Experimental stress during molt suggests the evolution of condition‐dependent and condition‐independent ornaments in the king penguin

Abstract Sexual selection and social selection are two important theories proposed for explaining the evolution of colorful ornamental traits in animals. Understanding signal honesty requires studying how environmental and physiological factors during development influence the showy nature of sexual and social ornaments. We experimentally manipulated physiological stress and immunity status during the molt in adult king penguins (Aptenodytes patagonicus), and studied the consequences of our treatments on colourful ornaments (yellow‐orange and UV beak spots and yellow‐orange auricular feather patches) known to be used in sexual and social contexts in this species. Whereas some ornamental features showed strong condition‐dependence (yellow auricular feather chroma, yellow and UV chroma of the beak), others were condition‐independent and remained highly correlated before and after the molt (auricular patch size and beak UV hue). Our study provides a rare examination of the links between ornament determinism and selection processes in the wild. We highlight the coexistence of ornaments costly to produce that may be honest signals used in mate choice, and ornaments for which honesty may be enforced by social mediation or rely on genetic constraints.


| INTRODUCTION
Understanding the evolution of conspicuous ornaments, often costly to produce and maintain, has been a long-standing focus of evolutionary biology (Andersson, 1994;Kuijper, Pen, & Weissing, 2012). Darwin (1871) laid the groundwork for this topic by observing that conspicuous ornaments could enhance access to sexual partners and reproduction, and that their evolution might be explained by sexual selection.
In the second half of the twentieth century, researchers realized that ornaments could also be used in competition for nonsexual resources, such as access to food and territories outside reproduction (Tobias, Montgomerie, & Lyon, 2012;West-Eberhard, 1983). West-Eberhard (1983) pointed out in her theory of social selection that ornaments can evolve whenever they enhance gene replication due "to differential success in social competition, whatever the resource at stake" (West-Eberhard, 1983). Consequently, sexual selection can be viewed as a "subset of social selection in which the resource at stake is mates" West-Eberhard, 1983). However, drawing the line between ornaments that are under sexual versus "nonsexual social" selection (hereafter "social selection") can be complicated as many ornaments have multiple functions. For example, ornaments used for courtship behavior may also be used for year-round territory defense (Tobias et al., 2012).
Regardless of function or selection process, ornament evolution may depend on how honestly they reflect the health, vigor, or dominance of their bearer (Tanaka, 1996;Zahavi, 1975 alternative mechanisms have been proposed to guarantee signal honesty: condition-dependence, social mediation, or genetic constraint (Hill, 2014). Whereas the condition-dependent signaling hypothesis states that individuals in better condition can afford to produce ornaments of higher quality highlighting the cost of production and/or maintenance (Hill, 2011), the social mediation and genetic constraint hypotheses are frequently invoked to explain the occurrence of trait production that is condition-independent (Hill & Brawner, 1998;Roulin, 2016). Under the social mediation hypothesis, the costs of bearing bright ornaments may be defrayed only after trait production, whereas the genetic constraint hypothesis implies that the ornament is not necessarily costly to produce, but it is nonetheless difficult or impossible to cheat (genetically linked) (Hamilton & Zuk, 1982).
We experimentally tested for condition-dependence in ornament production in a monomorphic seabird, the king penguin (Aptenodytes patagonicus), by manipulating birds' physiological status during molt. King penguins are brightly colored seabirds that breed in large colonies (>10,000 breeding pairs) throughout the subAntarctic islands. They display showy ornaments with three different modalities of color production (Table 1). Orange beak spots contain both exogenous carotenoid pigments (McGraw et al., 2007) and specialized stacks of elongated lamellae (resulting from cellular specialization) that also reflect structural UV colors (Dresp & Langley, 2006). In addition, king penguins possess orange auricular feather ornaments that contain an endogenously synthetized pterin pigment (Thomas, McGoverin, McGraw, James, & Madden, 2013). Experimental reductions of beak UV reflectance and auricular patch size have been shown to decrease the likelihood of pairing and thus the initiation of a reproductive event Nolan et al., 2010;Pincemy, Dobson, & Jouventin, 2009), demonstrating their important use in mutual mate choice. Correlative studies have also reported that individuals with larger auricular patches are more aggressive (Viera, Nolan, Côté, Jouventin, & Groscolas, 2008) enabling them to occupy more central breeding territories in the colony thought to be of greater reproductive value Viera et al., 2008). Those results suggest that auricular patch size may function as a social signal of dominance. Further, the color of yellow breast feathers (also containing the endogenously synthetized pterin pigment) has been related to innate immunity (Nolan, Dobson, Dresp, & Jouventin, 2006), suggesting that genes involved in the production of that specific pigment may also be linked to the immune system. There is also support for links between ornament colors and various condition indices, including body size and condition, stress status, and metabolic rate (Keddar, Couchoux, Jouventin, & Dobson, 2015;Schull et al., 2016;Viblanc et al., 2016) (see Table 1). Finally, the beak spot ornament is a dynamic signal, reflecting short-to medium-term physiological changes in parasite loads and fasting status . However, experimental studies are now required to test whether trait production (i.e., beak spot and auricular patch coloration and size) is condition-dependent. A positive association between indices of condition and social ornaments may be explained by social dominance and increased access to resources, rather than by a cost of ornament production itself (e.g., Gonzalez, Sorci, & De Lope, 1999).
In the king penguin, both the entire plumage and the yelloworange keratin beak plates are renewed each year during the molt, which occurs over a period of 3-4 weeks ; see Figure 1). This particular context of a complete renewal (i.e., production) of all ornaments over a short period of time provides an ideal opportunity to investigate the costs of production of ornaments used in sexual and social contexts, and thus to examine whether sexual and social ornaments differ in their costs of production. To do so, we experimentally subjected molting birds to increased chronic stress (elevated glucocorticoid levels) or to an immune challenge (lipopolysaccharide LPS injection). Both chronic stress and immune stress are energy costly processes that can divert resource investment from the production of showy ornaments (Faivre, Grégoire, Préault, Cézilly, & Sorci, 2003;Folstad & Karter, 1992 times lower than what is typically used in poultry studies (5 mg/kg; Cheng et al. 2004). Treated birds systematically showed signs of local inflammation (swelling) at the site of injection, confirming the treatment was effective. Control birds were injected with 1 mL physiological solution only (SHAM) and never showed signs of local swelling.

| Morphometric and ornamental measures
In both years, we obtained measurements of body mass and color ornaments at the beginning and end of the experiment. Both experimental and control birds were initially measured at the start of the molt before providing the implants, and were measured a second time when caught after returning from their postmolt foraging trip of ca. 15 days (9-25 days) in order to court and breed.
When initially caught, birds were transported to a nearby dry shelter (within 10 m of the colony), and their body mass was measured to the nearest 2 g using an electronic scale. Flipper length (indices of structural size) was also measured to the nearest 1 mm using a solid metal ruler. We then regressed body mass on flipper size (F 1,43 = 6.50, p = .013, R 2 = .07) and used the residuals as an index of body condition.
Colors reflected by the beak spot and the auricular patches were measured using a portable JAZ spectrophotometer (Ocean Optics Inc., Dunedin, FL, USA) containing a pulsed-xenon light with a spectral resolution of 0.3 nm across the spectral range of 320-700 nm, and was calibrated against a white standard (Ocean Optics Spectralon).
Measures were repeated 3 times on each ornament (on both sides of the bird) using a 200 μm fiber-optic probe with a 90° angle window.
Reflectance spectra of given ornaments were smoothed and averaged using an R script adapted from Montgomerie (2008). The obtained spectra (e.g., see Figure 2) were used to calculate mean brightness, hue, and chroma (see below) over the spectral range 320-700 nm, which corresponds to the full range of spectral sensitivity in birds (Cuthill, 2006). King penguin beak spots show a reflectance peak in UV violet (320-490 nm) and a plateau in the yellow-orange portion (491-700 nm) of the spectrum , and we calculated color variables separately over those two regions. In contrast, feathered auricular patches contain a pterin-based pigment (Thomas et al., 2013), only reflective above 450 nm. The spectral intensity, mean brightness (UV brightness and YO brightness ), was calculated by averaging reflectance over wavelengths: 320-490 nm and 491-700 nm for the beak, and 450-700 nm for the auricular patch (Montgomerie, 2006).
Hue is a measure of color appearance (e.g., "blue" and "yellow"). For the yellow-orange plateau portion of the spectrum, YO hue was calculated as the wavelength at which the reflectance was halfway between its maximum and minimum (Keddar, Andris, Bonadonna, & Dobson, 2013). For the UV violet color of the beakspot, UV hue was calculated as the wavelength of maximum reflectance between 320 and 490 nm.
Finally, chroma is a measure of color purity and was calculated within the region of interest (UV chroma and YO chroma ) as the difference between maximum and minimum reflectance over the mean reflectance for that particular region (formula S8; Hill & McGraw, 2006, p. 108

| Statistical analyses
All analyses were run in the statistical computing software R (v.

| Experiment 2: LPS immune challenge
At molt onset, both treated (LPS) and control birds were similar in terms of ornamental features (LMs; treatment: 0.89 < F 1,24 < 2.67, 0.120 < p < .360, Figure 5a SHAM higher CORT higher SHAM higher CORT higher size F I G U R E 4 Correlation of beak and auricular patch color variables, and auricular patch surface before (x-axis) and after the molt (y-axis), for the same king penguins (Aptenodytes patagonicus). Pearson correlation coefficients are given in the top right corner. CORT-and shamtreated birds are highlighted with red and black dots, respectively. A regression line is presented when the association is significant  and r = .44, t = 2.41, p = .024, N = 26, respectively); other relations were non-significant (0.123 < p < .698, N = 26) ( Figure 6). Initial body condition was related to auricular patch YO chroma (LM; F 1,24 = 4.67, p = .041). However, other parameters including the previous significant relation with auricular patch size (see Experiment 1) were not significantly related to initial body condition (LMs; 0.26 < F 1,24 < 4.67; 0.157 < p < .609). No significant links were found between body condition at birds' return from their postmolt foraging trip and ornamental features (LM; 0.12 < F 1,24 & F 1,23 < 1.37, 0.257 < p < .738).

| DISCUSSION
Ornaments can be used in sexual and/or social interactions (Hill, 2014;West-Eberhard, 1983) and, regardless of the selection process, are expected to honestly reflect individual health, vigor, or social status in competitive contexts (Tanaka, 1996;Zahavi, 1975). Ornament honesty can be enforced by production costs (condition-dependent signaling hypothesis; Hill, 2011), by the costs of bearing them in a social group (social mediation hypothesis; Rohwer, 1977), or by genetical constraints (i.e., linked to genes coding for immunocompetence; Whittingham, Freeman-Gallant, Taff, & Dunn, 2015). Here, we investigated ornament production costs in adult king penguins by manipulating physiological stress and immunity status during the molt. Our results showed that some ornamental features were strongly affected by our treatments (auricular patch YO chroma and beak spot UV chroma ), while we found no evidence of treatment effects on other ornamental features (auricular patch size and beak spot UV hue ). This demonstrates the coexistence of condition-dependent and condition-independent ornaments in adult king penguins.

| Condition-dependent ornaments
The condition-dependent signaling hypothesis points out that the honesty of ornaments can come from unavoidable physiological or developmental costs of production (thus, preventing dishonest signals; Hill, 2011). This hypothesis predicts the existence of a trade-off in energy or resource allocation between ornamentation and selfmaintenance. Thus, energetically challenging birds during ornament production (i.e., during the molt) by increasing the energy demand into self-maintenance is expected to highlight trade-off investments between those processes. Body condition of birds at the beginning of the molt was positively linked to the YO chroma of their auricular patches in both experiments, suggesting that color production of that ornament is under an energetic trade-off. Accordingly, ornaments that changed in response to CORT and LPS treatments showed a decrease in the purity of ornamental colors. Birds treated with CORT molted new auricular patches for which feather YO chroma was lower than controls. Similarly, individuals repeatedly injected with LPS molted beak spots and ear feather patches for which UV chroma and YO chroma , respectively, were both lower than in controls.
The so-called stress hormone CORT is under the regulation of the hypothalamic-pituitary-adrenal (i.e., HPA) axis, and its release modulates energy allocation, allowing individuals to overcome challenging environmental conditions (Angelier, Wingfield, Weimerskirch, & Chastel, 2010). At short time scales, CORT release induces a rapid mobilization of energy resources, preparing the organism to cope with challenging situations (Peckett, Wright, & Riddell, 2011).
Chronically elevated CORT leads to a reallocation of energy reserves into survival rather than other functions (Bókony et al., 2009;Spée et al., 2010), for example, by inhibiting vocalizations (Macdougall-Shackleton et al., 2009), coloration (Roulin et al., 2008), courtship (Moore & Miller, 1984), and breeding behavior ). In this study, penguins experimentally treated with a CORT implant may have invested less in the endogenous synthesis of pterins, resulting in a decreased allocation of this pigment to their ornaments reflected by lower YO chroma of auricular patch feathers.
Although changes in YO chroma have been directly linked to carotenoid pigment concentrations in feathers in other bird species (McGraw & Gregory, 2004;Saks et al., 2003), yellow-orange coloration in the F I G U R E 6 Correlation of beak, auricular patch color variables, and auricular patch size before (x-axis) and after the molt (y-axis), for the same king penguins (Aptenodytes patagonicus). Pearson correlation coefficients are given in the top right corner. LPS-and sham-treated birds are highlighted with red and black dots, respectively. A regression line is presented when the association is significant  (Thomas et al., 2013) is due to pterins (and not carotenoids). However, the costs associated with pterin endogenous synthesis and allocation remain unclear. Because pterins have been proposed to have antioxidant or immune functions (McGraw, 2005;Oettl, Greilberger, & Reibnegger, 2004;Weiss, Kennedy, Safran, & McGraw, 2011), those pigments could have been devoted to protection against CORT-induced oxidative stress (Costantini, Marasco, & Møller, 2011) at the expense of auricular patch coloration.
The immunocompetence handicap hypothesis (a subhypothesis of the condition-dependent signaling hypothesis) proposes that the production of ornaments comes at the expense of resistance to disease and parasites (Folstad & Karter, 1992). Several studies have shown positive links between immune efficiency and ornaments, notably ornamental coloration relying on exogenous pigment availability (Blount, Metcalfe, Birkhead, & Surai, 2003;, and reported a decrease in pigment-based coloration in immune-challenged individuals (Faivre, Grégoire et al., 2003). Thus, a trade-off in the allocation of pigments toward colorful ornaments or immune functions may explain the honesty of some ornamental features. In this study, LPS treatment stimulating the immune function of molting king penguins led to decreased purity in the yellow-orange color of auricular patch feathers (lower YO chroma ), which may result from the allocation of pterins to immune functions at the expense of auricular patch coloration. The LPS treatment also induced a decrease in purity of the UV coloration of the beak spot (UV chroma ). Conversely to YO chroma coloration of the auricular patch feathers, that is, due to pigments, beak spot UV reflectance in the king penguin has a structural basis resulting from crystal-like photonic microstructures in the horny layer of the beak (Dresp & Langley, 2006). Our results suggest that the production of that photonic structure may be costly, especially in response to an immune challenge.
A previous study where king penguins were treated with an antiparasitic solution outside the molting period (i.e., during breeding) showed a strong increase in beak spot UV hue and brightness and a weak decrease in UV chroma after parasite removal . Although those results support a link between beak spot UV coloration, immunity and parasitism, we would have expected an increase in UV chroma in response to parasite removal and/or a decrease in UV hue and brightness in response to LPS treatments (Griggio, Zanollo, & Hoi, 2010;Leclaire, Pauline, Chatelain, & Gasparini, 2014;Schull et al., 2016). Those results call for a deeper understanding of the physiological and structural mechanisms leading to changes in UV hue, brightness, and chroma, notably the importance of production and maturation of photonic structures during the molt versus maintenance and modulation of those structures afterwards (e.g., during breeding).

| Condition-independent ornaments
Previous studies have shown that environmental factors encountered during ornament production (parasites, food shortage, or other stressors) do not necessarily strongly impact their expression (Hill & Brawner, 1998). In other words, the production/expression of those ornaments may be viewed as condition-independent, raising questions about what factors enforce their honesty (Hill, 2013;Roulin, 2016). In contrast to auricular patch YO chroma, auricular patch size at the beginning of the molt was not consistently associated with birds' body condition (see differences between Experiments 1 and 2). One explanation could be that auricular patch size depends on the birds' condition at the time they are molted (i.e., birds' energy status during their production the previous year) but do not necessarily reflect the birds' condition 1 year later. However, the lack of association between ornamental features and condition at their return from the postmolt foraging trip in both experiments, suggests this is unlikely. Alternatively, a lack of consistent association suggests that auricular patch size may not be condition-dependent.
Indeed, we found no evidence that auricular patch size and beak spot UV hue were affected by CORT and LPS treatments. One hypothesis is that the honesty of condition-independent ornaments is enforced by social mediation (Rohwer, 1977;West-Eberhard, 1983). This hypothesis predicts that such ornaments should mirror social status and be constantly assessed during competitive interactions (Rohwer, 1977). Accordingly, auricular patch size in king penguins has been previously suggested to act as a social status badge, signaling individual competitiveness (Keddar, Couchoux et al., 2015;Viera et al., 2008). Individual penguins bearing larger auricular patches appear to be more aggressive and defend central breeding areas (Keddar, Couchoux et al., 2015;Viera et al., 2008) thought to be of higher quality (Bried & Jouventin, 2008) but see (Viblanc, Gineste, Stier, Robin, & Groscolas, 2014). The role of beak UV hue in social or sexual interactions is unknown. Experimental manipulation of auricular patch size (or beak UV hue) is now required to demonstrate whether bearing large auricular patches (or high UV wavelengths on the beak) entails socially mediated costs.
An alternative hypothesis for the evolution of conditionindependent ornaments in response to environmental challenges is that ornaments reflect the genetic quality of their bearer (i.e., genetical constraint hypothesis; Hill, 2014). Indeed auricular patch size and beak spot UV hue were not affected by CORT and LPS treatments.
Moreover, both traits showed similar expression before and after the molt. A similar strong relationship before and after the molt for beak spot UV hue was already observed in different individuals . Together, those findings indicate that the expression of auricular patch size and beak spot UV hue is maintained over time, which in turn suggests that their expression is largely determined by genetic or developmental factors encountered early in life that have lifelong consequences (e.g., silver spoon effects; Minias, Włodarczyk, Surmacki, & Iciek, 2015). However, the mechanisms linking individual genetic and epigenetic variation to trait production remain elusive. It has been recently proposed that ornament production could be constrained by the efficiency of vital cellular processes, with genes encoding for cellular metabolic pathways playing a key role in enforcing signal honesty (Hill, 2011(Hill, , 2014. Because the mitochondrion is the powerhouse of the cell, variation in mitochondrial and nuclear genes, epigenetic status, and mitonuclear interactions could account for interindividual genetic variation in ornament expression (Hill, 2014;. Work using larger sample sizes, individuals of both sexes, and investigating mitochondrial function during and outside the period of ornament production may provide insightful information on the honesty of UV signals in the king penguin, and more generally on the importance of mitochondrial efficiency as a mediator of signal honesty in the animal kingdom (Hill, 2011(Hill, , 2014.

| CONCLUSION
Taken together, our study provides evidence for the evolution of condition-dependent and condition-independent ornaments in the king penguin, and suggests that variation in the cost of ornament ex- This study also highlight that a unique ornament may evolve under non-mutually exclusive selective forces.

ACKNOWLEDGMENTS
We thank Emilie Lefol for her help in the field and two anonymous reviewers for their comments on the manuscript. This research was funded by the French Polar Institute (IPEV-Research Program 119) and the French National Center for Scientific Research (CNRS-