Independent sources of condition dependency and multiple pathways determine a composite trait: lessons from carotenoid-based plumage colouration

Authors

  • C. Romero-Diaz,

    Corresponding author
    • Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales (MNCN-CSIC), Madrid, Spain
    Search for more papers by this author
  • H. Richner,

    1. Institute of Ecology and Evolution, University of Bern, Bern, Switzerland
    Search for more papers by this author
  • F. Granado-Lorencio,

    1. Unidad de Vitaminas, Servicio de Bioquímica Clínica, Hospital Universitario Puerta de Hierro-Majadahonda, Madrid, Spain
    Search for more papers by this author
  • B. Tschirren,

    1. Institute of Evolutionary Biology and Environmental Studies, University of Zurich, Zurich, Switzerland
    Search for more papers by this author
  • P. S. Fitze

    1. Departamento de Biodiversidad y Biología Evolutiva, Museo Nacional de Ciencias Naturales (MNCN-CSIC), Madrid, Spain
    2. Fundación Araid, Zaragoza, Spain
    3. Instituto Pirenaico de Ecología (IPE-CSIC), Jaca, Spain
    4. Department of Ecology and Evolution (DEE), University of Lausanne, Lausanne, Switzerland
    Search for more papers by this author

Correspondence: Cristina Romero Díaz, Museo Nacional de Ciencias Naturales (MNCN-CSIC), c/ José Gutiérrez Abascal 2, 28006 Madrid, Spain. Tel.: +34 914111328; fax: +34 915645078; e-mail: cromero@mncn.csic.es

Abstract

Many colour ornaments are composite traits consisting of at least four components, which themselves may be more complex, determined by independent evolutionary pathways, and potentially being under different environmental control. To date, little evidence exists that several different components of colour elaboration are condition dependent and no direct evidence exists that different ornamental components are affected by different sources of variation. For example, in carotenoid-based plumage colouration, one of the best-known condition-dependent ornaments, colour elaboration stems from both condition-dependent pigment concentration and structural components. Some environmental flexibility of these components has been suggested, but specifically which and how they are affected remains unknown. Here, we tested whether multiple colour components may be condition dependent, by using a comprehensive 3 × 2 experimental design, in which we carotenoid supplemented and immune challenged great tit nestlings (Parus major) and quantified effects on different components of colouration. Plumage colouration was affected by an interaction between carotenoid availability and immune challenge. Path analyses showed that carotenoid supplementation increased plumage saturation via feather carotenoid concentration and via mechanisms unrelated to carotenoid deposition, while immune challenge affected feather length, but not carotenoid concentration. Thus, independent condition-dependent pathways, affected by different sources of variation, determine colour elaboration. This provides opportunities for the evolution of multiple signals within components of ornamental traits. This finding indicates that the selective forces shaping the evolution of different components of a composite trait and the trait's signal content may be more complex than believed so far, and that holistic approaches are required for drawing comprehensive evolutionary conclusions.

Introduction

The idea that colour ornaments are composite traits determined by different evolutionary pathways has become increasingly relevant (Badyaev et al., 2001; Badyaev, 2004; Grether et al., 2004; Jacot et al., 2010; Svensson & Wong, 2011). In particular, it has been proposed that carotenoid-based colouration, one of the best-known condition-dependent ornaments, is determined by at least four distinct components: pigment elaboration, patch area, pigment symmetry and patch area symmetry (Badyaev et al., 2001; Badyaev, 2004). Although research has mainly focused on these classic four components, few studies have investigated whether those components may be more complex, and whether independent condition-dependent pathways may determine their expression. For example, pigment elaboration, originally defined as ‘type and quantity of carotenoid pigments deposited in growing feathers’ and measured as colour hue (i.e. pigment hue, Badyaev et al., 2001), includes independent effects of pigments and feather background structure (Shawkey & Hill, 2005; Jacot et al., 2010). Here, we therefore use a more general terminology, that corresponds to this measure of pigment elaboration (i.e. colour hue), namely colour elaboration, which does not make any assumptions about how colouration is determined (note that colour elaboration strictly refers to the colour per se, excluding the extent or symmetry of the colouration, and is independent of the quantification method). Likewise, patch area (‘the area of plumage (i.e. number of feathers) with carotenoid pigmentation’, Badyaev et al., 2001) may depend on individual feather characteristics (e.g. feather width, feather length, feather shape), number of feathers and feather arrangement (Quesada & Senar, 2006). It has been observed that different aspects of fitness could be associated with individual components of colour elaboration (Badyaev et al., 2001). However, evidence that these components may be determined by independent condition-dependent pathways is indirect (Jacot et al., 2010; Matrkova & Remes, 2012). Thus, it remains unknown whether different components may be under independent selection, which could potentially explain why carotenoid-based colouration preserves an important amount of phenotypic variability (Tolle & Wagner, 2011).

Carotenoid pigments are the main sources of the red, orange and yellow colourations present in many taxa, including fish, amphibians, reptiles, mammals, birds, crustaceans and insects. The degree of carotenoid deposition is an important determinant of carotenoid-based colour elaboration (Hill, 1992, 2000; Hill et al., 1994, 2002; Saks et al., 2003; Shawkey & Hill, 2005). However, robust, experimental evidence for the condition dependency of structural aspects of ornamental feathers (i.e. feather characteristics) is to our knowledge absent. Moreover, only a handful of studies have simultaneously investigated the contributions of alternative colour components (see Svensson & Wong, 2011), and very few evidence exists that different components reflect different, independent sources of condition dependency. Therefore, it remains unknown whether and how different components and their combination affect colour elaboration. This, in turn, is the key to understanding how selection drives the evolution of colour displays.

In birds, carotenoid-based colouration has been shown to honestly signal health/immune status (Blount et al., 2003; Faivre et al., 2003; McGraw & Ardia, 2003), foraging ability (Hill, 1992; Hill et al., 2002), parasite load (Møller et al., 2000) and nutritional condition (Hill & Montgomerie, 1994; Hill, 2000) and it may provide advantages in intra- and intersexual selection (i.e. mating success, Hill, 1999) and interspecific interactions (i.e. protection from predation, Slagsvold & Lifjeld, 1985; Delhey et al., 2010). Carotenoids may also be used as antioxidants or immunostimulants, potentially leading to a trade-off between ornamental display and health (von Schantz et al., 1999; Lozano, 2001; Blount et al., 2003; McGraw & Ardia, 2003; Alonso-Alvarez et al., 2004; but see Navara & Hill, 2003; Fitze et al., 2007; Isaksson et al., 2007). According to theory (Møller et al., 2000; Blount et al., 2003; Faivre et al., 2003; McGraw, 2006) carotenoid availability is limited in nature (Slagsvold & Lifjeld, 1985; Hill, 1991, 1992; Andersson, 1994). As animals cannot synthesize them de novo, the production of carotenoid-based ornamentation is costly (Tschirren et al., 2003b) and so honesty assured (Zahavi, 1975).

Most research on the signalling properties of carotenoid-based plumage colouration has focussed on how condition-dependent colour variation is caused by differences in carotenoid concentration. In contrast, our experiment investigated whether different condition-dependent components exist (among components of carotenoid-based colour elaboration), and whether they may reflect alternative and independent evolutionary pathways. Using path analyses, we assessed the relative importance of different condition-dependent pathways and the relationships among different components. Unlike previous studies, we considered a wide range of potential colour determinants, including structural aspects of feather design and carotenoid concentration, and investigated how carotenoid supplementation and immune challenge affected these components, the different measures of feather colouration, and thereby colour elaboration.

The effect of pigment concentration on carotenoid-based colouration has been broadly studied (see above), but the role of structural contributions to plumage colour elaboration is incomplete. Among structural features, it has been demonstrated that feather overlap modifies colouration (Quesada & Senar, 2006). Evidence that colour elaboration is affected by condition-dependent variation in feather overlap, under natural conditions, is lacking. Similarly, two studies have shown that structural aspects may be condition-dependent, but it remains unclear exactly which components are those and which their determinants are (Jacot et al., 2010; Matrkova & Remes, 2012). Moreover, the activation of the immune system has been shown to alter colour expression, through a mechanism different from the proposed trade-off in carotenoid allocation between immune function and colouration (Fitze et al., 2007). The activation of the immune system generally negatively affects body, wing, and wing and tail feather growth rate (Klasing et al., 1987; Fair et al., 1999; Saino et al., 2002; Pap et al., 2011), but it remains unknown whether it affects ornamental plumage feathers and whether such differences may affect colour elaboration.

Here, we experimentally investigated which colour components are condition dependent and whether they are determined by independent pathways, using a 3 × 2 factorial design where we carotenoid supplemented and immune challenged great tit nestlings during early development. By using the proposed trade-off in carotenoid allocation between colouration and immune function (von Schantz et al., 1999), we tested which colour components induced changes in nestling plumage colouration. To pin down where the colour changes originated from, we used standardized photographic measurements taken on alive birds, spectrophotometric measurements taken on individual feathers, and measured several feather characteristics. We measured feather carotenoid concentration using high performance liquid chromatography (HPLC) and assessed structural feather components, namely feather length, feather opacity, feather development and barb density. Thereafter, we used multivariate path analyses to evaluate the relative contributions of these components to experimentally induced changes in carotenoid-based colouration and examined whether different environmental sources and condition-dependent pathways independently account for colour variation.

If carotenoid-based colouration (i.e. colour elaboration) of great tits is a complex composite trait consisting of different, independent condition-dependent components, we predicted that changes in plumage colouration would be the result of independent contributions of carotenoid concentration and structural features of feathers, affected by different environmental sources of variation.

Materials and methods

Species description

The great tit is a widespread small hole-nesting passerine that breeds in woodlands and gardens across all Europe. Males and females show yellow ventral feathers, a black breast stripe, black head and neck, prominent white cheeks and olive-green upperparts. Yellow breast colouration develops early in life (Fitze et al., 2003b) and depends on a combination of carotenoid pigments (lutein and zeaxanthin), mainly obtained through caterpillar ingestion (Slagsvold & Lifjeld, 1985; Partalli et al., 1987), and structural components (Jacot et al., 2010). Feathers develop in a sheath whose tip brakes open when the uppermost barbs are keratinized and their morphological development completed (Stettenheim, 2000). Lutein and zeaxanthin pigments are incorporated into the feather without metabolic transformation (Lucas & Stettenheim, 1972; Partalli et al., 1987). As in many other avian species (McGraw, 2006), only the distal end of the feather is coloured.

Experimental design

The experiment was carried out in 2001 in a great tit population breeding in nest-boxes in the Forst (46º54′N, 7º17′E/46º57′N, 7º21′E), a mixed deciduous forest near Bern, Switzerland. The experimental design and further methodological details are described elsewhere (Fitze et al., 2007). To assess whether one or multiple pathways affect carotenoid-based plumage colouration, we carried out an intranest experiment on nestling great tits testing for a trade-off in carotenoid allocation between colouration and immune function. Nestlings were randomly assigned to two crossed treatments, namely carotenoid supplementation and immune challenge, using a two-factorial design with three and two factor levels respectively.

Carotenoid supplementation

The carotenoid treatment comprised three treatment groups, consisting each of two randomly chosen nestlings per nest. A first group, the βLZ group (β-carotene, lutein, zeaxanthin), was fed 2.6 mg (± 0.25 mg) β-carotene beadlets (containing 8% β-carotene) and 17 mg (± 0.25 mg) lutein/zeaxanthin beadlets per feeding (containing 5.58% lutein and 0.44% zeaxanthin; Hoffmann–La Roche, Basel, Switzerland), which represents the carotenoids occurring in the natural diet of great tits (Partalli et al., 1987). A second group, the LZ (lutein, zeaxanthin) group, was fed 2.6 mg carotenoid free beadlets and 17 mg (± 0.25 mg) lutein/zeaxanthin beadlets per feeding, which represents the carotenoids present in great tit feathers (Partalli et al., 1987). Finally, the control group was fed with 19.6 mg (± 0.25 mg) beadlets per feeding containing no carotenoids. Nestlings were fed every second day, starting 4 days post-hatching, and thus several days before the first breast feathers appeared (P.S. Fitze, personal observation). Feeding treatment ended 14 days post-hatching. The supplemented amount of carotenoids and the carotenoid ratios were within the naturally ingested range (Tschirren et al., 2003a).

Immunization treatment

Four days after hatching, one randomly chosen nestling of each carotenoid supplementation group was immune challenged with an intramuscular injection of 50 mL human diphtheria-tetanus (DT) vaccine (Kinder, Merieux) and 50 mL 5% rabbit red blood cells in phosphate-buffered saline (PBS), hereafter referred to as immunized group (I). The other nestling was injected 100 mL PBS and served as a control for the immunization, hereafter referred to as control-injected group (CI).

Feather measurements

Color measurements

In the field, we photographed nestling breast plumage colouration fifteen days post-hatching under standardized conditions using a digital camera (Fitze & Richner, 2002). Photos allowed quantifying the colouration of the plumage (i.e. colour elaboration) integrating colouration originating from pigment concentration and structural components (i.e. feather shape, layers and arrangement). Photos were taken with standardized light exposure, photographic angle and object-objective distance. Mean RGB (red, green, blue) values were obtained of 10 square measurement areas, each consisting of 400 pixels, using Adobe Photoshop. Thereafter, we calculated HSB (hue, saturation and brightness) values following the algorithm described by Foley & van Dam (1984). For more detailed information on the applied methodology see Fitze & Richner (2002).

After taking a photograph, we collected 20 yellow breast feathers from the upper left breast of each nestling. Feathers were kept in hermetic plastic bags and stored in the dark until measurement. In the laboratory, we measured feather reflectance using an Ocean Optics USB4000 spectrophotometer (range 200–850 nm; Dunedin, FL, USA) with a light source (DT-MINI-2-GS) that provided light in the UV and visible range. We used a reflection probe (QR400-7-UV/VIS) fixed on a reflection probe holder (RPH-1) that excluded ambient light and allowed to measure reflectance at an angle of 90°. Reflectance was measured with respect to a white (WS-1, Ocean Optics) and a black standard (black photographic cloth with no light reflectance across all wavelengths). We placed five feathers on top of each other and used as a background the same black photographic cloth. For each nestling, we took five measurements of the feather tip on the dorsal side of the feather in an area of approximately 1 mm2 and for each measurement we alternated the order of the feathers in the pile (Jacot et al., 2010).

We computed the average reflectance of the five measurements and thereafter derived five indices describing the feather's colouration following Jacot et al. (2010). We calculated (1) ‘background reflectance’ corresponding to the absolute reflectance between 575 and 700 nm (R575–700 nm) and being a carotenoid-independent proxy of the feather's white background structure (Jacot et al., 2010), (2) ‘absolute carotenoid chroma’, a background structure-independent measure of carotenoid concentration (R400–515 nm / R575–700 nm) (Jacot et al., 2010), (3) UV-reflectance (R300–400 nm) (Bennett & Cuthill, 1994), the total amount of light reflected in the UV, (4) ‘RUVpeak, the wavelength of peak reflectance in the UV (Bleiweiss, 2005) and (5) ‘UV chroma’, corresponding to the proportion of light reflected in the UV while controlling for differences in background structure (i.e. R300–400 nm/R575–700 nm (Jacot et al., 2010) (Fig. 1). Photospectrometric measurements quantified feather colouration and thus colour variation arising from pigments and structural feather components, and they were independent of feather density.

Figure 1.

Reflectance spectra of breast feathers of great tit nestlings. Shown are measurement ranges of five indices describing feather colouration. Measures of total reflectance include (a) UV-reflectance measured between 300 and 400 nm, and (b) ‘background reflectance’ measured between 575 and 700 nm. Reflectance, independent of variance in the background structure, include ‘UV chroma’ (a/b) and ‘absolute carotenoid chroma’ (c/b), whereas λ at peak reflectance in the UV (d) corresponds to RUVpeak (e).

Structural measurements

To understand whether and how feather design affects feather colouration, we measured four structural components, namely feather length, barb density, feather opacity and developmental stage of all feathers used for the spectrophotometric analyses.

Feather length

Total feather length was measured manually with a ruler (± 0.5 mm) and corresponds to the straightened shaft length. We also measured the length of the different feather parts along the feather shaft (Fig. S1), including the length of the yellow, white and black coloured parts and of the calamus.

Barb density

For each of the measured feathers, we determined the barb density by counting the number of barbs in the uppermost 5 mm of the yellow tip of the feather. This area includes the spot where the spectrophotometric measurements were taken.

Feather opacity

All feathers used for the spectrophotometric measurements were individually photographed under standardized conditions using the same photographic set-up as for nestlings (Fitze & Richner, 2002). Briefly, feathers were put on black photographic cloth within a small box and pressed against a UV-photographic filter lens. This box was placed in a standard position inside a larger opaque camera box and photos with standardized light exposure and size were taken. Photos were imported into ImageJ (Rasband, 2005) and two different measures of feather opacity were obtained, 1) one-barb surface coverage and 2) the opacity of a feather area (Fig. S2). One-barb surface coverage measures the contribution of a single barb to feather opacity, while opacity of the feather area, hereafter referred to as ‘feather opacity’, corresponds to the surface proportion covered by the barbs and barbules of the measured feather area. Prior to the analysis, all photos were transformed into 8-bit black and white photos. To measure area opacity, we selected an area of 30 × 30 pixels within the yellow distal feather part. For all feathers, the centre of the square coincided with the point where the uppermost barb branched off from the rachis and the sides of the sampled square were aligned parallel to the shaft. We then used a grey threshold (for all feathers the same threshold) to determine the percentage of the 900 pixels covered by the feather barbs and barbules. For determining one-barb surface coverage, we selected an area of 30 × 30 pixels in the middle of the barb (between the shaft and the barb tip) where no other barbs overlapped. The square was parallel aligned with the ramus and it completely fell within the barb's contour line. Surface coverage was measured using the method applied for feather opacity.

Developmental stage

Since feather development may affect opacity and feather length, and thereby feather colouration, we assigned ‘development’ scores to each feather used for the spectrometric measurements. Feather development was measured using a discrete scale consisting of 5 levels ranging from 0 to 4 (i.e. 0 = undeveloped feather, 4 = completely developed). Developmental scores were independently attributed to the left and right side of the feather, according to the criteria shown in Fig. S3. Average developmental score was used for further analyses.

Carotenoid concentration

Feather carotenoid concentration was analysed by HPLC using a protocol adapted from Olmedilla et al. (1997). Briefly, 0.5–1.0 mg of feather tips was placed in 1.0 mL ethanol. The internal standard (retinyl acetate) was added and the mix was flushed with nitrogen and kept from light at 4 °C for 25 minutes. Then, the solution was placed in an ultrasound bath, with intermittent vortex, for 5 minutes. Double extraction was performed by adding 1 mL of distilled water and 2 mL of methylene chloride/hexane (1 : 5). Both organic phases were pooled and evaporated to dryness. The carotenoid residue was dissolved in tetrahydrofuran/ethanol (1/1) and thereafter injected into the HPLC.

The chromatographic system consisted of a Spheri-5-ODS column (Applied Biosystems, San-Jose, CA) with gradient elution of acetonitrile/methanol (85/15) for 5 minutes to acetonitrile/methylene chloride/methanol (70/20/10) for 20 minutes. Ammonium acetate (0.025 m) was added to the methanol. Detection was carried out by a photodiode array (Model 2996; Waters Associates, Milford, MA) set at 450 nm. This method allows to simultaneously detect trans-lutein, zeaxanthin, 13/15-cis-lutein, α-carotene, all-trans-β-carotene, 9-cis-β-carotene, 13/15-cis-β-carotene and several other carotenoids (Olmedilla et al., 1997). Identification of compounds was carried out by comparing retention times with those of authentic standards and on-line UV-visible spectra. Only lutein (average ± SE: 54.56 ± 34.37 μg g−1 feather) and zeaxanthin (12.76 ± 6.98 μg g−1 feather) were detected and feather carotenoid concentration (quantity of lutein and zeaxanthin/feather mass) was used in the analyses.

Statistical analyses

Treatment effects on the different components affecting plumage colouration were analysed with mixed-model anovas using JMP® statistical package (SAS Institute Inc., Cary, NC, USA) and R (R Development Core Team, 2005). Carotenoid supplementation, immunization treatment, and their interaction were modelled as fixed factors and nest was included as random factor. Residuals were tested for normality and homoscedasticity. If necessary, variables were transformed using logarithmic or arcsine square root transformations. Differences between treatment groups were analysed using post hoc LSMeans contrasts.

Using path analyses, we investigated the relative contribution of three exogenous experimental parameters (carotenoid supplementation, immunization and their interaction) on plumage colouration. All measured components were standardized and included in the path diagram. They were classified into five hierarchical levels, arranged from the left (lowest level) to the right (highest level, Fig. 2), where each hierarchical level was determined by hierarchically lower levels and thus by those presented on their left (Quinn & Keough, 2002). For total feather length and the length of the yellow, white or black coloured feather parts the hierarchy was not clear and thus we also modelled the backward effect. Similarly, for components of the same hierarchical levels, it was not clear whether and in which direction they affected each other and thus we allowed for effects in both directions. The resulting diagram (Fig. 2) shows all effects supported in ≥ 75% of all path models, including intermediate models resulting from backward elimination.

Figure 2.

Path diagram showing the relative importance of different additive and interactive pathways potentially explaining variation in carotenoid-based plumage colouration. Shown are statistically significant treatment effects (outermost left) acting on three hierarchical levels: two levels of colour components (left and middle) and one level of feather colouration (right), and their effects on plumage colouration (outermost right). By design, causal relationships exist between components of two different levels (unidirectional left to right arrows) and relationships between components of the same level are considered to be correlations (bidirectional arrows). Arrow width is proportional to the standardized partial correlation coefficients (β). For correlated variables, β can differ, depending on which one is hierarchically higher. Continuous lines depict positive and dashed lines negative effects. All effects that were statistically supported in ≥ 75% of the path models are shown.

The path diagram was based on ten randomly chosen nests (= 58 individuals). This was because HPLC analyses and structural feather measurements were based on this subset. Analyses on plumage colouration and feather colouration were conducted using both the subset and the full sample size of 295 individuals, from 54 nests. For comparisons between individuals belonging (1) or not (0) to the subset, we modelled subset as a factor. There were no significant differences in body size and body condition between subsets and no significant interactions between the applied treatments and subsets (all > 0.5).

Results

Interaction between carotenoid supplementation and immunization treatment

There was a significant interaction between carotenoid supplementation and immunization on plumage saturation, plumage brightness (Table 1, Fig. 3) and one-barb surface coverage (F2,43 = 3.75, = 0.032, 7.6% of variance explained). There was also a significant interaction in plumage saturation (F2,216 = 3.85 = 0.022) in the full data set. The effect of carotenoid supplementation depended on the immunization treatment. Immunized individuals of the βLZ group showed significantly reduced plumage saturation (F1,43 = 28.67, < 0.001) and plumage brightness (F1,43 = 5.98, = 0.019), but no differences existed between immunization groups in the LZ and C groups (> 0.1). There was a significant negative effect of immunization in the C group on one-barb surface coverage (F1,43 = 6.99, = 0.011), but no significant differences between I and CI nestlings existed in the βLZ and LZ group (> 0.1). There were no significant interactions in any of the other colour parameters, in feather carotenoid concentration, and in structural components (all  0.1).

Table 1. Effects of the mixed-model anova on plumage colouration (A) and feather colouration (B) of the subset sample (= 58). The test statistics are given. When significance was found, only results of the backward eliminated final model are shown and the percentage of variance explained (%) is given. The initial full model included two fixed factors: carotenoid supplementation and immunization, and their interaction
Colour parameterTreatment F d.f. P %
A. Plumage colouration
HueCarotenoid supplementation7.672,460.00117.2
SaturationCarotenoid supplementation28.702,43<0.00136.0
Immunization0.011,430.930 
Interaction4.54 2,430.0165.7
BrightnessCarotenoid supplementation2.212,430.1225.9
Immunization0.361,430.549 
Interaction4.172,430.02210.9
B. Feather colouration
Absolute Carotenoid chromaCarotenoid supplementation12.652,46<0.00118.7
UV-reflectanceCarotenoid supplementation 0.692,430.351 
Immunization0.321,430.574 
Interaction1.071.430.351 
UV chromaCarotenoid supplementation3.642,460.0341.6
RUVpeakCarotenoid supplementation0.082,430.919 
Immunization0.211,430.649 
Interaction0.412,430.668 
Background reflectanceCarotenoid supplementation0.122,430.885 
Immunization0.771,430.384 
Interaction1.152,430.327 
Figure 3.

Effects of the interaction between carotenoid supplementation (βLZ: β-carotene/lutein/zeaxantin; LZ: lutein/zeaxantin; C: control group) and immunization (I: Immunized; CI: control-injected group) on plumage hue (a) saturation (b) and brightness (c). Averages ± 1 SE are shown. Significant least-square mean contrasts are depicted by asterisks (*< 0.05; **< 0.01).

Carotenoid supplementation

Effects on plumage colouration

There was a significant effect of carotenoid supplementation on plumage hue and saturation, but not plumage brightness (Table 1, Fig. 3). Plumage hue was lower in the LZ compared to the βLZ (LSMeans contrast: F1,46 = 4.96; = 0.031) and the C group (LSMeans contrast: F1,46 = 15.21; < 0.001) and tended to be lower in the βLZ compared to the C group (LSMeans contrast: F1,46 = 2.99; = 0.080). Thus, LZ and potentially also βLZ nestlings produced plumages with more orange tones. Plumage saturation of the carotenoid supplemented groups was significantly higher than in the C group in both immunization groups (F1,43 ≥ 28.67; < 0.001) and it was significantly higher in the LZ group compared to the βLZ group, in the CI group (F1,43 = 4.42; = 0.041; also see Table 1). Similar results were found when using the entire data set (effects of carotenoid supplementation on hue: F2,216 = 38.89; < 0.001; saturation: F2,216 = 102.23; < 0.001; brightness: F2,216 = 2.09; = 0.126; Table S1).

Effects on feather colouration

Carotenoid supplementation significantly affected ‘absolute carotenoid chroma’ in both data sets (Table 1, F2,215 = 18.44; < 0.001). Both carotenoid supplementation groups (βLZ and LZ) showed reduced ‘absolute carotenoid chroma’ compared to the C group (LSMeans contrasts: all F1,46 ≥ 22.64; all < 0.001), indicating increased light absorption between 400 and 515 nm and thus more incorporated carotenoids (Jacot et al., 2010). Carotenoid supplementation did not affect ‘background reflectance’ (Table 1, F2,215 = 0.96; = 0.382). In the subset, carotenoid supplementation significantly affected ‘UVchroma’, but not UV-reflectance and ‘RUVpeak’ (Table 1). However, when using the full data set, carotenoid supplementation significantly affected all three variables (all F2,215 ≥ 5.38;  0.005, ≥ 2.4% variance explained, Table S1), as predicted by a previous study (Jacot et al., 2010). This indicates that detecting carotenoid effects on UV properties require large sample sizes because carotenoid reflectivity is relatively small in the UV wavelength. Group C showed significantly more UV-reflectance, higher ‘UVchroma’, and higher ‘RUVpeak’ than the LZ and the βLZ group (all LSMeans contrasts: < 0.05). There were no significant differences between the LZ and the βLZ group (all LSMeans contrasts:  0.1).

Effects on feather carotenoid concentration

Carotenoid supplementation significantly affected feather carotenoid concentration (F2,43 = 7.79; = 0.001, 16.5% of variance explained). Significantly higher concentration was observed in the carotenoid supplementation groups (LSMeans contrasts: βLZ vs. C: F1,43 = 13.31,  0.001, estimate βLZ: 18.945 μg g−1 ± 7.466 SE; LZ vs. C: F1,43 = 8.65, = 0.005, estimate LZ: 10.478 μg g−1 ± 7.584 SE) and there were no significant differences between the βLZ and LZ group (LSMeans contrast: F1,43 = 0.44; = 0.512).

Effects on feather design

None of the structural feather components was significantly influenced by the carotenoid treatment (all  0.1).

Immunization treatment

Immune challenged nestlings had significantly shorter breast feathers than control nestlings (6.8% of variance explained, Fig. 4) and tended to show reduced feather opacity (F1,43 = 3.65, = 0.061, estimate −1.030 ± 0.554). There were no significant effects of immunization on any of the other components (feather and plumage colouration in the subset and the complete data set, carotenoid concentration and other components of feather design, all  0.1).

Figure 4.

Effects of experimental immunization on total feather length. Immunized nestlings showed significantly shorter feathers.

Path analysis

Path analyses revealed that carotenoid supplementation affected plumage colouration through carotenoid concentration but also independently of it (Fig. 2). Increased carotenoid supplementation positively affected plumage saturation and feather carotenoid concentration, and negatively affected ‘absolute carotenoid chroma’, the latter being due to reduced light reflection in the 400–515 nm wavelength. In addition, it affected plumage saturation through its effect on carotenoid concentration, which negatively affected ‘absolute carotenoid chroma’. In line with the results of Jacot et al. (2010), the effect of carotenoid supplementation on ‘UV chroma’ observed in the mixed anovas (Table 1) was via ‘absolute carotenoid chroma’. Immunization negatively affected total feather length and feather opacity. There was a significant positive correlation between total feather length, and the length of the differently coloured feather parts, which in turn affected UV-reflectance and plumage hue, in the case of yellow length and both yellow and white lengths respectively. Reduced feather UV-reflectance affected ‘UV chroma’ and thereby plumage saturation. The interaction between carotenoid supplementation and immunization affected plumage brightness. Immune challenged individuals of the βLZ group (F1,43 = 11.93, = 0.001, estimate −0.014 ± 0.005) showed reduced plumage brightness, but not individuals of the other groups. There was also an interactive effect on feather ‘UV chroma’, which was reduced in the immune challenged βLZ group (F1,30 = 7.55, = 0.010, estimate −0.119 ± 0.042) but not in the other groups, and on ‘absolute carotenoid chroma’, which was higher in the immune challenged βLZ group than in all other groups (F1,30 = 5.61, = 0.024, estimate 0.137 ± 0.056). ‘UV chroma’ positively affected saturation and ‘absolute carotenoid chroma’, which affected plumage brightness and saturation.

Discussion

It has been proposed that carotenoid-based ornaments are determined by at least four distinct components: pigment elaboration, patch area, pigment symmetry and patch area symmetry (Badyaev et al., 2001; Badyaev, 2004). Recent studies suggest that these components may themselves be more complex than originally described (Jacot et al., 2010; Matrkova & Remes, 2012) but condition dependency of carotenoid-based ornaments have mainly focussed on how colouration reflects variation in one component (e.g. in carotenoid concentration; Saks et al., 2003; McGraw & Gregory, 2004; Senar et al., 2008). To date, there exists few and no direct evidence that different condition-dependent components may be affected by different, independent evolutionary pathways.

We carried out an experiment using great tits, testing whether different components of colour elaboration are determined by multiple and/or independent pathways of condition dependency and whether they mirror different sources of condition dependency. We analysed treatment effects on the different components affecting colouration and determined their relative contribution to intraspecific variance in colour elaboration using path analyses. Immune challenge reduced plumage saturation in the βLZ group (Fig. 3), which is in line with the proposed trade-off in carotenoid allocation, and suggests that nestlings with an activated immune system (those of the βLZ group, see Fitze et al., 2007) used less carotenoids for colouration. However, only pigment availability (carotenoid supplementation), but not immune challenge or their interaction, affected feather carotenoid concentration and ‘absolute carotenoid chroma’ (Table 1). Therefore, the interactive effect on colour elaboration (saturation) was carotenoid concentration independent. This confirms that the proposed trade-off between colouration and immune function for rare carotenoids does not account for reduced plumage saturation in nestlings with an activated immune system (see Fitze et al., 2007). Reduced plumage saturation could however be explained by shared pathways between trait production and vital cellular processes that determine condition (Hill, 2011). This shows that plumage colouration was the consequence of complex feather carotenoid concentration dependent and carotenoid-independent effects and thus that alternative pathways exist. Carotenoid supplementation led to higher feather carotenoid concentration, which decreased ‘absolute carotenoid chroma’ of the feathers, and thereby increased plumage saturation (Fig. 2). Moreover, independent of feather carotenoid concentration, carotenoid supplementation decreased ‘absolute carotenoid chroma’ and increased plumage saturation (Fig. 2; see arrows directly connecting treatment with measures of plumage colouration). This suggests that the effects of carotenoid supplementation (Table 1) are the result of alternative pathways. The interaction negatively affected plumage brightness (i.e. the βLZ, I group showed reduced brightness; Fig. 2) independently of the measured colour components, reduced ‘UV chroma’, which affected saturation, and positively affected ‘absolute carotenoid chroma’, which in turn increased plumage brightness and decreased plumage saturation. Therefore, the interaction had, at the same time, negative and positive effects on colour intensity (brightness), suggesting that its overall effect (Table 1) may depend on the strength of each pathway and thus it may not always lead to reduced colour intensity.

Given that the interaction affected ‘absolute carotenoid chroma’ and ‘UV chroma’, its effects on feather colouration were the consequence of chromatic colour changes, suggesting that the chromatic part of feather colouration is not exclusively carotenoid determined. This argument is supported by the fact that the immunized βLZ group exhibited higher ‘absolute carotenoid chroma’ than the control-injected βLZ group (Fig. 3). If solely carotenoid incorporation modified the chromatic part of the colouration, differences between the I and CI groups of the βLZ treatment would be the result of carotenoids and thus we would have also expected higher ‘UV chroma’ in the I compared with the CI group of the βLZ (Jacot et al., 2010; Fig. 1; see methods). In contrast, a negative interactive effect was observed on ‘UV chroma’ (Fig. 2). The fact that carotenoid supplementation and its interaction with immunization had carotenoid concentration-independent effects on plumage colouration, suggests that these experimental effects may have also been the result of other, here undetermined components affecting colouration, such as subjacent variation in skin colouration (Jourdie et al., 2004). Immunization reduced total feather length, which was, in turn, positively correlated with the length of the white and yellow feather parts, and thereby affected hue and saturation (Fig. 2). Differences in feather length could lead to different number of overlapping breast feathers, which has been shown to affect chromatic and achromatic colouration (Quesada & Senar, 2006). This indicates that immunization altered plumage colouration through an independent pathway, different from the pathways by which carotenoid supplementation or the interaction affected plumage colouration. Additional support for this alternative pathway stems from the lack of effect of carotenoid supplementation (and the interaction) on feather length and any other component of feather design.

The effect of the immune challenge on feather length (Fig. 4) is congruent with the effects observed on wing and tail length (Brommer et al., 2011; Maenniste & Horak, 2011; Pap et al., 2011) and showed that structural features of colour elaboration are under environmental control and thus that they are also condition-dependent traits, which contribute to environmentally induced carotenoid-based plumage colour variation (Fitze et al., 2003a). Interestingly, although immunization affected feather length, no effect of immunization on overall plumage colouration was found in the mixed-model anovas that do not take into account the hierarchy of the colour determinants shown in Fig. 2. This suggests that immunization effects on plumage colouration caused by reduced feather length may have been cancelled out due to opposing effects of other components of plumage colouration and thus that complex interactive effects may exist, that are not necessarily consistent across environments (Sillanpää et al., 2010). None of the treatments affected ‘background reflectance’ even though it has been shown to be partly determined by environment-related factors (Jacot et al., 2010; Matrkova & Remes, 2012). Thus, the ultimate determinants of feather background structure remain unknown.

The results stress the important, but not exclusive, contribution of carotenoids to condition-dependent colour elaboration. Although carotenoid supplementation and immunization affected plumage colouration with similar strength (i.e. had similar β's -arrow width-, Fig. 2), the former affected plumage colouration through several mechanisms, while immunization affected plumage colouration through feather length. Thus, condition-dependent colour elaboration is clearly not the sole result of differential carotenoid incorporation into the feathers, suggesting that environmental determination of this trait (e.g. Slagsvold & Lifjeld, 1985; Eeva et al., 1998; Hõrak et al., 2000; Møller et al., 2000; Fitze et al., 2003b; Tschirren et al., 2003a, b) is the result of simultaneous and alternative (pigmentary or not) pathways and complex interactions.

This shows that, at least in carotenoid-based colouration, colour elaboration is a more complex composite trait, affected by several different condition-dependent components and their associated alternative pathways, with independent sources of variation. This result also suggests that colour elaboration of other types of colouration, may be the result of independent pathways under differential control. For example, in melanin-based colouration, patch area may be the result of differential melanization and/or feather length. Our results may also have important implications for fish, reptiles and amphibians, where the basic chromatophore unit is composed of three different layers consisting of different types of pigments, which may bear different information and may have evolved independently (Grether et al., 2004). Evidence stems, for example, from the common lizard (Lacerta vivipara), whose carotenoid-based ventral colouration shows condition-dependent chromatic colour changes which are independent of carotenoid intake, suggesting that their carotenoid-based colouration may as well be a composite trait determined by different pathways (Fitze et al., 2009; San-Jose et al., 2012, in press). Carotenoid-based colourations observed in nature are ubiquitous and include, besides feather colouration, beak and leg colouration of birds, hair and skin colouration in other animals, and petal, leaf, stem and fruit colouration in plants. Thus, in both the animal kingdom and in plants, condition-dependent ornaments, whose expression is influenced by several factors, may indeed be the result of multiple condition-dependent components, determined by different condition-dependent pathways.

Our results support a new perspective on the evolution of colour traits, where selection may be acting to maintain a balance between the different components affecting the display (Grether et al., 2005; Svensson & Wong, 2011). On one hand, different selective pressures may affect different components of an ornamental trait, and those may thus evolve independently (e.g. feather length and carotenoid incorporation). On the other hand, different selective pressures may similarly affect different components, potentially leading to the incorrect conclusion that one selective pressure may be at the origin of their evolution. For example, in several species sexual selection has been suggested to favour individuals with yellower plumages since yellowness provides information about health status, body condition or parental quality (Dufva & Allander, 1995; McGraw & Ardia, 2003), and at the same time, selection for longer feathers provides an advantage in natural selection due to improved insulation/thermoregulation, which may indirectly lead to increased feather overlap, and thus, yellowness.

In conclusion, we show that inferring evolutionary explanations using only one component can lead to simplistic or inaccurate conclusions. Using an integrative approach, we provided the first experimental evidence for condition dependency of both, pigmentary and structural features of plumage colouration of great tits, and the existence of different, independent pathways shaping colour elaboration of carotenoid-based colouration, revealing a previously ignored level of complexity of colour composite traits. The condition-dependent nature of several of their components may favour the evolution of multiple-component signals. The existence of different condition-dependent mechanisms that independently affected colour elaboration may provide an explanation why different fitness optima may exist, why selection may not deplete genetic variance, and it may explain inconsistency among studies.

Acknowledgments

We thank Hoffmann–La Roche, Basel, for kindly providing carotenoids, K. Bernhard and A. Giger for their helpful advice and discussion, J.-D. Charrière and P. Fluri for providing bee larvae, and M. Walker for discussion on the experimental design.

C.R.D was supported by the FPU programme of the Spanish Ministry of Education. The experiment was financially supported by the Swiss National Science Foundation (grant 31-53956.98 to H.R.; grant PPOOP3_128375 to P.S.F) and conducted under a licence provided by the Ethical Committee of the Office of Agriculture of the Canton of Bern, Switzerland.

Ancillary