Environmental correlates of internal coloration in frogs vary throughout space and lineages

Abstract Internal organs of ectotherms have melanin‐containing cells that confer different degrees of coloration to them. Previous experimental studies analyzed their developmental origin, role in immunity, and hormonal regulation. For example, melanin increases with ultraviolet radiation (UV) and temperature in frogs and fish. However, little is known about how environmental variables influence the amount of coloration on organs among amphibian species over a large spatial extent. Here, we tested how climatic variables (temperature, UV, and photoperiod) influence the coloration of internal organs of anurans. We recorded the level of melanin pigmentation using four categories on 12 internal organs and structures of 388 specimens from 43 species belonging to six anuran families. Then, we tested which climatic variables had the highest covariation with the pigmentation on each organ after controlling for spatial autocorrelation in climatic variables and phylogenetic signal in organ coloration using the extended version of the RLQ ordination. Coloration in all organs was correlated with the phylogeny. However, the coloration of different organs was affected by different variables. Specifically, the coloration of the heart, kidneys, and rectum of hylids, Rhinella schneideri, some Leptodactylus, and Proceratophrys strongly covaried with temperature and photoperiod, whereas that of the testicle, lumbar parietal peritoneum, lungs, and mesenterium of Leiuperinae, Hylodidae, Adenomera, and most Leptodactylus had highest covariation with UV‐B and temperature. Our results support the notion that melanin pigmentation on the surface of organs of amphibians has an adaptive function conferred by the protective functions of the pigment. But most importantly, internal melanin seems to respond differently to climatic variables depending on the lineage and locality in which species occur.

heart, kidneys, and rectum is still not clearly understood (Colombo et al., 2011), but it is 1 6 1 probably related to the functions of the melanin molecule which mainly acts as antibiotic, in 1 6 2 light absorption (e.g., photoprotection), cation chelator, and free radical sink (Riley, 1997).
Additionally, the intensity of coloration on the heart and kidneys of hylids tends to be lower 1 6 4 than in the testicles of Leiuperinae, showing a phylogenetic signal. 1 6 5 We found that the amount of melanin on a given organ is determined jointly by its 1 6 6 physiology, environmental variables, and phylogenetic relationship (see also Provete et al., 1 6 7 2012). Melanocytes have distinct physiology depending on the external coloration of the 1 6 8 animal. For example, pigmented cells on the peritoneum of fish respond to hormones, such as 1 6 9 melatonin and epinephrine (Sköld et al., 2010). However, the aggregation or dispersion of 1 7 0 pigmented cells promoted by the hormone depends on the transparency of the animal (Sköld et  1  7  1 al., 2010). This demonstrates that the internal pigment cells can adapt to distinct situations, 1 7 2 behaving differently in animals depending on their cutaneous coloration (Sköld et al., 2010). In 1 7 3 addition, changes in internal color in transparent animals may be related to substrate adaptation 1 7 4 or social signaling (Sköld et al., 2010). These results reinforce the role of physiological 1 7 5 responses of pigmented cells. 1 7 6 As UV radiation and temperature can have deleterious effects, species that occur in 1 7 7 places with high incidence of these factors could have developed more melanin on the testicles 1 7 8 as to protect their germinal epithelium, since damages in the gametes can influence the 1 7 9 reproductive fitness of individuals. For example, hylodids have a large amount of melanin on 1 8 0 the testicles and are restricted to the Atlantic rainforest. This region has the same degree of UV 1 8 1 incidence of the northwest of São Paulo, where swamp frogs of the subfamily Leiuperinae 1 8 2 occur. As a consequence, these species developed similar strategies to deal with elevated UV-B 1 8 3 variation by having high amount of coloration on the testicles. Also, having a high amount of 1 8 4 melanin on the testicles may allow species to be active during the day, such as dendrobatids 1 8 5 ( Grant et al., 2006), or at dusk like Pseudopaludicola and some Physalaemus (Vasconcelos and 1 8 6 Rossa-Feres, 2005). Conversely, species lacking melanin on the testicles are mainly active at 1 8 7 night (e.g., Hylidae and Leptodactylidae; Vasconcelos and Rossa-Feres, 2005). 1 8 8 Therefore, the amount of internal melanin seems to be a key trait influencing anuran 1 8 9 species distribution throughout space, since it can protect internal organs against the deleterious 1 9 0 effect of high UV-B, temperature variation, and photoperiod. The anuran species used in this study were collected at night when calling, near 1 9 5 breeding sites in 26 localities in the states of São Paulo and Goiás, which are housed at the 1 9 6 collection of the Laboratório de Anatomia -UNESP. We used at least five adult males of each 1 9 7 species for the analysis of pigmentation. The specimens were anesthetized with 5 g/L of 1 9 8 benzocaine and dissected to expose the organs. All procedures followed the recommendations 1 9 9 of the COBEA (Brazilian College of Animal Experimentation) and the Ethics Committee of our 2 0 0 university (Protocol #70/07 CEEA). We also analyzed additional specimens from the amphibian Belussi et al., 2012) for species of the family Hylidae. In total, we had 388 specimens from 43 2 0 7 species belonging to six families. Species had different sample sizes because same of them were 2 0 8 widely distributed. Thus, we wanted to asses if intraspecific variation in coloration was 2 0 9 somehow related to spatial variation in coloration.
We recorded the distribution of visceral melanocytes in 15 organs or structures using a 2 1 3 Leica stereoscopic microscope (MZ16), coupled with an image capture system, namely: heart, 2 1 4 lungs, rectum, peritoneum, kidneys, testes, and intestinal mesenterium. For each individual, we 2 1 5 recorded the coloration on these organs/structures based on coloration intensity, following the 2 1 6 protocol of Franco-Belussi et al. (2009). Briefly, the intensity of organ coloration was divided 2 1 7 into four categories, ranging from absence to entirely colored, as follows: Category 0) absence 2 1 8 of pigment cells on the surface of organs, in which the usual color of the organ is evident; 2 1 9 Category 1) a few scattered pigment cells, giving the organs a faint pigmentation; Category 2) a 2 2 0 large amount of pigment cells; Category 3) a massive amount of pigment cells, rendering an 2 2 1 intense pigmentation to the structure, changing its usual color and superficial vascularization To minimize multicollinearity, we calculated the Variation Inflation Factor (VIF; Zuur 2 2 7 et al., 2010) and a pair-wise correlation for the organs. Cardiac blood vessels, Renal veins, and 2 2 8 Lumbar nerve plexus had a high VIF. Therefore, we excluded them from further analysis. We extracted the bioclimatic variables Bio2, Bio3, Bio4, Bio5, Bio6, Bio7, and Bio10 2 3 2 related to temperature from WorldClim (Hijmans et al., 2005) for the 26 localities. Data for 2 3 3 photoperiod (minutes of light-hours in the rainy season, when most species were collected) were 2 3 4 obtained from the Brazilian National Observatory (BRASIL 2016). Data for UV-B radiation 2 3 5 were extracted from a raster file (Beckmann et al., 2014). We standardized all variables to zero 2 3 6 mean and unit variance previously to analysis. Posteriorly, we tested for multicollinearity (Zuur 2 3 7 et al., 2010) and removed environmental variables with VIF > 10. The reduced variables were 2 3 8 Bio2, Bio7, Bio10, UVB, and photoperiod. We then tested for spatial correlation in the 2 3 9 environmental variables. All environmental variables were spatially autocorrelated, with 2 4 0 Moran's I varying between 0.154 and 0.485. The reduced matrix of environmental variables was 2 4 1 analyzed with a Principal Component Analysis (PCA; Legendre and Legendre, 2012).

4 2
The phylogeny for the species to which we had trait data was pruned from the dated 2 4 3 phylogeny of Pyron (2014) for amphibians. This phylogeny was inferred based on nine nuclear 2 4 4 genes and three mitochondrial genes for 3,309 species, with average 20% of completeness. To 2 4 5 this topology, we added each individual as a polytomy to its corresponding species with branch 2 4 6 length equal to unit ( Figure 6). Since our trait is categorical, we could not simply calculate its 2 4 7 standard error to account for intraspecific variation. Then, we extracted a distance matrix from 2 4 8 this phylogeny and calculated a Principal Coordinates Analysis (PCoA; Legendre and Legendre, 2 4 9 2012) to extract phylogenetic eigenvectors.
For the species composition matrix, the presence of each individual analyzed was 2 5 1 placed in rows and localities as columns. This matrix was analyzed with a Correspondence 2 5 2 Analysis (CA). 2 5 3 To model space, we built a neighbor matrix linking sites separated up to 318.88 Km 2 5 4 (based on the truncation distance of a Minimum Spanning Tree). Then, we computed a PCA 2 5 5 onto this neighbor matrix to use as spatial variables in the extended RLQ. 2 5 6 The trait matrix contained the coloration category for each individual (rows) in each 2 5 7 organ (columns). We then tested for phylogenetic correlation (phylogenetic "signal") in the 2 5 8 coloration of each organ by decomposing the trait diversity, calculated as Rao's entropy, along 2 5 9 the nodes of the phylogeny (Pavoine et al., 2010). The Rao's quadratic entropy only uses tree 2 6 0 topology to decompose trait diversity. Afterwards, we tested if the diversity of coloration 2 6 1 categories was biased towards the root of the phylogeny, or concentrated in a single or a few 2 6 2 nodes (Pavoine et al., 2010). In this context, a phylogenetic signal occurs when trait diversity is 2 6 3 skewed towards the root of the phylogeny, implying that all its descending lineages would have 2 6 4 similar values for that trait. We found a phylogenetic signal in the coloration of internal organs 2 6 5 when we consider them altogether (Table S1), but not separately. Then, we calculated a distance 2 6 6 matrix for traits based on the modified Gower similarity coefficient (Pavoine et al., 2009). 2 6 7 Posteriorly, we tested for a relationship between environmental variables and the coloration of 2 6 8 each organ using a multivariate version of the Fourth-corner analysis. Significance was tested 2 6 9 using the null model 4 (Dray and Legendre, 2008). All organs, but the pericardium had 2 7 0 significant relationship with environmental variables. Thus, we excluded this organ from further 2 7 1 analysis.