Nicola Saino, Dipartimento di Biologia, Università degli Studi di Milano, Via Celoria 26, I-20133 Milano, Italy. Tel.: +39 0250314808; fax: +39 0250314713; e-mail: firstname.lastname@example.org
Egg quality may mediate maternal allocation strategies according to progeny sex. In vertebrates, carotenoids have important physiological roles during embryonic and post-natal life, but the consequences of variation in yolk carotenoids for offspring phenotype in oviparous species are largely unknown. In yellow-legged gulls, yolk carotenoids did not vary with embryo sex in combination with egg laying date, order and mass. Yolk lutein supplementation enhanced the growth of sons from first eggs but depressed that of sons from last eggs, enhanced survival of daughters late in the season, and promoted immunity of male chicks and chicks from small eggs. Lack of variation in egg carotenoids in relation to sex and egg features, and the contrasting effects of lutein on sons and daughters, do not support the hypothesis of optimal sex-related egg carotenoid allocation. Carotenoids transferred to the eggs may rather result from a trade-off between opposing effects on sons or daughters.
Variation in the phenotype of the mother and in the environment she experiences can influence offspring phenotype beyond maternal contribution to the genetic constitution of her progeny (Mousseau & Fox, 1998a, b). Such nongenetic maternal effects may themselves be under the control of maternal genes, and are therefore expected to evolve under selective pressures acting both on the mother and the offspring (Mousseau & Fox, 1998a; Wolf et al., 1998; Parker et al., 2002; Müller et al., 2007). In oviparous animals that produce broods with more than one offspring, simultaneous growth of different ova may hinder the capacity of mothers to allocate optimal amounts of constituents to individual eggs (see Challenger et al., 2001; Young & Badyaev, 2004; Badyaev et al., 2006a), so that a mother may have to trade the resources transferred to one egg against those allocated to other simultaneously growing ova, or to her own maintenance.
Date of laying of the eggs and their position in the laying sequence constitute predictable gradients of rearing conditions (Badyaev et al., 2002) in several taxa, including birds (Lack, 1968; Perrins, 1970; Parsons, 1975a; Daan et al., 1990; Magrath, 1990; Ambrosini et al., 2006). Laying order in birds often parallels hatching sequence, and therefore predicts harshness of sib–sib competition, particularly for last-laid offspring (e.g. Parsons, 1975b; Royle & Hamer, 1998; Krebs, 1999). Predictable variation in rearing conditions related to seasonal effects or laying order (Badyaev et al., 2002) paves the way for the evolution of adaptive allocation of maternal effects. However, if sons and daughters differ in their growth requirements or susceptibility to post-natal conditions, sex and laying order may interact in determining optimal allocation strategies to the ova.
Egg carotenoids concentration may be predicted to covary with embryo sex because physiological carotenoid demands differ between male and female offspring, and mothers may discriminate between sons and daughters according to their susceptibility to parasites or other ecological factors (Tschirren et al., 2003), competitive ability (see Uller, 2006) or paternal quality (Burley, 1981). Badyaev et al. (2006b) showed that concentration of egg carotenoids varied with embryo sex according to different patterns in distant house finch (Carpodacus mexicanus) populations (see also Verboven et al., 2005; Bogdanova et al., 2006 for other studies showing sex-related variation; but see Saino et al., 2003a).
Experiments where carotenoid availability to mothers was manipulated (see above) allow to test how variation in dietary carotenoids available for transfer to the eggs, in combination with carotenoid effects on maternal physiology, affects offspring performance. However, if the question at hand is whether egg carotenoid content per se influences chick phenotype, an alternative and perhaps preferable approach is to manipulate egg carotenoid concentration directly, thus avoiding the confounding effects of dietary carotenoids on maternal state (Surai et al., 1998; Blount, 2004). The consequences of variation in the concentration of maternal substances have been analysed by inoculation into the yolk in several studies (Eising et al., 2001; Saino et al., 2003b, 2006; Müller et al., 2005; Navara et al., 2006; Rubolini et al., 2006a, 2007) but, to the best of our knowledge, only one study adopted this approach to manipulate yolk carotenoids (Saino et al., 2003b).
Variation in egg composition in relation to the sex of the embryo, and to laying date and order, may provide evidence for maternal adaptive allocation strategies, although such variation could also be a by-product of maternal physiology or ecological constraints rather than the expression of an adaptive strategy. Lack of variation in maternal allocation may also reflect an adaptive strategy if offspring of different sex or laying order do not differ in requirements for the specific material under scrutiny. In this latter case, supplementation of the eggs should result in similar (either positive or negative) consequences for the phenotypic quality of the offspring independently of other factors, including sex or laying order. However, contrasting effects of egg supplementation on chick phenotype and survival in relation to offspring sex or to the combined effects of sex and egg laying order, date or mass would not lend support to the idea of optimal maternal allocation of carotenoids to the eggs. This is the case because mothers would not adopt an allocation pattern of the same amount of carotenoids that could enhance the phenotypic quality of all the offspring.
In the first part of this study, we analyse variation in the concentration of carotenoids in yellow-legged gull (Larus michahellis) eggs in relation to embryo sex, laying order and date. In a previous analysis of yolk carotenoids concentration where the effect of sex was not considered (see Saino et al., 2008a), lutein concentration positively predicted tarsus length and body mass of the chicks up to 8 days after hatching. Zeaxanthin and dehydrolutein were found to have smaller and mostly nonsignificant predictive power of chick phenotype. Lutein concentration was also found to decline with laying date and order, and increased with egg mass.
We then performed an experiment where we supplemented freshly laid eggs with physiological doses of lutein (the most abundant carotenoid in our model species; Saino et al., 2008a), and analysed the effect of supplementation on chick phenotype and survival in relation to sex, laying date, position in the laying sequence and mass of the original egg. To the best of our knowledge, no published information exists on variation in the effects of carotenoid supplementation at any stage of the life cycle in relation to the combined effects of sex, laying date and order, and egg mass. Predictions about the consequences of lutein injection in the eggs in relation to these factors would therefore be speculative and based on largely untested assumptions.
Study species and field procedures
The yellow-legged gull is a semicolonial, monogamous species inhabiting mostly coastal habitats across the Mediterranean (Liebers et al., 2001). Clutch size ranges between 2 and 3 (rarely 1 or 4), with a modal size of 3 in our study population. Eggs are laid at 1–4 (most frequently 2–3) days intervals. Hatching occurs 25–32 days after laying of individual eggs, and is markedly asynchronous. The altricial, nidifugous chicks are fed by both parents and fledge at 35–40 days of age (Cramp, 1998). Both parents and chicks are euriphagous.
This study was carried out in the Comacchio lagoon (NE Italy, 44°20′N–12°11′E), at a colony comprising more than 400 breeding pairs, during March–June 2004, 2005 and 2007. In 2004 and 2005, we analysed variation in the concentration of carotenoids in relation to embryo sex in a nondestructive way (see Rubolini et al., 2006b; Saino et al., 2008a). In 2007, we performed a supplementation experiment by injecting lutein in the eggs. Date of laying of the first egg in the clutches we considered ranged between 16 March and 19 April. We did not consider the nests where re-laying occurred after all the eggs had disappeared, to reduce the risk of including replacement clutches. However, in the experimental study, different treatments were applied to eggs from the same clutch (see below), so that the results were probably not confounded by possible inclusion of replacement clutches.
The field methods for the correlational study (2004, 2005) of egg quality in relation to embryo sex are reported in Rubolini et al. (2006b) and Saino et al. (2008a) and we will only briefly summarize them here. We visited the colony daily or every second day. On the day when a new egg was found, it was marked with a water-proof marker, temporarily removed from the nest and taken to a nearby building for the extraction of yolk and albumen bioptic samples. The removed egg was replaced with a ‘dummy’ egg to avoid interference with parental behaviour. The egg was then carried back to its nest of origin within a few hours after removal. Chicks were assigned to their original egg by inoculating a drop of blue or green food dye in the pipping egg. A small sample of blood was taken for molecular sexing within the second day after hatching.
In the year when the lutein injection experiment was performed, for practical reasons before the onset of hatching, the colony was subdivided into two sectors that were visited alternately every second day. Each nest in a sector, however, was visited repeatedly during the same day to mark freshly laid eggs according to laying order. New nests were marked with a labelled stick, and the newly laid eggs with a water-proof marker. In six cases, two newly laid eggs were found at two consecutive checks of the same nest during a particular day, suggesting that the first egg we found had been laid in the preceding day. These eggs were given their real laying order but the same laying date for homogeneity with the other nests where eggs were assigned the laying date corresponding to the day when they were found. As nests were checked every second day and the newly laid eggs were injected on the same day when found, a maximum of 2 days elapsed between laying of an egg and the time when it was injected. On the day when a new egg was found, it was temporarily removed from the nest, replaced with a ‘dummy’ gull egg, and taken to a nearby building for treatment. Before injection, the eggs were weighed (nearest g) and kept with their acute pole pointing upwards for at least 15 min. A drill was opened through the eggshell close to the acute pole by means of a sterile pin after having carefully disinfected the eggshell region where the hole was to be drilled. Injections were made using a 1-mL sterile syringe mounting a 0.6 × 30 mm needle while the egg was held firmly with its longitudinal axis vertical. Immediately after injection, the hole was sealed with a minute drop of epoxidic glue and a small piece of eggshell superimposed to the hole. This procedure and needle length were chosen because in preliminary trials on 40 eggs using exactly the same procedure and needle length as those used for injection, we checked that the needle had reached the yolk by sucking the egg content into the syringe. The eggs were then taken back to their nest within 4 h, and the dummy eggs were removed.
In order to test the effect of an increase in the concentration of carotenoids in the yolk, we decided to inject lutein. This decision was justified by the fact that lutein is the most abundant carotenoid in yellow-legged gull eggs (Saino et al., 2008a). In addition, in a previous correlational study, we found that egg lutein concentration has stronger predictive power of chick phenotype than the other carotenoids, suggesting that it was a good candidate as a potential mediator of maternal effects (Saino et al., 2008a). Moreover, injection of a mixture of the carotenoids that were found in the eggs of this population would have been difficult to design because eggs vary extensively in both relative and absolute concentrations of carotenoids (Saino et al., 2008a). The clutches were assigned sequentially, according to the order in which they were started, to the following treatment schemes (nest, a-, b-, c-egg): nest 1, lutein injection (L), control injection (C), L; nest 2, C-L-C; nest 3, L-C-C; nest 4, C-L-L; and so forth with the following nests.
We aimed at increasing the concentration of lutein by 1 SD of the concentration recorded in the egg yolk. In order to decide the absolute amount of lutein to be injected in each individual egg, we had to take variation in total egg and thus yolk mass into account. Based on previous data on total egg and yolk mass from the same study colony, we knew that the yolk mass increases with total egg mass according to the following equation: yolk mass = 0.207 (0.026 SE) egg mass + 3.875 (2.228 SE); F1,46 = 64.57, P <0.001, r2 = 0.58). Because we decided to increase the concentration of lutein by 1 SD and the variance in lutein concentration could vary with egg features (e.g. mass), we tested if this was the case. Variation of the variance in lutein concentration in relation to egg mass was analysed by grouping the eggs in four 10-g wide classes of egg mass, encompassing the entire range of variation of egg masses in our study population. The variance in lutein concentration varied with egg (and thus yolk) mass, as determined by Levene test (P <0.05). We therefore estimated yolk mass based on the observed relationship with egg mass and injected an amount of lutein equal to that needed to increase the concentration by 1 SD of the concentration recorded within each class of egg mass. This procedure ensured that the lutein concentration in the yolk of each egg was increased by approximately 1 SD of the concentration recorded in the eggs of that particular, relatively narrow class of mass. The amount of lutein injected in the four classes of egg mass was (class of egg mass in grams: amount of lutein injected in micrograms): 66–75: 75; 76–85: 100; 86–95: 155; 96–105: 210. Lutein (FloraGLO® Lutein 20% Liquid in Safflower oil; Kemin Foods, Des Moines, Iowa) was diluted in safflower oil. We decided to inject the same total volume (=30 μL) of lutein plus control solutions in all eggs and therefore used four different lutein solutions at different concentrations. Control eggs were injected with safflower oil alone, and were otherwise subjected to the same manipulations as the eggs injected with lutein. It should be emphasized that assuming a normal frequency distribution of the concentration of lutein in the yolk, by injecting 1 SD of lutein content of the eggs, it can be estimated that approximately four of the chicks originating from lutein-injected eggs that reached age 4 days had a post-injection concentration of lutein in their egg exceeding the mean + 3 SD. We are therefore confident that our results refer to the effects of a physiological increase in egg lutein concentration. Modulation of the dose according to laying order and date, conversely, was not necessary because the variance in lutein concentration did not vary according to these factors (details not shown). It should also be stressed that 30 μL safflower oil contain approximately 0.02 SD of the mean content of vitamin E recorded in a sample of yellow-legged gulls (our unpublished data) and no selenium, suggesting that injection of the vehicle had a negligible effect in terms of supplementation of other important antioxidants (Surai, 2003).
All the nests where laying had been completed were visited every second day during the central part of the incubation period, and every day well before the expected hatching date to check for any sign of imminent hatching. Chicks were assigned to their original egg by injecting in the pipping egg a small drop of food dye (see above). Upon hatching, chicks were individually marked with elastic coloured plastic bands on both tarsi to allow individual identification.
For the purposes of this study, we measured egg mass upon injection (i.e. around laying), and chick body mass (to the nearest 1 g; measures expressed in grams) and tarsus length (to the nearest 0.1 mm; measures expressed in millimetres) at 4 and 8 days after hatching. We decided not to consider the growth of older chicks because maternal carotenoids derived from yolk are known to be important in determining chick carotenoid status mainly up to 1 week after hatching (Koutsos et al., 2003; Karadas et al., 2005). In addition, at older ages, vagile gull chicks can move to considerable distances from their nest and settle in areas with dense herbaceous vegetation where they can hardly be found. Thus, analyses of survival were restricted to age 8 days. In the analyses where we tested for the effects of egg treatment on survival, we only considered chicks that were not killed by rats, as determined by inspection of the carcasses that could be found, because we assumed that rat predation was independent of lutein treatment.
Based on our long-term experience of yellow-legged gull chicks (e.g. Rubolini et al., 2005), we are confident that we could identify whether a chick had hatched during the day of the visit or on the day before, after we had visited the nest, because recently hatched chicks retain a ‘wet’ dawn and are considerably less vagile than 1-day-old chicks. Exact age in days could therefore be assigned to each chick at the time when it was measured, and this was included as a covariate in the analyses run on chick phenotypic traits at ‘day 4’ or ‘day 8’. At the age of 8 days, we also performed an in vivo immune test [the phytohaemagglutinin (PHA), skin test], according to a standard protocol (Saino et al., 1997; Tella et al., 2002) (see also Saino et al., 2008a). The swelling response to subcutaneous injection of PHA (expressed in mm × 102) was assumed to represent a reliable indicator of the cell-mediated immunity (see Martin et al., 2006). The values of the wing web swelling index were highly repeatable within individuals, as determined by measuring twice the wing web of 11 individuals (F11,24 = 33.58, P <0.0001, r2 = 0.97). Blood samples were collected in capillary tubes by puncturing the ulnar vein on day 0 or 1 to perform molecular sexing (Rubolini et al., 2006b).
In the analyses, small differences in sample sizes among phenotypic traits within treatment by sex groups are mainly due to accidental reasons or, in the case of the wing web index (WWI), to failure in finding the chick the day after the immune challenge in order to measure wing web swelling.
We mainly relied on generalized linear mixed-effects models (GLMMs) as implemented by sas 9.0 or 9.1 (SAS Institute Inc., Cary, NC, USA) and spss 13.0 (SPSS Inc., Chicago, MA, USA) to analyse chick phenotypic traits and survival. An identity link function and normal error distribution was assumed in the analysis of phenotypic traits, and a logit link function and a binomial error distribution were assumed in the analysis of survival. In all analyses, nest was included as a random factor. Egg treatment and sex of the chick were considered as fixed-effect factors, and laying date of the original egg (1 = 1 January), laying order and egg mass (grams) were included as covariates. In the models, we initially included all the main effects together with the two- and three-way interactions between the factors and each of the covariates. Age (in days from hatching) was included as a covariate to account for small variation (± 1 days) in actual age with respect to the planned age of measurement (see above). Brood size was included to account for differences in sib-competition between broods of different size. Analyses at age 4 were run while entering brood size recorded at hatching and were re-run including brood size at age 4. Similarly, analyses at age 8 were run while including brood size at age 4 or brood size at age 8. The models obtained while including current brood size or brood size recorded at previous inspection of the nest gave qualitatively similar results (see Results). The models were subjected to a step-down simplification procedure whereby at each step, we excluded the term that had the largest associated P-value. Higher-order interactions were removed before the relevant lower-order interactions and main effects. The simplification procedure ended when all the terms that were still in the model were significant, or nonsignificant terms were involved in interaction terms that were retained in the model. When a significant three-way interaction between sex, treatment and a covariate emerged, the significance of the pairwise differences between the slopes of the sex by treatment groups was tested for lutein males vs. control males, lutein females vs. control females, lutein males vs. lutein females, and control males vs. control females by fitting an ‘equality of slopes’ model (see Littell et al., 1996). The significance of the differences between the groups was tested by computing the least square means (LSMs) at values of the covariate (for laying date and egg mass) equal to x − 2SD, x and x +2SD, where x is the mean value of the covariate recorded in the population and SD is the standard deviation (see also Littell et al., 1996). The mean (SD) values were = 96.65 (5.28) days after 1 January for laying date and 87.05 (6.44) g for egg mass. The tests for laying order were run at values of the covariate of 1, 2 or 3 corresponding to a-, b- or c-eggs. The significance of the deviation from 0 of the slope of the relationship between the phenotypic variables and the covariates for each sex, treatment or sex by treatment group was tested in separate GLMMs, including all the significant covariates that were included in the step-down model obtained from all data pooled. The effect of nest (random factor) was retained in all models. For brevity, the tests of the effect of nest are not reported in the tables. However, all chick phenotypic traits varied significantly among nests (P <0.05 in all cases). Estimates of statistical parameters are given with their SE in parentheses.
Covariation between concentration of carotenoids and embryo sex
In 2004 and 2005, we extracted a sample of yolk from freshly laid eggs to investigate the covariation between carotenoid concentration and egg features (egg laying date, order and mass). The results from these analyses are reported in a published paper where, however, the effect of sex of the embryo was not considered (Saino et al., 2008a). Here, we reanalyse those data to test for a covariation of carotenoid concentration in relation to sex and its combination with egg features, using results of molecular sexing analyses run in 2007 on the samples collected in 2004 and 2005.
The variance in yolk carotenoid concentrations did not differ between eggs with a male or a female embryo (Levene test, all P-values > 0.23). In mixed models with nest as random factor, we found no evidence for variation in the concentration of lutein, zeaxanthin or dehydrolutein (log-transformed data, see Saino et al., 2008a) in relation to the combined effects of sex and egg laying date, order or mass. In fact, interaction terms between sex and egg features were invariably excluded from the models after step-down selection of nonsignificant predictors (F-values for the effects of the interactions associated to P >0.29 for lutein, P >0.18 for zeaxanthin and P >0.28 for dehydrolutein; all analyses run on 113 eggs from 76 clutches with a proportion of males of 0.59). In step-down models, the main effect of sex was also found to be far from statistical significance (lutein: F1,96.3 = 0.004, P =0.95; zeaxanthin: F1,106.0 = 0.06, P =0.81; dehydrolutein: F1,111 = 1.04, P =0.31).
Sex ratio variation according to laying date and order
In 2007, we considered 190 clutches and broods that included at least one hatched chick. Mean clutch size was 2.87 (0.03; range 1–3) eggs. Among the 270 control eggs (n =96 a-, 91 b-, 83 c-eggs), 199 (=73.7%) hatched. Among the 272 lutein eggs (n =93 a-, 95 b-, 84 c-eggs), 195 (=71.7%) hatched. Among the 394 chicks that hatched, 372 (=94.4%) could be sexed. 22 chicks could not be sexed either because they were preyed upon before we could collect any blood or due to accidental loss of the blood sample. The sex ratio [M/(M + F)] was 0.513, which did not differ significantly from 0.50 (χ21 = 0.27, P >0.50). We tested whether laying date, laying order and egg mass predicted the sex of the hatchlings in a GLMM assuming a logit link-function and a binomial error distribution where sex was included as the two-state-dependent variable, nest was considered as a random factor and all the main effects and two-way interactions among the covariates were included as predictors. After step-down simplification of the model (see Methods), sex (coded as 0: male; 1: female) was significantly predicted by the two-way interaction between laying date and laying order [F1,368 = 6.48, P =0.011, slope = 0.072 (0.028)], but not by egg treatment or its interaction with other predictors (see Table S1 in Supporting Information). Figure 1 shows the predicted probabilities of chicks being male in relation to the combined effects of laying order and laying date. The probability that an a-egg produced a male steadily increased during the breeding season whereas the opposite was the case for c-eggs. The probability that chicks from b-eggs were male did not vary as the season progressed. In this analysis, the random effect of nest was nonsignificant (z =0.05, P =0.48), implying that there was no significant variation in the sex ratio among broods.
Body mass variation in relation to egg treatment and sex
The sample size for the chicks assigned to the four treatment by sex groups (control males; control females; lutein males; lutein females) and included in the analyses of body mass were: age 4: 91; 73; 71; 89 (total = 324); age 8: 70; 64; 59; 74 (total = 267). For tarsus length, sample sizes at age 4 were: 91; 75; 72; 88 (total = 326), while they were the same as for body mass at age 8. Sample sizes for immune response were: 67; 60; 55; 70 (total = 252). The numbers of chicks that did not survive until the age of 8 (listed as above) were: 35; 18; 25; 25 (total = 103). The percentages of chicks that survived until the age of 8 thus were: 66.7; 78.0; 70.2; 74.7 (overall = 72.2%).
There was an overall variation in the slopes of the relationships between body mass at age 4 and laying order in the four treatment by sex groups, as indicated by the significant three-way interaction (Table 1). Lutein males from a-eggs were significantly heavier than control males and lutein females (Fig. 2). No differences existed among the chicks of the four groups originating from b-eggs, whereas control males were larger than control females from c-eggs (Fig. 2). The relationship between body mass and laying order was significantly more steeply negative for lutein than control males (t273 = 3.18, P =0.002) and for control females than control males (t274 = 2.67, P =0.008), whereas the difference between the slopes was marginally nonsignificant between lutein males and lutein females (t266 = 1.94, P =0.053) being more steeply negative in males than females (see Fig. 2). The difference in the slopes for lutein and control females was nonsignificant (t294 = 1.41, P =0.16). In this analysis, body mass increased with egg mass [slope = 1.270 (0.200)], and declined with laying date [slope = −0.592 (0.250)] after controlling for concomitant effects (Table 1).
Table 1. Step-down generalized linear mixed-effects model of body mass and tarsus length of chicks at 4 and 8 days.
Original models included the effects of egg lutein treatment, sex, laying order, laying date, preincubation egg mass, brood size, age at measurement and interaction terms (see Methods). Terms that were excluded from all models are not listed. See text for parameter estimates of covariates and interaction terms.
Treatment × sex
Sex × laying date
Treatment × laying order
Sex × laying order
Treatment × sex × laying order
Treatment × sex
Treatment × laying order
Sex × laying order
Treatment × sex × laying order
Separate analyses of covariance on the four treatment by sex groups (see Statistical analyses) showed that body mass at age 4 significantly declined with laying order in chicks of both sexes originating from lutein eggs, and in control female chicks but not in male ones [slope for control males = −1.260 (3.368), t51.9 = 0.37, P =0.71; control females = −15.158 (2.911), t46.2 = 4.14, P <0.001; lutein males = −12.421 (2.698), t64.8 = 4.60, P <0.001; lutein females: −8.709 (3.122), t77.8 = 2.79, P =0.007] while controlling for the concomitant significant effects of egg mass and laying date.
At age 8, body mass of lutein chicks did not differ from that of control chicks independently of sex (Table 1). Body mass of male but not female chicks declined according to laying date of their original egg [slope for females = −0.890 (0.686), t96.6 = 1.30, P =0.20; males = −2.546 (0.782), t88.4 = 3.26, P =0.002], resulting in a significant difference in seasonal variation in body mass between the sexes (interaction term in Table 1). Chick mass increased with egg mass [slope = 2.241 (0.464)] and declined with laying order [slope = −14.100 (3.558)] (Table 1).
Tarsus length variation in relation to egg treatment and sex
The pattern of variation of tarsus length at age 4 in relation to sex, experimental treatment and the covariates closely matched that observed for body mass (Table 1). In particular, tarsus length covaried with laying order according to the combined effects of sex and treatment (Table 1). Tarsus length of control males declined with laying order significantly less steeply than tarsus length of control females (t283 = 2.56, P =0.011) and lutein males (t284 = 2.80, P =0.006), whereas there was no significant difference in the slopes between lutein males and females (t281 = 0.88, P =0.379) or between control and lutein females (t299 = 0.62, P =0.53). Pairwise differences between groups within positions in the laying sequence were significant between control and lutein males, control males being significantly smaller than lutein males when originating from a-eggs, whereas being significantly larger when originating from c-eggs (Fig. 2). In addition, among control chicks from c-eggs, males were larger than females. Tarsus length at age 4 was also significantly and positively predicted by egg mass [slope = 0.144 (0.020); Table 1].
In separate analyses on each treatment by sex group, there was no significant variation of tarsus length with laying order among control males [slope = −0.251 (0.263), t22 = 0.95, P =0.35], whereas tarsus got significantly shorter with increasing laying order in all the other groups [control females: slope = −1.290 (0.310), t47.5 = 4.17; lutein males: slope = −1.283 (0.273), t67.2 = 4.71; lutein females: slope = −0.980 (0.280), t84 = 3.29; P <0.001 in all cases].
The step-down GLMM indicated that the slopes of the relationships between tarsus length at age 8 and laying order differed between control and lutein chicks (Table 1). Both relationships were significantly negative [controls: slope = −1.006 (0.469), t118.5 = 2.15, P =0.034; lutein chicks: slope = −2.046 (0.382), t122.1 = 5.36, P <0.001]; hence, tarsus length declined with laying order in both experimental groups after controlling for laying date and egg mass, but the relationship was more steeply negative for lutein than control chicks. Pairwise comparisons between LSMs for the two groups at the three laying positions showed only nonsignificant trends for tarsus being shorter in control compared with lutein chicks from a-eggs [LSM for controls: 41.984 (0.435); lutein chicks: 42.884 (0.449); t201 = 1.61, P =0.11] and an opposite trend for chicks from c-eggs [LSM for controls: 40.124 (0.576); lutein chicks: 38.841 (0.490); t250 = 1.82, P =0.070]. This difference was far from significant among chicks from b-eggs [LSM for controls: 41.054 (0.300); lutein chicks: 40.863 (0.293); t194 = 0.52, P =0.61]. Thus, although the slopes of the relationships between tarsus length at age 8 and laying order significantly differed between control and lutein chicks, it cannot be concluded that these differences translated into significant differences in this body size index at any of the laying positions.
Tarsus length at day 8 declined with laying date [slope = −0.098 (0.045)] and increased with egg mass [slope = 0.193 (0.038)] (Table 1). A main effect of sex on tarsus length existed at age 8 (Table 1), with males having longer tarsi than females [LSM for males = 41.676 (0.302); females = 40.605 (0.312)]. Thus, sexual dimorphism in body size, as indicated by tarsus length, arises as early as age 8 in this species of gull.
In the analyses of body mass and tarsus length at both ages, brood size was never found to significantly predict chick phenotype (P-value associated with the effect of brood size at exclusion from the step-down models always > 0.19).
Variation in immune response in relation to egg treatment and sex
The wing web swelling response to PHA injection (WWI) showed an opposite pattern of covariation with original egg mass in the two experimental groups, as it increased with egg mass among controls, whereas it declined among lutein chicks (Table 2; Fig. 3). At an egg mass equal to the x − 2SD, WWI was significantly smaller in control than lutein chicks (P =0.012), whereas the reverse was true at egg mass equal to x +2SD (P =0.009). In this analysis, the effect of the interaction between sex and treatment was marginally significant (P =0.037; see Table 2) and arose because females had larger WWI than males among controls whereas the opposite was the case among lutein chicks [LSM for: control males = 88.44 (4.457); control females = 101.12 (4.662); lutein males = 97.18 (4.894); lutein females = 90.86 (4.335)]. However, the only significant pairwise contrast was that between control males and females (t235 = 1.99, P =0.048). WWI increased with body mass of the chicks, which was included because this index of immune response has been shown to covary with body mass (e.g. Moreno et al., 2005), whereas the effect of brood size at age 8 was negative and marginally nonsignificant (Table 2). Finally, WWI increased with date of laying of the original egg (Table 2). The same analysis run while including brood size at day 4 rather than 8 led to slightly different results, in that the effect of the interaction between lutein treatment and sex turned to marginally nonsignificant (F1,237 = 3.69, P =0.056) whereas the negative effect of brood size turned to highly significantly negative (Table 2). The results from this analysis for the other terms reported in Table 2 were qualitatively similar and are not reported in details. Taken together, these results thus suggest that the effect of lutein supplementation in the eggs on immune response depended on the sex of the chick, being larger in males, decreased with brood size and positively covaried with laying date and body mass of the chicks.
Table 2. Step-down generalized linear mixed-effects model of an index of immune response – the wing web swelling response to phytohaemagglutinin injection – recorded at the age of 9 days after hatching.
The original model included the effects of egg lutein treatment, sex, laying order, laying date, preincubation egg mass, brood size, age at measurement and interaction terms (see Methods). Terms that were excluded are not listed.
†Slope = 0.112 (0.052).
‡Slope = −6.731 (3.507).
*The analysis based on brood size at age 4 showed a highly significant negative effect [F1,156 = 7.54, P =0.007, slope = −8.118 (3.463)].
Treatment × sex
Treatment × egg mass
Body mass (age 8)†
Brood size (age 8)‡*
Chick survival at 8 days was analysed in a GLMM assuming a logit link-function and binomial error distribution. After step-down simplification of the initial model, we found a significant variation of the effect of egg treatment with sex and laying date of the original egg (F1,337 = 4.53, P =0.034; see Table S2 in Supporting information for the full final model). Although survival declined with laying date among control chicks of both sexes and among lutein males, survival of lutein females showed an opposite pattern (Fig. 4). Among individuals that hatched from eggs laid late in the season, lutein females were more likely to survive than lutein males (difference estimated at laying date = x + 2SD; P =0.026). Early in the season (laying date = x−2SD), lutein females had smaller survival than control females (P =0.044). There was a significant difference in the slope of the relationships between survival and laying date in male and female lutein chicks (t337 = 1.98, P =0.048), and a marginally nonsignificant difference between the slopes for control and lutein females (t337 = 1.88, P =0.061). The comparisons between control males and females, and between control and lutein males, gave nonsignificant results as well (P >0.23 in both cases).
The two main aims of our study were: (i) to test whether carotenoid concentration in the egg yolk of yellow-legged gulls varied according to sex, either per se or in combination with variation in laying date, laying order and egg mass; and (ii) to investigate the effects of a physiological increase in the concentration of egg lutein on the phenotype and survival of male and female chicks, in relation to egg features. We found no evidence that the sex of the embryo, per se or in combination with egg features, did predict carotenoid concentration of the egg. However, we found complex effects of supplementation of the eggs with lutein on chick phenotype and survival, depending on egg features, either per se or in combination with the effect of sex. Such effects are summarized below.
At the age of 4 days, the effects of egg treatment on body mass and size varied according to the combined effects of sex and laying order. After controlling for other sources of variation, control males were unaffected by laying order whereas control females had reduced mass and tarsus length as laying order increased. Among males, those from lutein-supplemented eggs had larger mass and longer tarsi than controls when originating from a-eggs, whereas they had shorter tarsi when originating from c-eggs. The decline of tarsus length with laying order at age 8 was unaffected by sex but was more steeply negative among lutein than control chicks, similarly to the differences in the trends at age 4 between treatments independently of sex. The effect of treatment on immune response at age 8 differed between sexes, being relatively larger in males than females, and declined with original egg mass, so that lutein chicks from relatively small eggs had higher immune response than controls and an opposite pattern existed among chicks from large eggs. Brood size did not predict body mass or size but negatively predicted immune response. Finally, survival of the chicks till age 8 also varied according to the effects of treatment and sex, in combination with laying date of the egg. Survival of control chicks and male lutein chicks decreased late in the season whereas it increased among lutein females.
The only major fitness component we could measure directly was survival in the post-hatching period. We found that egg treatment had a beneficial effect on females, compared to males, late in the season. Phenotypic variables could be measured on a sample of more than 250 chicks up to day 8 but not at later ages, because of practical limitations (see Methods). However, in a previous study of the same population (see Rubolini et al., 2005), we found that tarsus length around age 4 strongly and positively predicts tarsus length around age 20 days (partial correlation while controlling for exact age at measurements, rpar = 0.74, P <0.001, d.f. = 50, n = 54). In addition, mean tarsus length we recorded at 20 days of age [58.6 (5.20) mm; n = 54] was similar to that for Larus argentatus michahellis adults (unweighted mean of males and females = 62.4) reported by Cramp (1998), and not statistically different (z =0.099, P =0.921). This suggests that tarsus length at age 4 predicts tarsus length in fully grown individuals and, hence, that the effect of egg treatment in the present study could have consequences also for final body size attained by the chicks. In addition, immune response to PHA injection was found to predict survival in a diverse set of species (Møller & Saino, 2004; Cichon & Dubiec, 2005; Moreno et al., 2005), suggesting that the effects of egg treatment on immune response could have consequences for subsequent survival. Admittedly, however, we have no direct information on the long-term fitness consequences of the effects of lutein supplementation on chick morphology and immune response, which could hardly be obtained in this species with high post-natal dispersal and delayed sexual maturity (age at first breeding spans from 3 to 7 years) (Cramp, 1998). Thus, the experimental effects of lutein on chick phenotype could also have minimal or even no consequences for offspring fitness, implying that modulatory effects of lutein on chick growth and immunity depending on sex and other factors may not select for differential allocation strategies to the eggs. In addition, our conclusions obviously apply only to the specific set of characters and the post-hatching period we considered, and therefore provide only partial information on the effects of lutein on important traits. However, we showed differential effects of yolk lutein on the survival of sons and daughters depending on laying date, which indicates that lutein supplementation impacted offspring fitness.
We hypothesized that, under the assumption that mothers could adaptively adjust the amount of lutein they deliver to their growing ova, the natural pattern of variation in lutein concentration among eggs of different combinations of sex, laying order, laying date and mass should result in better offspring quality compared with alternative allocation patterns. If dietary carotenoids are limiting, experimental provisioning of lutein to the eggs should result in a generalized increase in offspring quality, independently of sex and original egg features. Our results markedly differ from this pattern. In fact, we found that chicks with different combinations of sex and egg laying order, date or mass responded differently to lutein supplementation. Notably, lutein supplementation did not have an invariably positive effect on chick phenotype and survival. Rather, lutein supplementation had positive effects for some combinations of sex and egg features, whereas the effect was reversed for other combinations. Hence, mothers could have benefited in terms of offspring quality and survival from reducing the amount of lutein allocated to particular eggs, and transferring the amount of lutein thus saved to other eggs, depending on laying order, date or mass. Because mothers seem not to be able to modulate lutein concentration in their eggs in relation to sex and other egg features, and because variation in lutein concentration has contrasting effects in relation to embryo sex and egg features, females of our model species may have to trade the benefits arising from larger yolk lutein concentration to particular eggs against the negative effects for other eggs. Obviously, testing the hypothesis of optimization of carotenoid allocation in this complex scenario is difficult, partly because of the intervention of potentially confounding effects. For example, hatching failures, which are quite frequent in the population we studied and were increased by injection procedures, and extensive chick mortality may have partly confounded the observed effects of laying order by influencing the number of competing siblings and the actual rank of the chicks within each brood. However, we partly controlled for this effect by including brood size in the models (see also below).
At a different level, matching of egg quality to embryo sex could have occurred if the patterns of allocation to the production of male and female offspring paralleled those of variation in egg lutein concentration. We have shown that mean concentration of lutein in the eggs declined during the breeding season and in c- compared to a- and b-eggs (Saino et al., 2008a). Mothers could have therefore produced more chicks of the sex that was favoured by hatching from an egg of a particular laying order or date. However, we found no evidence that this was the case because the proportion of male offspring did not vary according to the independent effects of laying date or laying order, but rather varied according to the combined (interaction) effects of these variables.
In our study, we also found that the effect of lutein injection on immune response differed between males and females. Immune response of males was increased whereas that of females was decreased by lutein injection in the original egg. Because of this differential effect, the significantly smaller immune response of males observed among the controls disappeared among lutein chicks. As immune response may predict viability (Møller & Saino, 2004; Cichon & Dubiec, 2005; Moreno et al., 2005), differential allocation of lutein by the mothers in relation to sex could have resulted in higher average viability of their offspring. Immune response of control chicks increased with the mass of the egg, whereas an opposite pattern of variation existed among chicks from lutein eggs. In a previous study, we found a positive covariation between lutein concentration and egg mass while controlling for the effects of laying date and order (Saino et al., 2008a). Thus, supplementation of lutein in small eggs may have generated a positive effect on immune response because it compensated for an original defect in lutein concentration in these eggs. Conversely, the negative effect of lutein supplementation on immune response by chicks from large eggs may indicate that a threshold in absolute lutein concentration exists whereby supplementation of lutein-rich, large eggs has detrimental rather than beneficial effects on offspring immunity.
An interpretation of the mechanisms that generated these differential effects of lutein injection on chick phenotype and survival in relation to sex and egg features would be premature because very little is known about physiological requirements and metabolism of carotenoids by male and female offspring in any bird species and, as far as we are aware, this is the first experimental study where the consequences of an increase in egg lutein concentration have been analysed in relation to sex and other concomitant effects in any species. In general, the effect of supplementation of lutein in the egg may vary depending on the concentration of other antioxidant compounds or in relation to any other egg component, including, for example, maternally derived or endogenous hormones (e.g. Badyaev et al., 2006b; Török et al., 2007). The variation in the effects of lutein we observed may thus depend on concomitant variation of other egg components and the differential effects that these factors have on male and female chicks. In addition, sex-related variation in survival with laying date among chicks from lutein eggs may depend on seasonal variation in extrinsic factors producing different effects on viability of female vs. male chicks.
The analyses, where no significant effects of the interactions between sex or treatment and the other covariates were found, showed consistent positive effects of egg mass and negative effects of laying order and laying date on chick phenotypic values, at both age 4 and 8 days, with the exception of immune response, which increased with laying date. These results are mostly consistent with those presented in previous studies of the same population of yellow-legged gulls (e.g. Rubolini et al., 2006b). The marginally significant effect of the interaction between sex and laying date independently of egg treatment we observed for body mass at age 8 showed that phenotypic quality of males declined whereas that of females did not vary during the breeding season, suggesting differential susceptibility of male and female offspring to ecological variation as the season progressed and/or differential effects of seasonal variation in parental quality.
In this study, brood size was surprisingly found not to affect body mass or tarsus length of the chicks until the age of 8 days. This might have occurred because competition for food, which is expected to increase with brood size, was not intense among young chicks, which require relatively small absolute amounts of food. However, immune response declined with brood size, in line with the expectation based on studies of other species (e.g. Saino et al., 1997). We speculate that a large brood size could negatively affect immune response because of the consequences of stressful interactions among siblings scrambling for food in large broods (Saino et al., 2003c; see also Tella et al., 2001). In addition, being in a large brood may reduce per capita availability of dietary components that specifically limit immune response, thus producing a negative effect of brood size on immune response and no effect on body mass and size.
In conclusion, the main findings of our study were that female yellow-legged gull do not allocate different amounts of lutein to their eggs in relation to embryo sex, and that an experimental increase in egg lutein concentration has contrasting effects on phenotype and viability of male and female chicks, in combination with the effects of laying date, order and mass of their egg. In addition, chick phenotypic quality varied with laying order, laying date and egg mass independently of sex. Finally, the proportion of male embryos varied markedly in relation to the combined effects of laying order and date. As we discussed, the complex effects we documented may have important implications for the interpretation of the evolution of reproductive strategies in our model species, and possibly in other species that exhibit similar sex-related maternal effects. When parents cannot adjust carotenoids concentration of their eggs in relation to the sex of their offspring, and laying order, laying date and egg mass, the differential effect of an increase in carotenoid concentration in relation to the combined effects of these factors can result in a suboptimal allocation of carotenoids to the eggs.
We thank the people who assisted during field work and the local authorities for permission to enter the study area. We are also indebted to R. Stradi and V. Bertacche for running the carotenoid analyses at the lab facilities. Two anonymous referees provided highly constructive comments that helped to improve the paper.