressure in different sites and evaluating how this might affect a range of traits associated with predator avoidance. Specifically, we quantified dorsal/ventral coloration, daytime resting place (i.e. the place where each frog was first spotted), and activity levels of natural populations of the frogs in 9 geographically distinct sites distributed across South Korea. We found the clearest differences in these traits between mainland and Jeju Island populations, with Jeju Island frogs exhibiting 1) less bright and less chromatic aposematic ventral coloration than mainland frogs, 2) lower activity during both day and night, and 3) a ten-dency to rest underwater. Follow-up predation experiments using clay frog models found that frogs in Jeju Island suffer higher preda-tion pressure than mainland frogs, which readily explains the divergence in behavioral and morphological traits among populations.

Key words: Amphibian, aposematism, coevolution, color, divergence, divergent selection.

INTRODUCTIONGeographical variation in coloration is a common, taxonomi-cally widespread phenomenon that has captured many biolo-gists’ attention for a long time (Wallace 1865; Dobzhansky 1933; Jablonski and Chaplin 2000). The interplay between natural and sexual selection is thought to have played a major role in gen-erating this variation (Endler 1995). In many cases, sexual selec-tion exerts directional selection on coloration and promotes the evolution of sexually attractive (usually bright and colorful) col-oration, while predation pressure that favors camouflage balances out and suppresses the expression of bright coloration, thereby

generating diversity in coloration dependent on local predation pressure (Endler 1995; Zuk and Kolluru 1998). Each of the above studies serve to demonstrate that when predation pressure differs between regions then it can exert divergent selection, leading to geographical variation in color (Stuart-Fox et  al. 2004; Vignieri et al. 2010).

Local adaptations of the prey to avoid predation have been reported in numerous studies (Endler 1995). For example, lar-vae of the streamside salamander that coexist with fish predators exhibit lighter coloration which matches the coloration of their background compared to the larvae which inhabit predator-free sites (Storfer et  al. 1999). Differential selection by predators can also affect the expression of aposematic coloration: indeed, theory and empirical studies have demonstrated that spatial heterogene-ity of predator communities (or predators from different localities

Address correspondence to Y. Jang, Division of EcoScience, Ewha Womans University, Seoul, 120–570, Republic of Korea. E-mail: jangy@ewha.ac.kr.

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Kang et al. • Predator induced geographical traits variation in B. orientalis

who have learned to avoid distinct signals) can generate a geo-graphical mosaic of selection, promoting variation in aposematic signals (Endler and Mappes 2004; Chouteau and Angers 2011; Nokelainen et al. 2014).

In amphibians, there is considerable evidence that both natu-ral selection mediated by predators and sexual selection have affected the formation of geographical color variation (Rudh et al. 2007; Maan and Cummings 2009; Noonan and Comeault 2009). Accumulated evidence suggests that the divergence in frog col-oration often does not correlate with genetic distance and occurs faster than neutral processes which indicate that local selection pressure has shaped the generation of color pattern diversity in frogs (Rudh et al. 2007; Amézquita et al. 2009; Brown et al. 2010). To date, however, the majority of studies on geographical varia-tion in coloration and the selective pressure that generated this diversity in amphibians have largely focused on the poison frog family Dendrobatidae, while other systems have only rarely been investigated. Furthermore, most of the studies that identify the geographical variation have focused on a single trait variation (col-oration), and have not explored other traits (e.g., behaviors) which might have co-evolved with coloration (Rudh et  al. 2011; Rudh et  al. 2013). Therefore, to broaden our insights into this biologi-cally widespread phenomenon, it is crucial to explore the patterns of geographical variation in multiple (possibly correlated) traits in more taxa and evaluate the role of local adaptation in shaping this variation.

In this study, we examined the extent of geographical variation in multiple traits most likely involved in anti-predator defense and how predator selection has affected this variation. We have used the oriental fire-bellied toad, Bombina orientalis (Boulenger 1890; Amphibia, Anura, Bombinatoridae), as a model system for several reasons. First, B. orientalis shows 2 different functional color patterns that have evolved for anti-predator purposes: their dorsal color patterns resemble those of their natural backgrounds which pos-sibly make the frogs become cryptic, while their ventral color pat-terns are conspicuously colored and used as an aposematic signal to advertise their defensive chemicals (Bajger 1980). Second, there exist high degrees of natural variation in both dorsal and ventral color patterns (Figure  1). These characters make B.  orientalis ideal for studying adaptive coloration and how local adaptation shapes geographical patterns of variation in coloration. The phylogeo-graphic genetic structure of this species has been revealed recently (Fong et al. 2016), and it indicates that genetically different groups coexist in South Korea. However, natural (geographical) varia-tion in phenotypes, which may be directly under local selection have never been investigated to date. Our goals in this study are to 1) characterize the geographical patterns of natural variation in both dorsal and ventral coloration, 2) investigate whether behaviors that are associated with anti-predator defense co-vary with color-ation, 3)  examine phenotypic differences between mainland and island populations where genetic divergence has occurred (Fong et  al. 2016), and 4)  test whether local adaptation (induced by dif-ferential predator pressure) could be responsible for the observed divergence in morphological and behavioral traits.

To achieve our goals, we sampled natural populations of frogs in geographically distinct sites. We included 8 mainland sites dis-tributed across the mainland of South Korea and one island site. The island population is considered to be founded after a single dispersal event from the mainland that occurred in the region of 4.32 Ma ago (Fong et  al. 2016). For frog traits, we measured a suite of (inter-related) phenotypic characteristics including 1)  the

dorsal and ventral coloration, 2) daytime resting place (i.e. the place where each frog was first spotted), and 3)  day/night activity lev-els. To elucidate the drivers of this variation, we examined whether populations that exhibited different suites of phenotypic traits also experienced different predation pressure by conducting a clay frog model predation experiment. We predicted that frogs experienc-ing higher predation threats would have attributes that help the frogs 1)  reduce the encounters with predators (reduced activities and hiding habits) and/or 2)  increase the survivorship once they encountered a predator (brighter ventral aposematic signal). To test these predictions, we first investigated geographical variation in coloration and resting behavior across all 9 study sites. To further focus our study, we then compared the activity levels and predation pressures between the sites where the most dramatic behavioral and morphological differences were found.

MATERIALS AND METHODSStudy subject

B. orientalis is a semi-aquatic frog species commonly found in Korea, northern-east China, and adjacent parts of Russia. Like other Bombina species, this frog is semi-aquatic and inhabits a variety of different habitats including forests, river valleys, and swamps. Normally, B.  orientalis is found close to water. The adults emerge from late April and are active until they hibernate in late October in South Korea. They can live up to 30 years in captivity. The sexes are similar in appearance, but males have nuptial pads on their forefeet during the mating season and knobby backs. As noted, there exists high degree of natural variation in dorsal/ventral color

127º E

50 km

38º N

36º N

34º N

GP

GS AD

dorsal

ventral

Mainland

Island

dorsal ventral

WSJJ

SCYA

JI

GN

129º E

Figure 1Map of the study sites. The frog images on right show example images of dorsal/ventral coloration of mainland (upper) and Jeju Island populations (lower), respectively. Mainland populations showed high variation in dorsal coloration from green/dark patterned to dark-brown (upper 2 dorsal images) while the Island population showed dark-brown dorsal coloration exclusively. Images were standardized (linearized and equalized, see methods), thus look darker than the real specimen looks to human eyes, especially dorsal coloration.

1123

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Behavioral Ecology

patterns: their dorsal color pattern ranges from bright-green to dark-brown with dark spots and their aposematic ventral surface is covered to different extents with bright red-orange coloration with dark mottles (Figure  1). Birds and mammals are considered the main predators of adult B. orientalis. Under an imminent threat, they show “unken reflex” behavior which can be characterized by arching of their body to reveal previously hidden aposematic ven-tral coloration, a behavior that is often accompanied by the release of toxins through their skin (Bajger 1980).

Survey of geographical variation in coloration and behavior

To evaluate the extent of geographical variation in traits, we sur-veyed frogs in 9 different sites in South Korea: Gapyeong (GP), Gangneung (GN), Goesan (GS), Andong (AD), Jeonju (JJ), Woolsan (WS), Sooncheon (SC), Youngam (YA), and Jeju Island (JI). These sites were approximately evenly spread throughout South Korea (Figure  1, see Supplementary Table S1 in Supplementary mate-rials for GPS coordinates). For each site, we searched for frogs in local creeks. B. orientalis is the only Bombina sp. that inhabits in South Korea, so it is readily distinguishable from other species of frogs. Whenever a frog was detected, we noted the place that the frog was first spotted, captured it using latex-gloved hands, and took a pho-tograph of its dorsum immediately (Sony alpha 65 with SAL1855 lens, F5.6, ISO 400, manual shutter speed for controlling expo-sure; Sony, Tokyo, Japan). A  standard color checker (X-rite color checker passport, X-Rite, Kentwood, MI, USA) was always placed next to each frog for later processing of the images. All photos were taken as a RAW format set with a slight under-exposure to avoid saturation of the images (Stevens et al. 2007). The first place a frog was found was coded as either “on a substrate”, “under a rock”, “floating on water”, and “underwater” (Figure  2). We note here that when a frog was “on a substrate”, potential substrate materials included leaf litter, rocks, vegetation, and clay, but because B.  ori-entalis is active during daytime and moves between substrates fre-quently, we used a pooled category for all substrate types.

Following discovery of each frog and photography, we measured the reflectance of ventral coloration 3 times in random red areas of the frog using a spectrometer (see below). We also measured snout-vent-length (SVL) and identified the sex of each frog. Each individual was sexed based on the morphological differences: males

had nuptial pads on forefeet and knobby backs. Sex was clear for most frogs except for 11 individuals which were excluded from the analysis whenever sex was included as a predictor variable. After all our examinations were completed, we released the frogs at their collected spots. A total of 248 frogs from all populations (including Jeju Island) were sampled between 19th August and 2nd October 2014 during 0900–1800. We additionally sampled 34 individuals from Jeju Island in a different site in June 2016 for the first location to increase the power of the analysis and to confirm the behavioral consistency between populations within Jeju Island.

We used photography to quantify dorsal coloration yet spectrom-etry for ventral coloration for several reasons. When measuring dorsal coloration, it was important to measure the color immedi-ately after the capture because, like other frog species, dorsal color-ation of B. orientalis can change rapidly (unpublished data, Kang C). Since spectrometry requires setting-up time and handling of frogs for a prolonged time, photography was considered the appropri-ate method to rapidly quantify the natural variation. Reassuringly, spectrometry of dorsal coloration showed that the dorsum of B. orientalis reflected a negligible amount of light in the UV region (less than 5% reflectance across 300–400  nm), so standard pho-tography captured most of the variation in dorsal coloration in animal-visible spectrum (400–700 nm). On the other hand, ventral coloration of B. orientalis did not change rapidly and showed a sec-ond peak in the UV region (see Supplementary Figure S1). Since standard photography cannot capture variation in the UV region, we used a spectrometer (Jaz, Ocean Optics, Dunedin, FL) with a 99% reflectance standard (Lab-sphere, North Sutton, NH) to mea-sure the relatively fixed ventral coloration.

Image and spectral data processing

Each image was saved as RAW files and converted to 8-bit TIFF images using Sony Image Data Converter. To generate standard-ized color images that are independent of light conditions and device, we first corrected non-linear responses of the camera to the light intensity (a process called linearization), equalized the 3 RGB channels in respect to grey colors in the color checker (i.e. we ensured that a standard grey color had the same pixel value for the 3 channels), and rescaled images in a way that the brightest grey color in the color checker (which showed 99% reflectance across visible spectrum) had maximum RGB values, following Stevens et al. (2007). We then took 3 random areas from each frog’s dorsum image and measured median RGB values. We always measured brighter patches of dorsum unless frogs showed dark-brown color with no apparent patterns in that case we measured random dark areas. We then converted the measured RGB values into Lab color space (Wyszecki and Stiles 1982) to calculate brightness (i.e. the degree of dark or light), chroma (i.e., the degree of purity/vivid-ness of a color of a given hue), and hue (i.e. the attribute of a color that is discernible as red, blue, or yellow independent of brightness and chroma) using the following equations:

brightness = L,

chroma a b= + ,2 2

hue

ba

=tan .1−

From the reflectance data, the following colorimetric variables were calculated for each frog using “pavo” package in R (Maia et  al.

32%

11%

84%

3% 2%

35%

33%

Mainland(N = 165)

Jeju Island

Resting place

(N = 64)

floating on wateron a substrateunder a rockunderwater

Figure 2Comparison of the place where frogs were first spotted in daytime between mainland (pooled) and Jeju Island (JI) populations.

1124

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Kang et al. • Predator induced geographical traits variation in B. orientalis

2013): brightness was calculated as the mean relative reflectance over 300–700 nm range; chroma was calculated as (reflectancemax − reflectancemin)/(mean brightness); hue was calculated as the wave-length of maximum reflectance.

Activity level survey

Following the survey of geographical variation in coloration and resting behavior, we compared the activity levels of the frogs col-lected from different sites for 24  h. This investigation was con-ducted at only 3 sites to narrow down the number of sites to a manageable number within our time-scope: we chose GP and AP as representative sites for mainland, and JI as an island site. Our main goal for this survey was to compare mainland and island populations since the preceding survey of coloration and resting behaviors showed the most dramatic differences between mainland and island populations, and revealed only minor differences among mainland populations. The GP and AP sites were chosen because habitats in these 2 sites are most similar to JI site (creeks in a moun-tain, although most study sites shared these habitat characteristics), thereby reducing any effect of different predator types induced by differences in habitat type. In addition, the 2 sites were away from human activity which was crucial for the clay frog model experi-ment (see below) conducted simultaneously. For each population, we collected 5 male and 5 female frogs and brought them to indoor semi-aquatic terrariums (a facility at Ewha Womans University for mainland frogs, a facility at Jeju National University for Jeju frogs) that consisted of water, rocks, and moss which mimics the natural habitats of B.  orientalis (Edmonds 2011). Water was dechlorinated in the outside under the sun for 24 h before being used. Each sex was kept separately in 2 different terrariums (40  cm × 46  cm × 62  cm, 5 frogs in each terrarium) which were placed adjacent to each other. Juvenile crickets were provided as a food source ad libi-tum. First, the frogs spent 24 h in the terrariums to be acclimatized to the environment. B. orientalis is known to behave well in captivity, and our observations suggest that they had no problem exhibiting a range of behaviors including foraging, calling, mate searching, and mating (when 2 sexes were mixed). Then we video-recorded their behaviors 24  h from the above with natural circadian light-ing (approximately 14:10  h light:dark for all populations) without any disturbance. Other environmental conditions were maintained similarly between each test in indoor conditions (mean daily tem-perature varied only slightly between each test from 23.2  °C to 24.4  °C, humidity was a less concern because the terrarium was semi-aquatic where half of the surface was filled with water). After the experiment, all tested frogs were released to their locations of capture. This experiment was performed on 19th May (GP), 2nd June (JI), and 26th June (AD), 2016.

Predation experiment

To test whether differential predation pressure can help explain the observed geographical divergence in behavioral and morphologi-cal traits, we conducted a field predation experiment using frog clay models to compare the predation risk on B. orientalis between main-land and island sites. To reduce the number of sites to a manageable number, we conducted predation trials in 3 different sites (with 2 replications in each site) where geographical trait divergence was the most prominent based on the results of the coloration and behavior survey. We again chose GP and AD sites as representatives of main-land sites, and JI site as an island site. To make frog models, we used

sulfur-free plasticine clay (Gloriaclay, Sungnam, Kyungki-do, South Korea) using the modified methods of Yeager (2011). First, we cre-ated a hand-crafted initial clay model of B. orientalis, then made sili-con molds of frogs using silicon and catalyst mix (10:1). Using these silicon molds, we made each frog model by filling red colored plasti-cine clay into a mold, took it out, and painted the dorsal part of the model with non-toxic acrylic paints (Chromacryl, NSW, Australia). Since B.  orientalis showed natural variation in dorsal coloration, we prepared 2 types of frog models which differed in dorsal color pat-tern: 1)  dark-brown non-patterned form by mixing “black” and “forest green” color with 1:5 ratio, and 2)  green/black patterned form by mixing “forest green” and “green light” with 1:5 ratio for the background color, then overlaid dark spots using the dark-brown color (see Supplementary Figure S2 for example photos). Both dor-sal coloration and size of the frog model (48 mm SVL, the actual SVL of frogs ranging from 30.61 to 56.79  mm) were within the natural variation of B.  orientalis: brightness, chroma, and hue were 15.41, 20.44, and −0.66 for the green/black patterned model, and 3.23, 0.53, and −0.11 for the dark-brown model respectively, when the actual ranges of each colorimetric in frogs were 0.97 to 34.99, 0.27 to 47.35, and −1.55 to 1.56).

We conducted 6 predation trials (2 trials for each site) and deployed 100 frog models (50 dark-brown and 50 green/black pat-terned) for each trial. Because of the nature of field predation trials in geographically different sites, the full control of external condi-tions (such as the exact timing of a season or weather) was impos-sible. To reduce the confounding effects, we blocked the treatment sites (AD, GP, JI) and randomly allocated the order of testing sites within each block. For each trial, frog models were placed on a rock at the edge of the creeks where B.  orientalis is commonly found in mainland. After 24 h from deployment, we scrutinized any signs of predatory attack on the models and collected all back. On the sec-ond trial at the same site, we visited different, but adjacent creeks to conduct the trial. All trials were conducted between 7th June and 24th July, 2016 on non-rainy days.

Data analysis

All statistical analysis were conducted in R 3.1.2 (open-source software, R Developmental Core Team). To test the association between the place where each frog was first spotted and the site from which these observations were drawn, we used Pearson’s chi-square test of independence. Two sites (SC and YA,) were excluded from this analysis since the first place a frog was found was not properly recorded in these locations. Therefore, a classification matrix of 7 different sites × 4 first place spotted categories were used for the analysis.

To compare the colorimetrics (brightness, chroma, and hue) of dorsal coloration among frogs in geographically distinct popula-tions, we used permutation analysis of variance using “lmPerm” package (Wheeler and Torchiano 2016) because the data did not meet the assumptions of parametric tests. To compare the variance of dorsal colorimetrics among populations, we used the Brown–Forsythe test. However, because frogs can change dorsal coloration dependent on environmental conditions such as background color or temperature (King et al. 1994; Kang et al. 2016), the results of dorsal color comparisons should be interpreted with caution and may reflect the conditions of microhabitat, not the geographi-cal pattern. Therefore, we do not present the results in the main text but include it in Supplementary Information for the interested reader.

1125

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Behavioral Ecology

To test whether the colorimetrics of ventral coloration differed between study sites, we employed multivariate analysis of vari-ance (MANOVA) using the range of colorimetric endpoints as response variables, geographical site, sex, and their interaction as fixed explanatory variables. We also included SVL as an additional response variable to test whether there exists differences in size among populations in different sites. When significant effects were detected, we further employed univariate Anova to identify which colorimetrics showed significant differences. We additionally exam-ined whether frogs in different sites could be discriminated based on their colorimetrics using linear discriminant function analysis. Because of the large number of levels in treatment (9 different sites which would give the results of 36 comparisons per test), post hoc comparisons were not employed here which would dramatically increase the type-1 error (or decrease the power of the analysis if adjustment is applied), rather we discuss the results based on main results and data visualization.

For the analysis of activity levels, individual tracking was impos-sible from the video because frogs repeatedly (especially JI popu-lation) went underwater and disappeared from sight. Instead, for each study site, we measured “per capita movement per hour” and “per capita advertisement call by male frogs per hour” over 1-h periods. A movement was defined as “any movement of a frog that accompanied change in its position”. The change in its position was recorded only when a frog completely moved away from its cur-rent position by jumping (or swimming in the water), thus when it moved farther than its body size. When measuring calling activity, we only measured advertisement calls (a call that male frogs make to attract females), and not releasing calls (a call that is made when a frog has been grabbed by another frog) (Zeyl and Laberge 2011). We defined daytime as from 6 am to 7 pm, nighttime as from 8 pm to 5 am (excluded the transition time), and used this time category, site, and their interaction as explanatory variables, activity levels as response variables to employ MANOVA.

For the predation trial analysis, the survivorship of frog models (i.e. the presence/absence of signs of predator attack) was analyzed using generalized linear mixed model (GLMMs) with binomial error structure. Whether each replica survived for 24 h without any signs of attack was used as a binary response variable, study sites and color of the replica was used as explanatory variables with the interaction effect, and each block was set as a random effect. P val-ues were adjusted to control for false discovery rates (Benjamini and Hochberg 1995). All tests were 2-tailed.

Ethical statement

The experiment presented here met the ABS/ASAB ethical guide-lines for the use of animals in research. The protocols for this study were approved by Ewha Institutional Animal Care and Use Committee (IACUC14-111) in 2014, and Carleton Animal Care Committee (No. 104357) in 2016.

RESULTSGeographical variation in the place where frogs were first spotted

We found a significant association between study sites and the place where frogs were first spotted (χ18

2 = 304.69, P < 0.001; see

Supplementary Figure S3 for the graphs for each population). In this analysis, the pivotal difference was found between mainland and Jeju Island populations (chi-square test between pooled main-land vs. Jeju Island population; Figure 2; χ3

2 = 183.82, P < 0.001).

Although the spotted place also differed among mainland popula-tions (χ10

2 = 304.69, P < 0.001), all frogs were found either “floating”

or “on a substrate” or “under a rock” (Figure 2 and Supplementary Figure S3). However, most frogs in Jeju Island were found under-water, usually motionless under rocks or leaf litter (Supplementary Video S1) while all frogs in mainland were found actively moving on land. The first spotted places were not different between 2 dif-ferent JI sites (96 and 76% frogs were found underwater for each site respectively, χ3

2 = 10.976, P = 0.08). We note here that frogs on

the mainland were readily detected without searching because most frogs actively responded to human approach by either jumping away, swimming away, or escaping to a nearby refugee. However, most frogs in JI sites were spotted only after removing a rock or vegetation under water (see Supplementary video S1 for examples) and frequently did not move once they were revealed.

Geographical variation in ventral coloration

Our fitted MANOVA model showed significant effects of site (Wilk’s λ = 0.27, approximated F8,219 = 10.39, P < 0.001) and sex (Wilk’s λ = 0.92, approx. F1,219 = 4.56, P = 0.001), but no signifi-cant interaction (Wilk’s λ  =  0.83, approx. F8,219=1.25, P  =  0.16) on the colorimetrics and SVL of the frogs. In terms of colorimet-rics, the ventral coloration of frogs differed in terms of bright-ness (Figure  3a, F8,219  =  6.22, P  <  0.001), chroma (Figure  3b, F8,219  =  21.72, P  <  0.001), but not hue (F8,219  =  1.13, P  =  0.33) among sites. Notably, both brightness and chroma were lower in the JI population than the mainland populations. In other words, the ventral red coloration of B.  orientalis was brighter and more vivid in mainland than in JI. Although there existed some overlap, the discriminant function analysis separated the JI population from mainland populations (Figure 3c) based on ventral coloration, while there were considerable overlaps in the characteristics of mainland populations. Discriminant functions 1 and 2 separated 86% and 11% of the data respectively. Both LD1 and LD2 discriminated among sites based on brightness (correlation with LD1 = 0.32; with LD2  =  0.93) and chroma (LD1 correlation  =  0.82; LD2 correla-tion = −0.56), but not hue (correlations < 0.09).

There were no differences between the sexes in terms of bright-ness (F1,219 = 0.06, P = 0.80), chroma (F1,219 = 3.30, P = 0.07), but significant differences were found in hue (F1,219 = 13.40, P < 0.001) in that female ventral coloration was more shifted towards longer wavelengths, thus more reddish (hue for males: 663.44  ±  1.44, females: 672.68 ± 2.11, mean ± SEM). Interaction effects were not significant in all analyses (all P > 0.09). Populations also differed in terms of SVL (Figure 3d, F8,219 = 5.39, P < 0.001). Notably, JI frogs were small compared to the mainland frogs.

Comparison of activity levels between mainland and Jeju Island populations

Overall, the activity levels were different between day and night (Wilk’s λ = 0.76, approx. F1,64 = 9.82, P < 0.001), and among sites (Wilk’s λ = 0.49, approx. F2,64 = 13.10, P < 0.001). We also detected a significant interaction effect (Wilk’s λ = 0.77, approx. F2,64 = 4.28, P = 0.002). Specifically, in terms of the number of movement per hour, activity levels were not significantly different between daytime and nighttime (Figure 4a and b, F1,64 = 0.42, P = 0.51), but different among populations (F2,64 = 31.13, P < 0.001) in that JI populations were significantly less active than the other 2 mainland popula-tions (JI vs. AD: z = 6.65, Padj < 0.001; JI vs. GP: z = 6.62, Padj < 0.001). We found no differences between AD and GP populations

1126

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Kang et al. • Predator induced geographical traits variation in B. orientalis

(a) (b)

(c) (d)

DayNight

DayNight

20

10

Act

ivity

leve

l (m

ovem

ent)

0 Act

ivity

leve

l (m

ovem

ent)

0

Act

ivity

leve

l (ca

lling

)

0

1

2

3

4

5

AD GP JI

IslandMainland

AD GP JI

IslandMainland

5

10

15

***

******

***

0 5 10 15

Time of a day

LocationADGPJI

20

Act

ivity

leve

l (ca

lling

)

0

5

10

0 5 10 15

Time of a day

20

Figure 4Comparisons of activity levels in terms of movement (a and b) and male advertisement calls (c and d). AD and GP populations were used as representative populations of mainland. (a) and (c) show the change in activity levels in 24 h with 0 representing midnight local time. Each line represents the mean activity level per frog per hour for each population. (b) and (d) show the mean number of male calling (error bars represent ± SEM) for each population in daytime (defined as between 6 am and 7 pm) and nighttime (8 pm and 5 am). ***P < 0.001.

20

(a) (b)

(c) (d)

3.0

2.5

2.0

19V

entr

al b

righ

tnes

s

Ven

tral

chr

oma

18

17

16

AD GN GP GS JJ

Mainland Island

48

46ADGNGPGSJJSCWSYAJI

Snou

t-ve

nt le

ngth

(mm

)

44

42

40

location

Mainland Island

SC WS YA JI AD GN GP GS JJ SC WS YA JI

Mainland Island

AD−6

−2

0LD

2

2

4

−4 −2 0

LD1

2 GN GP GS JJ SC WS YA JI

Figure 3Comparisons of the colorimetrics of ventral coloration (a and b) and snout-vent length (d) among studied populations. Hue was not depicted because the differences were non-significant among populations. Points and error bars indicate mean ± SEM in (a), (b), and (d). (c) Plot of linear discriminant function LD1 and LD2 from discriminate function analysis on the colorimetrics of ventral coloration. Points indicate the position of each individual frog, ellipses show 95% confidence ellipses for each population. Although some overlaps exist, the JI population was largely separated from mainland populations while there were considerable overlaps in characteristics among mainland populations.

1127

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Behavioral Ecology

(z = 0.09, Padj = 0.92). We also found a significant interaction effect (F2,64  =  3.05, P  =  0.05), but data visualization suggests that the interaction effect (higher nocturnal activity in AD/JI population than GP population) is not prominent and appears weaker than the main effect of site (Figure 4b).

Calling activities were different among sites as well (Figure  4c and d; F2,64 = 11.00, P < 0.001) in that male frogs in mainland pop-ulations called more frequently than JI frogs (JI vs. AD: z = 4.69, Padj < 0.001; JI vs. GP: z = 3.72, Padj < 0.001). There were no dif-ferences between AD and GP populations (z  =  1.02, P  =  0.30). Calling activities were higher during nighttime than daytime (Figure 4d, F1,64 = 15.81, P < 0.001) only for mainland populations, but consistently low in JI population (interaction effect: F2,64 = 3.23, P = 0.04).

Frog model predation results

Among 600 deployed frog models, 45 models were attacked (7.5% of total). Of the attacked models, 44 models were found with signs of avian attacks, one showed a sign of rodent attacks (see Supplementary Figure S2 for example photos). We found a sig-nificant effect of site on survivorship of the frog models (Figure 5, χ2

2  =  15.33, P  <  0.001). Post hoc comparisons revealed that the

frog models were attacked more frequently in Jeju Island than the other 2 mainland sites (JI vs. AD: z = 2.29, Padj = 0.03; JI vs. GP: z  =  3.63, Padj < 0.001). There were no significant differences in the frog model survivorship between AD and GP sites (z  =  1.86, Padj = 0.06). We found no effect of frog color type on survivorship (χ1

2 = 2.93, P = 0.08) and no significant interaction effect of color

and site on survivorship (χ12 = 1.37, P = 0.59).

DISCUSSIONThrough a series of surveys and experiments, we examined the geographical patterns of variation in behavioral and morphologi-cal traits in B.  orientalis, and tested whether local predation pres-sure could be implicated in the observed geographical divergence in these traits. Our results clearly reveal the geographical pat-terns of variation in multiple traits, which were most dramatic when comparing characteristics between mainland and Jeju Island

populations. Jeju Island frogs can be characterized by 1) less bright and less chromatic aposematic ventral coloration, 2) less active dur-ing both day and night compared to mainland populations and 3) a tendency to rest underwater which was never found in mainland frogs. Although it is possible that the differences could be due to microhabitat differences rather than geographical differences, dorsal coloration also differed between mainland and Jeju Island populations in that frogs in Jeju Island exhibited consistently dark colored dorsum without bright-green patterns while frogs in main-land showed high variation in their dorsal coloration which poten-tially matches the coloration of their microhabitat (Supplementary Figure S4 and Supplementary Table S2). The results of predation experiments further suggest that differential predation between Jeju Island and mainland is a key driver that has shaped the divergence in traits in B. orientalis.

Given the clear differences, it is natural to ask what factors drove the diversification in behavioral/morphological traits between island and mainland. The recent phylogeographical analysis of B. orientalis suggests that Jeju Island population diverged from main-land populations by a single dispersal event (Fong et al. 2016). So, a founder effect followed by local adaptation to the island conditions (Foster 1964) may account for these evolutionary changes (Kolbe et  al. 2012). We note that the island frogs tended to be smaller, a feature commonly attributed to different selection pressures on body size in mainland and island populations (Whittaker and Fernández-Palacios 2007). In general theory, individuals in island population evolve larger size when predation pressure is relaxed, but they evolve smaller size when resources are limited (Whittaker and Fernández-Palacios 2007). In B. orientalis, our predation experi-ment showed that frogs in Jeju Island suffer higher predation pres-sure than mainland frogs (on average, the number of attacked frogs were 2.6 times higher on Jeju Island than on the mainland). We speculate that the higher predation risk in Jeju Island limited foraging activities and consequently resulted in a body size reduc-tion. However, further investigation would be necessary to disen-tangle the mechanisms that involved in body size reduction in Jeju Island frogs.

One evolutionary scenario for the observed phenotypic differ-ences between Jeju Island and mainland populations is that the higher predation risk in Jeju Island first drove the frogs to evolve underwater hiding behavior during daytime. Then this behav-ioral change initiated shifts in associated traits (Huey et  al. 2003) in that it subsequently, or synchronously if the 2 traits are linked (Armbruster and Schwaegerle 1996), affected their activity levels to avoid predation. With regard to the shift in coloration, we pro-vide 2 non-mutually exclusive hypotheses for the expression of less pronounced ventral aposematic coloration in Jeju population. First, less pronounced ventral aposematic coloration can be simply a con-sequence of hiding underwater and being less active which would decrease the time for foraging and the consumption of carotenoid-rich diet that are essential for the color expression (Umbers et  al. 2016). Alternatively, or in addition, the evolution of underwater hiding behavior might decrease the encounter rates with predators which resulted in relaxed selection on aposematic coloration (Lahti et al. 2009). Laboratory rearing experiments could be conducted to evaluate the plausibility of both hypotheses. If the dietary hypoth-esis were true, we would expect that the eggs (or early stages of tad-poles) from mainland and Jeju island would express similar ventral coloration when they become adults if they were reared under the same diet conditions. However, if the coloration was under relaxed selection, we expect genetic changes so that frogs from Jeju Island

20 n.s.

AD GP

Mainland Island

JJ

***

*

10

% A

ttac

ked

0

Figure 5Percentage of the attacked frog clay models after 24  h of deployment in predation experiments. Total 600 frog models (200 for each site) were deployed at 2 mainland sites (AD and GP) and Jeju Island (JI). Bars and error bars represent mean ± 95% binomial confidence intervals. Frog models were significantly more attacked in JI site than the 2 mainland sites (see results for statistics). There were no difference between GP and AD sites. n.s., non-significant; *P < 0.01, ***P < 0.001.

1128

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Kang et al. • Predator induced geographical traits variation in B. orientalis

would express less pronounced aposematic coloration despite the same diet.

Although our predation trials did not quantify the predation threats underwater (a practically challenging experiment since painted plasticine clay models can lose their color and shape in water), we consider that predation on adult B.  orientalis under-water is likely to be low at all sites because, to our knowledge, 1)  there are no aquatic predators that are sufficiently large to consume a B.  orientalis in a creek habitat, 2)  underwater forag-ing is limited to only a few uncommon species of avian pred-ators, such as Cinclus pallasii or Halcyon coromanda, in our study sites, and 3)  whenever B.  orientalis was found underwater in Jeju Island, they were usually found under rocks or leaf litter which further reduces their chance of being detected by preda-tors. Likewise, we could not rule out the possibility that, as in all natural systems, multiple frog models were attacked by the same individual predator. However, like all field experiments of this nature, we feel that our results reflect overall predation pres-sure regardless of the identities of the predators involved. Some may also argue that underwater resting behavior has evolved for other functions such as mating or foraging, but an adult B.  orientalis (both mainland and Jeju island frogs) mainly feeds on terrestrial invertebrates (Koh et  al. 2007) and finds a mate by calling, so underwater resting spot appears to have evolved mainly  to  avoid  terrestrial predators, with other functions less likely.

In addition to the recent evidence of genetic diversification of Jeju Island population (Fong et  al. 2016), our study further reveals the diversification of multiple phenotypic (morphological and behavioral) traits, most likely due to differential predation risk between mainland and Jeju Island. However, our under-standing of this diversification process is far from complete and many questions remain to be tested. We encourage future studies to tackle questions such as the role of gene/environment on the expression of warning coloration in different forms of B. orientalis, and the role of local selection and founder effects in the diversifi-cation process.

SUPPLEMENTARY MATERIALSupplementary data are available at Behavioral Ecology online.

FUNDINGThis work was supported by National Institute of Ecology Korea (NIE No. 2014-0004), National Research Foundation of Korea (NRF-2012R1A2A2A01046977, NRF-2017R1A2B2003579), and the Natural Sciences and Engineering Research Council of Canada Discovery Accelerator Program.

We thank Y. Kim, Y. Bae, K. Heo for their assistance in field studies, and J.W. Lee for spectrometer use. We also thank 2 anonymous reviewers for the constructive comments on the manuscript.

Data accessibility: Analyses reported in this article can be reproduced using the data provided by Kang et al. (2017).

Handling editor: Luke Holman

REFERENCESAmézquita A, Lima AP, Jehle R, Castellanos L, Ramos Ó, Crawford AJ,

Gasser H, Hödl W. 2009. Calls, colours, shape, and genes: a multi‐trait

approach to the study of geographic variation in the Amazonian frog Allobates femoralis. Biol J Linn Soc. 98:826–838.

Armbruster WS, Schwaegerle KE. 1996. Causes of covariation of pheno-typic traits among populations. J Evol Biol. 9:261–276.

Bajger J. 1980. Diversity of defensive responses in populations of fire toads (Bombina bombina and Bombina variegata). Herpetologica. 36:133–137.

Benjamini Y, Hochberg Y. 1995. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Ser B. 57:289–300.

Boulenger GA. 1890. A list of the reptiles and batrachians of Amoorland. Ann Mag Nat Hist. 5:137–144.

Brown JL, Maan ME, Cummings ME, Summers K. 2010. Evidence for selection on coloration in a Panamanian poison frog: a coalescent-based approach. J Biogeogr. 37:891–901.

Chouteau M, Angers B. 2011. The role of predators in maintaining the geographic organization of aposematic signals. Am Nat. 178:810–817.

Dobzhansky T. 1933. Geographical variation in lady-beetles. Am Nat. 67:97–126.

Edmonds D. 2011. Frogs and toads: a complete guide to fire-bellied toads, horned frogs and 40 other species. Neptune (NJ): TFH Publications Inc.

Endler JA. 1995. Multiple-trait coevolution and environmental gradients in guppies. Trends Ecol Evol. 10:22–29.

Endler JA, Mappes J. 2004. Predator mixes and the conspicuousness of aposematic signals. Am Nat. 163:532–547.

Fong JJ, Li PP, Yang BT, Zhou ZY, Leaché AD, Min MS, Waldman B. 2016. Influence of geology and human activity on the genetic structure and demography of the Oriental fire-bellied toad (Bombina orientalis). Mol Phylogenet Evol. 97:69–75.

Foster JB. 1964. Evolution of mammals on islands. Nature. 202:234–235.Huey RB, Hertz PE, Sinervo B. 2003. Behavioral drive versus behavioral

inertia in evolution: a null model approach. Am Nat. 161:357–366.Jablonski NG, Chaplin G. 2000. The evolution of human skin coloration. J

Hum Evol. 39:57–106.Kang C, Kim YE, Jang Y. 2016. Colour and pattern change against visually

heterogeneous backgrounds in the tree frog Hyla japonica. Sci Rep. 6:22601.Kang C, Sherratt TN, Kim YE, Shin Y, Moon J, Song U, Kang JY, Kim

K, Jang Y. 2017. Data from: differential predation drives the geographi-cal divergence in multiple traits in aposematic frogs. Dryad Digital Repository. http://dx.doi.org/10.5061/dryad.k6916.

King RB, Hauff S, Phillips JB. 1994. Physiological color change in the green treefrog: responses to background brightness and temperature. Copeia. 1994:422–432.

Koh Y, Jang M, Oh H. 2007. Comparison of food habits between the two populations of the fire-bellied toad, Bombina orientalis, in Korea. Korean J Environ Ecol. 21:461–467.

Kolbe JJ, Leal M, Schoener TW, Spiller DA, Losos JB. 2012. Founder effects persist despite adaptive differentiation: a field experiment with liz-ards. Science. 335:1086–1089.

Lahti DC, Johnson NA, Ajie BC, Otto SP, Hendry AP, Blumstein DT, Coss RG, Donohue K, Foster SA. 2009. Relaxed selection in the wild. Trends Ecol Evol. 24:487–496.

Maan ME, Cummings ME. 2009. Sexual dimorphism and directional sex-ual selection on aposematic signals in a poison frog. Proc Natl Acad Sci USA. 106:19072–19077.

Maia R, Eliason CM, Bitton P, Doucet SM, Shawkey MD. 2013. pavo: an R package for the analysis, visualization and organization of spectral data. Methods Ecol Evol. 4:906–913.

Nokelainen O, Valkonen J, Lindstedt C, Mappes J. 2014. Changes in preda-tor community structure shifts the efficacy of two warning signals in Arctiid moths. J Anim Ecol. 83:598–605.

Noonan BP, Comeault AA. 2009. The role of predator selection on poly-morphic aposematic poison frogs. Biol Lett. 5:51–54.

Rudh A, Breed MF, Qvarnström A. 2013. Does aggression and explorative behaviour decrease with lost warning coloration? Biol J Linnean Soc. 108:116–126.

Rudh A, Rogell B, Håstad O, Qvarnström A. 2011. Rapid population divergence linked with co-variation between coloration and sexual display in strawberry poison frogs. Evolution. 65:1271–1282.

Rudh A, Rogell B, Höglund J. 2007. Non-gradual variation in colour morphs of the strawberry poison frog Dendrobates pumilio: genetic and geographical isolation suggest a role for selection in maintaining poly-morphism. Mol Ecol. 16:4284–4294.

Stevens M, Parraga CA, Cuthill IC, Partridge JC, Troscianko TS. 2007. Using digital photography to study animal coloration. Biol J Linnean Soc. 90:211–237.

1129

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022

Behavioral Ecology

Storfer A, Cross J, Rush V, Caruso J. 1999. Adaptive coloration and gene flow as a constraint to local adaptation in the streamside salamander, Ambystoma barbouri. Evolution. 53:889–898.

Stuart-Fox DM, Moussalli A, Johnston GR, Owens IP. 2004. Evolution of color variation in dragon lizards: quantitative tests of the role of crypsis and local adaptation. Evolution. 58:1549–1559.

Umbers KDL, Silla AJ, Bailey JA, Shaw AK, Byrne PG. 2016. Dietary carotenoids change the colour of Southern corroboree frogs. Biol J Linnean Soc. 119:436–444.

Vignieri SN, Larson JG, Hoekstra HE. 2010. The selective advantage of crypsis in mice. Evolution. 64:2153–2158.

Wallace AR. 1865. I. On the phenomena of variation and geographical dis-tribution as illustrated by the Papilionidæ of the Malayan region. Trans Linn Soc Lond. 25:1–71.

Wheeler B, Torchiano M. 2016. lmPerm: Permutation tests for linear mod-els [2016 October 10]. Available from: http://cran.r-project.org/web/packages/lmPerm/.

Whittaker RJ, Fernández-Palacios JM. 2007. Island biogeography: ecology, evolution, and conservation. Oxford (UK): Oxford University Press.

Wyszecki G, Stiles WS. 1982. Color science. New York (NY): Wiley.Yeager J. 2011. A new technique for the production of large num-

bers of clay models for field studies of predation. Herpetol Rev. 42:357–359.

Zeyl JN, Laberge F. 2011. Multisensory signals trigger approach behaviour in the fire-bellied toad Bombina orientalis: sex differences and call speci-ficity. Zoology (Jena). 114:369–377.

Zuk M, Kolluru GR. 1998. Exploitation of sexual signals by predators and parasitoids. Q Rev Biol. 73:415–438.

1130

Dow

nloaded from https://academ

ic.oup.com/beheco/article/28/4/1122/3860981 by guest on 02 June 2022