Strength and cost of an induced immune response are associated with a heritable melanin-based colour trait in female tawny owls

Authors


*Correspondence author. E-mail: jgaspari@snv.jussieu.fr

Summary

  • 1Melanin pigments provide the most widespread source of coloration in vertebrates, but the adaptive function of such traits remains poorly known.
  • 2In a wild population of tawny owls (Strix aluco), we investigated the relationships between plumage coloration, which varies continuously from dark to pale reddish, and the strength and cost of an induced immune response.
  • 3The degree of reddishness in tawny owl feather colour was positively correlated with the concentration of phaeomelanin and eumelanin pigments, and plumage coloration was highly heritable (h2 = 0·93). No carotenoids were detected in the feathers.
  • 4In mothers, the degree of melanin-based coloration was associated with antibody production against a vaccine, with dark reddish females maintaining a stronger level of antibody for a longer period of time compared to pale reddish females, but at a cost in terms of greater loss of body mass.
  • 5A cross-fostering experiment showed that, independent of maternal coloration, foster chicks reared by vaccinated mothers were lighter than those reared by nonvaccinated mothers. Hence, even though dark reddish mothers suffered a stronger immune cost than pale reddish mothers, this asymmetric cost was not translated to offspring growth.
  • 6Our study suggests that different heritable melanin-based colorations are associated with alternative strategies to resist parasite attacks, with dark reddish individuals investing more resources towards the humoral immune response than lightly reddish conspecifics.

Introduction

Melanin is the most abundant and widespread pigment in animals and is responsible for inter- and intraspecific variation in coloration (Majerus 1998). In vertebrates, melanin pigments exist in two forms, yellow to red phaeomelanins and black eumelanins, produced de novo in specialized cells (i.e. the melanocytes). Inter-individual variation in the amount and type of melanin-based coloration is regulated in part by mutations in the melanocortin-1 receptor gene (Mc1-R) and/or by variation in post-transcriptional processing of melanocortin hormones (α-MSH, β-MSH and ACTH) or their antagonist agouti protein that binds to Mc1-R (Slominski et al. 2004). Nowadays, despite the fact that our understanding of the biochemical pathways of melanin production is making quick progress (Bennett & Lamoreux 2003), the adaptive function of melanin-based colours remains poorly known (Hill & McGraw 2006; Hoekstra 2006). The finding that melanin-based colours frequently covary with life-history traits has led to the hypothesis that such traits can signal to conspecifics alternative strategies adapted to particular ecological or social conditions (e.g. in the presence of parasites, review in Roulin 2004a). Although variation in the degree of melanin-based coloration is frequently found among individuals of the same population, reports of how such variation in colour relates to physiological responses are rare (e.g. Jones, Leith & Rawlings 1977; Hill 1991; Roulin 2004b).

In the present study, we tested the hypothesis that variation in melanin-based coloration is associated with different strategies to cope with parasites. We considered the tawny owl (Strix aluco) because this species varies continuously in melanin-based plumage coloration from dark to pale reddish, which can be easily scored by the human eye (Galeotti & Sacchi 2003; Roulin et al. 2003; Brommer, Ahola & Karstinen 2005). This variation is often referred to as ‘colour polymorphic’ and occurs throughout Europe (Galeotti & Cesaris 1996). Previous studies in Switzerland have reported that dark reddish individuals produce heavier offspring than pale reddish individuals in some years, while the reverse is true in other years (Roulin et al. 2003, 2004). In a Finnish population, a long-term monitoring of reproductive success showed that pale owls achieve a higher lifetime reproductive success than dark ones (Brommer et al. 2005). The ecological factors mediating covariation between plumage coloration and fitness parameters are unclear, but recent observations in an Italian population showing that reddish owls are infested by more blood parasites than pale reddish owls (Galeotti & Sacchi 2003) point out that the degree of melanin-based coloration may correlate with alternative parasite resistance strategies. In response to parasites, vertebrates have evolved a wide array of physiological and behavioural defences against parasites, among which the immune response is one of the most efficient, but potentially costly strategies (Lochmiller & Deerenberg 2000; Bonneaud et al. 2003). Hence, we experimentally investigated whether the capacity to mount an immune response and its associated costs with the degree of reddishness. We hypothesized that melanin-based colour variants reflect alternative strategies adapted to resist parasite attacks. We therefore predicted that the degree of reddishness covaries positively or negatively with both the intensity of immune responses and their associated costs (we do not have an a priori prediction on the direction of the relationship). Indeed, two alternative scenarios may explain the higher parasite load of dark than pale reddish Italian owls. First, reddish owls may have a weaker immune system and thereby be more intensely infested by parasites. Alternatively, reddish owls may exploit environments where parasites are more abundant, and as a consequence, reddish owls may have evolved relatively potent immune responses.

We investigated whether the degree of melanin-based coloration is associated with strategies adapted to resist parasite attacks by injecting breeding female owls either with a vaccine or with a sterile saline solution as a control. As we expected to find that the intensity of the immune response towards the vaccine would covary with the degree of female reddishness, we examined whether such an immune challenge would entail concomitant costs to females in terms of body mass loss. We also examined whether the melanin-based dependent cost of mounting an immune challenge towards the vaccine was shared with offspring, measured in terms of depressed chick mass. Because we were interested in assessing whether mounting an immune response towards a vaccine negatively affected parental care, we cross-fostered offspring between nests to ensure that offspring genotypes were randomly assigned to foster parents.

Materials and methods

study organism

The tawny owl is a philopatric monogamous species laying one to nine eggs in early spring (January to April) and incubate them for c. 28 days (Roulin et al. 2003). Both parents feed the young up to fledging, which occurs at c. 30 days of age, and fledglings stay on the parental territory up to independence, which occurs at c. 104 days of age (Sunde 2008).

study area, measurement of plumage coloration and feather concentrations of phaeomelanin and eumelanin pigments

The study was carried out in 2005 in a study area of 911 km2 in western Switzerland where we established 366 nest boxes between November 2004 and February 2005. In March 2005, nest boxes were visited every 28 days to determine the presence of incubating females. At first capture, individuals were marked by ringing them with a numbered aluminum ring. The back is the part of the body that displays the highest variation in reddishness, and hence we collected three feathers to be stored in the dark until later determination of their coloration. Feathers were stuck individually with adhesive tape onto a black paper, placed in a black box equipped with a fluorescent tube (8w/20-640 bl-super), and individually photographed with a digital camera (Dimage A200, Konika Minolta, Osaka, Japan) fixed at a distance of 27 cm to the feather. Pictures were imported in the software adobe photoshop to measure hue, saturation and brightness on 10 randomly chosen locations. After 1 h, values of the same feather started to change as the digital camera became to warm, and thus we made pictures during sessions of 1 h to avoid potential timing effects. Measurements of these three colour attributes were repeatable between feathers of the same individual (comparing mean values of each of the three feathers, repeatabilities = 0·86, 0·88 and 0·73 for hue, saturation and brightness, respectively, all P values < 0·0001). Similarly, measurements of 16 feathers taken on two different days were repeatable (r = 0·93, 0·97 and 0·69 for hue, saturation and brightness, respectively, all P values < 0·0001). We obtained a coloration score by calculating mean hue, saturation and brightness values over the 30 measurements per individual (10 measurements performed randomly on each of three feathers). As they were highly intercorrelated (Pearson correlation: r = −0·93 for hue and saturation, r = 0·81 for hue and brightness and r = −0·77 for saturation and brightness; P values < 0·0001), we extracted the first component (PC1) of a principal components analysis which explained 87% of the total variance (loading factors for hue, saturation and brightness were 0·59, −0·58 and 0·55, respectively).

The method of measuring coloration with the software adobe photoshop from pictures has been preferred to measurements taken with a reflectance spectrophotometer because it enables us to better target the area of the feathers which vary in melanin coloration (feathers are not homogenous in coloration). Spectral analyses for a subsample of the female feathers (i.e. after having photographed the feathers, part of them was destroyed to measure melanin concentration) showed however that feather mean spectral hue (r = −0·61, n = 77, P < 0·0001) and spectral brown chroma (r = −0·85, n = 77, P < 0·0001) correlated with coloration scores obtained with the software adobe photoshop (PC1), thus confirming that the two methods provide similar results. Reflectance spectra were collected using an Ocean Optics S2000 spectrophotometer (Dunedin, FL, USA) and a ‘Deuterium-Halogen 2000’ light source (Mikropack, Ostfildern, Germany), and spectral scores were calculated following methods described in Montgomerie (2006). PC1 scores also correlated strongly with indices of coloration previously assigned in the field (r = 0·89, n = 89, P < 0·0001; colour scores in the field varied from 1 (dark red) to 5 (pale red); see Roulin et al. 2003 for methods). Therefore, PC1 values derived from colour measurements made using adobe photoshop provided our score of female plumage coloration, which is used in all subsequent analyses. A low PC1 score indicates that birds are dark reddish and a high PC1 score that they are pale reddish.

We also collected two downy feathers from the back of 285 cross-fostered nestlings from 85 nests (coloration of both biological and foster parents was available for 65 of these 85 nests). We obtained a score of chick plumage coloration in the same way as for adults with the first component of a principal components analysis explaining 78% of the total variance (loading factors for hue, saturation and brightness were 0·61, −0·61 and 0·50, respectively). To compute the heritability of plumage coloration, we used 42 nestlings that were recaptured at adulthood and, therefore, for whom we had a colour score at adulthood as for their parents. Thus, plumage heritability was calculated as a strict comparison between adult–adult plumage traits rather than between nestling down feathers and adult feathers. Note that coloration of down feathers and adult feathers in 41 individuals were correlated (r = 0·71, P < 0·0001, coloration score of one individual was missing at nestling age) indicating that coloration measured at the nestling stage is finally a good surrogate of coloration at adulthood.

K. Wakamatsu identified the concentration in phaeomelanin and eumelanin pigments in one back feather of 15 adults using high-performance liquid chromatography (HPLC) previously described in Wakamatsu & Ito (2002; see also Roulin et al. 2008). J.D. Blount analysed tawny owl feathers for carotenoid pigments in 13 individuals using standard methods described in Roulin et al. (2008). As no carotenoids were detected in tawny owl feathers, we only tested whether plumage coloration was related to the concentration of phaeomelanin and eumelanin pigments.

experimental procedure

Just before hatching, we exchanged complete clutches between 51 pairs of nests based on the criteria that clutches were initiated on a similar date (Pearson correlation: r = 0·88, n = 51, P < 0·0001, average absolute difference between matched nests: 4·1 days ± 0·6) and were of similar size (r = 0·36, P = 0·008, average absolute difference between matched nests: 0·8 egg ± 0·09), and that coloration of breeding females from the same pair of nests used to cross-foster clutches were not correlated (r = −0·09, P = 0·53). Female coloration was not correlated with laying date, clutch size and mean egg size of biological (i.e. before the cross-fostering) and foster eggs (i.e. after the cross-fostering) (all P values > 0·36). For the 65 females of known age, their coloration was not correlated with their age (Pearson correlation, r = −0·17 P = 0·17). We randomly assigned females belonging to the same pair of nests used in cross-fostering experiments to one of two experimental groups, vaccinated (n = 54 females) and control (n = 48 females). There were no significant differences between vaccinated and control females in plumage coloration, clutch size, mean egg size, laying date, mass at injection, tarsus, wing or tail length (Student's t-tests, all P values > 0·14). Vaccinated females were injected subcutaneously in the neck with 0·1 mL of a vaccine solution of antigens of four human diseases (Tetravac vaccine, Aventis Pasteur MSD, Zug, Switzerland; which contained 80 U.I mL−1 of tetanus toxoid, 60 U.I mL−1 of diphtheria toxoid, 50 µg mL−1 of pertussis toxoid, 50 µg mL−1 filamentous hemagglutinin, 80 U mL−1 of D antigen of the poliovirus type 1, 16 U mL−1 of D antigen of the poliovirus type 2 and 64 U mL−1 of D antigen of the poliovirus type 3) while control females were injected in the same manner with 0·1 mL of a phosphate buffer saline solution (PBS). To control for possible pre-exposure to Tetravac antigens, we collected 50 µL of blood from the brachial vein with a heparinized capillary (Microvette CB 300 LH, Sarstedt, Sevelen, Switzerland) just before Tetravac and PBS injection to measure the initial level of antibodies against Tetravac. Between days 6 and 38 post-injection (mean ± SE = 18 ± 1), we collected a second blood sample in females to measure the production of anti-Tetravac antibodies. This was carried out in 89 females that did not fail between these two time points (pre- and post-injection). Nest failure was not associated with female coloration score (logistic regression, inline image n = 102, P = 0·93) or the immune challenge (inline image n = 102, P = 0·89). Blood samples were immediately centrifuged and plasma kept on ice in the field before being stored at the end of the day at −20 °C until laboratory analyses in autumn 2005. Each female was captured two times, i.e. a first time during incubation to inject Tetravac or PBS and a second time to take a blood sample post-injection (we recaptured 44 vaccinated females and 39 control females of 89 females for which at least one egg hatched). The time period between the two captures did not differ between vaccinated and control females (Student's t-test, t81 = 1·25, P = 0·22) and was not correlated with female plumage coloration (r = −0·07, P = 0·54). We tested for a cost of vaccination in females by comparing the change in body mass of Tetravac and PBS females (measured to the nearest 1 g) between the day of injection and the day when we collected a second blood sample. At the time when females were blood sampled to measure antibody production, we weighed each chick (11·7 ± 0·2 days of age, range: 2–19) to the nearest 1 g, measured wing length to the nearest millimetre and tarsus length to the nearest 0·1 mm, collected a blood sample for molecular sexing (Py et al. 2006) and two downy feathers for scoring coloration (as above). Eighty-nine out of the 102 experimental nests produced young (47 vaccinated nests with 164 chicks and 42 control nests with 152 chicks). We captured 65 of the 89 males during the rearing period (37 males for which their partner was Tetravac injected and 28 males for which their females was in the control treatment), which enabled us to test whether males paired with vaccinated and control females differed in body mass (we had only one measure per male). In the two treatments, male plumage coloration was not correlated with the coloration of their partner (all P values > 0·10). Furthermore, males paired with females assigned to the Tetravac and PBS treatments did not differ in coloration (Student's t-test, t63 = 0·05, P = 0·96).

assessment of humoral immune response

Antibody concentration in blood plasma was determined using a sandwich elisa (Crowther 2001). For the solid phase, we used a microtitre plate (Corning 96 well flat-bottom, Sigma, Buchs, Switzerland). Each well was coated with a Tetravac vaccine dilution (200 µL diluted in 10 mL of PBS) and incubated for 2 h at room temperature (RT). The plates were then washed five times with PBS-tween 0·05%, hereafter denoted PBS-tween (Tween 20, Reactolab, Servion, Switzerland). We saturated each plate with 200 µL of PBS-tween containing 5% of milk (PBS-milk, blotting grade blocker non-fat dry milk, Bio-Rad, Reinach BL, Switzerland) for 2 h at RT and washed again. Then, 100 µL of the diluted plasma (1:100) were randomly distributed into the wells and incubated overnight at 4 °C. After washing, 100 µL of peroxydase-conjugated rabbit anti-chicken IgG (1:3000, Sigma, A-9046) in PBS-milk were added and left for 2 h at RT. After washing five times, 100 µL of peroxydase substrate (o-phenylenediamine dihydrochlorides, 0·4 mg mL−1, Sigma) were added for 15 min at RT and then stopped using 50 µL of hydrochloric acid (HCl 1M). Optical density (OD; relative measure of anti-Tetravac antibody concentration) was read at 490 nm. As a standard, several positive samples in serial dilutions were measured on each plate. After calibration with the standard, there was a high repeatability between OD values of the same samples within (repeatability = 0·94, F19,20 = 35·15, P < 0·0001) and between plates (repeatability = 0·98, F21,22 = 119·42, P < 0·0001). Antibody production was estimated as the difference in OD values in blood samples taken before and after vaccination. For 15 samples, the amount of plasma was too low to quantify the antibody concentration. In total, we obtained a measure of antibody production for 68 of the 83 recaptured females (31 control females and 37 vaccinated females). Note that for each individual, we measured antibody production (based on two blood samples) rather than the dynamics of antibody production (which requires three or more blood samples).

statistical analyses

We investigated the relationships between plumage coloration (PC1 values) and antibody production in breeding females using a multiple regression analysis including the number of days after vaccination as a covariate. ancovas were used to test the effects of vaccination on the change in body mass of breeding females, and on the body mass of breeding males, and on chick body mass and tarsus length. We controlled for the non-independence of data collected for siblings by incorporating nest identity as a random factor. Chick age and sex were included as two cofactors in our analyses of chick development. Nonsignificant interactions were removed step by step from the model. Chick date of birth was estimated based on the size of their wings measured at our first visit to the nest box after hatching. As we used wing length to estimate chick age, and chick age was entered as a covariate in our analyses, we did not examine the effect of female vaccination on chick wing length. Statistical analyses were performed using the SAS system (version 9·1; SAS Institute Inc, Cary, NC, USA). Means are quoted ± SE, statistical tests are two tailed and P values less than 0·05 are considered significant. For illustration, we split females into dark and pale reddish morphs according to median coloration score; however, we carried out all the statistical analyses with the continuous female coloration score.

Results

feather concentrations of phaeomelanin and eumelanin pigments

Phaeomelanin and eumelanin pigments were detected in all feathers. Feather concentrations of eumelanin (15 013 ng mg−1± 2176) and phaeomelanin pigments (13 811 ng mg−1 ± 2035) were not significantly correlated (r = −0·33, n = 15, P = 0·22). The degree of reddishness was positively correlated to the amount of both phaeomelanin and eumelanin pigments stored in feathers (multiple regression: phaeomelanin: F1,12 = 95·04, P < 0·0001; eumelanin: F1,12 = 23.16, P = 0·0004), with phaemelanin pigments accounting for 68% of the total variance in plumage coloration and eumelanin pigments for 21%. Additional multiple regressions performed on the hue, saturation and brightness revealed similar relationships with melanin pigments (All P < 0·0001 for phaeomelanin and P < 0·004 for eumelanin). This suggests that the different coloration parameters were all associated in the same way with concentration in melanin pigments.

origin-related effects on melanin-based coloration

Mean coloration scores of chicks were strongly related to coloration scores of their biological parents (multiple regression analysis: biological mother: F1,60 = 9·19, P = 0·004, biological father: F1,60 = 6·82, P = 0·01) but not to colour scores of foster parents (same model, foster mother: F1,60 = 0·00, P = 0·96, foster father: F1,60 = 2·66, P = 0·11). Therefore, resemblance in melanin-based coloration between related individuals is due to origin-related factors and is not inflated by the environment. In addition, mean chick coloration score was not affected by the vaccination of foster females (Student's t-test: t83 = 0·25, P = 0·81) suggesting that melanin-based coloration is not affected by the quality of rearing conditions (see below). Based on the coloration of 42 cross-fostered offspring (28 origins) recaptured at adulthood, we obtained a heritability estimate of 0·76 (SE = 0·16) (linear regression mid-offspring by mid-parents coloration: F1,27 = 23·25, P < 0·0001, Fig. 1). Without the outlier that can be seen in Fig. 2, which may result from an extra pair fecundation (EPF rate is equal to 0·7% in this population, Saladin et al. 2007), the heritability is 0·93 (SE = 0·13) (linear regression mid-offspring by mid-parent coloration: F1,26 = 49·13, P < 0·0001). Among 132 female and 90 male tawny owls captured in our population in 2005, coloration was not sexually dimorphic (Student's t-test, t220 = 0·59, P = 0·56).

Figure 1.

Relationship between mean phaeomelanin-based coloration of cross-fostered offspring and mean coloration of biological parents (h2 = 0·93 ± 0·13, without the outlier which probably resulted from an extra pair copulation). All individuals have been colour scored at adulthood (PC1).

Figure 2.

Relationship between antibody production and number of days between the time when females were vaccinated and when a blood sample was collected to assay antibody production for dark (full line) and pale reddish (dashed line) vaccinated breeding females. For illustration, we allocated females to dark and pale reddish morphs according to median coloration score (female coloration was nevertheless introduced as a continuous variable in the statistical analyses; the distribution of female coloration was not significantly different to the normal distribution, Kolmogorov–Smirnov: P = 0·15). Antibody production is computed as the difference between OD values obtained with elisa before and after vaccination.

melanin-based coloration and antibody production

Immune challenge with Tetravac was successful (Student's t-test on the production of anti-Tetravac antibodies by vaccinated and control females: t66 = 11·53, P < 0·0001) since only vaccinated females mounted a significantly detectable immune response against Tetravac (mean production of anti-Tetravac antibodies in control and vaccinated females = −0·007 ± 0·007 vs. 0·193 ± 0·015). Preliminary statistical analyses restricted to vaccinated females showed that female antibody production over time was best described by a linear decrease (F1,35 = 14·81, P = 0·0005) rather than by a quadratic relationship (F1,34 = 0·38, P = 0·54). This suggests that female anti-Tetravac antibody production probably reached a peak between 7 days and 12 days after vaccination followed then by a gradual decrease in antibody production (Fig. 2). Additional modelling where antibody production was fitted following a linear decrease over time revealed that the dynamics of antibody production was associated with the degree of reddishness, as shown by the significant interaction between female coloration and number of days after vaccination (interaction: F1,33 = 4·88, P = 0·03; coloration: F1,33 = 1·47, P = 0·23; number of days after vaccination: F1,33 = 22·47, P < 0·0001). Pale reddish females more rapidly decreased their production of anti-Tetravac antibodies over time, compared to dark reddish females (Fig. 2). These results were not confounded by female body mass measured on the day of injection and the breeding date as the production of anti-Tetravac antibodies was not significantly correlated with female body mass at injection and hatching date, alone or in interaction with female coloration (similar multiple regressions, all P values > 0·53).

melanin-based coloration and associated costs of immunity

Immune challenge with Tetravac affected differentially the change in body mass of dark and pale reddish breeding females, as shown by a significant interaction between female coloration and vaccination (ancova on change in female body mass; coloration: F1,79 = 4·36, P = 0·04, vaccination (i.e. Tetravac vs. PBS): F1,77 = 0·48, P = 0·49, number of days after injection: F1,77 = 2·79, P = 0·10, body mass at injection: F1,77 = 0·19, P = 0·003, interaction vaccination by coloration: F1,77 = 5·40, P = 0·02). When vaccinated, dark reddish females tended to loose body mass while pale reddish females tended to gain mass, as shown by the positive relationship between coloration and body mass change in vaccinated females (multiple regression on body mass change; coloration: F1,41 = 8·76, P = 0·005, number of days after injection: F1,41 = 0·09, P = 0·76, body mass at injection: F1,41 = 7·27, P = 0·01; Fig. 3). This relationship was not detected in control females (similar analysis; coloration: F1,36 = 0·01, P = 0·93, number of days after injection: F1,36 = 5·49, P = 0·02, body mass at injection: F1,36 = 3·88, P = 0·05; Fig. 4).

Figure 3.

Relationship between phaeomelanin-based plumage coloration and body mass change in breeding females injected with a saline solution (i.e. control mothers) (full regression line) or with a complex of four antigens (i.e. mothers vaccinated with Tetravac) (dashed regression line).

Figure 4.

Mean (± 1 SE) difference in body mass (a) and tarsus length (b) of cross-fostered chicks in relation to their age and the treatment of their foster mother (vaccinated versus control).

Male body mass was not associated with the immune challenge nor with plumage coloration of their female partner (ancova on male body mass: number of days after injection of their female partner: F1,60 = 0·03, P = 0·85, female coloration: F1,60 = 0·081, P = 0·78, female vaccination: F1,60 = 0·73, P = 0·40, interaction female vaccination by female coloration: F1,60 = 0·09, P = 0·76).

Foster chicks reared by vaccinated females gained body mass and tarsus length at slower rates than foster chicks reared by control females (Table 1, Fig. 4). The effect of vaccinating breeding females on chick body mass and tarsus length was stronger at the beginning of life, as indicated by the significant interaction between chick age and maternal treatment (Table 1, Fig. 4). Because all the chicks were swapped in the present study, future studies are required to establish effect of maternal immune challenge on nestlings which have not been cross-fostered. In addition, a higher proportion of foster chicks reared by control females (29 out of 152, 19·1%) were found breeding in 2006 and 2007 than foster chicks reared by vaccinated females (16 out of 164, 9·8%; variation in recruitment rate (i.e. number of recruits per brood/number of fledglings per brood) was analysed using a general linear model with a binomial error structure and a logit-link function: F1,87 = 5·66, P = 0·02).

Table 1.  Mixed model ancova with chick body mass and tarsus length as the dependent variable in separate models, foster mother coloration, initial clutch size (i.e. before cross-fostering), brood size and chick age as covariates, chick sex and maternal treatment (i.e. vaccinated with Tetravac or PBS injected) as cofactors. We included the nest as a random effect nested within maternal treatment (as indicated by the brackets) as siblings sharing the same nest were not statistically independent. Interactions are indicated with the symbol x. F values are given for fixed effects and Wald Z-values for random effects
 Chick body massChick tarsus length
F or Z statisticd.f.PF or Z statisticd.f.P
Fixed effects
 Brood size   0·641·144  0·42   0·231·144  0·63
 Initial clutch size   4·611·144  0·03   1·991·144  0·16
 Chick age1368·671·79< 0·00012031·621·79< 0·0001
 Chick sex   3·591·144  0·06   4·831·144  0·03
 Foster mother coloration   0·041·144  0·84   1·211·144  0·27
 Maternal treatment   7·561·84  0·007   6·821·84  0·01
 Chick age × sex  20·301·144< 0·0001   6·741·144  0·01
 Chick age × maternal treatment   5·391·79  0·02   6·181·79  0·01
Random effects
 Nest(treatment)   0·00   1·12  0·13
 Nest(treatment) × chick age   3·84< 0·0001   1·85  0·03

The lack of significant interaction between foster mother coloration and vaccination on chick body mass and tarsus length (P value > 0·24; Table 1) indicates that the negative impact of the immune challenge in their ability to care for foster offspring was not associated with mother plumage coloration. Furthermore, body mass change in foster mothers was not correlated with body mass and tarsus length in their foster offspring (P values > 0·64), and was thus removed from the final model (Table 1). In addition, we did not find any interaction between vaccination of foster mother and coloration of chick and biological parents on body mass and tarsus length (similar model to that in Table 1, all P values > 0·29).

Discussion

melanin-based coloration in tawny owls

The variation in plumage coloration in tawny owl is explained to a large extent by the level of phaeomelanin pigments (68% of total variance) contained in feathers and to a lesser extent by the level of eumelanin pigments (21% of total variance) contained in feathers. Carotenoids were absent from feathers, showing that this class of pigments which is often associated with condition-dependent expression does not influence plumage coloration in this species. As previously found in the barn owls (Tyto alba, h2 = 0·81, Roulin & Dijkstra 2003) and in the Alpine swift (Apus melba, h2 = 0·78, Bize et al. 2006), the melanin-based coloration in tawny owls is strongly heritable (h2 = 0·93; Fig. 2).

melanin-based coloration and antibody production

In agreement with the hypothesis that in the tawny owl, melanin-based coloration correlates with parasite resistance (Galeotti & Sacchi 2003), our results show that in females the degree of reddishness is associated with the production of antibodies against nonpathogenic antigens, dark reddish females maintaining high concentration of antibodies for a longer period of time after vaccination than pale reddish females. Two alternative scenarios may explain this difference in antibody production between females displaying a melanin-based trait to a different extent. First, pale reddish females may have a more efficient immune response than dark reddish ones if, for instance, antibodies produced by pale reddish females bind and help neutralize foreign antigens more efficiently than antibodies of dark reddish females (i.e. greater antibody affinity). This hypothesis predicts that pale reddish females were exposed to Tetravac for a shorter period of time, and therefore that pale reddish females had to maintain a potent humoral immune response for a shorter period of time after having been vaccinated compared to dark reddish females. Second, dark reddish females may have been selected to produce greater amounts of antibodies for a longer period of time compared to pale reddish females if, for instance, dark reddish females inhabit environments where parasites are more abundant or virulent than in habitats occupied by pale reddish females. In any case, both scenarios are consistent with the finding that in Italy, dark reddish tawny owls hosted more endoparasites (Haemoproteus) than pale reddish birds (Galeotti & Sacchi 2003). Indeed, dark reddish owls are predicted to be infested by a higher number of parasites than pale reddish individuals if the immune response of dark reddish birds is less efficient at clearing parasite infestation and/or if dark reddish birds are naturally exposed to more parasites than pale reddish birds. Future studies are required where differently coloured birds are experimentally infested with endoparasites to tease apart which mechanism (differences in immune clearing capacity vs. differences in exposure to parasites) can account for the patterns of infestation observed in natural populations.

melanin-based coloration and costs of immunity

Change in female body mass was significantly associated with coloration only among vaccinated individuals but not among control ones. After being immune challenged by nonpathogenic antigens, pale reddish females gained in mass while dark reddish females lost mass. Exposure to nonpathogenic antigens has been previously demonstrated to induce body mass loss in the house sparrow (Passer domesticus) and in the common eider (Somatoria mollissima), and this body mass loss has been interpreted as a cost of immunity (Bonneaud et al. 2003; Hanssen 2006). In this context, the greater body mass loss in vaccinated dark than pale reddish owls suggests that darker reddish birds paid a higher cost of mounting an immune response. We found also that chicks were smaller and had a lower body mass when reared by vaccinated mothers compared to control mothers, but only soon after hatching. Indeed, body mass differences were not detectable in older chicks, suggesting that individuals can compensate for a poor start to life by increasing their growth rate later on, despite potential long-term fitness costs that this may incur (Metcalfe & Monaghan 2001). In agreement with this idea, the proportion of recruits was higher for foster chicks reared by control females than for foster chicks reared by vaccinated females. This result supports the general idea that an activation of the immune system in females incurs fitness costs for females and young dependent on maternal care (Ilmonen, Taarna & Hasselquist 2000; Bonneaud et al. 2003; but see Williams et al. 1999; Verhuslt, Riedstra & Wiersma 2005). As we performed a cross-fostering experiment soon before egg hatching, our results suggest that retardation of growth in chicks reared by vaccinated mothers is the consequence of reduced food provisioning (rather than differences in incubation constancy) either because mothers ate more prey items stored by males in their nest box (Galeotti 2001) or because the immune challenge may have reduced their capacity to hunt. The fact that offspring mass reduction induced by the vaccination of their foster mother was not associated with the coloration suggests that the immune challenge impaired maternal care similarly in differently coloured females. This scenario is based on the assumption that males did not compensate for the greater costs of vaccination incurred by dark reddish mothers by provisioning their brood at a higher rate than broods of pale reddish mothers. Although being not conclusive, the available data are consistent with this assumption because male body mass was not correlated with the coloration of their vaccinated partner. Our study therefore suggests that after an immune challenge, dark and pale reddish females are adopting alternative strategies with dark reddish individuals investing more effort in antibody production at the expense of body maintenance compared to pale reddish females.

proximate mechanism of the covariation between melanin-based coloration and immunocompetence

From a proximate point of view, at least two nonmutually exclusive mechanisms may account for a correlation between melanin-based coloration and immunity. First, dark and pale reddish individuals may exploit alternative habitats, which is often the case in colour polymorphic species (Galeotti & Rubolini 2004; review in Roulin 2004b). Because the abundance of parasites can vary among habitats (Moyer, Drown & Clayton 2002) and the immune system is plastic and inducible in the presence of parasites, antibody production by dark and pale reddish owls may differ because the intensity of parasite exposure differs between habitats. In agreement with this mechanism, in an Italian population darker reddish owls were found more often in territories with extensive woodlands and high levels of parasites (Galeotti & Sacchi 2003). Research is in progress to test whether in our Swiss population, dark and pale reddish owls occupy different habitats.

Second, melanin-based coloration can covary with the capacity of an individual to mount an immune response if these two phenotypic traits are genetically correlated. For example, genes coding for melanin-based coloration may pleiotropically affect immune responses. Accordingly, pharmacological research on the melanocortin systems have established that melanocortin receptors, their agonists melanocortin hormones (MSHs and ACTH) and antagonist agouti protein have key regulatory effects on the production of melanin pigments, energy balance and immunity (Catania et al. 2004; Slominski et al. 2004; Boswell & Takeuchi 2005; Ducrest, Keller & Roulin 2008). Experiments performed in barn owls (T. alba) have demonstrated that the offsprings’ capacity to mount an immune response (Roulin et al. 2000) and to resist parasites (Roulin et al. 2001) are associated with eumelanin-based coloration of their biological mother and not to the coloration of their foster mother. Experiments carried out in the tawny owl and barn owl provide support to the linkage disequilibrium hypothesis between melanin-based coloration and the immune system.

adaptive function of melanin-based coloration

From an ultimate point of view, the finding that immune responsiveness correlates with female plumage coloration in the tawny owl may suggest that melanin-based colours have an adaptive function in sexual selection, assuming that individuals can detect variation in coloration in conspecifics (Hamilton & Zuk 1982; Roulin & Bize 2007). Indeed, it has recently been shown that visual cues can be used by nocturnal birds during mate choice (Roulin 1999; Penteriani et al. 2007). If melanin-based coloration indicates a female's ability to produce antibodies against infectious pathogens, males might preferentially mate with darker reddish females in parasite-rich habitats. Higher antibody levels in females could be advantageous if darker reddish females pass onto their offspring genes that confer particular efficiency in resisting pathogen attacks, or a higher transmission of maternal antibodies via the egg yolk (Gasparini et al. 2001, 2002) providing chicks with higher resistance against pathogens (Gasparini et al. 2006). Therefore, males could indirectly enhance the immune resistance of their offspring by pairing preferentially with darker reddish females. Experiments are called for to investigate the role of female melanin-based coloration in the process of male mate choice.

Acknowledgement

This study was supported by the Swiss National Science Foundation (grant n° PPOOA-102913 to A. Roulin and PPOOA-109009 to P. Bize). J.D. Blount was supported by a Royal Society University Research Fellowship. We thank Philippe Christe, Godefroy Devevey and Joël Meunier for helpful comments on first drafts and Alan Juilland for assistance during the fieldwork. The experiment was under legal authorization of the ‘service vétérinaire du canton de Vaud’ (n° 1508).

Ancillary