Carotenoid intake during early life mediates ontogenetic colour shifts and dynamic colour change during adulthood

Carotenoids play an import role as one of the most prevalent pigments in animals. Carotenoid-based coloration accounts for striking sexually and naturally selected colour adaptations. Several anurans (frogs and toads) change body coloration either slowly and permanently between life stages (ontogenetic colour change), or rapidly and temporarily within minutes or hours (dynamic colour change). We investigated ontogenetic colour change from orange to green morphs in the Wallace's ﬂ ying frog, Rha- cophorus nigropalmatus , and tested the in ﬂ uence of dietary carotenoids on colour change during postmetamorphic development. At the age of 9 months, while all individuals still possessed orange-red body coloration, a 20-week-long feeding experiment was performed by supplying the frogs with either no carotenoid supplements or dietary carotenoids once or four times per week. A high carotenoid diet resulted in a faster increase in green colour chroma as well as higher levels of green and carotenoid chroma of back coloration. Less or no carotenoid supplementation led to an increase in UV-blue chroma, contributing to a dull turquoise appearance often observed in captive-bred and captive-raised anurans. In addition, we showed for the ﬁ rst time that Wallace's ﬂ ying frogs also perform dynamic colour changes. We tested dynamic changes triggered either by 2 min tactile handling or varying 1 h dark and light conditions. Our results demonstrate that a high carotenoid diet facilitates rapid and reversible change of body coloration in response to a tactile stressor, an adaptation absent in frogs receiving no carotenoids. Dynamic colour changes were likewise observed in response to changing light conditions presumably camou ﬂ aging individuals and providing protection from UV irradiation. The ontogenetic and dynamic pigmentation changes are discussed in relation to mechanism and as a likely strategy to avoid predation both at different life stages and in different environments.

Carotenoids are one of the predominant pigments used in animal coloration. All carotenoids have a chromophore in the centre of the molecule that absorbs light from 400e500 nm and is responsible for the appearance of yellow, red and orange colours (Bjørnland, 1997). Carotenoid-based coloration has received most attention in the context of sexually selected signals (Blount & McGraw, 2008;Svensson & Wong, 2011), such as visual signals during maleemale competition (Hamilton et al., 2013), species recognition (Blount & McGraw, 2008;Macedonia & Stamps, 1994) and courtship displays (Hill, 1991;Maoka, 2011). However, several carotenoid-based colours and patterns have evolved under natural selection because they promote predator avoidance strategies, for example aposematism (Bezzerides et al., 2007) and crypsis (Blount & McGraw, 2008;Delhey et al., 2010). Certain defence colorations may only be effective in specific habitats (Endler, 1978;Endler, 1993b;Merilaita et al., 2001;Price et al., 2019) or against particular predators (Barnett et al., 2018;Higginson & Ruxton, 2009). These relationships might change with age and/or across habitats (Endler, 1978;Rojas et al., 2018;Toledo & Haddad, 2009), for example, caterpillars of the genus Saucrobotys change from crypsis to aposematism as a response to shifting foraging demands during ontogenetic development (Grant, 2007). Natural and/or sexual selection for certain colour traits at differing life stages may result in ontogenetically changing pigmentation as species use different behavioural strategies to enhance survival and reproduction. Little is known about the underlying mechanism influencing this physiological and coloration demands. Dull or miscolorations resulting from restricted carotenoid availability are observed in captive-bred and captive-raised, but also wild species, for example ornamental fish (Sathyaruban et al., 2021), birds (Sumasgutner et al., 2018) and anurans (Frost & Robinson, 1984;Ogilvy et al., 2012). However, husbandry reports demonstrate successful reproductive events of miscoloured rhacophorids (Dunce, 2016). In such conditions, limited carotenoid resources might be primarily used for health and reproductive demands, for example egg production (Svensson & Wong, 2011), and are no longer available in sufficient amounts to invest in pigmentation. Duller or degraded colorations could be a first indication of nutritional problems and carotenoid deficiency. In the false tomato frog, Dyscophus guineti, dietary carotenoids increased both orange body coloration and circulating retinol concentrations, suggesting improved health (Brenes-Soto & Dierenfeld, 2014). Commercially available carotenoid supplements provide an important dietary component that could counteract health issues and promote correct pigmentation. Some studies in anurans have revealed positive effects of dietary carotenoid supplements on general health and/or coloration (Brenes-Soto & Dierenfeld, 2014;Dugas et al., 2013;Umbers et al., 2016) and suggest a critical effective period for their uptake during postmetamorphic development (Ogilvy et al., 2012).
Over recent decades, knowledge about nutritional requirements in husbandry, zoo-based amphibian research and conservation initiatives have become more important for a wide range of amphibian species due to rapidly declining habitats (Browne et al., 2011;Jacken et al., 2020;McFadden et al., 2008). Optimal husbandry conditions are, however, difficult to determine (Livingston et al., 2014), highlighting the immense difference between what animals can tolerate to survive and what they need for successful reproduction. Despite the clear importance of mimicking natural habitat conditions (temperature, humidity, light), there is still a lack of knowledge regarding feeding regimes in captivity compared to resources utilized in the individuals' natural habitat.
Wallace's flying frog, Rhacophorus nigropalmatus, undergoes a striking and irreversible colour change from orange-red juvenile to emerald-green adult morph (Ready, 2009). In the current study we investigated the ontogenetic colour change during the first 9 months and tested the influence of carotenoids on colour change which, despite its conspicuousness, has not been explored in detail. Captive-raised adults often exhibit lighter and turquoise body colorations (Ready, 2009) compared to wild conspecifics. We tested the hypothesis that carotenoid pigments in chromatophores supplied during postmetamorphic development affect body colour. We predicted that carotenoid resources during this life stage facilitate storage and accumulation of carotenoid pigments in the skin to establish the species' wildtype coloration as adults. Additionally, we explored the species' ability to dynamically change colour, an adaptive behaviour only briefly described for a related species, Rhacophorus smaragdinus, after handling an individual (Deepak et al., 2019). Based on the assumption that tactile stress effects dynamic colour changes we tested the behaviour and its dependency on previous carotenoid diets, predicting that carotenoids support sufficient accumulation and rearrangement of pigments for rapid changes. In a final step, we investigated a yet undescribed colour change of the frogs' complete and partial body in response to environmental light. We hypothesized that dynamic body coloration changes could be a beneficial camouflage strategy in fluctuating light environments of rainforests and predicted rapid and reversible changes when exposing the frogs to varying light and dark conditions.

Study Animals and Site
Rhacophorus nigropalmatus is a large nocturnal treefrog (males 79e89 mm; females 89e100 mm) that is native to the rainforest of South East Asia (Inger, 1966;Inger & Stuebing, 2005). These frogs are able to glide and manoeuvre easily through the understory of rainforests (Emerson & Koehl, 1990), which suggests that they live in the canopy of primary forests. During breeding, small groups have been observed in vegetation overhanging forest pools, in which females lay eggs protected by foam nests. Adults have green backs, white bellies and vocal sacs, yellow flanks and prominent black and yellow webbing between the toes (Inger & Stuebing, 2005). Juveniles look very different from adults, displaying an orange-red body coloration (Haas et al., 2021).
The study was conducted with a captive population of Rh. nigropalmatus at the Vienna Zoo, Austria. The Rainforest House in the Zoo houses adult frogs raised in captivity with an average snoutevent length of 85 mm (mean ± SD ¼ 85.28 ± 3.65 mm, N ¼ 9) and body mass of 37 g (mean ± SD ¼ 36.65 ± 6.91 g, N ¼ 9). Forty-eight Rh. nigropalmatus juveniles were bred and raised for the first time at the Zoo.

Husbandry
Frogs were obtained as juveniles and were 7 years old at the start of the study. Adults were fed three times per week with crickets, Acheta domesticus, and Turkestan cockroaches, Shelfordella lateralis. Feeder insects were dusted with either Amphib (Herpetal, Keweloh Animal Health GmbH & Co KG, Neuenkirchen, Germany), cuttlefish bone or Cal Plus LoD (Repashy Ventures Inc., Oceanside, CA, U.S.A.) once a week and were fed with each of the (multi-) carotenoid supplements Superload (Repashy Ventures Inc.), Superpig (Repashy Ventures Inc.) and spirulina algae (Blattner Heimtierfutter, Kempten, Germany) once per week. Carotenoid food supplementation of Superload and Superpig started 6 months before the reproductive event described below and adult individuals received no additional dietary carotenoids during earlier development.
On 12 August 2019 a brief but heavy thunderstorm in Vienna caused a drop in air pressure. This was a possible stimulator of breeding behaviour and, on the following day, a foam nest was discovered in the breeding terrarium. We immediately removed the two adult males and one female from the tank. Twenty-four hours later, the first tadpoles could be observed moving inside the nest and after 3 days 95 tadpoles had dropped out of the nest into the 20 cm deep water below (temperature 25e26 C). The tadpoles were transferred to five separate aquaria (60 Â 30 cm and 30 cm deep) equipped with air pumps as well as stones and clay pots as shelter. Water temperature was 25e26 C. Tadpoles received a diverse diet of Pleco tablets (Tetra GmbH, Melle, Germany), Sera Vipachips, Sera Catfish Chips, Sera Vipan Staple Food and Sera spirulina flakes (Sera GmbH, Immenhausen, Germany). In addition, over the 2-month larval stage, algae were offered once, blanched stinging nettle, Urtica dioica, twice and pollen granulate (Blattner Heimtierfutter) four times. The survival rate was 97.9%.
The first froglets were found 37 days after tadpole hatching and, within 2 weeks, the majority had metamorphosed. Forty-eight juveniles successfully metamorphosed and survived the first 2 weeks of development. As information about nutritional requirements is limited for most amphibian species and only general qualitative and quantitative guidelines are available (Ferrie et al., 2014;Pramuk & Gagliardo, 2012), the diet of the juveniles up until 9 months of age consisted of daily feedings alternating between fruit flies, Drosophila melanogaster, crickets, Turkestan cockroaches, silverfish, Lepisma saccharina, and waxmoths, Galleria melonella. Once a week Superpig and spirulina algae were fed to feeder insects 24 h prior to feeding. In addition, feeder insects were dusted three times a week with Amphib, cuttlefish bone, Vitamin D3 or Cal Plus LoD alternating every 2 days.

Ontogenetic colour change
To determine coloration changes of Rh. nigropalmatus, a spectrometer (JAZ series; Ocean Optics, Dunedin, FL, U.S.A.) was used to measure the spectral reflectance of the back, flank and throat of a randomly selected group of 12 juveniles during the first 9 months of development. The spectrometer had an integrated pulsed xenon light source (Jaz-PX) with a spectral response of 190e1100 nm. The reflectance data were collected for 300e700 nm and expressed as the percentage of reflectance relative to a white standard (WS-1 Diffuse Reflectance Standard, Ocean Optics). To reduce specular reflection, we used a customized probe holder to keep the reflection probe at a distance of 5 mm and at an angle of 45 to the surface of the frog's skin. The probe holder touched the frog's skin, preventing stray light from entering. For each measurement on the back, flank or throat three reflectance scans were averaged. Measurements were performed with the same group of individuals every 2 months starting 1 month after metamorphosis. We measured snoutevent length (SVL; ±0.1 mm) and body mass (±0.1 g) of each individual. Additionally, we took dorsal pictures of individuals immediately after metamorphosis (N ¼ 7) until 14 months of age (N ¼ 26) on millimetre-paper to quantify white spots (>1 mm) on the individual's back.

Feeding experiment
To experimentally test the impact of a carotenoid-supplemented diet on the coloration of Rh. nigropalmatus, we randomly assigned 26 individuals at the age of 9.5 months to one of three experimental groups: group 1 (N ¼ 8, G1) did not receive carotenoid supplements; group 2 (N ¼ 9, G2) received carotenoid supplements via gut-loaded crickets once a week; and group 3 (N ¼ 9, G3) received carotenoid supplements via gut-loaded crickets four times a week. Gut-loaded crickets were fed with cucumbers covered with the carotenoid supplement Superpig (0.7 g) containing the carotenoids astaxanthin, capsanthin, capsorubin, beta carotene, alpha carotene, beta cryptoxanthin, zeaxanthin, neoxanthin, cucurbitaxanthin, violaxanthin, lutein, echinenone, canthaxanthin and lycopene. This comprehensive carotenoid supply was chosen as no knowledge of the frogs' natural carotenoid profile is available and several carotenoids might contribute to pigmentation. Additional ingredients were Phaffia yeast, paprika, marigold flower, algae meal (Chlorella), rosehips, hibiscus flower, dried watermelon and turmeric. The general diet apart from the carotenoid supplement remained the same as described above.
The feeding experiment started after a baseline reflectance measurement of the back, flank and throat of the 26 test individuals. Reflectance and morphometric (SVL and body mass) measurements were performed every 2 weeks after the start of group diets for a period of 20 weeks until 14 months of age. One measurement had to be excluded due to incorrect data (errors in the reflection measurement). Reflectance measurements of back, flank and throat were also collected from nine 8-year-old adult individuals maintaining their regular diet with carotenoid supplements once per week for later comparison.

Dynamic colour change
Following the feeding experiment, we tested reversible and rapid colour change triggered by tactile handling in 14-month-old subadult Rh. nigropalmatus. The spectral reflectance of the back, flank and throat of eight adults and the 26 subadults, previously assigned to three different supplementary diets, were measured in three steps: (1) the baseline coloration was measured immediately after taking the individual out of the terrarium (handling time less than 30 s); (2) the next measurement was made after a 2 min handling period (touching, holding and placing the individual on a scale); and (3) the last measurement was recorded after a 30 min recovery period of resting on a plant in the terrarium.
In a further experiment, we tested whether different light conditions elicit dynamic colour change in two set-ups. In set-up 1, adults (N ¼ 6) and subadults from feeding group G3 (N ¼ 6) were placed in four empty terraria (three individuals per terrarium) during daytime (the natural resting period for the frogs). After a 1 h acclimatization period with regular daytime light conditions (Lucky Reptile Bright Sun UV 35W Model #BSJ-35), two terraria (one containing three adults, the other three subadults) were shaded with a black board allowing only ambient room light (Osram Lumilux Cool Daylight, L36W/865, 325 LM) to enter the front of the terraria. The other two terraria remained under daytime light conditions. Reflectance measurements of the backs of all 12 individuals were taken after 1 h without handling the individuals by placing the probe of the spectrometer on the back of the individuals while they were at rest. Subsequently the conditions were switched by removing the black boards and placing them on the previously lit terraria. The formerly shaded terraria now received regular daytime light conditions. Reflectance measurements were repeated after 1 h for all 12 individuals.
In the second set-up, we shaded parts of the individuals' body by loosely taping a leaf to the terrarium wall, covering the side or back of the frogs without touching them. We tested five adults and five subadults (G3) under regular daytime light conditions. After 1 h, reflectance measurements were collected from the lightexposed and shaded backs of all individuals.

Data Analysis
Three overall parameters of coloration were extracted from the reflectance spectra using Avicol software v6 (Gomez, 2006): brightness, hue, and chroma. Brightness corresponds to the total reflectance, calculated as the surface area under the spectral curve. Hue corresponds to the notion of colour and we calculated it as maximal reflectance searched in the range 300e700 nm and/or the wavelength of the maximal slope because several reflectance measurements lacked distinct peaks at specific wavelengths (Endler, 1990;Gomez, 2006;Gomez et al., 2009). Chroma is a measure of saturation of colour judged as a proportion of the brightness of a similarly illuminated area, or colour signal intensity describing the concentration of the reflectance around a wavelength. We analysed different parameters depending on the evaluated body part (Andersson et al., 2002). To describe the frogs' back reflectance and colour alterations due to diet we analysed UV-blue and green chroma as well as maximal average yellow chroma (Bl evin, 2018;Gomez, 2006), referred to below as carotenoid chroma. Carotenoid chroma is a measure linked to the incorporated level of carotenoids as they maximally absorb around 450 nm (Andersson et al., 2002;Andersson et al., 2007;Isaksson et al., 2008). UV-blue chroma is expressed as the difference between the reflectance of 300e450 nm divided by the total reflectance (R (300e450 nm) /R (300e700 nm) ). Green chroma is computed as the proportion of the total reflectance that occurs over the range of 500e560 nm (R (500e560 nm) /R (300e700 nm) ). Carotenoid chroma is the difference of the maximal reflectance between 500 and 700 nm and the reflectance at 450 nm divided by the average reflectance from 300 to 700 nm ((R max(500e700 nm) ÀR 450 )/R AV ) and reported as absolute value. From flank measurements we extracted green chroma and carotenoid chroma. To express ontogenetic changes from orange-red to bright white throat coloration we analysed UV chroma, the chroma over the range 300e400 nm (R (300e400 nm) / R (300e700nm) ) and carotenoid chroma. In addition, we analysed chroma in the range 400e500 nm (R (400e500 nm) /R (300e700 nm) ) to measure all wavelengths as reflected in white coloration.
To test differences in colour parameters during the ontogenetic development, the feeding experiment and the dynamic colour change experiment, we used KruskaleWallis tests to compare test groups and generalized linear mixed models (GLMMs) to compare parameters of the same test groups at differing measurement points. The GLMMs allow repeated measurements of the same individual to be fitted in the model as random variables. We used a GLMM with a normal distribution and identity link function. For post hoc tests, we used Student's t statistic with sequential Bonferroni correction for alpha. The statistical assumptions for GLMM analysis were met (KolmogoroveSmirnov test). For statistical analysis we used the program R (v3.4.2; R Core Team 2021) and SPSS version 26 (IBM SPSS Statistics, Armonk, NY, U.S.A.).
To describe colour change over the first 9 months of development we compared brightness, hue and chroma colour parameters of juveniles with GLMMs. The colour parameters were entered as dependent variables, with measurement point in time as predictor variables. We entered the identities of measurement point (individuals) as nested random variable to correct for repeated measurements of the same individuals. Similarly, we compared the number of dorsal spots during ontogenetic development. The numbers of spots were entered as dependent variables, with measurement point as predictor variable. The identities of measurement point (individuals) were entered as nested random variables.
To investigate the influence of carotenoid supplements we tested colour parameters of back, flank and throat between the three test groups and additionally compared coloration of adults. We used KruskaleWallis tests with Bonferroni-corrected pairwise comparisons to analyse group differences in the first baseline measurement and the last measurement after 20 weeks. Similarly, we compared size and weight of the three test groups. To test differences and similarities in colour change during the feeding experiment within each test group, we used GLMMs entering colour parameters as dependent variables, with the point of measurement as predictor variable. We entered the identities of measurement point (individuals) as nested random variables to correct for repeated measurements of the individuals.
Dynamic changes in colour parameters were compared between the treatments with GLMMs: baseline, handling and 30 min resting for each of the three feeding groups and the adults. For each test group and the adults, we entered the respective colour parameters as dependent variables, with treatment as predictor variable. We entered the identities of treatment (individuals) as nested random variables.
Further GLMMs were used to test dynamic colour change of shaded or lit terraria or body parts of adults and subadults. We entered the colour parameters as dependent variables, with the treatment (dark/light) as predictor variable. We entered the identities of treatment (terraria (individuals)) as nested random variables.

Ethical Note
Frogs were handled carefully and handling time was kept to a minimum ( 2 min). In between reflectance and morphometric measurements, frogs were able to move freely around their terraria, feed ad libitum and otherwise behave normally without interference from the experimenters. The study was designed to improve the welfare of captive-bred anurans and investigates the importance of carotenoids as dietary component during development. All experiments reported in this article complied with the current laws of the country in which they were performed (Austria) and required no institutional or governmental reviews. We followed the ASAB/ ABS Guidelines for the treatment of animals in research.

Feeding Experiment
At the beginning of the feeding experiment, 9.5-month-old juveniles had orange-red body coloration and across-group comparisons of back, flank and throat colour parameters showed no differences (KruskaleWallis tests: P > 0.05). By the end of the experiment, at the age of 14 months, the subadults had developed their adult body coloration (green backs, yellow flanks, white throats; Table 1, Fig. 4). The body size and weight were similar between the groups at the beginning and end of the experiment (KruskaleWallis tests: P > 0.5) and was thereby unaffected by supplementary carotenoids. No difference in whitish dorsal spot number was found between the groups during the feeding experiment from 9.5 to 14 months (KruskaleWallis tests: P > 0.5).
Within test group comparisons showed that the hue of the frogs' backs before and after 20 weeks of food treatment decreased similarly in G1 (GLMM: F 9,70 ¼ 2.484, P ¼ 0.016), G2 (F 9,80 ¼ 3.518, P ¼ 0.001) and G3 (F 9,80 ¼ 3.403, P ¼ 0.001). Back brightness varied during the experiment, but no differences between the beginning and end of the experiment were found in any test group (GLMM: P > 0.05 for all groups).

Throat colour changes
Comparison between test groups showed no differences in throat colour parameters after the feeding experiment (KruskaleWallis tests: P > 0.05; Table 1). Consistent with the white appearance of throats at the age of 14 months, we found similar colour changes within all groups: maximal slope and carotenoid chroma decreased (GLMMs: P < 0.001), UV chroma and chroma 400e500 nm increased (GLMMs: P < 0.001), whereas brightness did not change (GLMMs: P > 0.05).
The flanks of adult individuals also showed no differences in brightness or green chroma (KruskaleWallis tests: P > 0.05) compared to the three test groups. However, the carotenoid chroma of the flanks of adults was higher than that of G1 and G2 and similar to G3 (KruskaleWallis test: H 3 ¼ 16.608, P ¼ 0.001; pairwise comparison, adults versus G1: P ¼ 0.001; adults versus G2: P ¼ 0.017; adults versus G3: P ¼ 0.462; Table 1, Fig. A3b).

DISCUSSION
Rhacophorus nigropalmatus has a striking orange-red coloration in postmetamorphic development. In addition to the conspicuous coloration, several whitish spots form on the juvenile's back. A high carotenoid diet during postmetamorphic development resulted in a faster increase in green colour chroma and higher levels of green and carotenoid chroma of the back coloration (Fig. 5). Less or no carotenoid supplementation led to an increase in UV-blue chroma, contributing to a dull turquoise appearance often observed in captive-bred and captive-raised anurans. The frogs additionally were able to perform dynamic colour changes in response to tactile stress and changing light conditions. Our results demonstrate that a high carotenoid diet facilitates rapid and reversible change in body coloration, an adaptation absent in frogs receiving no carotenoids.
Insects, birds and Old World anthropoids have the ability to see red coloration (Bennett et al., 1994;Gerl & Morris, 2008;Surridge et al., 2003). Therefore, juvenile coloration in Rh. nigropalmatus can be perceived by predators and cannot be considered to contribute to camouflage like the cryptic back patterns and stripes observed in other rhacophorids (Biju et al., 2013). Natural selection for crypsis in the dappled light of the understory of primary rainforests should favour dark colours whereas the canopy promotes green body coloration (Endler, 1993a;Th ery, 2001). We suggest juvenile coloration could resemble a defence strategy by masquerading as a conspicuous, inanimate part of the environment (Ferreira et al., 2019;Toledo & Haddad, 2009). Several fruit-eating bats, as well as the Asian glossy starling, Aplonis panayensis, produce orange-red droppings. We propose that conspicuously coloured juvenile frogs mimic these animal droppings (Toledo & Haddad, 2009), which acts as an antipredator strategy mainly described for adults of some anurans such as Theloderma asperum (Jablonski & Hegner, 2015) and Dendropsophus marmoratus (Toledo & Haddad, 2009). White spots in Rh. nigropalmatus, only maintained during the early life stage, change the appearance of the frog's body which hence resembles the Asian crab spiders (Phrynarachne spp.) commonly known as 'bird dropping crab spiders' (Koh & Bay, 2019). The diurnal, immotile sleeping behaviour of the frogs on leaves and the indistinct body shape with tightly folded extremities reinforces the efficacy of masquerading as animal droppings. Morphological features, such as full webbing and accessory skin flaps, allowing the frogs to glide away and hence survive a predation attempt (Emerson & Koehl, 1990), develop throughout the first year of ontogenesis and are not present in juveniles.
During the carotenoid feeding experiment, all juveniles changed from orange-red to a body coloration similar to that of adults with green backs, yellow flanks and white throats, regardless of the test group (Figs. 1, 4). Carotenoids influenced the temporal occurrence  . Rhacophorus nigropalmatus subadult after partial shading. Bright-green body part was shaded with a leaf for 1 h; dark-green body part was light exposed.
of colour change and amounts of green and carotenoid chroma in the test groups. A high-carotenoid diet in G3 resulted in a faster increase in green chroma and higher levels of green and carotenoid chroma of the back coloration, whereas less or no carotenoid supplementation led to an increase in UV-blue chroma, contributing to the dull turquoise appearance of G1 (Fig. 4a). Similarly, the yellow flank stripes of G3 showed higher carotenoid chroma than G1. Overall, carotenoids contribute to the green body coloration in anurans (Michaels et al., 2015;Ogilvy et al., 2012) and effectively absorb light at wavelengths of 400e500 nm (blue-green). Green coloration results when light hits the amphibian skin and the iridophores reflect blue wavelengths, which passes through the overlying carotenoid-containing yellow/red xanthophores (Bagnara et al., 1968;Bagnara et al., 1969). Xanthophores also contain pteridines, which are produced by the chromatophores. Together with carotenoids, they are the main factors producing yellow body coloration (Duellman & Trueb, 1986;Steffen & McGraw, 2009;Umbers et al., 2016;Weiss et al., 2012). Our results are consistent with previous studies demonstrating that dietary carotenoids do not affect morphometrics, but increase yellow (e.g. false tomato frog) and green (e.g. red-eyed tree frog) pigmentation in anurans (Brenes-Soto & Dierenfeld, 2014;Ogilvy et al., 2012). Comparisons of the adults' back coloration showed similarities to that of G1 in UV-blue and carotenoid chroma. The adults at the Vienna Zoo were not raised on a high-carotenoid diet and never had the emerald-green appearance observed in wild populations in Borneo (D. Preininger, personal observation). Additional carotenoid supplementation in the adult population started 6 months prior to the reproductive event. This result suggests a paramount importance of carotenoid availability during postmetamorphic development and the likely inability to fully restore ontogenetic deficits in later life stages. Flank colorations are more difficult to interpret in this manner, as the adults' yellow coloration was similar to that of G3. The yellow flanks might constitute a sexually selected colour signal and/or could enhance close distance visual detection and recognition of conspecifics under low light conditions (Gomez et al., 2009;Robertson & Greene, 2017;Taylor et al., 2011). Several species demonstrate preferences for carotenoid pigmentations (McGraw & Ardia, 2003;Saino et al., 2000) conveying potential information about signaller quality and thereby influencing receivers. Limited carotenoid resources need to be balanced between competing coloration and physiological demands creating a trade-off (Svensson & Wong, 2011). Building on these assumptions, adults displaying degraded body colours would experience predation risks due to camouflage mismatch as a major cost from a limited carotenoid supply, while presumably sexually selected flanks would still favour investment in carotenoid-based signals. While predation risk is negligible in captivity, determining nutritional requirements remains a key element in successful breeding management especially in light of the rising number of ex situ conservation programmes (Jacken et al., 2020). Carotenoid availability in postmetamorphic frogs directly influenced skin colorations and we suggest a carotenoid-based diet contributes to the amphibian's welfare and might further prevent potential fitness consequences of captivity.
The behavioural relevance of dietary carotenoids was evident when we exposed the frogs to a mild stressor. Our results suggest that early life acquisition of carotenoids is essential to mediate the robustness of the adaptive behaviour of rapidly and reversibly changing body coloration in later life stages. The high carotenoid supplement group darkened significantly after handling and tended to recover after a resting period, unlike the other groups. A similar rapid colour change was observed in R. smaragdinus changing from green to brown in a few minutes, and reversing colour change after 25 min, a suggested camouflage strategy observed while handling the individual to obtain photo images (Deepak et al., 2019). Rapid colour changes have also been described in leaf frogs (Phyllomedusidae) and were suggested to act as a defence strategy, by increasing the contrast of the back coloration to the aposematic coloration of the flanks (Machado et al., 2015). In some species of the genus Hyla, stress leads to an aggregation of xanthophore pigments, a dispersion of melanin and a decrease in lightness, but epinephrine induces the opposite colour reactions (Nielsen & Dyck, 1978). Melanin and carotenoids are the two pigments responsible for most animal coloration serving several functions including crypsis and protection from sunlight (Alton & Franklin, 2017;Blount & McGraw, 2008;Byron, 1982;Licht & Grant, 1997). While melanin is synthesized within the body from amino acids, carotenoids need to be obtained by ingestion (Feltl et al., 2005;Ziegler, 2003). We propose that dietary carotenoids in Rh. nigropalmatus are necessary for accumulation and rearrangement of pigments inside skin chromatophores to mediate adaptive dynamic colour changes. Stress responses prompting quick colour change after surviving a predator attack while actively switching light environments by gliding through the forest understory might greatly improve crypsis and predation avoidance.
Further support for dynamic camouflage comes from experimental exposure to 1 h of light and shade conditions which reversibly influenced body coloration of Rh. nigropalmatus adults and subadults (Fig. 6). All individuals exhibited brighter backs in shaded conditions compared to lit conditions. Rapid changes in pigmentation in response to environmental light cues suggest that frogs can assess the brightness differences with skin and/or eye pigment cells, to achieve full or partial body colour changes in heterogeneous environments. In this case, light could act as a stimulus for the pituitary to release melanocyte-stimulating hormone, resulting in the dispersion of melanin and thereby elicit dynamic coloration changes (Camargo et al., 1999). In contrast, exact delimitations between coloration differences on the frog's back in response to partial shading suggest that light-sensitive pigment cells in frog skin such as melanopsin control pigmentation (Provencio et al., 1998). Rapid and reversible changes resulting from minor temperature variations are questionable; slight temperature changes could have arisen in our set-up, but the experimental room was maintained at a fairly constant temperature of 25 C during experiments and thermoregulation has less importance in tropical environments (Duellman, 1999). Pigmentation could also depend on the light regime of the frog's habitats (e.g. understory versus canopy) and environmental conditions, such as temperature or UV radiation (Endler, 1993a;Kang et al., 2017;Th ery, 2001). Chromatophores expand and accumulate in response to UV radiation and provide protection from the damaging effects of UV radiation, which is especially beneficial in the canopy (Blaustein & Belden, 2003;Garcia et al., 2004). Skin darkening has also been suggested to result in an increased absorption of UV radiation by the skin and a decreased transmission of UV radiation to deeper tissues (Blaustein & Belden, 2003;Coohill et al., 1970;Coohill & Fingerman, 1975). Studies in several animal taxa have documented the efficiency of a carotenoid-rich diet in photoprotection via the filtering of harmful radiation (Britton, 2008), but experimental studies in anurans addressing this question are scarce.
We suggest that accumulation of skin carotenoids and dispersal of melanin support rapid and reversible skin coloration and act as both dermal defence against UV radiation and camouflage strategy. Dynamic colorations allowed individuals to change their appearance based on differing light conditions and in response to tactile stress. The cellular and/or endocrine control and mechanisms have yet to be explored in detail. Colour-changing species also provide an opportunity to assess how changes in individual appearance influence predation risk. We should be cautious about drawing conclusions on wild populations from a captive population, but the observed adaptations warrant future studies. Investigations of predation events and visual environments, coupled with measurements of background matching, dynamic colour change and masquerading in the natural habitat could help to explain how differing strategies benefit predator avoidance and might challenge our current view of the adaptive function of coloration in amphibians.

Conclusion
Across several animal taxa, carotenoids are the main components of colour production depending on the individuals' nutrition. Successful husbandry and breeding of amphibians in conservation and research programmes requires knowledge of nutritional demands, which is still scarce and often species specific. Carotenoids are fundamental dietary components and their exclusion in the diet of frogs had effects on natural coloration and the stress response. Pigmentation changes likely contribute to predator avoidance, which is certainly inconsequential in captive populations. Nevertheless, we suggest dull body colorations are the first indication of nutritional deficiencies that could subsequently lead to poor fitness in captive populations. Effectively compensating for a lack of carotenoids during postmetamorphic development might be difficult at later life stages. In our experiments, commercially available carotenoids supplied via gut-loaded feeder insects provided a sufficient diet for frogs to invest in skin pigmentation. However, it remains questionable what can be considered a deficient, adequate or excessive quantity of carotenoids. High amounts of supplements might be harmful for individuals (Svensson & Wong, 2011) or even toxic (Britton, 1995;Burton & Ingold, 1984). There are certainly consequences at both ends of the continuum, and it remains immensely important to understand gradual variations in dietary supplements to reveal what animals need in contrast to what they can tolerate. Such conclusions can only be drawn when investigating dietary components in natural habitats. Considering the current alarmingly rapid species and environmental declines, this research is desperately needed.

Author Contributions
S.S. and D.P. planned the experiments. S.S., S.C. and D.P. carried out the experiments. S.S. and D.P. analysed the results. All authors contributed to the interpretation of the results. S.S. and D.P. wrote the manuscript. All authors provided critical feedback and helped improve the manuscript.  Adult G1 G2 G3 Figure A3. Comparisons of colour parameters of (a) back, (b) flank and (c) throat between Rhacophorus nigropalmatus adults and juveniles after a 20-week carotenoid supplement feeding experiment. Group 1 received no supplements (G1, N ¼ 8), group 2 received a carotenoid supplement once per week (G2, N ¼ 9) and group 3 received a carotenoid supplement four times per week (G3, N ¼ 9). Box plots represent median (centre horizontal line), the 25th to the 75th percentile (box), and the 5th and 95th percentiles (whiskers); dots denote outliers (also see Table 1).