Early maternal, genetic and environmental components of antioxidant protection, morphology and immunity of yellow-legged gull (Larus michahellis) chicks


Nicola Saino, Dipartimento di Biologia, Università degli Studi di Milano, via Celoria 26, I-20133 Milano, Italy.
Tel.: +39 02 50314808
e-mail: nicola.saino@unimi.it


Maternal effects mediated by egg quality are important sources of offspring phenotypic variation and can influence the course of evolutionary processes. Mothers allocate to the eggs diverse antioxidants that protect the embryo from oxidative stress. In the yellow-legged gull (Larus michahellis), yolk antioxidant capacity varied markedly among clutches and declined considerably with egg laying date. Analysis of bioptic yolk samples from clutches that were subsequently partially cross-fostered revealed a positive effect of yolk antioxidant capacity on embryonic development and chick growth, but not on immunity and begging behaviour, while controlling for parentage and common environment effects. Chick plasma antioxidant capacity varied according to rearing environment, after statistically partitioning out maternal influences mediated by egg quality. Thus, the results of this study indicate that egg antioxidants are important mediators of maternal effects also in wild bird populations, especially during the critical early post-hatching phase.


Phenotypic variation within populations consists of genetic components and environmental effects, which, in the broadest sense, embrace all nongenetic sources of variation (Falconer & Mackay, 1996). Maternal effects (see Wade, 1998 for a discussion) are regarded as a peculiar form of variation where offspring phenotype is influenced by maternal environment and phenotype rather than by ecological conditions experienced by the offspring themselves (Mousseau & Fox, 1998). The peculiarity of maternal effects is that they may have both environmental and heritable genetic components (Mousseau & Fox, 1998). The interest in early maternal effects has recently increased with the appreciation that they are ubiquitarian, constitute a major source of phenotypic variation, and can therefore have profound influences on the course of evolutionary processes (Mousseau & Fox, 1998).

An important pathway whereby variation in maternal environment and phenotype translates into variation in offspring phenotype occurs via the egg. In fact, early maternal effects mediated by egg mass and composition have been documented in a variety of taxa and can have pervasive effects on offspring ontogeny (e.g. Williams, 1994; Bernardo, 1996; Fox & Mousseau, 1998).

Egg size variation within populations or species has been suggested to have large additive genetic components (Christians, 2002), although studies of heritability of this trait have seldom taken maternal effects into account. However, egg size has also been demonstrated to depend on maternal condition and environmental factors before and during laying in diverse taxa (Fox et al., 1997; Weigensberg et al., 1998; Christians, 2002; Saino et al., 2004). Egg size predicts offspring growth rate and physiology during the early life stages, and may have carry-over effects into adulthood, by influencing adult body size and fecundity in a variety of taxa (e.g. Hutchings, 1991; Kaplan, 1992; Bernardo, 1996; Azevedo et al., 1997; Price, 1998; Fox, 2000; Svensson & Sinervo, 2000; Torres-Vila & Rodriguez-Molina, 2002; Fischer et al., 2003; Maruyama et al., 2003; Tamada & Iwata, 2005). Variation in egg size can have consequences for offspring phenotype because the amount of egg materials affects development (Sinervo & Huey, 1990; Sinervo, 1993; Finkler et al., 1998; Einum, 2003; Jardine & Litvak, 2003; Ferrari et al., 2006). However, large eggs may be laid by high quality females which are able to allocate to their eggs a disproportionate amount of particular components that have a major effect on the progeny phenotype (Lipar & Ketterson, 2000; Eising et al., 2001; Saino et al., 2003, 2005; Rubolini et al., 2005, 2006a,b; see review in Groothuis et al., 2005 for studies of steroid hormones). Egg size and content of quantitatively minor components (e.g. antioxidants, hormones and immune factors) are likely to have both independent and combined effects on offspring phenotype.

Antioxidants, in particular, are a functionally defined, diverse class of compounds which protect biological molecules from oxidative stress that can either be produced by the organism itself (e.g. cholesterol, uric acid and glutathione) or have to be acquired with the food (e.g. ascorbate, carotenoids, vitamin A, tocopherols; Surai, 2003). Molecules with high oxidant potential (such as free oxygen and nitrogen radicals and other reactive oxygen metabolites) are normally produced and released in the body of organisms as by-products of metabolism or during immune response (Chapple, 1997; Halliwell & Gutteridge, 1999; Surai, 2003), and can damage DNA, proteins, lipids and carbohydrates, therefore having negative effects on tissues integrity and organismal vital processes (Halliwell & Gutteridge, 1999; Surai, 2003).

In birds, antioxidants are transferred to the egg by mothers (Surai, 2003), and such allocation may have to be traded against allocation to maternal maintenance (see Blount et al., 2004; Blount, 2004 for a review). Indeed, some studies of birds in the wild suggested that antioxidants are limiting to laying females, as supplementation with antioxidants results in larger transfer to the eggs, with beneficial consequences for selected offspring traits (Biard et al., 2005). In addition, injection of physiological amounts of antioxidants into the egg can enhance a major component of the acquired immune system and having apparently no adverse consequences on other offspring traits (Saino et al., 2003). Finally, the concentration of antioxidants in the egg yolk is higher than in maternal tissues, implying that mothers actively allocate these compounds in egg tissues (Sunder & Flachowsky, 2001; Surai, 2003).

Yolk antioxidants are incorporated within the developing embryo and chick tissues (e.g. liver, blood), and may contribute substantially to offspring antioxidant defences (Surai & Sparks, 2001; Surai et al., 2001a,b, 2003; Kang et al., 2003; Koutsos et al., 2003; Karadas et al., 2005). Studies of poultry and fish in captivity have shown that egg antioxidants positively influence egg hatchability, the rate of body mass and osteometric growth, immunity, behaviour and viability at least in the early post-hatch life (Lin et al., 2004; see Surai, 2003 for a review), which is a critical period for chick survival. Furthermore, studies of mammals and birds have shown that antioxidants, including vitamin A, E and carotenoids, can reduce brain malformation during early ontogenetic stages (see review in Ramakrishna, 1999), thus providing a potential link between antioxidant capacity (AOC) and behavioural performance of the offspring. Proper antioxidant protection by egg antioxidants may thus be crucial for early survival, potentially affecting the somatic and neurobehavioural development of the offspring as well as the development of other major functions, including immunity.

However, no study has experimentally investigated the genetic, environmental and early maternal components of antioxidant protection of the offspring in any vertebrate species in the wild. The aim of the present study of the yellow-legged gull was to analyse offspring phenotypic variation in relation to genetic effects and environmental factors including nest of rearing and maternal effects mediated by AOC of the egg yolk and egg mass. To this end, we performed a partial cross-fostering experiment whereby we reciprocally swapped recently laid eggs between pairs of synchronous clutches and analysed plasma AOC, T cell-mediated immune response, morphology and begging behaviour of the chicks in relation to parentage and nest of egg incubation and rearing, reflecting environmental effects acting on the embryos and chicks (such as microhabitat conditions and parental care and behaviour). Typical cross-fostering experiments (e.g. Merilä & Fry, 1998) do not allow differentiation between early maternal and origin effects. In order to test for the effect of variable AOC of the eggs while controlling for common origin effects, we extracted bioptic samples from the yolk of freshly laid eggs and analysed the statistical effect of AOC per unit yolk mass and egg mass on chick phenotype. This procedure allowed us not only to analyse the parentage (nest of origin) and environmental (nest of rearing) components of phenotypic variation on chick traits, as carried out in other studies of avian species (e.g. Merilä & Fry, 1998; Meriläet al., 1999; Christe et al., 2000), but also to test for maternal effects mediated by egg size and AOC whereas simultaneously taking into account sources of phenotypic variation because of parentage and rearing environment. We predicted that chicks hatched from eggs with the largest AOC would generate offspring that grew faster and had larger T cell-mediated immune response. In addition, we predicted that plasma AOC of the chicks in the early post-hatching period would positively covary with AOC of the yolk of the original egg. In the yellow-legged gull, egg size markedly declines with laying order (see Results), suggesting that maternal investment in egg production declines in late-laid eggs. We therefore also investigated whether AOC per unit volume of yolk and total antioxidant capacity (TAOC) of the yolk (computed using estimates of yolk mass based on its isometric relationship with egg mass), also varied with laying order of the eggs.

To our knowledge, the approach of measuring overall AOC in specific tissues, rather than the concentration of individual antioxidants, has never been adopted in ecological or evolutionary studies of wild vertebrates. The quantification of the TAOC was obtained by means of a commercial kit (see Methods). The procedure consists of a reduction reaction of an oxidant agent by the antioxidants contained in the test tissue sample, and was originally developed for the assay of TAOC in human tissues (e.g. Cornelli et al., 2001; Trotti et al., 2001). However, it is also useful in the analyses of animal tissues (see Ballerini et al., 2003), because it is simply based on a chemical reaction. Therefore, throughout this study, we will assume that this measure provides an index of the overall protection from oxidative stress.


Study species, field procedures and cross-fostering experiment

The yellow-legged gull is a large (800–1500 g), ground-nesting, semi-colonial gull species of the Mediterranean region, which belongs to the herring gull complex (Liebers et al., 2001). Nests consist of a small cup of various materials and contain a maximum of three large eggs (70–105 g), which are laid at 1–3 days interval. Egg size decreases monotonically with laying order (see Results). Eggs are incubated for 26–30 days. The semi-precocial chicks remain around the nest for the first 5–10 days of life, after which they can wander considerable distances from the natal territory (Cramp, 1998). Eggs hatch asynchronously (see Rubolini et al., 2005), resulting in marked within-nest hierarchies where earlier hatched chicks become dominant over later hatched ones and have better survival prospects (Parsons, 1975; Hillström et al., 2000).

The study was conducted at a large monospecific colony (>300 breeding pairs) in the Comacchio lagoon (NE Italy, 44°20′N–12°11′E), during 2004–2005 (spring). The colony was visited daily or every second day (depending on weather conditions), starting from the last week of March, when the first eggs are laid in the study population. Nests were marked the day when the first egg was found, and all eggs were univocally marked as a-, b- or c-eggs according to position in the laying sequence. When a new egg appeared, it was brought to a nearby building for the yolk biopsy (see below), and temporarily replaced with a dummy yellow-legged gull egg. We did not visit the colony during inclement weather or during the central hours of the day in sunny days, to avoid excessive cooling or overheating of eggs and chicks. During all field procedures, care was taken not to shake the eggs and not to expose them to direct sunlight or rain. All eggs were removed for a maximum of 5 h before being taken back to their original nest. Meanwhile, they were always kept protected and maintained at ambient temperature and humidity.

On the visit following clutch completion, we performed a partial cross-fostering experiment, where eggs of the same laying order were reciprocally swapped between pairs (=dyads) of synchronous nests (with respect to laying date of the last egg). Dyads of nests were assigned sequentially to a predetermined cross-fostering scheme, where in the first dyad the a-egg was swapped between the two nests, in the second dyad a- and b-eggs were swapped, in the third dyad the b-egg was swapped and so on, until all six possible combinations (i.e. abc, abc, abc, abc, abc, abc, where the italics denote that the egg was swapped) were applied. If an egg disappeared before the day of cross-fostering or if only two eggs were laid, we randomly exchanged one egg between nests. Nests where only a single egg was laid or where a single egg remained were not cross-fostered and were excluded from the analyses.

Nests were checked daily around the time of hatching to record the first signs of eggshell breaking by the embryo. In this species, 1 or 2 days elapse between the appearance of the first signs of hatching and the appearance of the chick. In order to properly assign chicks to their original egg, when the embryo produced a small opening in the eggshell, we injected a minute amount of blue or green food dye. This resulted in faint blue or green markings on the chick's down, which usually disappeared during the first few days after hatching. By this procedure, all the chicks could be unequivocally assigned to their egg of origin (see also Rubolini et al., 2005). On the day they were first found, at the average age of 0.63 (0.04 SE) days (age 1 hereafter), chicks were measured (body mass, to the nearest 1 g and tarsus length, to the nearest 0.1 mm) and marked with combinations of coloured elastic plastic bands on tibiotarsi.

Within the first 2 days after hatching, we recorded the intensity of the begging behaviour directed to parents by means of a standard protocol (see Rubolini et al., 2005 for a detailed description of the procedure). Briefly, chicks were individually placed in their nest and were presented with a realistic plastic head of an adult gull, to which the chicks respond by vigorously pecking at the red spot on the lower mandible, a behaviour which acts as a releaser for food regurgitation by parents. As a measure of the intensity of the begging display, we recorded the number of distinct pecks directed to the dummy head during 1 min trials.

At an average 4.46 (0.03 SE) days of age (hereafter age 4 for simplicity), we again measured body mass and tarsus length, and took a first sample of blood (150 μL) from the ulnar vein. Blood samples were kept cool until plasma was separated from blood cells (within a few hours from collection) by centrifugation (10 min at 11 000 rpm) and stored at −20 °C for subsequent analyses. The procedure was repeated on average at day 8.41 (0.04 SE) (hereafter age 8 for simplicity). At this time, we also started a standard in vivo test to measure the intensity of the T cell-mediated immune response (Lochmiller et al., 1993; Saino et al., 1997; Tella et al., 2002). The wing web of the right wing was injected subcutaneously with 0.2 mg of phytohaemagglutinin (PHA) dissolved in 0.05 mL phosphate-buffered saline (PBS) and the same amount of PBS was injected in the left wing web to serve as a control. We measured the thickness of both webs at the site of inoculation prior to the inoculation and 24 h later using a pressure-sensitive micrometer. The difference in change in thickness between the right and the left wing webs was used as an index of T cell-mediated immune response (Rubolini et al., 2005). In all the analyses, mass data were expressed in ‘g’ and tarsus length in ‘mm’.

Egg biopsies

In order to measure the yolk AOC, we obtained a small bioptic sample of the yolk. Before taking the bioptic sample, all eggs were weighted to the nearest 1 g by means of a Pesola spring balance. Biopsies were performed the day of laying by means of 2.5 mL disposable sterile syringes mounting a 21-gauge, 40-mm-long needle. Eggs were left with the acute pole upward for approximately 10–15 min before the procedure, to allow the yolk to reach a standard position within the egg. The acute pole of the egg was then carefully cleaned and disinfected, and a small hole was drilled by means of a sterile needle at approximately 1 cm from the pole. The needle of the syringe was then inserted for 3/4 of its length into the egg, with holding the needle tip slightly pointing towards the vertical axis of the egg. This depth and inclination were chosen because we estimated that the needle tip would reach the middle of the yolk of an average egg during preliminary trials on partially open eggs. A small amount (approximately 100–200 mg; corresponding to <0.8% of an average yolk mass) of yolk was then sucked into the syringe. We then gently extracted the needle and sealed the hole by glueing a small piece of gull eggshell over it. The yolk sample contained in the syringe was then immediately transferred into a vial and stored at −20 °C within a few hours of extraction until laboratory analyses. Although obviously not all yolks could be sampled in the same position, because of the large variation in egg size and shape in this species, we checked for the reliability of the AOC of the bioptic sample with respect to the AOC of the whole yolk in random sample of 20 eggs that were dissected after the biopsy (see below). We estimated that the biopsy procedure per se caused an increase of hatching failures of approximately 23% in our sample of eggs (hatching success for eggs subjected to biopsy was 55%, whereas that of unmanipulated eggs in the study population was 78%, see Rubolini et al., 2005). However, hatching failures should not have biased our analyses of the effects of antioxidants on chick traits because there was no difference in yolk antioxidant concentration between eggs of the same nest of origin that hatched and those that did not (mixed-model anova with nest of origin as a random effect factor, F[1,128] = 0.39, P = 0.53).

Antioxidant capacity

Antioxidant capacity of the yolk was measured using commercial OXY-Adsorbent kits purchased from Diacron s.r.l. (Grosseto, Italy). In principle, the OXY-Adsorbent test allows the colorimetric assessment of the capacity of a test sample (e.g. blood) to prevent oxidation by the hypochlorous acid (HClO), which has high oxidant potential and occurs as a natural oxidant in biological fluids. The test sample is exposed to oxidation by a HClO solution of known titre. Excess HClO, which is not reduced by the sample antioxidants, is then exposed to an alkyl-substituted aromatic amine solubilized in a chromogenic solution. Such amine is oxidized by residual HClO and transformed into a pink-coloured derivative. Antioxidant capacity of the test sample is directly related to the amount of HClO, which is neutralized by the test sample and, thus, negatively related to the amount of the aromatic amine that is oxidized by HClO, which can be determined by colorimetry. Antioxidant capacity (i.e. the capacity of neutralizing oxidant compounds) can thus be expressed in terms of moles of neutralized HClO per unit volume of the test sample.

In practice, we extracted 5–8 mg (=5–8 μL assuming a 1 : 1 mass/volume ratio) from egg bioptic samples or 10 μL from chick plasma samples. Before extraction of the test subsamples, the original sample was carefully mixed (plasma) or sonicated (yolk). These samples were diluted in bidistilled water to obtain a dilution of 1 : 200 (yolks) or 1 : 100 (plasma). The reason why we did not always start from the same mass of yolk is that, owing to yolk viscosity, we could not always extract a predetermined amount. Yolk samples were therefore weighed (Sartorius CP124S, accuracy of 0.0001 g) and the volume of distilled water due to be added was decided according to the mass of the subsample of yolk extracted. Yolk solutions were sonicated for 2 min at 15 kHz. Both yolk and plasma solutions were vortexed for 20 s. About 10 μL of these solutions were added to 1 mL of the R1 (according to kit terminology) oxidant solution in ad hoc cuvettes, gently mixed by inversion three times, and left at 37 °C for 10 min. Whereas 10 μL of R2 reagent were then added to the cuvettes, which were gently mixed, producing the chromogenic reaction. Absorbance of the solutions was measured approximately 10 s after adding the R2 reagent at 546 nm, according to kit instructions, using a dedicated photometer (Free, Diacron s.r.l.). Antioxidant capacity was expressed as [(AbsB − AbsS)/(AbsB − AbsStd)] × [Std], where AbsB is the absorbance of the ‘blank’ solution, consisting of 1 mL of reagent R1 (see above) with 10 μL of bidistilled water added and processed as a normal test sample, AbsS is the absorbance of the test sample, AbsStd is the absorbance of a standard human serum sample of known anti-HClO AOC provided by the kit producer and Std is the AOC (in μmol mL−1) of the standard serum. Antioxidant capacity per unit volume of yolk or plasma (AOC hereafter) was expressed in μmol mL−1.

Mean intra-assay coefficient of variation of AOC for yolks was 5.8% (1.2 SE; five assays, 33 measures of 15 samples assayed in duplicate or triplicate) whereas that for plasma was 6.2% (1.1 SE; five assays, 34 measures of 17 samples assayed in duplicate). Mean interassay coefficient of variation for yolks was 8.2% (1.0 SE; five samples assayed in duplicate to quadruplicate) whereas that for plasma was 7.1% (1.2 SE; five samples assayed in duplicate or triplicate).

As the relationship between yolk and total egg mass was isometric [type II log–log regression of yolk mass on egg mass: slope = 0.997 (0.083 SE), R2 = 0.635; H0: β = 1, HA: β = 1; t50 = 0.04, n = 55, P > 0.90], we obtained an index of total antioxidant potential as: [AOC × Unincubated egg volume (in mL, assuming 1 g = 1 mL) × 0.29 (= mean yolk/total egg mass in a sample of 55 eggs)]. TAOC was calculated because it provides an indication of total maternal investment for antioxidant allocation to the egg, whereas AOC is a measure of the antioxidant protection provided to the offspring per unit yolk mass. TAOC was expressed in mmol.

Antioxidant capacity in the biopsies could not reflect AOC per unit volume in the whole yolk if, for example, AOC varies with position in the yolk and biopsies from individual eggs were obtained from different parts of the yolk. This could be the case, despite the fact that we standardized all the field procedures (see above), because of variation in the size and shape of the eggs and in the size of the yolks. We therefore tested whether AOC in the biopsies predicted overall AOC of the yolk by extracting biopsies by the same procedures outlined above from eggs that were subsequently dissected to measure AOC per unit volume of carefully homogenized yolks (n = 20 eggs). We first extracted a bioptic sample and then extracted a sample from the entire yolk after homogenization and sonication. The correlation between AOC in the biopsy and in the sample from the entire yolk was positive and highly significant (r = 0.83, P < 0.0001). A type II regression of AOC measured in a sample extracted from the entire yolk (y-variable) on AOC in the biopsy had an estimated slope of 0.81 (95% CI: 0.59–1.03) and an intercept of 95.57 (95% CI: −7.4 to 198.53). Thus, the confidence interval of the slope included 1 and that of the intercept included 0, indicating that the measure of AOC in the yolk was proportional to AOC measured in the biopsy and the slope of this relationship did not significantly differ from 1.

Statistical analyses

Variation in egg mass and AOC or TAOC at the time when biopsies were taken, i.e. before cross-fostering, was analysed in mixed-model analyses of covariance where clutch (=nest) of origin was included as a random factor, laying order was included as a factor rather than as a continuous covariate to account for nonlinear/nonmonotonic variation in features of consecutive eggs, and laying date was considered as a continuous covariate. As in previous studies of this population of gulls, in the analyses we included laying order of the original egg, which was known for all nestlings, rather than hatching order, which could not be determined on a few occasions when two or more chicks were found to have hatched on the same visit to the colony (see also Rubolini et al., 2005). However, hatching order closely reflects laying order in the study population (Rubolini et al., 2005). It should be noted that the analyses of egg mass variation in relation to nest of origin and laying order and date were based on a larger sample than those of AOC because we also considered a sample of unmanipulated clutches studied in 2004 in the same area (see also Results).

We relied on linear mixed-model analyses of variance (e.g. Merilä & Fry, 1998) to investigate the effects of the nest of origin, reflecting parentage effects, and the nest where the eggs were incubated and the chicks were reared, reflecting environmental and parental effects during incubation on time to hatching, chick plasma AOC and morphology. In these analyses, the effect of nest of origin or nest of rearing and their interaction were nested within the effect of dyad, because reciprocal cross-fostering occurred within dyads of nests, and were considered as random effects (see Merilä & Fry, 1998 for a similar design). Further details on the structure of specific models are provided throughout the Results section. In the analyses concerning chick traits, chick age at measurement was included in the models as a covariate to account for small variations in age at measurement because of practical reasons (see Study species, field procedures and cross-fostering experiment). Mixed-model analyses of variance, including nest of origin as a random factor and examining the sources of variation in yolk AOC, TAOC, egg mass and egg mass variation between consecutive eggs (see Results), were subjected to a step-down simplification procedure; nonsignificant terms (P > 0.05) were sequentially removed from the model, starting from interaction terms, until minimal adequate models, including only significant terms, were obtained (Crawley, 1993). Throughout the Results section, the significance of variance components explained by random effects was tested by means of a likelihood ratio test, calculated as the difference in the −2 × log-likelihood (−2LL) scores for a full model vs. a reduced model, i.e. a model without the effect of interest. In hierarchical mixed-models including dyad, nest of origin and nest of rearing as random effects, the full model was tested against a reduced model that excludes the effect under test and/or the interaction term, whereas the effect of dyad, the highest level in the hierarchy, was tested against a model without random effects. The difference in −2LL follows a χ2-distribution with degrees of freedom equal to the number of factors by which the full and the reduced models differ, and P-values are one-tailed (see Littell et al., 1996 for further details). All full and reduced models had the same fixed effect structure. Parameter estimates for all mixed analyses of variance models were obtained by the maximum likelihood method (results were qualitatively unchanged whether restricted maximum likelihood was adopted; details not shown), and degrees of freedom for fixed effects were estimated by the Sattertwaithe's approximation. Variance components are expressed as the percentage of variance explained by a given random effect when all other random effects were included in the model. All analyses were performed using the sas system (ver. 9.0).


Variation of egg antioxidant capacity

Antioxidant capacity per unit volume of yolk (AOC) and yolk TAOC were measured on bioptic samples from 209 eggs (sample of a-eggs: 73, b-eggs: 77, c-eggs: 59) belonging to 107 clutches.

Analyses of variance showed that a highly significant among-clutches variation existed in both AOC (F[106,102] = 2.85, P < 0.0001) and TAOC (F[106,102] = 2.80, P < 0.0001), with nest of origin accounting for 74.7% of the variance in AOC and 74.4% of the variance in TAOC. Thus, most of the variation in AOC observed in our study population occurred at the among-clutches rather than at the within-clutch level. A mixed-model anova with nest of origin as a random effect and laying order as factor showed that AOC did not vary with laying order (F[2,132] = 1.45, P = 0.24, Fig. 1), whereas TAOC declined with laying order (F[2,125] = 5.80, P = 0.004, Fig. 1). In these analyses, nest of origin explained a significant amount of variance in AOC or TAOC (variance explained >48%; inline image > 30.0, P < 0.0001).

Figure 1.

 Mean (+SE) mass, antioxidant capacity per unit yolk volume (AOC) and total antioxidant capacity of the yolk (TAOC) of eggs from 334 clutches (egg mass) or 107 clutches (antioxidants) in relation to laying order (a-eggs = first laid eggs). The size of the samples of eggs is shown (see also Methods for the discrepancy between the number of clutches and the number of first eggs in the sample).

Mixed-model analyses of variance of AOC with nest of origin as a random factor where we initially included laying order, egg mass and laying date and two-way interactions as predictors showed no effect of laying order and all two-way interactions between covariates. The final model resulting from the step-down exclusion of nonsignificant predictors showed that AOC was negatively predicted by laying date [F[1,117] = 7.86, P = 0.0059, coefficient = −4.744 (1.693); Fig. 2] and egg mass [F[1,208] = 6.60, P = 0.0109, coefficient = −3.407 (1.327); Fig. 3], implying that large eggs that were laid late in the season had smaller AOC per unit volume of yolk than small and early laid eggs. The slope of the relationship between AOC and laying date yields a decline in AOC by approximately 15.9% of the mean AOC (=830.80 μmol mL−1, SD = 140.08, n = 209) or 0.94 SD over the 28 days timespan of laying dates (26 March–22 April) covered by the present study. The nest of origin explained a significant proportion of the variance (=44%; inline image = 19.4, P < 0.0001).

Figure 2.

 Yolk antioxidant capacity (AOC) in relation to laying date of individual eggs. Simple linear regression line is shown (see also Results).

Figure 3.

 Yolk antioxidant capacity (AOC) in relation to mass of unincubated eggs. Simple linear regression line is shown (see also Results).

Total antioxidant capacity was found to increase with egg mass, after removal of the nonsignificant egg mass × laying order interaction (Table 1; Fig. 4). This result was not necessarily expected because AOC could decrease with egg mass, as we in fact observed in the present data set. In addition, it decreased with laying date, as shown by the negative values of the parameter estimated for eggs of all laying orders (Table 1; Fig. 5). However, the relationship between TAOC and laying date was steeper for a- than for b- and c-eggs (Table 1). In this model, the effect of laying date could mask the effect of laying order, which was, however, observed in the single-variable model (see above).

Table 1.   Minimal adequate mixed-model anova of total antioxidant capacity (TAOC) in relation to laying date, egg mass (continuous covariates) and laying order (fixed effect factor).
Source of variationNum d.f.Den d.f.FP-valueEstimated parameters (SE)
  1. Nest of origin was included as a random factor to link eggs from the same clutch, and explained a significant proportion of the variance (=43%; inline image = 18.5, P < 0.0001). Sample size is 73 a-, 77 b- and 59 c-eggs.

Laying order21490.690.5013 
Egg mass118218.03<0.00010.171 (0.040)
Laying date11048.380.0046 
Laying date × laying order21353.410.0360a-eggs −0.166 (0.080) b-eggs 0.005 (0.074)
     c-eggs −0.075 (0.067)
Figure 4.

 Total yolk antioxidant capacity (TAOC) in relation to laying date of individual eggs. TAOC was calculated as AOC × egg mass × 0.29 (=proportion of yolk relative to total egg mass). Simple linear regression line is shown (see also Results).

Figure 5.

 Total yolk antioxidant capacity (TAOC) in relation to egg mass. TAOC was calculated as AOC × egg mass × 0.29 (=proportion of yolk relative to total egg mass). Simple linear regression line is shown (see also Results).

Time elapsed between laying of consecutive eggs (i.e. between a- and b-eggs or between b- and c-eggs) did not predict variation in AOC or TAOC in mixed-model analyses of variance with laying sequence as a two-levels fixed factor (indicating whether the difference in AOC or TAOC referred to the difference between a- and b-eggs or between b- and c-eggs) and nest of origin as a random factor (details not shown).

Variation in egg mass and time to hatching

An analysis based on 334 clutches and including data for 332 a-, 334 b- and 286 c-eggs showed that most of the variation in egg mass occurred at the among-clutches level (one-way anova, F[333,618] = 3.13, P < 0.0001, R2 = 62.8%). The size of the sample of eggs included in the analyses of egg mass variation was much larger than that used for AOC analyses because we also included eggs from unmanipulated nests measured in the same colony for other purposes during 2004.

In a mixed-model anova of egg mass in relation to laying order, laying date and their interaction, where nest of origin was included as a random factor, egg mass varied according to laying order (F[2,786] = 296.91, P < 0.0001; Fig. 1), after removal of the nonsignificant laying order × laying date interaction. Bonferroni post hoc tests showed that a significant monotonic decline of egg mass occurred with laying order (P < 0.0001 in all pairwise comparisons). Egg mass declined with laying date [F[1,377] = 6.64, P = 0.010, coefficient = −0.123 (0.048)], and the slope of the relationship yields an overall decline of 3.69 g during the 30-days period of laying as we investigated, corresponding to 4.3% of mean values and 48.6% of the SD in egg mass (=85.18, SD = 7.59, n = 952 eggs).

Variation of egg mass was significantly predicted by time elapsed since laying of the previous egg in a mixed-model anova with nest of origin as a random factor, laying sequence as a factor (see previous paragraph) and time (in days) since laying of the previous egg as a covariate. The difference in mass between each egg and the preceding one increased with time since laying of the previous egg [F[1,602] = 7.05, P = 0.008, coefficient = 0.582 (0.219)]. In addition, mass decline between a- and b-eggs was smaller than between b- and c-eggs (F[1,602] = 103.36, P < 0.0001).

Time elapsed from laying to hatching (expressed in days; duration of incubation hereafter) was analysed in a mixed-model anova in relation to nest of origin and nest where individual eggs were transferred after cross-fostering (nest of incubation). The random effects of nest of origin, nest of incubation and their interaction were nested in the effect of dyad (see Statistical analyses). Laying order, egg mass and AOC were included as fixed effects. AOC significantly and negatively predicted the duration of incubation [F[1,178] = 5.69, P = 0.018, coefficient = −0.0015 (0.0007)], which increased with egg laying order (F[2,109] = 84.10, P < 0.0001). Among random factors, only dyad explained a significant proportion of the variance in time to hatching (variance = 37.3%; inline image = 30.5, P < 0.0001), whereas all other random effects were nonsignificant (P > 0.06). The analysis of duration of incubation by the same model described above while including TAOC rather than AOC was not run because TAOC was computed as the product of AOC and estimated yolk mass based on egg mass (see Methods), which, in turn, was already included in the model as a covariate.

Plasma antioxidant capacity in the chicks

Yolk AOC was significantly larger than plasma AOC at age 4 (t-test for paired data; t179 = 65.42, P < 0.0001; Fig. 6), and at age 8 (t-test for paired data; t117 = 42.66, P < 0.0001; Fig. 6). However, AOC was smaller at age 4 than at age 8 (t-test for paired data; t113 = 12.56, P < 0.0001; Fig. 6). The comparison between yolk AOC and plasma AOC at age 4 based only on chicks that could also be sampled at age 8 gave similar results (t-test for paired data; t112 = 52.12, P < 0.0001).

Figure 6.

 Mean (+SE) antioxidant capacity (AOC) in the egg yolk and in the plasma of 4- or 8-day-old chicks. The size of the samples is shown. Mean values computed over yolk biopsies and plasma samples collected at day 4 from chicks that could also be sampled at age 8 were very similar to presented, which are based on the maximum sample available.

We investigated whether chick plasma AOC was related to egg features, including yolk AOC, egg mass and chick age at blood sampling, in mixed-model analyses of variance. AOC at age 4 was not predicted by egg AOC or other egg features and was unaffected by nest of origin or rearing (Table 2). At age 8, AOC varied significantly according to nest of rearing (Table 2). In addition, plasma AOC at age 8 was found to be larger in chicks from a- compared with b- or c-eggs and was negatively predicted by yolk AOC (Table 2).

Table 2.   Mixed-model anova of antioxidant capacity of the chicks’ plasma in relation to nest of origin or rearing and egg features including yolk antioxidant capacity (AOC), egg mass, laying order and age at measurement (see Results).
Source of variationStatistics
 χ2d.f.P-valueVariance (%)
Age 4
Random effects
 Nest of origin (dyad)010.500
 Nest of rearing (dyad)2.710.05118.7
 Nest of origin × nest of rearing (dyad)010.500
Fixed effectsFd.f.P-valueEstimate (SE)
 Laying order0.732, 1240.48 
 Yolk AOC0.061, 1730.800.004 (0.015)
 Egg mass0.711, 1620.40−0.258 (0.307)
 Chick age0.981, 1760.324.05 (4.09)
 χ2d.f.P-valueVariance (%)
Age 8
Random effects
 Nest of origin (dyad)0.610.223.5
 Nest of rearing (dyad)10.71<0.00156.5
 Nest of origin × nest of rearing (dyad)0.710.2116.2
Fixed effectsFd.f.P-valueEstimate (SE)
  1. The variance explained by random effects is calculated from the variances of the model including all random effects simultaneously, whereas the χ2-statistics of the variance components for nest of origin, nest of rearing and their interaction are obtained from the difference in the −2 × log-likelihood of a model vs. that of a reduced model excluding the effect under test and/or the interaction term (the effect of dyad, the highest term in the hierarchy, is tested against a model without random effects; see Methods for further details). The analysis at age 4 is based on 180 chicks from 43 nests of origin and 37 nests of rearing. The analysis at age 8 is based on 118 chicks from 30 nests of origin and 23 nests of rearing.

  2. *Post hoc Bonferroni test: chicks from a-eggs > (b-eggs = c-eggs).

 Laying order4.592, 44.10.015* 
 Yolk AOC6.501, 86.40.013−0.075 (0.029)
 Plasma AOC at age 41.371, 57.40.2460.158 (0.135)
 Egg mass0.111, 65.20.7380.196 (0.582)
 Chick age4.201, 81.70.044−16.17 (7.89)

Chick begging behaviour, growth and immunity in relation to egg mass and laying order and yolk antioxidant capacity

Offspring phenotype was analysed in relation to parentage and rearing environment in mixed-model analyses of variance where the effects of nest of origin or rearing and their interaction where nested within the effect of dyad (see Methods). In these analyses, we included yolk AOC as a covariate because we aimed at testing the effect of yolk antioxidants on chick phenotype. In addition, we entered egg mass because it is known to predict chick morphology (Williams, 1994) and laying order as a factor to control for the effect of variation in egg traits (other than antioxidants) on chick traits (e.g. hormones, Royle et al., 2001). Finally, we included chick age to account for the effect of small variations in age at measurement because of practical reasons (see Methods). For brevity, in this section we will only report the details of the analyses where the effects of nest of origin or rearing, egg mass or AOC were found to be significant. We did not repeat the analyses of chick phenotypic traits while entering TAOC rather than AOC in the models presented in Table 3 because TAOC was computed based on egg mass, which was entered in the models as a covariate.

Table 3.   Mixed-model anova of chick phenotype at three successive ages (see Methods) in relation to nest of origin (Origin) or rearing (Rearing) and egg features including yolk antioxidant capacity (AOC), egg mass and laying date, and age at measurement (see Methods).
Source of variationAge 1Age 4Age 8
 χ2d.f.P-valueVariance (%)χ2d.f.P-valueVariance (%)χ2d.f.P-valueVariance (%)
Tarsus length
Random effects
 Dyad010.50 00.110.40 02.110.07 6.3
 Origin (dyad)010.46 0010.50 0010.50 0
 Rearing (dyad)010.50 01.010.16 4.22.910.04523.7
 Origin × rearing (dyad)1.510.1135.31.710.1030.11.010.1623.5
Fixed effectsFd.f.P-valueEstimate (SE)Fd.f.P-valueEstimate (SE)Fd.f.P-valueEstimate (SE)
 Laying order2.832, 125 0.063 7.752, 123<0.001  7.972, 69.3<0.001 
 Yolk AOC4.861, 188 0.0290.001 (0.001)3.841, 174 0.0520.002 (0.001) 0.601, 119 0.440.001 (0.002)
 Egg mass66.51, 174<0.0010.085 (0.010)28.81, 160<0.0010.118 (0.022)12.21, 102<0.0010.135 (0.039)
 Chick age64.81, 184<0.0010.938 (0.117)36.91, 173<0.0011.698 (0.280)26.01, 127<0.0012.571 (0.504)
 χ2d.f.P-valueVariance (%)χ2d.f.P-valueVariance (%)χ2d.f.P-valueVariance (%)
Body mass
Random effects
 Dyad3.410.03312.0010.500 0.71 0.20 0
 Origin (dyad)010.50 0010.500 01 0.50 0
 Rearing (dyad)010.50<0.00152.7
 Origin × rearing (dyad)0.210.34 7.41.910.0834.7 01 0.50 0
Fixed effectsFd.f.P-valueEstimate (SE)Fd.f.P-valueEstimate (SE)Fd.f.P-valueEstimate (SE)
  1. See Table 2 and Methods for details of tests of random effects and variance components. The analysis at age 1 is based on 188 chicks from 45 nests of origin and 38 nests of rearing. The corresponding figures at age 4 are 182, 42 and 37, and at age 8 are 130, 29 and 24.

 Laying order1.592, 155 0.21 5.292, 114 0.006 8.212, 68.6<0.001 
 Yolk AOC2.191, 175 0.140.004 (0.002)4.661, 179 0.032 0.020 (0.009)0.851, 113 0.36 0.021 (0.023)
 Egg mass1481, 176<0.0010.722 (0.059)9.51, 157<0.001 1.096 (0.202)16.41, 103<0.001 1.84 (0.45)
 Chick age25.41, 179<0.0013.420 (0.679)9.471, 172<0.00117.95 (2.552)45.51, 124<0.00141.8 (6.21)

Begging rate recorded within 2 days after hatching was not affected by nest of origin or rearing and by egg features (details not reported). At hatching (age 1), body mass and tarsus length at hatching were not affected by parentage or nest of rearing or their interaction (Table 3). Yolk AOC significantly and positively predicted tarsus length whereas the effect of AOC on body mass was positive but not significant (Table 3). As expected, egg mass positively influenced both tarsus length and body mass (Table 3). However, laying order did not affect phenotype at hatching and controlling for egg mass, which showed a marked decline with increasing order of laying (Table 3; see also above).

At age 4, tarsus length and body mass were also not affected by nest of origin, rearing environment or their interaction (Table 3). The effect of yolk AOC was positive and significant on body mass whereas it was positive and marginally nonsignificant on tarsus length (Table 3). Both body mass and tarsus length positively covaried with original egg mass and were influenced by egg laying order. In fact, at Bonferroni post hoc tests, the tarsi of chicks from c-eggs were significantly shorter than those of chicks from a- or b-eggs (P < 0.007 in both comparisons) whereas chicks from a- and b-eggs did not differ significantly (P > 0.99). Body mass of chicks from c-eggs was smaller than that of chicks from b-eggs (P = 0.005). Body mass of chicks from a-eggs was not significantly different from that of chicks originating from b- or c-eggs (P > 0.16).

At the age of 8 days, body mass but not tarsus length significantly varied in relation to rearing environment (Table 3). AOC did not predict nestling phenotypic values, which significantly increased with mass of the original egg (Table 3). Laying order was also found to affect chick phenotype. Chicks from c-eggs were lighter and smaller than those from a- or b-eggs (post hoc Bonferroni tests; P < 0.005 in all comparisons) whereas no differences existed between chicks from a- and b-eggs (P > 0.99 for both variables).

Finally, T cell-mediated immune response varied among rearing environments (variance explained = 31.6%; inline image = 3.4, P = 0.032), but was not affected by nest of origin, the interaction between rearing and origin or any of the egg features (details not shown).

The same models presented in Table 3 for age 4 or 8 were also run while including AOC at age 4 or, respectively, at both ages 4 and 8. Plasma AOC did not predict nestling phenotype at both ages (details not shown). However, these analyses confirmed the positive effects of yolk AOC at age 4 on tarsus length (F[1,142] = 3.91, P = 0.050) and body mass (F[1,173] = 5.54, P = 0.020).


Early maternal effects mediated by egg quality are currently thought to be a major source of offspring phenotypic variation whereby mothers can adaptively adjust their allocation to reproduction in relation to the reproductive value of their current progeny, under the constraints imposed by intrinsic factors (e.g. health and general state) and ecological conditions. Such effects, while playing a potentially major role in the generation of phenotypic variation and in population evolutionary dynamics, can result in nongenetic phenotypic correlations between siblings and can therefore confound heritability estimates based on resemblances between genetic relatives (Mousseau & Fox, 1998).

In this study of the yellow-legged gull, we aimed at evaluating the genetic and environmental components of chick morphology, begging behaviour, immunity and antioxidant protection, while statistically partitioning out maternal influences mediated by egg mass and AOC of the yolk of the original egg. The main findings were that: (i) most of the variation in egg AOC and egg mass occurred at the among-clutches level; (ii) egg AOC declined with laying date and (iii) egg mass negatively predicted AOC whereas it positively predicted TAOC (AOC of the entire yolk, TAOC). In addition, (iv) time elapsed from laying to hatching of individual eggs was negatively predicted by yolk AOC; (v) maternal provisioning of antioxidants to eggs, as measured by AOC per unit volume of yolk, positively predicted chick size and mass in the critical early post-hatching period, independently of the effects of original egg mass, parentage or rearing environment and (vi) plasma antioxidant protection of the chicks at the age of 8 days varied with nest of rearing, after controlling for nest or origin and maternal influences mediated by egg mass and yolk AOC. Below we will discuss these main findings, together with other minor results.

Determinants of variation in egg antioxidant capacity and mass

Most of the variance in yolk AOC and egg mass was accounted for by the nest of origin. Previous studies of wild birds have also reported extensive variation in the concentration of specific antioxidants among clutches laid by different females (Saino et al., 2002; Blount et al., 2004; Verboven et al., 2005). Thus, mothers differ considerably in the level of antioxidant protection they provide to their developing offspring. The variation in egg mass is also consistent with the majority of the studies of birds, which indicate that egg size shows considerable variation among females (Christians, 2002).

The seasonal decline in yolk antioxidant capacity (both AOC and TAOC) and the concomitant decline in egg mass suggests that variation in egg quality among clutches could at least partly be due to a decline in mean phenotypic condition of laying females as the season progresses. Alternatively, seasonal variation in AOC and egg mass could reflect a seasonal shift in optimal decisions on allocation of antioxidants and resources to the eggs. In fact, high concentrations of particular antioxidants in the yolk enhance the resistance of the developing poultry embryos to thermal stress and may improve post-hatching performance (Sahin et al., 2002; Kucuk et al., 2003). Similarly, larger eggs that originate heavier chicks, have lower cooling rates than smaller eggs because of their lower surface-to-volume ratio (Kendeigh et al., 1977). Indeed, weather conditions, and air temperature in particular, can vary considerably in our study area over the study period (i.e. late March–late April). Thus, mothers that lay their eggs early in the season may deliver more antioxidants and resources to their eggs, compared with those that lay late, in order to buffer the negative effects of environmental stress on the developing embryo or young chick.

Seasonal variation in yolk AOC relative to overall variation observed in the population was considerably larger than variation in egg mass. This may suggests that phenotypic plasticity in yolk antioxidant transfer may be larger than plasticity in egg size because selection for high allocation of antioxidants to the eggs may be weaker than selection for production of large eggs, because overall egg size may be more influential on offspring performance than AOC.

Antioxidant capacity was significantly negatively predicted by egg mass, whereas TAOC increased with eggs mass. If high AOC per unit embryo (and chick) mass enhances chick performance, these results suggest that a trade-off may be operating between the advantages of laying large eggs, which produce large chicks, and the cost of relatively low AOC per unit chick mass. In addition, these findings may indicate that large eggs are more costly to mothers than small ones also because they are provided with a larger amount of antioxidants than smaller ones.

Antioxidant capacity did not vary with laying order of the eggs. Previous studies of the lesser black-backed gull (Larus fuscus) and of the barn swallow (Hirundo rustica) have shown that the concentration of maternal carotenoids in egg yolks declines with egg laying order (Blount et al., 2002; Saino et al., 2002). Concentration of selected antioxidants, such as carotenoids, may thus not reflect overall antioxidant protection per unit yolk volume that is transferred to the egg. The decline in TAOC with egg laying order was thus due to a decline in yolk mass with laying order and implies that the last eggs in a clutch require a smaller physiological effort by the mothers, in terms of allocation of antioxidants, compared with first eggs.

Time elapsed between laying of consecutive eggs in a clutch positively predicted change in mass (but not in AOC) between the first and the second egg in the pair. Female yellow-legged gulls therefore seem to experience a trade-off between laying consecutive eggs at short time intervals, thus reducing the risk of egg predation, and egg mass, which positively predicts chick mass. The mechanism that generates the positive relationship between the duration of inter-egg intervals and egg mass may depend on larger biosynthesis and accumulation of materials due to be delivered to the egg as the time interval between laying of consecutive eggs increases. The differential effect of inter-egg interval on egg mass and AOC may result from the fact that females cannot markedly affect the composition of the yolk of any given egg during the days immediately preceding laying of that particular egg because accumulation of yolk material (the rapid yolk development period) occurs during a period of several days before laying (Astheimer & Grau, 1990; Ruiz et al., 2000). Conversely, females may tune allocation of materials to the albumen, which is accumulated over a short time interval before laying (Ruiz et al., 2000), and larger inter-egg intervals may allow larger accumulation of albumen material.

Offspring phenotypic variation in relation to parentage, rearing environment and maternal influences mediated by egg antioxidants and mass

The duration of incubation was shorter for increasing values of yolk AOC, while taking into account the effects of nest of origin and nest of incubation. This result is in accordance with the observed positive effects on skeletal growth and body mass (see below), and suggests that yolk antioxidants may directly enhance embryonic development (see Surai, 2003), because the rearing environment made no significant contribution to the variance in duration of incubation.

Chick body mass and tarsus length showed significant variation among nests of rearing only at day 8 post-hatching. Nest of rearing effects at earlier ages, and the effects of parentage at all ages were nonsignificant, suggesting that these components of chick phenotypic variation were relatively small. Variation in plasma AOC at age 8, but not at age 4, showed a significant environmental variation, and may therefore depend on rearing conditions, possibly in terms of dietary intake and nest microhabitat, which can affect, for example, exposure to parasites. In addition, chicks hatched from first laid eggs had larger plasma AOC values at age 8 than chicks hatched from later laid eggs, independently of original egg mass and other potentially confounding variables, suggesting that variation in egg characteristics with laying order (other than yolk antioxidants or egg size, e.g. Royle et al., 2001) could influence the chick's plasma antioxidant protection during development. Laying order also affected body mass and tarsus length at ages 4 and 8 whereas controlling for the effect of egg mass, with chicks hatched from last laid eggs showing generally lower phenotypic values than other chicks. Thus, variation in egg quality with laying order seems to have persistent effects on a suite of chick traits.

We found positive effects of yolk AOC on tarsus length of the chicks at hatching and at day 4 after hatching (though the latter was marginally nonsignificant), and on body mass at day 4, whereas these effects vanished at the age of 8 days. These results were obtained while taking into account the effects of parentage and nest of rearing, and controlling for egg mass, which was also found to positively predict chick mass and size at all ages. Although this covariation does not imply a causal relationship, our results support the hypothesis that egg antioxidants are important mediators of early maternal effects in that they may enhance offspring phenotypic quality, at least soon after hatching, when mortality appears to be maximal (personal observation).

Egg AOC did not predict plasma AOC at day 4 post-hatching, but negatively predicted plasma AOC at age 8. The negative covariation between plasma AOC at day 8 and yolk AOC is intriguing, in that it suggests that chicks may face a trade-off between growth and plasma antioxidant protection. In fact, the positive association between yolk antioxidants and chick tarsus length and body mass at earlier ages corroborates the idea that chick antioxidants of maternal origin can influence growth, at least during the early post-hatching phase (see Surai, 2003), but perhaps at the expense of later plasma antioxidant protection. However, antioxidants in birds are mainly stored in organs such as the liver (Surai et al., 1999; Surai, 2003), and therefore variation in circulating antioxidants may be buffered or confounded by mobilization of these compounds from the liver.

Finally, we tested whether AOC in the original egg also predicted other chick traits, which can affect survival in the early post-hatching life, such as begging behaviour and immunity, but could find no evidence for such relationships. Among birds, specific yolk antioxidants (e.g. carotenoids) have been experimentally shown to enhance immune response (Blount et al., 2003; Saino et al., 2003). Therefore, specific antioxidants have immunostimulating or immunomodulating properties (Chew & Park, 2004), whereas TAOC, which reflects the action of diverse classes of antioxidant compounds, appears not to have consequences for the component of the acquired immunity we measured.

In conclusion, this study indicates for the first time in any wild bird population that variation in plasma antioxidant defences of the chicks has an environmental component. Furthermore, yolk AOC predicted duration of incubation and growth of the chicks in the early post-hatching stages independently of egg mass and rearing environment, suggesting that allocation of antioxidants to the eggs is an important form of early maternal effect in avian species.