Stress‐induced changes in color expression mediated by iridophores in a polymorphic lizard

Abstract Stress is an important potential factor mediating a broad range of cellular pathways, including those involved in condition‐dependent (i.e., honest) color signal expression. However, the cellular mechanisms underlying the relationship between stress and color expression are largely unknown. We artificially elevated circulating corticosterone levels in male tawny dragon lizards, Ctenophorus decresii, to assess the effect of stress on the throat color signal. Corticosterone treatment increased luminance (paler throat coloration) and decreased the proportion of gray, thereby influencing the gray reticulations that produce unique patterning. The magnitude of change in luminance for corticosterone‐treated individuals in our study was around 6 “just noticeable differences” to the tawny dragon visual system, suggesting that lizards are likely to be able to perceive the measured variation. Transmission electron microscopy (TEM) of iridophore cells indicated that luminance increased with increasing density of iridophore cells and increased spacing (and/or reduced size) of crystalline guanine platelets within them. Crystal spacing within iridophores also differed between skin colors, being greater in cream than either gray or yellow skin and greater in orange than yellow skin. Our results demonstrate that stress detectably impacts signal expression (luminance and patterning), which may provide information on individual condition. This effect is likely to be mediated, at least in part, by structural coloration produced by iridophore cells.

fined as "the capacity to maintain vital cellular processes" (Hill, 2011).
Consequently, color signals may be reliable indicators of an individual's condition and ability to tolerate stress. Evidence for this hypothesis derives from experiments showing that an increase in glucocorticoids can affect the size and intensity of sexually selected color signals (Fitze et al., 2009;Kemp & Rutowski, 2007;Kemp, Vukusic, & Rutowski, 2006;Lendvai, Bókony, Angelier, Chastel, & Sol, 2013;Roulin et al., 2008;Saino et al., 2013;San-Jose, Granado-Lorencio, Sinervo, & Fitze, 2013;Steffen & McGraw, 2009). However, whether stress affects processes of pigment production and/or deposition, the structural properties of cells and tissues, or both is largely unknown. The few studies that disentangle environmental effects, including stress, on pigments versus structural components of the same color patch have shown the latter to be partly or wholly responsible for environmentally induced variation (Jacot, Romero-Diaz, Tschirren, Richner, & Fitze, 2010;Kemp & Rutowski, 2007;. In poikilothermic vertebrates (fish, amphibians, and reptiles), coloration is produced by the arrangement and interaction of three primary pigment cell (chromatophore) types that contain either light-absorbing pigments or light-reflecting crystalline guanine platelets (Bagnara & Matsumoto, 2006;Bagnara, Taylor, & Hadley, 1968). The three chromatophore types typically appear in distinct layers, forming the dermal chromatophore unit (Bagnara et al., 1968). The three layers of chromatophores comprise: the xanthophores, containing red to yellow pigments such as carotenoids and pteridines; the iridophores, containing light-reflecting and light-scattering purine (mostly guanine) crystals; and a layer of dermal melanophores, containing eumelanin pigments that absorb the remaining light and affect the intensity of the color produced (Bagnara & Matsumoto, 2006) (Figure 2). Epidermal melanophores, similar in appearance and function although of a smaller size than dermal melanophores, also appear close to the epidermis (Bagnara & Matsumoto, 2006). Variation in the size, shape, orientation, and distribution of crystalline platelets within iridophores can produce a range of structural blue, green, gold, and red coloration, as well as altering the color produced by pigmentation (Kuriyama, Miyaji, Sugimoto, & Hasegawa, 2006;Morrison, 1995;Morrison, Rand, & Frost-Mason, 1995;Morrison, Sherbrooke, & Frost-Mason, 1996;Saenko, Teyssier, van der Marel, & Milinkovitch, 2013;Teyssier, Saenko, van der Marel, & Milinkovitch, 2015). For example, changes in the spacing of platelets rapidly produce color change in tree lizards (Morrison et al., 1996) and chameleons (Teyssier et al., 2015). Furthermore, in common lizards (Lacerta vivipara), corticosterone manipulation affects carotenoidbased color expression by altering iridophore-mediated "background reflectance," but not carotenoid deposition in the skin .
The tawny dragon (Ctenophorus decresii) is a small agamid lizard inhabiting rocky outcrops of South Australia (Figure 1a). Males have conspicuously colored throats that they emphasize during male-male competition and courtship displays (Stuart-Fox, Moussalli, Johnston, & Owens, 2004). Males in northern populations are polymorphic for throat color, with four discrete morphs: gray (dark gray and cream reticulation with no orange or yellow), yellow (varying amounts of yellow with cream and dark gray reticulations), orange (varying amounts of orange with cream and dark gray reticulations), and orange+yellow (orange center surrounded by yellow) (Figure 1b-e). The color morphs differ in behavior (Yewers, Pryke, & Stuart-Fox, 2016) and associated endocrine levels (Yewers, Jessop, & Stuart-Fox, 2017). Although color morphs and color expression (amount of orange or yellow) are heritable, there is nevertheless substantial variation between individuals in the extent of coloration and patterning within each morph category (Rankin, McLean, Kemp, & Stuart-Fox, 2016). Such color variation may explain the ability of C. decresii males to distinguish between familiar and unfamiliar rivals, allowing them to avoid costly fights (Osborne, 2005). Thus, color conveys a range of information in this species, from F I G U R E 1 (a) The tawny dragon, Ctenophorous decresii, a small agamid lizard inhabiting rocky outcrops of South Australia. Photograph by Adam Elliott (b-e) The four throat color morphs of northern male C. decresii. ((b) "yellow" shows yellow coloration with gray patterning. (c) "orange" shows orange coloration with gray patterning. (d) "orange+yellow" shows a central orange patch surrounded by yellow coloration with gray patterning. (e) "gray" shows cream coloration with gray patterning) behavioral strategy to individual identity; yet the extent to which color expression is condition-dependent or influenced by stress remains unknown. In this study, we artificially elevated corticosterone levels and examined the effect on the chromatic and luminance properties. To identify cellular mechanisms, we quantified the effect of iridophore density and platelet spacing from transmission electron microscopy on perceived skin colors (orange, yellow, dark gray, and cream). Lizards were misted and provided with live crickets three times per

| Corticosterone implantation
Lizards were divided into two treatment groups, "cort" (n = 12) and "control" (n = 11). Those in the "cort" group had corticosterone levels artificially elevated using surgical implants, while "control" lizards were left untreated. Our goal was to ensure that the two groups differed in circulating corticosterone; therefore, "control" lizards were not sham-implanted, to ensure that "control" lizards would not exhibit an increase in corticosterone levels due to the stress of minor surgery. We did not include a third sham-implanted treatment group due to the limited number of individuals available for this study. Nevertheless, it is possible that differences in color could be due to unmeasured effects of the minor surgical procedure rather than to corticosterone increases. However, this seems highly unlikely given that the small, subcutaneous incision caused no bleeding or inflammation and given the large differences in circulating corticosterone between treatment and control groups (see Section 3). The incision was closed with medical adhesive silicone (Dow Corning, Midland, MI, USA). After implantation, the animals were monitored until they showed normal breathing, righting and palpebral, toe-pinch and tail-pinch responses. intra-and interassay coefficients of variation were 1.85% and 18.11%, respectively, and sensitivity was 30 pg/ml.

| Measuring color expression
Color expression of each lizard was measured over four weeks through weekly digital images of the throat taken under standardized light conditions and spectral reflectance measurements. Images were used to quantify pattern (the proportion of gray, orange, and yellow on the throat), while spectral measurements were used to quantify color.
Images were taken using a Canon PowerShot SX1-IS digital camera set to macro mode and calibrated with respect to radiance and light intensity (Stevens, Párraga, Cuthill, Partridge, & Troscianko, 2007). Reflectance measurements were taken weekly using an Ocean Optics JAZ spectrometer and PX-3 Pulsed Xenon light source connected to a probe via a bifurcated fiber-optic cable. Measurements were of an oval point sample 3 × 4 mm, and taken relative to a 99% Spectralon diffuse white reflectance standard (Labsphere, NH, USA).
Measurements were taken of multiple regions of the throat at a 45° angle in the range of 300-700 nm, the visual spectrum for most lizards (Loew, Fleishman, Foster, & Provencio, 2002;Olsson, Stuart-Fox, & Ballen, 2013;Yewers et al., 2015). Each measured region was classified as being either dark gray, orange, yellow, or cream based on its spectral properties and appearance. We calculated chromatic (color) and achromatic (luminance) contrasts against a neutral background (white-uniform 99% reflectance) for each region. This enabled us to assess both the nature and degree of change. We applied the receptor noise limited (RNL) model Vorobyev, Osorio, Bennett, Marshall, & Cuthill, 1998) to estimate the perceptual distance, measured in units of Just Noticeable Differences (JNDs), between skin color and the neutral white background, based on the likely visual system of C. decresii. We assumed an irradiance spectrum of full sunlight (Rottman, Floyd, & Viereck, 2004). Photoreceptor spectral sensitivities of UVS λ max = 365 nm, SWS λ max = 440 nm, MWS λ max = 493 nm, LWS λ max = 571 nm were based on information for the closely related ornate dragon, Ctenophorus ornatus (Barbour et al., 2002), as well as C. decresii visual systems, which includes a UVS cone (Yewers et al., 2015). We assumed that photoreceptor noise (ω i ) for the LWS photoreceptor was 0.1 and derived the (ω i ) of remaining photoreceptor classes using a ratio of 1:1:3.5:6 based on the relative photoreceptor frequencies in Barbour et al. (2002). Achromatic contrast was calculated based on the LWS photoreceptor, assuming ω I = 0.05 and was log-transformed for statistical analyses. Full details of visual modeling calculations are given in Teasdale et al. (2013) and McLean, Moussalli, and Stuart-Fox (2014). Achromatic contrast values were log-transformed to meet model assumptions.

| Transmission electron microscopy sample preparation
Four weeks after treatment, the animals were humanely killed according to ethics guidelines using pentobarbitone. We randomly chose a subset of 13 individuals from which we dissected 24 tissue samples (3 mm 2 ) representing the different skin colors (seven orange, eight yellow, four dark gray, and five cream). As pigment cell densities, and especially iridophores, are likely to differ between skin colors, we focused on skin color rather than morph, treating the same skin color from different morphs as equivalent (there is no difference in the spectral properties of orange or yellow between morphs). Nevertheless, we ensured that morphs were approximately equally distributed between treatment groups.
Samples were fixed in 2.5% glutaraldehyde in phosphate-buffered saline (PBS) for 4 hr at room temperature. Fixed tissue samples were rinsed three times in PBS for 10 min each, before being dehydrated in increasing concentrations of ethanol consisting of 10%, 30%, 50%, 70%, 90%, 100%, and 100% anhydrous ethanol for 60 min each.
Following dehydration, the cells were infiltrated with increasing concentrations of LR White resin in ethanol consisting of 25%, 50%, 75%, and 100% resin for 6 hr each step. After a second change of 100% resin, the samples were embedded in fresh resin in gelatine capsules and allowed to gently sink to the bottom. The gelatine capsules were capped to exclude air and the resin polymerized in an oven at 60°C for 24 h.
The embedded tissues in resin blocks were sectioned with a diamond knife on a Leica Ultracut S microtome and ultrathin sections (90 nm) were collected onto formvar-coated 100 mesh hexagonal copper grids. The sections on grids were sequentially stained with saturated uranyl acetate for 10 min and Triple Lead Stain for 5 min (Sato, 1968) and viewed in an FEI Tecnai Spirit transmission electron microscope at 120 kV. Images were captured with a Gatan Eagle digital camera, at a resolution of 2048 × 2048 pixels.

| Iridophore density
Images were taken at low magnification (1650×) to measure the density of iridophores in different skin colors up to 15 μm from the epidermis. At this magnification, iridophores and melanophores could be easily recognized but xanthophores could not be reliably distinguished from bundles of collagen (Figures 2, S1). Furthermore, melanophores were relatively rare. We measured the length of intersections of iridophores with a series of 15μm vertical transects (n = 539, 15-36 transects per skin sample), spaced 5 μm apart within each analyzable section of the sample. To determine iridophore density, we measured the proportion of each transect that intersected with iridophores, deriving a median proportion across the 15-36 transects, depending on the size of the sample, for statistical analysis.

| Iridophore platelet spacing
High magnification (6500× and 11,000×) images were taken to measure platelet spacing within iridophore cells. In some lizard species (e.g., spiny lizards, Morrison et al., 1995;whiptail lizards, Mathger, Land, Siebeck, & Marshall, 2003;plestiodons, Kuriyama et al., 2006; geckoes, Saenko et al., 2013;and chameleons, Teyssier et al., 2015), iridophore cells show a traditional lattice structure, where nearidentical platelets are evenly spaced throughout the cell. This is not the case in tawny dragon skin tissue, where there is too much variation to meaningfully characterize platelet arrangement based on measurements of platelet size and the intervening spaces. Instead, we measured the area of a series of 1 × 1 μm plots (n = 600, 21-28 per sample) taken up by platelets compared to cytoplasm ( Figure S2).
As the size of intact platelets and the holes left behind by platelets wholly or partially damaged by sample preparation were comparable, both were classified as iridophore platelets. Using ImageJ, the identified iridophore platelets were outlined in white and, following threshold adjustment, selected and measured using the wand tool ( Figure S2). We used the median area of iridophore platelets per sample (from the 21-28 1 × 1 μm plots per sample) for statistical analyses. This is equivalent to % of the two-dimensional area occupied by platelets relative to cytoplasm.

| Statistical analysis
We analyzed the effect of implants on circulating corticosterone using To measure changes in throat coloration in response to corticosterone treatment, we analyzed the proportion of gray, orange, and yellow using general linear mixed models (GLMM), with time, treatment, and their interaction as fixed factors. The achromatic and chromatic contrast of specified points on the throat were analyzed using GLMMs with sample color, time, and treatment, as well as the interaction between treatment and time, and color and treatment as fixed factors. In all models, ID was included as a random factor to account for repeated measures over time.
To assess the difference between skin colors (orange, yellow, dark gray, and cream) in the median proportion of iridophores intersected by 15μm transects perpendicular to the epidermis, we used a GLMM with sample color as a fixed factor and ID as a random factor to account for 11 instances where more than one sample was taken from one individual. We also assessed the effect that the median proportions of iridophores within a sample have on sample achromatic and chromatic contrasts using GLMMs with individual ID as a random factor.
To analyze the spacing of iridophore platelets, which may alter to produce changes in color (Morrison et al., 1996;Teyssier et al., 2015), we used a GLMM with the median area of iridophore platelets as a dependent variable, sample color as a fixed factor and lizard ID as a random factor. The effect that iridophore platelet area had on achromatic and chromatic contrast was analyzed using a GLMM with achromatic or chromatic contrast as the dependent variable, sample color, platelet area, and their interaction as fixed factor and ID as a random factor.
All analyses were performed in SAS 9.4 (SAS Institute Inc.; PROC MIXED). Residual plots for all models were checked for normality to ensure model assumptions were met and effects were considered significant at p < .05.

| Effect of corticosterone on skin color
Segmentation analysis of weekly photographs showed an effect of corticosterone treatment on the proportion of throat colors (Tables 1   and 2). There was a significant treatment by time interaction for the proportion of gray and orange throat coloration ( Table 2). The proportion of gray throat coloration decreased significantly in corticosteronetreated but not control individuals (Table 1). This was because the proportion of both yellow and orange in corticosterone-treated individuals increased, while the proportion of orange in control individuals decreased (Table 1). However, the change in the proportion of yellow or orange coloration individually for either treatment group was not significant (Table 2).
Chromatic contrast decreased at week 1 then increased to preimplantation levels at week 4 for both treatment groups (  (Table S1).

| Relationship between iridophore density and skin color
Iridophores were the most prevalent chromatophore from 5.6 to 15 μm from the epidermis, peaking at 7.2 μm where they intersected with 30.6% of transects. There were no significant overall differences between skin colors in the median proportions of iridophores that intersected with the 15μm transects (F 3,7 = 1.96, p = .21; Figure 4a).
T A B L E 1 Percentages of gray, orange, and yellow throat coloration, and achromatic and chromatic contrast values at three time points. Values are least-squared means ± SE. Control and cort values were not significantly different for any component of color at any one time point; however, gray coloration significantly decreased over time in cort-treated individuals and not control individuals. The proportions of gray, orange, and yellow throat colors were derived from segmentation analysis of digital images. Chromatic contrast and achromatic contrast values were derived from spectral measurements However, there was a significant positive relationship between the median proportion of iridophores intersected by the 15μm transects and achromatic contrast (F 1,9 = 11.64, p = .008; Figure 5a), but not chromatic contrast (F 1,9 = 1.70, p = .22).

| Relationship between iridophore spacing and skin color
Iridophore platelet area within 1 × 1 μm plots differed significantly between skin sample colors (F 3,584 = 11.9, p < .0001) (Fig. S3). Platelets in cream samples took up 28.8 ± 1.62% of 1 × 1 μm plots, a significantly smaller percentage than in dark gray or yellow samples (cream vs. dark gray: t 583 = −4.47, p < .0001, cream vs. yellow: t 583 = −3.57, p < .0004) (Figure 4b), indicating that the platelets in cream skin were either smaller on average and/or more widely spaced. Iridophore cells in orange samples had a significantly lower area taken up by platelets than in yellow samples (orange vs. yellow: t 583 = −3.78, p = .0002).
There was a weak trend for a relationship between achromatic contrast and median iridophore platelet area (F 1,9 = 3.37, p = .1).
Increasing contrast (i.e., decreasing luminance) was associated with an increase in platelet area (Figure 5b). Platelet area did not affect the chromatic contrast of skin samples (F 1,9 = 0.61, p = .46).

| DISCUSSION
Our results demonstrate that stress alters ornamental color expression in male tawny dragons. Throat color became paler (i.e., higher luminance), and the proportion of gray decreased in corticosteronetreated but not control individuals. Analyses of TEM images indicated that luminance is associated primarily with the density of iridophore cells, but also the spacing of crystal platelets within them. As expected, luminance increased with an increasing median proportion of iridophores and a decrease in the area taken up by crystals, most The effect of sample color on the density of iridophores, based on the proportion of a 15μm transect (N = 539) that intersects with iridophore cells (proportion of transects). (b) Differences in the area of 1 × 1 μm plots taken up by platelets in iridophore cells between sample colors. Data are least-squared mean ± SE. Different letters (a, b, and c) indicate statistically different values F I G U R E 5 (a) Correlation between median proportion of iridophores per transect and achromatic contrast against a white background (i.e., higher achromatic contrast = lower luminance, lower achromatic contrast = higher luminance) (R 2 = 0.36). (b) Correlation between iridophore platelet spacing (the median area of platelets within 1 × 1 μm plots) and achromatic contrast (R 2 = 0.14, NS) likely caused by an increase in spacing between crystals and/or a decrease in their size. The area taken up by crystals within iridophores also differed between skin colors, being greater in cream than either dark gray or yellow skin and greater in orange than yellow skin. We cannot directly associate corticosterone-induced changes in skin color with iridophore variation because there was no difference in the color (chromatic or achromatic contrast) of corticosterone-treated and control individuals in the subset of samples randomly chosen for TEM analyses. However, our results suggest that effects of stress on ornamental coloration in this species are likely to be mediated, at least in part, by effects on the structural color produced by iridophores.
The magnitude of change in achromatic contrast for corticosteronetreated individuals in our study was around 6 JNDs, suggesting that lizards are likely to be able to perceive the observed variation.
Furthermore, corticosterone had greatest impact on the proportion of dark gray reticulations that produce the unique patterning on tawny dragon throats. General lightening of throat coloration may also reduce the apparent contrast between different throat colors, causing gray patterning to appear less prominent. Therefore, the combined change in color and pattern that we observed in response to stress could be sufficient to convey information to mates and rivals on individual condition. Similarly, stress has been shown to alter both the size and luminance of dark coloration in birds Roulin et al., 2008) and other lizard species .
Changes in the density of iridophores and/or the spacing of crystalline platelets within them may mediate the stress-induced change in luminance that we observed. Paler samples (higher luminance) intersected a higher proportion of iridophore cells, in which platelets were more widely spaced and/or smaller. Changes in the spacing of platelets can occur rapidly in response to increased osmotic pressure in chameleons (Teyssier et al., 2015) and temperature in tree lizards (Morrison et al., 1996), causing coloration to become lighter or bluer, respectively. However, unlike chameleons or tree lizards, platelets in tawny dragon skin tissue were highly irregular in size, shape, and orientation (although quasi-ordered), which would produce incoherent scattering of light across all visible wavelengths (Vukusic, Hallam, & Noyes, 2007). Our data suggest that one or both mechanisms (change in iridophore density and change in crystal platelet spacing) may be involved in the change in luminance. The relative importance of these mechanisms may depend on time scale; for example, change in platelet spacing could occur relatively rapidly in response to acute stress, while changes in pigment cell density (or pigment synthesis/deposition) in response to chronic stress may occur over a longer time scale of days, weeks, or months.
Elevation of corticosterone levels appeared to have little effect on the intensity or proportion of orange or yellow coloration. Effects of stress on red to yellow coloration may depend on pigments used to generate these colors. The two classes of pigments used to generate red to yellow coloration in poikilothermic vertebrates are carotenoids and pteridines (Bagnara & Matsumoto, 2006 (Braasch, Schartl, & Volff, 2007;Svensson & Wong, 2011;Ziegler, 2003). Stress limits the size and color of pteridine-based orange ornamentation in striped plateau lizards (Weiss, Foerster, & Hudon, 2012;Weiss et al., 2013). Carotenoid-based coloration, however, is highly variable in its response to corticosterone which may have no effect (Fitze et al., 2009; in lizards), a negative effect (Loiseau, Fellous, Haussy, Chastel, & Sorci, 2008; in birds) or a positive effect (McGraw, Lee, & Lewin, 2011;Kennedy, Lattin, Romero, & Dearborn, 2013; in birds) although this may vary taxonomically. This may be because the relationship between stress and carotenoid-based color expression is influenced by the competing physiological functions of carotenoids and consequent trade-offs in carotenoid allocation (Fairhurst, Dawson, van Oort, & Bortolotti, 2014). When orange or yellow coloration is produced by both carotenoid and pteridine pigments, as is the case in tawny dragons (McLean et al., 2017), it is possible that their relative contribution could vary, depending on stress, although this idea has not been investigated to our knowledge. Furthermore, the corticosterone dose used in our implants was relatively low compared to those used by Weiss et al. (2013), and may have been insufficient to cause changes in either pteridine synthesis or carotenoid deposition.
Levels of circulating corticosterone post-implantation were also much lower than those found in wild tawny dragon populations (Jessop et al., 2016). Despite these elevated stress levels, tawny dragons in the wild are able to produce conspicuous coloration, which indicates that there may be other environmental effects on color expression and/or significant differences in the ability of captive and wild animals to cope with stress. Alternatively, captive tawny dragons that have reduced breeding opportunities may more readily allocate resources away from producing ornamental coloration. Ctenophorus decresii is polymorphic for male throat coloration, but color morphs do not differ in their response to capture stress (Yewers et al., 2017) despite their differences in aggressiveness and boldness (Yewers et al., 2016). Consistent with this, color morphs did not differ in effects of stress on color expression, with effects observed across skin colors within and between morphs.
Thus, stress may be a general mechanism influencing signal honesty in this species, irrespective of morph-specific behavioral strategies.
The study of the cellular mechanisms underlying color expression is important to understand both the information conveyed by signals and mechanisms maintaining their honesty. Stress may be a particularly important mechanism mediating condition-dependent (i.e., honest) color expression, particularly structural coloration (Kemp et al., 2006;McGraw, Mackillop, Dale, & Hauber, 2002;. Our results demonstrate that stress impacts ornamental coloration in a detectable way that may provide information on individual condition by influencing overall luminance as well as patterning. This effect is likely to be mediated, at least in part, by changes in iridophore cells producing structural coloration. Disentangling the effects of stress on the complex systems of color production involves an integrative approach investigating pigment cell structures, the phys-