The oxidative cost of reproduction depends on early development oxidative stress and sex in a bird species

Abstract

In the early 2000s, a new component of the cost of reproduction was proposed: oxidative stress. Since then the oxidative cost of reproduction hypothesis has, however, received mixed support. Different arguments have been provided to explain this. Among them, the lack of a life-history perspective on most experimental tests was suggested. We manipulated the levels of a key intracellular antioxidant (glutathione) in captive zebra finches (Taeniopygia guttata) during a short period of early life and subsequently tested the oxidative cost of reproduction. Birds were allowed to mate freely in an outdoor aviary for several months. We repeatedly enlarged or reduced their broods to increase or reduce, respectively, breeding effort. Birds whose glutathione levels were reduced during growth showed higher erythrocyte resistance to free radical-induced haemolysis when forced to rear enlarged broods. This supports the hypothesis predicting the occurrence of developing programmes matching early and adult environmental conditions to improve fitness. Moreover, adult males rearing enlarged broods endured higher plasma levels of lipid oxidative damage than control males, whereas adult females showed the opposite trend. As most previous studies reporting non-significant or opposite results used females only, we also discuss some sex-related particularities that may contribute to explain unexpected results.

1. Introduction

The cost of reproduction is a central issue in evolutionary ecology [1]. It has been hypothesized that oxidative stress could be a key component of such a cost, explaining the negative correlation between fecundity and longevity (here as the ‘oxidative cost of reproduction’ (OCR) hypothesis [2–4]). Oxidative stress is considered one of the main mechanisms promoting ageing in animals (e.g. [5,6]). It is a disturbance in the pro-oxidant/antioxidant balance in favour of the oxidants, leading to a disruption of redox signalling and control and/or molecular damage [7,8]. Reproduction could increase oxidative stress by consuming antioxidants otherwise allocated to self-maintenance (i.e. a resource allocation trade-off [9,10]) or as an inevitable consequence of a higher cell metabolism rising reactive oxygen species (ROS) levels (an ‘obligatory’ cost [3,11]).

However, the hypothesis that oxidative stress is a cost of reproduction has been challenged due to inconsistent results reported by manipulations of reproductive effort [4,12]. Experimental work mostly performed on female rodents (although examples in other species exist) reveals that reproductive effort may even decrease oxidative damage ([13], in birds [14]). This contrasts with first experiments more than 10 years ago supporting OCR in fruit flies [2,15] and birds (zebra finches, Taeniopygia guttata [3,9]). Many different arguments have been provided to explain this discrepancy, such as the lack of appropriate measures of oxidative stress and the necessity of using many different markers [12]. Oxidative damage measures (mostly lipid and protein oxidation levels) have been preferred to assess oxidative stress as they should reveal the net result of the imbalance, being thus easier to interpret than antioxidant or pro-oxidant levels [4]. Nonetheless, from an evolutionary perspective, the direct link between the level of any oxidative damage biomarker and longevity or fecundity (i.e. fitness) remains barely tested and demonstrated (but see [16]).

It has been argued that, among iteroparous species, manipulations (cross-fostering) to enlarge or reduce broods or litters are more suitable than comparing reproductive versus non-reproductive individuals because only the first effectively prevents animals to adjust their reproductive effort according to their expectancies of future breeding output ([4] but see [12]). We must note that most support for the idea that reproductive effort reduces oxidative stress comes from studies comparing breeders and non-breeders [13]. Furthermore, it has also been highlighted that most recent work testing the hypothesis involved young animals without previous reproductive experience. This could have allowed them to invest more in antioxidant protection to guarantee survival and subsequent breeding effort (argued by Speakman & Garratt [12]). Therefore, a life-history approach (that includes past history and future prospects) is required to correctly test the OCR hypothesis. In fact, it is not only the future reproductive prospects that are relevant, but also previous investment because it may determine the individual's capacity/strategy to face a reproductive challenge. In this context, the influence of early developmental conditions on the OCR has been virtually ignored.

It is well known that environmental conditions during early life can determine the way in which individuals endure the challenges of adulthood. This has broadly been addressed in a biomedical context within the metabolic ‘programming’ or ‘imprinting’ concepts (e.g. [17–19]). Adverse early environments may not only imply a constraint for adulthood (e.g. [19]) but also serve to adapt individuals to future conditions. An archetypical example is that of mammal fetuses exposed to limitations in energy availability during development, which generates adult phenotypes apparently adapted to environments with scarce resources (‘thrifty phenotypes’ [20]). In evolutionary terms, it has been suggested that fitness returns should be highest when both developmental and adult environments match (see ‘environmental matching’ [21] and also ‘predictive adaptive response’ hypotheses, e.g. [22]). In this context, the reproduction may here be understood as a more or less stressful state that an organism must face according to its own developmental history (e.g. [23,24], but see [13]). In this case, we hypothesize that an early episode of oxidative stress could programme the organism to better face increased oxidative stress linked to reproductive effort.

We not only tested the OCR hypothesis, but also, for the first time, the impact of oxidative stress endured during development on the cost of reproduction. The study involved a captive zebra finch (T. guttata) population. Half of the birds underwent a transient artificial reduction in their capacity to synthesize the most important intracellular antioxidant (i.e. glutathione) during growth [25,26]. In adulthood, the birds were allowed to mate and breed freely in an outdoor aviary. Brood size was then repeatedly manipulated and the levels of oxidative damage (plasma lipid peroxidation), several blood antioxidants as well as the resistance of erythrocytes to ROS-induced haemolysis were measured before and after the breeding period. We must note that the last parameter (e.g. [3]) has been ignored in all the reviews on the OCR [4,12,13,27] as it is not considered to be a marker of oxidative damage. Nevertheless, this variable could indirectly reveal oxidative damage in protein and lipids as both affect the integrity of cell membranes (e.g. [28,29]). Furthermore, this parameter has consistently supported the OCR hypothesis in two bird species [3,30,31], being also positively correlated with survival during more or less long life periods [32–34]. Here, we predict that birds rearing enlarged broods should suffer higher oxidative damage, lower antioxidant values and reduced erythrocyte resistance to ROS. Additionally, we predict that adverse oxidative stress conditions during growth could constrain the capacity to face the cost, thus suffering higher oxidative stress levels. Alternatively, if early oxidative stress enabled animals to endure the oxidative challenge of reproduction, the opposite could also be expected.

2. Material and methods

The experiment lasted 1 year approximately (electronic supplementary material, figure 1S). Eighty randomly formed zebra finch pairs were placed in breeding cages and bred over a five-month period producing 409 nestlings (F1 generation). The early environment of F1 birds was manipulated when nestlings reached a minimum body mass of 3 g (mean ± s.e.: 4.82 g ± 0.03). Half of the nestlings in each brood were randomly assigned to a treatment receiving dl-buthionine-S,R-sulfoximine (BSO; Sigma, ref. B2640) diluted in sterilized physiological serum (50 mg ml⁻¹; n = 206) and the other half (n = 203) received sterilized physiological serum only (controls), all of them by means of four subcutaneous injections of 0.06 ml. Thus, BSO chicks received a total of 12 mg BSO. BSO is a specific inhibitor of the first enzyme in the glutathione biosynthesis pathway (see the electronic supplementary material; also [25]). The treatments were randomly allocated in each brood (electronic supplementary material). A blood sample (approximately 100 µl from the brachial vein) was collected for each bird 8 days after the start of the treatment. The effect of the treatment on total glutathione (tGSH) levels was significant in all subsamples of birds analysed in the study (always: p < 0.001; n = 165, d = 0.73; n = 96, d = 0.83), BSO-treated birds always showing lower tGSH values in erythrocytes than controls (means ± s.e.: 4.07 ± 0.158 and 4.73 ± 0155 µmol g⁻¹ for n = 165; 3.96 ± 0.187 and 4.67 ± 0.185 µmol g⁻¹ for n = 96; both groups respectively, see also [25]).

Males and females were housed separately in different indoor aviaries at independence (approximately 40 days old). Birds were again placed in individual cages just after sexual maturity (i.e. 100 days old approximately) to perform a four-week long experiment where males were housed in front of males or females in order to test the influence of adult social conditions on sexual signalling and physiology (i.e. [26]). Both males and females in this study were involved in that experiment. Here, the influence of the social treatment of males was tested in all the statistical models (below), but the experimental manipulation did not affect the results (tables 1S–3S and text in electronic supplementary material).

(a). Reproduction in the outdoor aviary

After the social context manipulation, a random subset of males and females (n = 165) was released in a large outdoor aviary (6.20 × 12 × 3 m). We only released these birds to maintain an adequate population density (less than one bird per square metre; electronic supplementary material). The birds were released in two events separated by 40 days. Blood was sampled and body mass and tarsus length measured just before being released. Moreover, bill hue (a sexually dichromatic trait, males being redder than females) was estimated from digital pictures (electronic supplementary material). The early treatment and sex were balanced in the released subsample (electronic supplementary material).

The outdoor aviary included 194 wood nestboxes and a commercial mix of seeds, a supplement favouring breeding, water and coconut fibre for nest construction were provided ad libitum throughout the study (electronic supplementary material). Birds were individually marked with a numbered metal ring and a combination of two coloured plastic rings on the opposite tarsus. A microchip was also fixed to the metal ring to allow identification in the nestboxes (electronic supplementary material).

Reproduction was freely allowed for seven months approximately (first round: 231 days, second round: 191 days; electronic supplementary material, figure 1S), when all the birds were recaptured, weighed and blood sampled (electronic supplementary material). Inspection of nestboxes and manipulations (below) were made every other day throughout the study. Parental identity was assigned by identification of coloured rings and microchip readings (electronic supplementary material). One of the parents was not identified in six broods (i.e. 0.7% of broods) due to ring and chip loss. Although paternity on the whole of the sample was not confirmed by molecular analyses, we must consider that only the cost of chick rearing was tested here. However, we cannot fully discard the case where biased extra-pair fecundity could have influenced certainty of paternity among mates that in turn would have reduced parental effort in a biased way.

At the end of the study, 277 broods were produced. Each F1 bird reared (mean and s.e.) 3.33 ± 1.74 broods (range: 0–9). Six individuals did not breed and three females only produced eggs (electronic supplementary material). Hence, 156 out 165 released birds (95%) produced some hatchlings.

A number of reproductive variables were recorded and tested for each F1 bird throughout the reproductive period. As these variables did not influence the main hypotheses addressed in the study (the OCR and the influence of early oxidative stress on that cost), the results are only shown in the electronic supplementary material.

(b). Manipulating the cost of reproduction

In order to disrupt the natural positive covariation between parental quality and reproductive investment, brood size was repeatedly manipulated throughout the study. The first manipulation started 46 days after the first release round and, hence, 6 days after the second one. Some birds that were subsequently manipulated had produced a brood before their first cross-fostering (n = 14), but their treatments and sexes were fully balanced (electronic supplementary material and dataset in repository [35]). Among birds producing hatchlings (i.e. n = 156), 129 birds (83%) endured at least one brood size manipulation. Each cross-fostering manipulation involved a different number of broods (mean: 3.90 ± s.e.: 1.79, range: 2–9; N = 148 broods). Cross-fostering manipulations were made when nestlings were 2 days old approximately (mean: 1.62 ± 1.06, range: 0–5), being performed for 120 consecutive days.

The cross-fostering manipulations aimed to (i) increase/decrease the original breeding effort of the parents, (ii) maintain the same first treatment (brood size enlargement or reduction) for parents with several manipulations, (iii) distribute siblings in at least two broods of different treatments, and (iv) avoid leaving any chick in their own nest. The two last points were necessary for future quantitative genetic analyses on F2 birds. In some cases, we could not avoid leaving some chicks (1–2) in their own nests (point 4), although the proportion of original chicks in a brood did not differ between brood size manipulation treatments (electronic supplementary material). Therefore, we aimed to meet all these rules but prioritizing the first two. The unavailability of synchronous broods did not allow the same brood size manipulation (enlargement or reduction) to be consistently maintained in all the broods of 18 birds. These birds were excluded from the statistical analyses of reproductive costs (electronic supplementary material for further information). Brood size was enlarged by one to three additional chicks or decreased by one or two chicks. For the sake of simplicity, the different manipulations were clustered in two categories: enlarged or reduced broods. The number of removed or added chicks (i.e. −2 to +3) did not differ between early treatments (p > 0.20; see electronic supplementary material for the test of other potential biases). The resulting broods were approximately two chicks in reduced broods (mean ± s.e., range: 2.00 ± 0.05, 1–4 chicks) and four in enlarged broods (4.33 ± 0.12, 2–7 chicks).

The clutch size of the first cross-fostering event did not differ among birds receiving enlarged or reduced broods, and there were no significant interactions with early treatment or sex (all p-values more than 0.59; mixed models with Poisson error and log link). However, birds with the largest original brood sizes (more than or equal to 5; n = 15) in their first cross-fostering event were erroneously assigned to reduced broods, generating a bias: the original brood size significantly differed or tended to significantly differ between birds enduring enlarged versus reduced broods (males: F_1,12 = 5.43, p = 0.038; females: F_1,12 = 4.36, p = 0.059; mixed models with Poisson error and log link). Nonetheless, original brood sizes in subsequent breeding events did not differ between birds assigned to enlarged or reduced broods (all p-values more than 0.48). To avoid the bias in the first manipulated brood, the statistical analyses of the cost of reproduction did not include these 15 birds (eight males and seven females; in that case both p > 0.22; also electronic supplementary material). The tests on potential initial biases (above) reported identical non-significant results when these birds were removed.

To summarize, 96 individuals were finally analysed for the cost of reproduction, 52 birds rearing enlarged broods and 44 birds with reduced broods (also electronic supplementary material). As expected, the manipulation exerted a significant effect on these birds regarding the number of nestlings and fledglings (14 days old) reared in the first exchange and in the total number of all the exchanges (always p < 0.001; first event: Poisson mixed models; sum from all the exchanges: normal mixed models after log-transformation). To validate our results all the tests were repeated on a reduced subsample from only those birds producing the modal brood size (i.e. three nestlings) in their first cross-fostering event (20 male and 22 female parents). Most of these analyses provided similar results (electronic supplementary material).

(c). Antioxidant levels

Glutathione was quantified in blood pellets by following Griffith's method [36] with modifications (electronic supplementary material; [25,37]). Results are given in millimolar per gram of pellet. Total carotenoids in plasma were quantified by using a spectrophotometric method, using lutein as a standard (electronic supplementary material, [37]). Finally, a technique often called total antioxidant status (TAS) was used to estimate the availability of non-enzymatic hydrosoluble antioxidants (electronic supplementary material). As the idea that this measure assesses all the antioxidants is questionable, we will only use a generic ‘plasma antioxidants' term (PLAOX, [25,37]). The procedure is based on Miller's method [38].

(d). Plasma triglycerides and uric acid levels

The glycerol phosphate oxidase/peroxidase method and the uricase/peroxidase method were used to measure triglyceride and uric acid levels, respectively, by means of commercial kits (electronic supplementary material).

(e). Oxidative damage in plasma lipids

The protocol described in Agarwal & Chase [39] with modifications by Nussey method [40] was followed to assess the amount of a product of lipid peroxidation named malondialdehyde (MDA; [7]) by means of HPLC (electronic supplementary material).

(f). Red blood cell resistance to haemolysis under oxidative stress in adults

The resistance to haemolysis under oxidative stress in adult birds was assessed by measuring the time needed to haemolyse 50% of the erythrocytes exposed to a controlled ROS attack. The principle is to submit whole blood to thermo-controlled ROS (mostly alkoxy radical [RO–] and peroxy radical [ROO–]) aggression by adding 2,2-azobis-(aminodinopropane) hydrochloride (AAPH), measuring the decrease in optical density of the suspension. The longer the time required to haemolyse 50% of erythrocytes, the stronger the resistance to oxidative stress (full description in [25]).

(g). Statistical analyses

The statistical models reported here did not correspond to any test published from the same zebra finch population in previous studies (i.e. [25,26]). A subsample of those birds is used and a new life period studied.

Linear mixed models were used to analyse the influence of early development treatment, brood size manipulation and sex on different variables after the breeding period. To detect potential biases on reproductive parameters (above), models testing treatments and interactions on initial variability (at the time of release in the outdoor aviary) were performed. The models were always carried out by using the GLIMMIX procedure in sas software 9.3 (SAS Institute, [41]), fixing the distribution error (e.g. binomial, Poisson, normal) and their corresponding links (logit, log and identity, respectively). The default normal distribution and identity link were used when not explicitly mentioned. As some F0 parents produced more than one brood, the identity of the brood where the F1 bird was born, nested into the identity of the cage where it was reared, was included as a random factor [25]. The releasing round and the laboratory session (the latter only for blood variables) were also added as random factors.

Effects on final levels were analysed by including the initial value as a covariate to control for any initial bias. Other covariates were tested to avoid confounding effects (electronic supplementary material). Thus, plasma triglyceride and uric acid levels were added to control for positive relationships of recent food intake with plasma MDA and PLAOX models, respectively [26,37,42]. When used as covariates, physiological variables were recalculated as their residuals from mixed models controlling for random effects [25]. Finally, the tarsus length was tested as a covariate to model size-corrected body mass (i.e. body condition). Alternative models that excluded covariates other that initial values gave similar results.

A backward stepwise procedure (p < 0.05) was adopted starting from saturated models to find the best-fitted model, but other procedures (forward, AIC) were also tested and gave similar results (also electronic supplementary material). Random factors and initial value covariates were, however, maintained for consistency, when reporting the results. LSD post hocs were used and Cohen's d effect sizes reported for significant pairwise comparisons. Bill hue, MDA, triglycerides and initial PLAOX were all log-transformed for normality. We reported all the terms in the model when factors and/or interactions had a p < 0.05. When non-significant, factor and interaction tests are reported just before being removed. Means and s.e. are least square means and s.e. obtained from the best-fitted model.

3. Results

(a). Initial variability

Initial variability was tested on all the birds and only on those involved in successful brood size manipulations (i.e. n = 96; electronic supplementary material). Briefly, we did not find any difference in body mass or condition, even though females tended to be heavier than males among birds whose brood size was manipulated (p = 0.054; d = 0.40; electronic supplementary material, tables 2S and 4S). No difference in bill hue other than sexual dichromatism was found (electronic supplementary material, tables 2S and 4S).

With respect to antioxidants, tGSH values did not differ between early treatments, but females always showed higher tGSH levels than males (p < 0.030; n = 165: d = 0.51; n = 96, d = 0.46; electronic supplementary material, tables 1S and 2S). Moreover, control females showed larger plasma carotenoid levels than BSO females (p = 0.003; d = 0.65), whereas no difference was found in males (p = 0.986). All differences vanished when considering the brood size manipulation subsample (p > 0.10; electronic supplementary material, table 2S).

With regard to oxidative damage (triglyceride-controlled MDA), differences between early treatments (d = 0.43) and sexes (d = 0.65) were found (electronic supplementary material, table 1S), whereas the interaction between the two factors was not statistically significant (p > 0.10). BSO birds endured more oxidative damage than control birds and females showed higher MDA values than males (electronic supplementary material). However, when the analyses were performed on the subsample of birds rearing manipulated broods only sex retained its significant effect (d = 0.55; electronic supplementary material, table 2S). The same results were found when MDA was not controlled for triglyceride variability (electronic supplementary material). No factor or interaction influenced both the uric acid-corrected PLAOX values and red blood cell resistance to ROS (electronic supplementary material, tables 1S and 2S; all p > 0.099).

(b). Early development and reproductive effects on oxidative stress

Here, only models testing birds engaged in successful brood size manipulation are presented. The final body mass was only affected by the sex of the birds (table 1), females being heavier than males (14.46 ± 0.314 and 13.99 ± 0.317 g, respectively; d = 0.46). The same was found when body mass was corrected for body size (table 5S and electronic supplementary material; d = 0.65).

Table 1.

Mixed models testing the impact of the inhibition of glutathione synthesis during development and brood size manipulation on body mass and physiological variables (n = 96). The best-fitted model after a backward stepwise procedure or instead the model including the last factor being retained just before being removed (p < 0.05) are shown.

dependent variable	terms in the model	slope	s.e.	F	d.f.	p-values
body mass	sex			5.01	1,89	0.028
	last laying date	0.015	0.006	5.31	1,89.3	0.024
	initial values	0.592	0.053	126.9	1,89	<0.0001
MDA	sex			8.44	1,72.1	0.005
	brood size manipulation			2.23	1,76.5	0.140
	sex × brood size manipulation			19.16	1,76.3	<0.0001
	triglyceride levels	0.468	0.056	70.01	1,77.5	<0.0001
	last laying date	−0.001	0.001	4.82	1,75.7	0.031
	initial values	−0.245	0.142	2.96	1,77.7	0.089
PLAOX	sex			0.06	1,81.3	0.814
	brood size manipulation			0.02	1,84.6	0.884
	sex × brood size manipulation			3.99	1,81.8	0.049
	uric acid levels	0.047	0.006	57.71	1,81.9	<0.0001
erythrocyte resistance to ROS-induced haemolysis	early treatment			0.61	1,76.8	0.436
	sex			13.80	1,85.6	0.0004
	brood size manipulation			0.58	1,86.6	0.450
	sex × brood size manipulation			7.05	1,86.7	0.009
	early treatment × brood size manipulation			18.23	1,86.6	<0.0001
	initial values	−0.284	0.125	5.20	1,83.6	0.025

tGSH levels were only influenced by the sex (electronic supplementary material, table 5S), females showing higher values than males (4.583 ± 0.076 and 4.202 ± 0.080 µmol g⁻¹, respectively; d = 0.83). Similarly, sex was the only factor influencing plasma triglycerides (electronic supplementary material, table 5S), with females reporting higher levels than males (2.35 ± 0.064 and 2.24 ± 0.064 mg /dl⁻¹, respectively; d = 0.56).

Triglyceride-controlled MDA showed a significant sex×breeding manipulation interaction (table 1). Among males, those with enlarged broods endured more oxidative damage (p < 0.001; d = 1.26), whereas the opposite was found among females (p = 0.042; d = 0.62; figure 1a). The results were similar when the triglyceride and last laying date covariates were removed from the model (electronic supplementary material, table 5S).

Figure 1.

Final levels of (a) plasma triglyceride-corrected lipid peroxidation, (b) uric acid-corrected plasma antioxidants (PLAOX) and (c) erythrocyte resistance to haemolysis depending on the sex. White and grey bars represent birds that reared reduced and enlarged broods, respectively. Least-squared means ± s.e. from the models. MDA levels were log-transformed to reach normality. See the model description in Results.

A significant sex × breeding manipulation interaction was also found on uric acid-corrected PLAOX (table 1). No post hoc test was significant (p > 0.13), but the sign of the differences (figure 1b) mimics that of MDA values (figure 1a).

Similar to MDA and PLAOX, the erythrocyte resistance showed a significant sex×brood size manipulation interaction (table 1 and figure 1c). Among males, those whose brood size was enlarged showed a stronger resistance (p = 0.019; d = 0.72), whereas females did not show any difference (p = 0.168). Moreover, among enlarged brood birds, erythrocytes of males were more resistant than those of females (p < 0.0001; d = 1.35), the difference being absent among reduced brood individuals (p = 0.487).

Interestingly, the resistance to oxidative stress also showed a highly significant early treatment×brood size manipulation interaction (table 1 and figure 2). Among controls of both sexes, enlarged brood birds were less resistant to oxidative stress than those whose brood was reduced (p = 0.012; d = 0.74). By contrast, enlarged brood individuals were more resistant to oxidative stress than the reduced brood ones among BSO birds (p = 0.001; d = 1.06). According to brood size manipulation, reduced brood control birds also showed more resistant erythrocytes than reduced brood BSO birds (p = 0.018; d = 0.73). In enlarged broods, erythrocytes from controls were less resistant to oxidative stress than erythrocytes from BSO birds (p = 0.0003; d = 1.07).

Figure 2.

Final levels of erythrocyte resistance to haemolysis in both sexes depending on the early treatment. White and dark grey bars represent birds that reared reduced and enlarged broods, respectively. Least-squared means ± s.e. from the model.

4. Discussion

Our results show that the early oxidative status can influence the cost of reproduction by altering the resistance to oxidative stress. Males and females whose capacity to synthesize intracellular antioxidants (glutathione) was transitorily reduced during development had erythrocytes that better resisted ROS when rearing enlarged broods compared with those whose glutathione values were unaltered. Furthermore, our results, as a whole, support the OCR but reveal interesting sex-related differences.

To the best of our knowledge, this is the first demonstration that early oxidative stress conditions can determine the way in which the OCR is managed. In a larger sample of individuals, including those considered here, we previously showed (i.e. [25]) that low early glutathione levels alter the adult phenotype (heavier females, and more intense sexual signals, i.e. redder bills), suggesting a better capability for early breeding. We hypothesized that, under natural conditions, protein availability in the diet may constrain the capacity to synthesize glutathione (protein, and particularly, some amino acids used as glutathione substrate; [25] and references therein). Glutathione levels would, in turn, influence adult phenotype to face adverse adult conditions via early reproduction, perhaps at the expense of adult oxidative status and late survival [25]. How this could occur is unknown but we must consider that glutathione is deeply involved in redox signalling, in that it is also conjugated with thiol protein residues (glutathionylation) to influence pathways from energy metabolism to apoptosis and is able to enter into the cell nucleus probably altering gene expression (see references in [25]). Here, in a subsample of birds, however, the age of early reproduction was not influenced by early glutathione status (electronic supplementary material), probably as a consequence of birds being engaged in another experiment just after sexual maturity (social environment manipulation; [26] and electronic supplementary material). This could have forced birds to breed some time later, reducing inter-individual variability in age at first reproduction. Moreover, in contrast with any possible cost of preparation for early breeding, BSO birds did not suffer a stronger cost of increased reproductive effort (figure 2).

This result provides support to the environmental matching hypothesis [21,22]. It suggests that phenotypic traits expressed as a consequence of adverse developmental conditions allow (‘programme’) an individual to cope with poor conditions in adulthood such that they are even able to outperform individuals that endured better developmental conditions. This would be the case if we assume that enduring high breeding effort due to a badly adjusted brood size (Lack's clutch size hypothesis; [43]) represents a poor quality environment. Even though the environmental matching hypothesis is still the subject of intense debate (e.g. [22,44]), our results are consistent with it. We are aware of a single study providing some support to the environmental matching hypothesis in terms of oxidative stress [45]. Zebra finches enduring similar accessibility to dietary micronutrients (mostly antioxidant vitamins) during both growth and sexual maturity had better enzymatic (erythrocyte glutathione peroxidase) and non-enzymatic (similar to PLAOX) antioxidant defences than those experiencing upward or downward changes in micronutrient availability. Nonetheless, the impact of reproduction in this context was untested [45].

Another question is why the environmental matching effect was found for a single blood biomarker. Here, we can first note that captivity conditions are usually considered benign, providing ad libitum food and protection. This has been highlighted as an argument trying to explain inconsistencies in studies testing the OCR [4]. Nonetheless, we cannot discard effects on other unassessed variables and tissues. Anyway, only those variables linked to survival or fecundity should have some significance in evolutionary terms. In this context, we must remember that this biomarker has been positively correlated with survival in three avian studies, including zebra finches [32–34], suggesting that the detected effect (figure 2) can indeed be translated to fitness.

Our results also show significant differences between sexes regarding management of oxidative stress derived from reproductive effort. As far as we know, this is the first experimental evidence of opposing effects between sexes in an oxidative damage biomarker (i.e. MDA). In fact, experimental studies testing OCR have mostly focused on one sex (mostly females; e.g. [13]). The only experimental studies simultaneously testing males and females (i.e. in the same manipulation) are limited to birds. Thus, under free-living conditions, enlarged brood female barn swallows (Hirundo rustica) showed lower plasma antioxidant (similar to PLAOX) capacity than reduced brood females and males, but no effect was found in oxidative damage assessed as plasma hydroperoxides [46]. Christe et al. [31] showed that both male and female free-living great tits (Parus major) forced to rear enlarged broods suffered a reduction in the same erythrocyte resistance measured here. In captive zebra finches, Reichert et al. [47] did not find sex-related oxidative costs, but reduced telomere length in both sexes, a trait indirectly associated with oxidative stress [48]. In earlier work on the same species, Wiersma et al. [9] reported lower absolute activity levels of an antioxidant enzyme (superoxide dismutase) in muscles of males after brood size enlargement, whereas Alonso-Alvarez et al. [3] found lower erythrocyte resistance to ROS-induced haemolysis also in males when breeding effort was increased in a single reproductive event. Thus, zebra finch studies, including the present one (figure 1), seem to indicate that males can endure a larger oxidative cost than females when brood size is enlarged. In fact, females rearing enlarged broods reduced lipid oxidative damage levels, an apparently beneficial effect of reproduction only supported by studies comparing breeding versus non-breeding animals (mostly in female mammals [13]). The only support for this counterintuitive effect in birds come from Costantini et al. [14] showing that captive breeding canaries (Serinus canaria) of both sexes endured lower plasma levels of protein and non-protein carbonyls (oxidative damage biomarkers) than birds prevented from reproducing, but not detecting any significant effect from brood size manipulations.

A higher OCR in male zebra finches seems to contradict Trivers's theory [49] proposing that females from iteroparous species with biparental care should pay a larger cost from current reproduction. The initial investment in larger gametes would require females to allocate more resources to following reproductive investments compared to males, and hence, to pay a larger reproductive cost (but see e.g. [50]). From a mechanistic point of view, endocrinology could, however, support a higher oxidative cost for males. Female sexual steroids (oestrogens) are often considered as molecules with antioxidant properties [51], whereas male testosterone may in some cases be associated with high oxidative stress [52]. This might explain why males showed lower glutathione (antioxidant) values in erythrocytes before reproduction than females, but not why males also showed lower plasma MDA values. Moreover, if oxidative stress due to reproductive effort leads to accelerated ageing, and hence, reduced longevity as proposed in the OCR hypothesis (e.g. [15,32]), why do males commonly outlive females among avian species? We must note that this has been attributed to the costs of reproduction being greater in female birds [53–55]. Nonetheless, a recent meta-analysis of experimental studies increasing brood or clutch size in wild populations of 19 bird species (mostly passerines such as zebra finches) showed a negative effect on male, but not female, survival [56]. Sex-related differences in lipid oxidative damage (figure 1) could contribute to explain this finding. Santos et al. [56] hypothesized that female birds probably work close to their maximum provisioning capacity at natural levels of parental demand, whereas males can invest more when they detect a chance to improve their reproductive success. This may have led males to pay a higher oxidative cost.

In spite of suffering higher oxidative damage due to the breeding effort, males apparently mounted a compensatory (hormetic) response (e.g. [57]). Males rearing enlarged broods showed an apparent (but non-significant) increase in PLAOX levels and significantly higher levels of erythrocyte resistance to oxidative stress (figure 1). However, we did not find a negative correlation between these markers and MDA supporting a causal mechanism (i.e. if an MDA-triglyceride residual covariate is added to PLAOX and erythrocyte resistance models, the term always has p > 0.10; also electronic supplementary material). Nonetheless, these two variables could have served to successfully fight-off oxidative damage in other non-assessed molecules. PLAOX indicates hydro- (but not lipid-) soluble antioxidant levels and has previously been shown to be unrelated to MDA values in zebra finches [37], whereas resistance to ROS-induced haemolysis seems to be not only related to oxidized lipid levels in the cell membrane, but to oxidized protein [28,29]. Accordingly, the lower MDA values of enlarged brood females could also be due to a compensatory (or also ‘shielding’; see [13]) response engaging other antioxidant mechanisms devoted to lipid protection.

Again, whatever the proximate mechanism explaining sex-related oxidative costs, their relevance obviously depends on the final impact on survival and lifetime fecundity (as fitness proxies). Although the majority of our birds are still alive in our aviaries (60%), some conclusions could be drawn from studies testing the link between biomarker levels and subsequent reproduction/longevity. Nevertheless, longitudinal studies are scarce, with the exception of biomedical literature positively linking oxidative damage, disease and infertility in mammals, mostly humans (e.g. [58,59]). In birds, a single study has reported a negative association between plasma MDA levels and survival (recruitment in wild shag Phalacrocorax aristotelis nestlings); [16]), whereas erythrocyte resistance to ROS, as mentioned above, has been linked to survival/longevity repeatedly.

Finally, the impact of reproductive effort on oxidative stress markers was independent of body mass as no experimental effect on this trait was found. Moreover, body mass change during the breeding period was unrelated to MDA, PLAOX or erythrocyte resistance (body mass change as a covariate in these models: p > 0.25; also electronic supplementary material). This supports the hypothesis that the OCR is not the result of a resource allocation trade-off between reproduction and self-maintenance involving energy stores (i.e. [3,12]).

In conclusion, our findings as a whole support the OCR, but reveal that early developmental conditions may influence early oxidative stress (particularly glutathione levels). This should be considered when trying to understand and experimentally test the OCR. This may easily be attained in captivity studies where developing conditions are known, but is a challenge for studies on wild populations. In addition, the different strategies followed by males and females when facing the cost of reproduction should be taken into account as a way to understand and establish the real impact of oxidative stress in the trade-off between reproduction and longevity.

Ethics

This research project was approved by the animal experimentation committee of the University of Castilla La Mancha under license number CEEA: 1201_08.

Data accessibility

Dataset available at Dryad: http://dx.doi.org/10.5061/dryad.dj3v1.

Authors' contributions

A.A.R.-H., G.S. and C.A.-A. formulated the ideas. A.A.R.-H. and C.A.-A. conceived the study, performed the experiment and analysed samples and data; A.A.R.-H. and C.A.-A. wrote the paper. G.S. reviewed the text.

Competing interests

We declare we have no competing interests.

Funding

A.A.R.-H. was funded by a Formación de Personal de Investigación grant (BES-2010-035013; Ministerio de Economía y Competitividad, MINECO, Spanish Government). Financial support was obtained from projects CGL-2009-10883-C02-02, CGL2012-40229-C02-01 and CGL2015-69338-C2-2-P (MINECO, Spain).