Interactive effects of early and later nutritional conditions on the adult antioxidant defence system in zebra finches

During early life, organisms have to cope with a wide range of environmental factors that can influence their development and, in the longer term, affect their fitness (Burton and Metcalfe, 2014; Gluckman et al., 2007; Monaghan, 2008). Phenotypic plasticity during this early stage of life enables organisms to potentially tailor their phenotypes to the prevailing environmental conditions (Callahan et al., 1997; Monaghan, 2008; West-Eberhard, 2003). Phenotypic traits such as growth rate, stress responses, oxidative status and metabolism are very sensitive to variations in environmental conditions (Garland and Kelly, 2006; Metcalfe and Alonso-Alvarez, 2010; Surai, 2002). However, although the ability to cope with environmental variation is crucial in fitness terms, our knowledge of the potential costs of such early phenotypic changes is still in its infancy (Costantini, 2014; Saastamoinen et al., 2010).

An important component of the phenotype of an organism that can have substantial fitness consequences is its antioxidant defence system (Catoni et al., 2008; Cohen et al., 2008; Metcalfe and Alonso-Alvarez, 2010; Monaghan et al., 2009). This includes a variety of endogenously produced antioxidants plus antioxidants obtained from the diet, and acts to detoxify reactive oxygen species (ROS) that otherwise can cause oxidative damage to a range of biomolecules (Halliwell and Gutteridge, 2007). Many physiological functions depend on the subtle balance between ROS and antioxidant defences (Finkel and Holbrook, 2000; Halliwell and Gutteridge, 2007; Monaghan et al., 2009); while ROS are used as signalling molecules within cells, overexposure to ROS can lead to cell and tissue oxidative damage and the eventual development of degenerative diseases, premature ageing and death (Finkel and Holbrook, 2000; Kirkwood and Austad, 2000); however, the relationship between ROS and ageing is still under debate (e.g. Speakman and Selman, 2011).

There is a large literature showing that the vertebrate antioxidant system shows a high degree of plasticity and can be influenced by a wide range of environmental factors, particularly the diet (Surai, 2002,, 2006). Dietary micronutrients such as vitamin A, E or B that cannot be synthesised de novo are very important exogenous lipophilic and hydrophilic antioxidants in vertebrates (Costantini, 2014; Monaghan et al., 2009; Surai, 2002). Essential minerals such as selenium, copper or manganese are also involved in the absorption, recycling and formation of enzymatic and non-enzymatic antioxidants (e.g. glutathione peroxidase or vitamine E; Halliwell and Gutteridge, 2007; Surai, 2002,, 2006). Furthermore, both vitamins and essential minerals play an important direct role in epigenetic regulation of gene expression (Hardy and Tollefsbol, 2011; Rao et al., 2012) and can influence key life-history traits such as growth, sexual development and reproduction in a wide range of taxa (reviewed by Catoni et al., 2008).

Most of the studies that have investigated the early-life influence of dietary micronutrients on antioxidant protection have focused on a single developmental period, usually during pre- or early postnatal development (de Ayala et al., 2006; Hall et al., 2010; Marri and Richner, 2014; Noguera et al., 2011b). However, the diet during later developmental periods is also relevant since there may be compensation for earlier periods of undernutrition if conditions subsequently improve (Metcalfe and Monaghan, 2001). Hence, not only nutritional conditions during early post-natal development (e.g. the growth period) but also later development (e.g. during sexual maturation and early adulthood) might alter the development of antioxidant defences and potentially, a wide range of phenotypic and life history traits (i.e. plumage colouration, growth, sexual maturation and reproduction; Catoni et al., 2008; Hõrak et al., 2000).

Here, we use zebra finches (Taeniopygia guttata) to examine how the availability of micronutrients at two stages of development influences antioxidant defences and the level of oxidative damage individuals incur. We manipulated the availability of dietary micronutrients within the natural range during both the growth period (the first 40 days of life) and the period of sexual maturation and early adulthood (40–90 days of age). In order to see the extent to which the micronutrient intake during the growth phase influenced the antioxidant defence system during sexual maturation, half of the birds were switched from the high or low micronutrient treatment to the opposite treatment during sexual maturation in a full 2×2 design. Birds receiving the lower level of dietary micronutrients during the first period would be expected to show higher levels of oxidative damage unless they were able to make compensatory adjustments when dietary conditions improved.

During the experiment, body mass (M_b) increased with age but differed between sexes (supplementary material Table S5); during the growth period both sexes showed a similar M_b (LSD post hoc tests, day 1: t₁₁₇=−0.180, P=0.906; day 20: t₁₁₇=−1.116, P=0.262; day 40: t₁₁₆=−0.249, P=0.803) but during the period of sexual maturation females were, on average, 10.7% heavier than males (day 65: t₁₁₅=3.549, P<0.001; day 90: t₁₁₄=4.396, P<0.001). None of the other considered variables, including the micronutrient treatments, had any significant effect on M_b (supplementary material Table S5, Fig. S1).

Pectoral muscle and body fat stores were significantly correlated with M_b throughout the experiment (pectoral muscle: estimate=0.030; body fat: estimate=0.153; supplementary material Table S5) but only fat stores significantly changed with age (supplementary material Table S5, Fig. S2). Body fat increased between 20 and 65 days (LSD post hoc tests, 20–40 days: t₂₃₈=−25.243, p<0.001: 20–65 days: t₂₃₃=−3.178, P=0.002) but significantly declined between 65 and 90 days of age (t₂₃₁=−3.792, P<0.002). On average, females also accumulated 7.35% more body fat than males during the experiment (estimate=0.199; supplementary material Table S5). No other variables, including the micronutrient treatments, had any significant effect on pectoral muscle and body fat stores (supplementary material Table S5).

The three-way interaction between the two dietary treatments and age had a highly significant effect on TAC levels (Table 1). Thus, at 20 days of age, the birds that were being reared on a low micronutrient diet (‘L’ group) had a lower TAC level than those reared on the high micronutrient diet (‘H’ group; LSD post hoc test: t₁₁₇=−2.396, P=0.018; Fig. 1). However, birds in the L group then showed an increase in TAC, so by the end of the growth period (40 days) the two dietary groups did not significantly differ in TAC (t₁₁₆=0.664, P=0.664; Fig. 1). Interestingly, the birds' TAC level during the period of sexual maturation (40–90 days) was influenced by the micronutrient treatment received during both the growth period and the period of sexual maturation (Table 1): only the birds that were on the same diet throughout development (L/L or H/H) increased their level of TAC during the period of sexual maturation (L/L: t₈₂=−2.93, P=0.019; H/H: t₈₄=−2.99, P=0.014). As a consequence, at 90 days, these birds had higher TAC levels than those that had experienced a switch to the opposite micronutrient diet at 40 days of age (L/L vs L/H: t₅₃=2.465, P=0.017; L/L vs H/L; t₅₇=2.426, P=0.018; H/H vs H/L: t₅₅=2.351, P=0.022; H/H vs L/H: t₅₁=2.424, P=0.019; Fig. 1). Our results also showed that females on average had 15.4% higher TAC levels than males throughout development (estimate=0.255; Table 1). M_b had no significant effect on non-enzymatic antioxidant capacity, nor did the remaining two- and three-way interactions initially included in the model (Table 1).

GPx activity was also affected by the triple interaction among our dietary treatments and age (Table 1). During the growth period, GPx increased with age (Table 1; Fig. 2), but was not significantly affected by the micronutrient treatment received during that period (Table 1). However, as for TAC, there was an interaction between the two treatment periods (growth and sexual maturation) during the period of sexual maturation (Table 1; Fig. 2). In the group that received a low micronutrient diet during the growth period, only those birds that continued with the same micronutrient diet (L) during sexual maturation did not experience a significant reduction in GPx activity levels (LSD post hoc test, L/L between 40–90 days: t₈₄=−1.034, P=0.304; L/L vs L/H at 90 days: t₅₅=2.919, P=0.005). In contrast, GPx activity in birds reared on a high micronutrient diet during the growth period was not significantly influenced by the dietary treatment received during the period of sexual maturation (H/H vs H/L at 90 days: t₅₈=0.135, P=0.893). There was no significant effect of sex or M_b, nor were any other two and three-way interactions significant (Table 1).

The level of oxidative damage to lipids (MDA) increased over the course of the experiment (Table 1), but levels were not significantly affected by our dietary treatments or their interactions with age and/or sex (Table 1). Neither M_b nor any interactions initially included in the model were significant (Table 1).

Both mixed-model and repeatability analyses indicated that GPx and MDA levels were significantly repeatable within individuals (GPx: F_109,220=2.743, P<0.001, r=0.37; MDA: F_109,220=2.078, P≤0.001 r=0.26; but see also estimates for random factors in Table 1, main text); the same was not true for TAC level (F_109,220=1.210, P=0.123, r=0.07). With regard to the covariance among measures, TAC, GPx and MDA levels were not significantly correlated with each other at 20 days of age (supplementary material Table S6). However, at the end of the growth period (40 days), TAC and MDA levels were weakly but significantly correlated (Pearson's r=0.28, P=0.002). In addition, at the end of sexual maturation, we also found a (weak) negative correlation between GPx and MDA level (Pearson's r=0.21, P=0.020).

The data sets had a high power (≥89.81% in all analyses) to detect medium to large effect sizes (r≥0.3; i.e. Cohen, 1992), although they provided lower power (18–20%) to detect small effects (r≤0.1).

The main aim of this study was to assess whether several components of the antioxidant system are influenced by variation in the micronutrient content of the diet at different developmental periods (i.e. during growth and sexual maturation). The dietary manipulations had no significant effect on rates of growth or final size attained, showing that the diets were different only in micronutrients and not the macronutrients that fuel growth. There were also no significant differences throughout the experiment in the level of oxidative damage sustained by the zebra finches, at least so far as we could tell from our measures. However, we did find effects of micronutrient intake on the activity of the non-enzymatic component of the antioxidant defences. During the middle of the growth period, the non-enzymatic antioxidant defence capacity was lower in those birds experiencing the low dietary micronutrient diet, but by the end of this period (at 40 days) these differences were no longer evident. Interestingly, we found that the diet during the growth period influenced antioxidant defences during sexual maturation. The only birds that showed a steady increase in their level of non-enzymatic antioxidant defences (TAC) throughout development were those that had experienced the same dietary treatment (low or high) throughout development; a switch in diet during development (whether from low to high or vice versa) resulted in a failure to increase this aspect of the antioxidant system. Similarly, the most sustained increase in levels of enzymatic antioxidant defences (GPx) was in the group that experienced the ‘low’ diet throughout development, with the biggest drop in GPx occurring in those switched from the low to the high micronutrient diet. Together, these results suggest that the capacity of zebra finches to respond to variations in the intake of dietary micronutrients during later development may be determined by their nutritional experience early in life. Furthermore, the results raise the possibility that the balance between endogenous and exogenous antioxidant activity might be shaped by the micronutrient environment experienced in early life.

The reduced level of TAC shown at 20 days of age by zebra finches reared on the low micronutrient diet supports previous studies that have emphasised the important role that dietary micronutrients play in the body's antioxidant defences (Evans and Halliwell, 2001; Surai, 2002,, 2006). The observed reduction in antioxidant capacity was probably the result of deficiencies in dietary antioxidant vitamins such as vitamin A or E, which have important and recognised antioxidant properties in birds (Surai, 2002) and were the most abundant in the micronutrient supplement. The lower availability of other micronutrients such as selenium (Se) could also have had an important role, since these trace elements are an integral part of selenoproteins such as glutathione peroxidase (Surai, 2006), although there was no evidence of a stimulatory or inhibitory effect on GPx activity in any of the groups of birds. While this may indicate that dietary micronutrients primarily influence the non-enzymatic antioxidant defence system, we cannot rule out the possibility that our dietary treatment had an effect on other (non-measured) antioxidant enzymes (e.g. catalase or gluthatione reductase; Shireen et al., 2008).

Interestingly, birds on the low micronutrient diet initially had a reduced TAC but this capacity then increased significantly over time. This compensatory response occurred in the absence of any increase in the intake of dietary micronutrients, but could have occurred through at least three mechanisms. Firstly, low levels of circulating antioxidants could have triggered the up-regulation of endogenous non-enzymatic antioxidants such as vitamin C, which can occur via ROS-sensitive signalling transduction pathways when oxidative stress levels increase (Chatterjee et al., 1975; Surai, 2002). Secondly, previously stored antioxidants may have been mobilised; powerful antioxidants such as vitamin E or A are often stored in the adipose tissue or internal organs such as the liver, and may be mobilised when the body antioxidant requirement increases (Surai, 2002). However, it seems unlikely that animals on the low micronutrient diet could have built up stores of surplus antioxidants, and fat reserves remained similar in both experimental groups, both of which suggest that any antioxidant mobilisation would have been on a small scale. Lastly, given the highly plastic nature of the digestive system of birds (McWilliams and Karasov, 2001), it is possible that the birds in the low micronutrient group increased their absorption and assimilation efficiency of dietary antioxidants over time. However, the assimilation of dietary micronutrients in zebra finches has previously been thought to be promoted by a high rather than a poor micronutrient diet during this growth period (Blount et al., 2003). Therefore, the results seems to better support the idea that environmentally induced oxidative stress may have triggered the up-regulation of endogenous nonenzymatic antioxidant defences (Dimova et al., 2008), which, in turn, would explain why birds in the low diet did not suffer higher levels of oxidative damage than those in the high diet.

In contrast to our predictions, we did not find evidence of improved antioxidant defences when birds were switched from a low to a high micronutrient diet; indeed, birds that started life on a particular diet (whether low or high) actually had better antioxidant defences if they continued with that same diet throughout sexual maturation. This suggests that early micronutrient conditions might have shaped the birds' antioxidant phenotype in such a way that prepared them for a similar (micronutrient) environment in future, an idea that fits with the ‘environment-matching’ hypothesis of developmental plasticity (often also called the ‘predictive adaptive response’ hypothesis; Gluckman et al., 2005). The evolution of environment-matching responses relies on the ability of developing organisms to capture accurate information about the quality of the existing environment, and on early environmental conditions being accurate predictors of adult environmental conditions (Monaghan, 2008; Nettle et al., 2014). This is more likely to be the case in fast-developing species in which development through to adulthood may occur before the environmental conditions have radically altered (e.g. less than 2 months from birth to sexual maturity in zebra finches). This might explain why studies that support the environment-matching hypothesis tend to be of fast-developing species (e.g. Oksanen et al., 2012; Saastamoinen et al., 2010) rather than those that develop slowly (Douhard et al., 2014; Hayward et al., 2013). However, given the small number of studies carried out so far, more work is needed to examine this further.

In contrast to expectations, the pattern of oxidative damage (as assessed by levels of MDA) did not vary significantly between diet treatments despite differences in levels of antioxidant protection. These results, however, do not necessarily indicate that there were no differences in oxidative stress between groups. On one hand, supplementation studies often show that a higher availability of micronutrients does not always reduce the level of all biomarkers of oxidative damage (Kim et al., 2013; Lucas et al., 2014; Noguera et al., 2011b), given that they may vary between tissues and organs (Veskoukis et al., 2009). Such variability can therefore make it difficult to reach conclusions relating to oxidative status when measurements are of damage to only a single tissue or type of macromolecule (Monaghan et al., 2009; Selman et al., 2012). By extension, the same problem can be applied to the use of one single biomarker of antioxidant defences. However, this was not a problem in this study since we have shown that diet micronutrients induced a similar effect in both non-enzymatic and enzymatic antioxidant defences. On the other hand, the lack of a significant effect of our dietary treatments on MDA levels might reflect a limitation of statistical power, if for instance, our dietary treatment had a very small effect on the MDA level (Cohen, 1992). Hence, more work is needed to determine how early and later environmental conditions interact to affect different biomarkers of oxidative stress.

In conclusion, this study found that a low availability of dietary micronutrients during the growth period of zebra finches initially compromised their non-enzymatic antioxidant defences, but the birds later responded by up-regulating their endogenous antioxidants defences. Most importantly, those individuals that experienced similar nutritional conditions (regardless of the micronutritional content) during both postnatal growth period and the period of sexual maturation had better enzymatic and non-enzymatic antioxidant defences than those that experienced a change (either upward or downward) in their micronutrient intake. These data support the hypothesis that nutritional conditions experienced in early life act as predictors of the later-life environment, and affect the later regulation of antioxidant defences. Future studies should investigate whether environmental stressors other than nutrition (e.g. population density, predation risk, competition) may induce similar responses in the antioxidant defence system and to what extent such physiological responses affect different components of fitness.

In September 2012, 50 pairs of adult zebra finches (Taeniopygia guttata Reichenbach 1862) from our outbred stock population (∼1500 birds) at the University of Glasgow were allowed to breed under standardised conditions, each pair in a separate breeding cage (60×50 cm and 50 cm high). Paired birds were all experienced breeders and of different pedigree. During the experiment, the birds were housed in standard conditions, in a room maintained at 20.5±2°C under full spectrum, artificial light (16 h:8 h light:dark cycle). The birds’ diet prior to the hatching of the chicks comprised an ad libitum supply of mixed seeds (common millet, yellow millet and canary seed in an approximate ratio of 3:1:1; Johnson & Jeff, U.K.), oyster shell grit, cuttlefish and water. Once a week the birds also received Calcivet calcium supplement (Vetafarm, Wagga Wagga, NSW, Australia), a protein-conditioning supplement (J. E. Haith, Cleethorpes, UK) and fresh vegetables. One day after the experimental chicks hatched, all pairs continued receiving an ad libitum supply of mixed seeds (see below), oyster shell grit, cuttlefish and green vegetables but without any calcium or protein supplement. At hatching, all chicks (N=126 from 36 families) were weighed using an electronic balance (±0.01 g) and marked with a unique nest colour and individual colour combination by dyeing their toes with permanent marker pens. Each brood of chicks was coupled with another brood with a similar hatching date (±1 day) and clutch size; one of the two pairs was then assigned to follow a ‘high’ dietary micro-nutritional treatment and the other a ‘low’ treatment (‘H’ or ‘L’, respectively; details below). One day after the chicks hatched, half of the chicks within each brood were cross-fostered between the two nests to account for family (genetic) and environmental (rearing) effects (see statistical section). Consequently, half of each genetic brood was reared under H and half under L micronutrient treatment but all chicks in a rearing nest received the same treatment. Chick M_b did not significantly differ between experimental groups either at hatching (day=0; F_1,124=0.573, P=0.451) or after being cross-fostered (day=1; F_1,122=0.141, P=0.708).

From hatching, all chicks were fed with a low micronutrient diet (common and red millet in a 1:1 ratio) specially formulated to reduce the daily intake of the main dietary micronutrients (i.e. Vitamin A, B, C and E) within the natural range of variation in captive conditions (see below). Chicks allocated to the H group received individually an extra dose of diet micronutrients every 3 days, diluted in water and via oral administration from hatching to 40 days old (hereafter ‘growth period’). The main components of the commercial micronutrient supplement were vitamins A, B, C, D, E and K, and essential minerals such as selenium, magnesium or copper (see Noguera et al., 2015 and below for further details). The micronutrient supplement was adjusted to double the natural daily intake of main micronutrients that birds would obtain from the low antioxidant diet. Chicks in the ‘H’ group received an increasing dose of micronutrient supplement in order to match their increasing size (see below for a detailed description of dosage). Nestlings in the L group were manipulated in the same way as chicks in the H group, but they received only the vehicle (water). Chick survival (119 of 126; 94.4%) at the end of the dietary treatment (40 days old) was not significantly different between experimental groups and sexes [generalised linear mixed-effects model (GLMM); dietary treatment: F_1,121=0.00, P=0.93; sex: F_1,120=0.00, P=0.96; dietary treatment×sex: F_1,119=0.00, P=0.96].

When the broods reached 40 days of age, the parents were removed from the cages and half of each brood was assigned to follow with the same micronutrient treatment as during the growth period (low to low or high to high), and half to change to the opposite treatment (low to high or high to low) until they were 90 days of age. This second treatment corresponded with the period of sexual maturation and early adulthood in this species (Zann and Bamford, 1996) (hereafter ‘sexual maturation period’). This experimental design resulted in four treatment groups that differed in the micronutrient availability the birds received during their growth period and the period of sexual maturation (i.e. L/L, L/H, H/H or H/L). Survival of the birds during the period of sexual maturation (117 of 119; 98.3%) was not significantly different among experimental groups and sexes (GLMM; growth dietary treatment: F_1,116=0.298, P=0.586; sexual maturation dietary treatment: F_1,117=0.311, P=0.578; growth dietary treatment×sexual maturation dietary treatment: F_1,115=0.233, P=0.630).

We used blood samples to examine the oxidative status of the birds to avoid terminal sampling and to enable effects on later life performance to be assessed in future studies. During the experiment, all birds were blood sampled at 20, 40 (the end of the growth period) and 90 days of age (the end of sexual maturation). Blood samples were maintained on ice and then centrifuged to separate plasma from red blood cells (RBCs). Several aliquots were made for both plasma and RBCs samples and were stored at −80°C. Birds were weighed (±0.01 g) at 1, 20, 40, 65 and 90 days old, and their fat reserves and the size of pectoral muscle assessed at 20, 40, 65 and 90 days old. Fat reserves were assessed by visual inspection using an 8-point scale (Kaiser, 1993) and pectoral muscle profiles were scored on a 4-point scale following Bairlein (1995).

Many seeds present in commercial seed mixtures, such as pearl or yellow millet, are especially rich in micronutrients such as vitamin E (supplementary material Table S1; Royle et al., 2013). Because zebra finch parents feed their chicks (Zann and Bamford, 1996), so preventing us from independently controlling the amount of micronutrients each chick ingested, all broods were fed with a low-micronutrient seed mix specially formulated to reduce the daily intake of several antioxidants and vitamins to what we presumed would be the lower end of the natural range in captive conditions (supplementary material Table S2), but every 3 days those chicks in the H treatment also received an oral dose of a micronutrient supplement. This supplement was made from 1 g of a commercial micronutrient complex (Magic Antistress Mix/Performax, Feed-Food Ltd, UK), specially formulated for bird requirements, mixed with 17.6 mg of water-soluble vitamin E (50% w/w) and diluted in water (5 ml); the H chicks received an increasing dose of this suspension (6.9 µl between 1 and 5 days of age, 17.1 µl between 6–10 days, 25.28 µl between 11 and 15 days, and 28.1 µl between 16 and 40 days) in order to match their increasing size. The main components of the commercial micronutrient supplement were vitamins A, D₃, C, E, B and K and essential minerals Se, MgSO₄, MnSO₄ and ZnSO₄ (supplementary material Table S3). Given that vitamin E and C were qualitatively and quantitatively the most important micronutrients present in the micronutrient complex, the dosage was adjusted to be approximately double the daily amount of vitamin E and C estimated to be consumed by chicks on the low-micronutrient diet (supplementary material Table S4). These calculations assumed that birds consumed the two types of seeds present in the low micronutrient seed mix (common and red millet, 1:1) in direct proportion to their availability (McGraw et al., 2002) and that chicks had a daily seed consumption of 0.74 g when aged 1 to 5 days, 1.83 g when aged 6–10 days, 2.7 g when aged 11–15 days and 3 g from day 16 onward (Zann and Bamford, 1996). Similar manipulations of the micronutrient intake have been successfully used to quantify the effect of diet quality on development, sexual maturation and fitness in this species (i.e. daily increase in micronutrients of between 4 and 71% of the base level for birds receiving the dietary supplement; calculation based on our own data and the data provided in Birkhead et al. (1999), Blount et al. (2003), Royle et al. (2003). Broods in the L group were handled every 3 days in the same way as the H birds and received an oral dose of the same amount of water but lacking the micronutrient complex.

The total non-enzymatic antioxidant defence capacity was measured in plasma using the method described by Erel (2004). The main antioxidants contributing to this assay are hydrophilic antioxidants such as the –SH group of proteins, uric acid and vitamin C, and to a lesser extent, vitamin E and others hydrophobic antioxidants. In brief, plasma samples (5 µl) were diluted 1:2 in ultra-pure water and allowed to react with a coloured 2,2′-azino-bis-(3-ethylbenzothiazoline-6-sulphonic acid) radical cation (ABTS) in a 96-well microplate. The ABTS is decolourised by antioxidants according to their concentration and antioxidant capacity (Erel, 2004). This change in colour was measured as a change in absorbance at 660 nm (Thermo Scientific Multiskan Spectrum spectrometer, Fisher Scientific UK Ltd, Loughborough, UK) and the assays were calibrated with Trolox. Levels of non-enzymatic plasma antioxidant capacity were expressed as millimoles of Trolox equivalent per litre. Analyses were carried out in duplicate and were repeatable at each age (20 days: r=0.75, F_119,120=4.506, P<0.001; 40 days: r=0.75, F_116,117=5.565, P<0.001; 90 days: r=0.76, F_111,112=6.228, P<0.001; inter-assay CV=5.73%). Repeatability estimates were calculated for all analyses as described in Lessells and Boag (1987). During the TAC and GPx assays (see below), samples from the same individual were run on the same plate and the plate included the same number of individuals from each one of our four experimental groups in order to distribute the measurement error equally among treatments and ages.

Enzymatic antioxidant defences were assessed by analysing RBC activities of GPx, the most important and widespread intracellular antioxidant enzyme (Halliwell and Gutteridge, 2007). GPx catalyses the reduction of hydroxyperoxides and its main function is to protect against the damaging effect of endogenously formed hydroxyperoxides. GPx activity was measured using a commercial kit (Cayman Chemical, Ann Arbor, MI, US; Catalog Number 703102), and following the manufacturer's instructions. Briefly, this assay measures GPx activity by way of a coupled reaction with glutathione reductase. Glutathione is oxidised during reduction of an organic hydroperoxide (cumene) by GPx, and is recycled to its reduced state by glutathione reductase and NADPH. The oxidation of NADPH to NADP⁺ decreases absorbance at 340 nm which is directly proportional to the GPx activity of the sample. During the assays, all samples (RBC lysate) were diluted 1:25 with dilution buffer prior to analysis, and reactions were carried out in a 96-well plate and the absorbance monitored at 340 nm (Thermo Scientific Multiskan Spectrum spectrometer, Fisher Scientific UK Ltd, Loughborough, UK) at five time points spanning 5 min. All samples were run in duplicate and activity measures were expressed as nmol min⁻¹ mg⁻¹ protein. GPx concentrations were highly repeatable at each age (20 days: r=0.90, F_118,119=19.019, P<0.001; 40 days: r=0.89, F_116,117=17.750, P<0.001; 90 days: r=0.91, F_116,117=21.630, P<0.001; inter-assay CV=5.79%).

The level of oxidative damage to lipids was assessed by quantifying in duplicate the concentration of plasma malondialdehyde (MDA) by high-performance liquid chromatography, following the method described by Karatas et al. (2002) but modifying the volume of sample (10 µl) and reagents as describe previously (Noguera et al., 2011a). MDA is one of the major secondary oxidation products of polyunsaturated fatty acids (Gutteridge and Halliwell, 1999). The absorbance of the eluent was monitored at 254 nm and quantified relative to external standards (calibration curves, R²=0.999). Lipid peroxidation was expressed as micrograms of MDA per millilitre of plasma and showed a high repeatability at each age (20 days: r=0.99, F_117,118=329.80, P<0.001; 40 days: r=0.99, F_117,118=838.521, P<0.001; 90 days: r=0.99, F_114,115=1621.415, P<0.001).

All statistical analyses were carried out using IBM SPSS 22 for Windows. The effect that the micronutrient treatments had on M_b, fat stores and pectoral muscle was analysed using linear mixed effect models (LMM) or generalised lineal mixed-effect model (GLMM) as appropriate. The models included age, sex, micronutrient treatment during the growth period (H or L) and micronutrient treatment during sexual maturation (H or L) as categorical fixed factors as well as all their possible interactions. For the models analysing pectoral muscle and fat stores, M_b was also included as a covariate. Bird identity was included as a random factor in all models.

The effect of our micronutrient treatments on antioxidant defences (TAC and GPx activity in RBCs) and oxidative damage level (plasma MDA) were analysed by using LMM. The models included age, sex, micronutrient treatment during growth period (H or L) and micronutrient treatment during sexual maturation (H or L) as categorical fix factors as well as all their possible interactions. Bird M_b was included as covariate in all models. In all models, bird identity was included as a random factor to account for the non-independence of measures from the same individual. For the models analysing TAC and GPx, the identity of the microplate where the samples were assayed was also included as a random factor.

The identity of the two nests between which chicks were switched (‘cross-foster dyad’) and the identities of biological and rearing nest (nested within the cross-foster dyad) were initially also included as random effects in the growth (M_b, pectoral muscle and fat stores) and oxidative stress (TAC, GPx and MDA) models to control for non-independence of chicks from the same genetic or rearing family. However, these random factors had no significant effect and hence, were removed in subsequent analyses. All models were simplified by removing non-significant terms (in a backward deletion procedure), starting from four-way interactions; reported significance levels are those estimated when non-significant terms were dropped from the model. The Satterthwaite approximation was used for the estimation of denominator degrees of freedom, and post hoc comparisons were carried out using Fisher's least significant difference (LSD) test. Sample sizes in some analyses can slightly differ because a small number of chicks died over the course of the experiment (9 out of 126) and because in a few cases there was insufficient blood in a sample to perform all the lab analyses. Data are presented as means±s.e. and significant level was set at 0.05. The statistical power of the tests was analysed by using GPOWER v3.1 freeware (Faul et al., 2009), following Cohen's (1992) conventions.

The within-individual consistency over time was assessed by calculating repeatability estimates (r) for TAC, GPx and MDA following the method described in Lessells and Boag (1987). These analyses only included birds for which TAC, GPx and MDA levels were available at all ages (N=110). In addition, we also assessed the covariance among these oxidative stress markers by calculating the Pearson's correlation coefficients separately for each age (20, 40 and 90 days old). Results of these analyses together with power estimates (see above) are reported in the text.

We are grateful to all the animal husbandry and veterinary staff for their invaluable help. We thank P. Surai for his advice on the micronutrient treatment and for providing us with the micronutrient complex. We also thank two anonymous reviewers for their very helpful comments on the manuscript. The study was carried out under the UK Home Office Project Licence No. 60/4109.