1Reactive oxygen species, produced as a by-product of normal metabolism, can cause intracellular damage and negatively impact on cell function. Such oxidative damage has been proposed as an evolutionarily important cost of growth and reproduction and as a mechanistic explanation for organismal senescence, although few tests of these ideas have occurred outside the laboratory.
2Here, we examined correlations between a measure of phospholipid oxidative damage in plasma samples and age, growth rates, parasite burden and investment in reproduction in a population of wild Soay sheep on St. Kilda, Scotland.
3We found that, amongst females of different ages, lambs had significantly elevated levels of oxidative damage compared to all other age classes and there was no evidence of increasing damage with age amongst adult sheep.
4Amongst lambs, levels of oxidative damage increased significantly with increasing growth rates over the first 4 months of life. Neither mean damage nor the effect of growth rate on damage differed between male and female lambs.
5Amongst adult female sheep, there was no evidence that body mass, current parasite burden or metrics of recent and past reproductive effort significantly predicted oxidative damage levels.
6This study is the first to examine age variation in an assay of oxidative damage and correlations between oxidative damage, growth and reproduction in a wild mammal. Our results suggest strong links between early conditions and oxidative damage in lambs, but also serve to highlight the limitations of cross-sectional data for studies examining associations between oxidative stress, ageing and life history in free-living populations.
Future fitness costs of current life history decisions are a central assumption of theories of the evolution of life histories and ageing (Williams 1957; Kirkwood & Rose 1991; Roff 1992; Stearns 1992). However, our understanding of the mechanisms underlying such costs remains limited, particularly in wild populations (Partridge & Gems 2002; Reznick 2005; Harshman & Zera 2006). Costs of raised investment in growth, immune function and reproduction have traditionally been considered to result from diversion of limited somatic resources away from somatic maintenance functions or future reproduction (Reznick 1992; Stearns 1992). Alternatively, by-products of normal metabolism and immune activity may themselves directly damage the soma, in which case selection acts on the fitness trade-off between damage accumulation and repair (Kirkwood 1977; Harshman & Zera 2006). Oxidative damage to cells and their macro-molecular components may occur as a by-product of normal metabolism: it has been implicated as a proximate mechanism responsible for organismal senescence and an important cost of raised investment in metabolically demanding activities such as growth, reproduction and immune function (Beckman & Ames 1998; von Schantz et al. 1999; Golden et al. 2002; Alonso-Alvarez et al. 2004; Costantini 2008; Monaghan et al. 2009). To date, few studies have examined this potentially important mechanism underpinning trade-offs between growth, reproduction and survival outside of artificial laboratory conditions, despite the importance of placing laboratory findings in a more functional ecological and evolutionary context (although see Costantini et al. 2006; Rubolini et al. 2006; Torres & Velando 2007; Bizé et al. 2008 for examples in wild birds). Here, we examine associations between an assay of plasma phospholipid oxidative damage and age, early growth rates and reproductive history in a wild population of Soay sheep (Ovis aries).
Reactive oxygen species (ROS) are primarily produced as a by-product of aerobic metabolism in mitochondria and can cause oxidative damage to cells and tissues (Harman 1956; Beckman & Ames 1998). The potential for ROS-induced damage can be reduced by a suite of enzymatic antioxidants, endogenous non-enzymatic antioxidants, diet-derived compounds and various repair mechanisms (Beckman & Ames 1998). According to the free radical theory of ageing, oxidative damage to cells determines the rate of organismal senescence, i.e. the decline in survival probability and reproductive performance in old age (Harman 1956; Beckman & Ames 1998). This mechanistic theory of ageing complements the disposable soma theory of ageing, which states that senescence occurs due to an adaptive failure of repair and replacement mechanisms as natural selection weakens with age (Kirkwood & Rose 1991; Kirkwood & Austad 2000). Here, it is the failure of antioxidant defence and oxidative damage repair systems to deal with ROS-induced cellular damage that ultimately leads to senescence (Beckman & Ames 1998). The free radical theory of ageing remains controversial. Although there is some empirical support from model laboratory systems and humans showing increases in oxidative damage and declines in antioxidant defence capacities with age (Sohal et al. 1993; Sohal et al. 1994a; Rao et al. 2003; Lombard et al. 2005; Alonso-Alvarez et al. 2006), there is also contradictory evidence from dietary and genetic manipulations of model organisms and from comparative work across species of differing lifespan (Barja 2004; Sanz et al. 2006; Doonan et al. 2008).
Conditions in very early life may also profoundly alter the balance between ROS production and antioxidant defence and repair capacity. Hatching or birth in vertebrates represents a profound change in aerobic conditions and a potential oxidative insult to the developing organism. Evidence from humans, domestic mammals and wild birds supports this idea by revealing increased ROS and antioxidant levels around birth or hatching relative to later in development (Drury et al. 1998; Gaál et al. 2006; Rubolini et al. 2006). Work in wild bird and lizard populations shows that a high proportion of variation in the oxidative state of eggs and hatchlings can be attributed to their nest of origin or mother's identity (Costantini & Dell’Omo 2006b; Rubolini et al. 2006; Olsson et al. 2008). This suggests that early environmental conditions and parental investment may have profound effects on ROS production and/or antioxidant defence capacity during development. However, the longer term consequences of such variation in environment or parental investment on adult antioxidant capacity, oxidative state and later senescence have hardly been studied (although see Blount et al. 2003). Most studies of oxidative stress in wild vertebrates have been conducted on birds and have tended to focus on variation in levels of antioxidants or measures of total antioxidant capacity in developing offspring and/or their parents (Costantini 2008). Few studies to date have examined variation in oxidative damage across individuals of different ages in wild populations (although see Torres & Velando 2007).
Laboratory experiments on vertebrate and invertebrate systems have been able to relate experimentally raised investment in growth and reproduction to increased levels of ROS, increased oxidative damage and reduced antioxidant availability (Wang et al. 2001; Alonso-Alvarez et al. 2004; Wiersma et al. 2004; Alonso-Alvarez et al. 2007; de Block & Stoks, 2008). Researchers have argued that increased investment in growth or reproduction should result in raised metabolic rate and thus increased ROS production, which, if not accompanied by an up regulation of antioxidant defences, or repair and replacement systems, will result in oxidative stress and loss of cellular homeostasis which may represent a substantial fitness cost to the organism in due course (Alonso-Alvarez et al. 2004; Costantini 2008; Monaghan et al. 2009). It should be noted, however, that such a simple relationship between metabolic rate and ROS production is currently under challenge: mitochondrial uncoupling can lead to associations between metabolism, ROS production and lifespan that are opposite to those predicted above (Brand 2000; Speakman et al. 2004; Selman et al. 2008). What is quite clear, however, is that further work is required to understand the links between life history and oxidative stress and that studies in natural settings are important to place laboratory findings within a more meaningful ecological and evolutionary context (Costantini 2008; Selman et al. 2008). To our knowledge, no study to date has addressed these issues in a free-living mammal population.
Here, we use data collected as part of a long-term individual-based study of Soay sheep (Ovis aries), on the island of Hirta, St. Kilda (Fig. 1; Clutton-Brock & Pemberton 2004). Plasma samples were collected from a cross-sectional sample of 150 sheep caught in August 2007 and assays of a marker of phospholipid oxidative damage were conducted. Based on the free radical theory of ageing, we would predict that oxidative damage should be increased amongst older individuals. Life history theory would predict that oxidative stress is a cost of metabolically intensive activities such as growth, reproduction and immune responses and we accordingly expected levels of oxidative damage to correlate positively with growth rates, parasite burden and reproductive investment. Here, we initially examined correlations between assayed levels of phospholipid oxidative damage and age. We then examined how (i) early life growth rates, birth mass, parasite burden, twin status and sex influenced damage amongst lambs; (ii) prior reproduction, parasite burden and body mass influenced damage in yearling females; and (iii) how body mass, parasite burden, and recent and past reproductive investment influenced damage amongst adult female sheep.
Materials and methods
study population and sampling
The Soay sheep living in Village Bay on the island of Hirta, St. Kilda have been individually studied since 1985 (Clutton-Brock & Pemberton 2004). In this population, around 30% of females breed as yearlings and then typically give birth each year from 2 years of age onwards, producing either singletons or twins (around 80% of females breed in an average year and 13% of litters are twins; Clutton-Brock & Pemberton 2004). Soay lambs develop fast and are typically weaned by mid-June, although they may continue to be suckled throughout the summer (Clutton-Brock & Pemberton 2004). Most lamb mortality occurs during the winter months, but an estimated 24% of lambs die within the first 6 months of life with around 85% of this neonatal mortality occurring within a month of birth (Overall et al. 2005). The environmental conditions experienced by sheep and the consequent levels of over-winter mortality vary markedly between years dependent on population density and winter weather conditions (Clutton-Brock & Pemberton 2004). Each year, individuals are caught and measured during the lambing season (March–April) and during an annual summer catch (August). Lambs are caught within a few days of birth, tagged for future identification, and weighed and measured. Around 90% of lambs born in the study area are caught each spring and daily monitoring of the population means that the day of birth of these lambs is known precisely (Clutton-Brock & Pemberton 2004). Each August, as many sheep in study population as possible are rounded up in a series of temporary traps, caught and processed over the course of a 2-week period. Unknown individuals are tagged and individuals are weighed and blood and faecal samples are taken.
The following measures of size, growth, reproductive investment and parasite burden were used in this study.
Birth weight (kg; lambs only): estimated by correcting capture mass for age at capture during the lambing season (following Overall et al. 2005). Using data from 164 lambs caught and weighed March–April 2007, we found no evidence of any interactions between age at capture and lamb sex (F(df=1) = 2·13, P = 0·15) or whether or not the lamb was a twin (F(1) = 0·63, P = 0·43). We therefore back-calculated an estimated birth weight based on the regression slope of age at capture on capture mass (0·076 kg day−1 ± 0·019 SE).
Lamb growth rate (kg day−1; lambs only): the difference between a lamb's mass at August capture and April capture divided by the time elapsed between the captures (in days).
Faecal egg count (‘FEC’): Strongyle faecal egg counts are performed on faecal samples shortly after collection (Gulland & Fox 1992). Strongyle faecal egg count (FEC) was estimated as the number of eggs per gram using a modified McMaster technique (following Gulland & Fox 1992). On St. Kilda, five nematode species contribute to this count, the most abundant being Teladorsagia circumcincta, Trichostrongylus axei and Trichostrongylus vitrinus (Craig et al. 2006).
August mass: body mass (kg) measured during the August catch-up.
Reproductive investment: Since lactation has been shown to be the most costly aspect of reproduction in ungulates (Clutton-Brock et al. 1989; Festa-Bianchet 1989) and the vast majority of females produce at least one offspring per year from 2 years of age onwards (Clutton-Brock & Pemberton 2004), we used the number of offspring produced by a female that survived until 1 August as an index to capture variation in annual reproductive investment. Note that 85% of neonatal mortality occurs within a month of birth (Overall et al. 2005) and so this is likely to reflect the number of offspring (0, 1 or 2) for which a female paid the full costs of lactation. We used two measures of reproductive investment in adult female sheep: reproductive investment in the year of measurement (2007; ‘recent reproductive investment’) and the sum of annual reproductive investment measures across all previous years of a female's life (‘total reproductive investment’).
Age at first reproduction: around 90% of females start to breed as either yearlings or 2-year olds. Age at first reproduction was therefore treated as a two-level factor: bred as a yearling or started breeding in second year or later.
sampling and measuring oxidative damage
Our 2007 samples were collected on three separate days in August using five different traps. 2007 was not an especially harsh year relative to the extremes the Soay sheep have experienced across our study period. Overwinter population densities were moderate and there was no population crash over the 2006/07 winter period, prior to our samples being taken. Demographic indices likely to be indicative of environment stress experienced by lambs and adults in the months prior to sampling also suggest that conditions in 2007 were relatively moderate. The proportions of lambs surviving from birth to 1 October in the 2007 cohort was 0·77, which was close to the study average (1985–2008: average = 0·80, range = 0·49–0·92), as was the mean back-calculated birth weight (1·82 kg in 2007; 1985–2008: average = 1·92, range = 1·06–2·38).
We used blood samples collected during the August 2007 catch to assay plasma levels of malondialdehyde (MDA) – a widely applied indicator of oxidative damage to phospholipid molecules (e.g. Agarwal & Chase 2002; Magwere et al. 2006) – and relate these assays to age, early growth, reproductive performance, parasite burden and body mass. At capture, blood was taken using lithium heparin tubes and refrigerated at 4 °C within 2 h of collection. Samples were centrifuged within 24 h of collection at 3000 r.p.m. for 10 min and plasma was drawn off. It was then stored at –20 °C, and the time between collection and freezer storage was noted for all samples. We tested whether the interval in minutes between plasma collection and freezing was a significant predictor of MDA (due to potential artefactual oxidative damage) in the statistical analyses (see below). Samples were then kept at –20 °C until the MDA assays were performed 4 months later (in December 2007). We assayed MDA in samples from 39 female lambs and 45 male lambs, which were caught both in August and around birth in 2007. Since relatively few samples were available for males of different adult age classes due to male-biased dispersal and juvenile mortality, we decided to focus on assaying older females only. We assayed samples from 20 female yearlings, and 46 adult females aged 4 years or more, including nine unusually old females ( ≥ 9 years old).
Plasma concentrations of MDA were measured using high performance liquid chromatography (HPLC) with fluorescence detection, as described by Agarwal and Chase (2002) with some modifications. All chemicals were HPLC grade, and chemical solutions were prepared using ultra pure water (Milli-Q Synthesis; Millipore, Watford, UK). Sample derivatization was done in 2 mL capacity screw-top microcentrifuge tubes. To a 50 µL aliquot of sample or standard (1,1,3,3-tetraethoxypropane, TEP; see below) 50 µL butylated hydroxytoluene solution (0·05% w/v in 95% ethanol), 400 µL phosphoric acid solution (0·44 M) and 100 µL thiobarbituric acid (TBA) solution (42 mM) were added. Samples were capped, vortex mixed for 5 s, then heated at 100 °C for 1 h in a dry bath incubator to allow formation of MDA-TBA adducts. Samples were then cooled on ice for 5 min, before 200 µL n-butanol was added and tubes were vortex mixed for 20 s. Tubes were then centrifuged at 15 338 × g and 4 °C for 3 min, before a 50 µL aliquot of the upper (n-butanol) phase was collected and transferred to an HPLC vial for analysis. Samples (20 µL) were injected into a Dionex HPLC system (Dionex Corporation, California, USA) fitted with a 5 µm ODS guard column and a Hewlett-Packard Hypersil 5 µ ODS 100 × 4·6 mm column maintained at 37 °C. The mobile phase was methanol-buffer (40:60, v/v), the buffer being a 50 mM anhydrous solution of potassium monobasic phosphate at pH 6·8 (adjusted using 5 M potassium hydroxide solution), running isocratically over 3·5 min at a flow rate of 1 mL min−1. Data were collected using a fluorescence detector (RF2000; Dionex) set at 515 nm (excitation) and 553 nm (emission). For calibration a standard curve was prepared using a TEP stock solution (5 µM in 40% ethanol) serially diluted using 40% ethanol. We assayed a randomly selected set of 14 blood samples in duplicate to determine repeatability. The correlation between assays was very high (r = 0·85) and the repeatability (calculated following Lessells & Boag 1987) of measurements was 0·62. Although this repeatability was moderate, in a linear regression relating the two assays the estimated elevation was not significantly different from 0 (a = 0·23 ± 0·18 SE, t(df=12) = 1·32, P = 0·21) and the slope was not significantly different from 1 (b = 0·77 ± 0·14 SE, t(12) = 1·68, P = 0·12).
All analyses were run in R version 2·6·1 (R Core Development Team, 2005). We began by testing whether variation in the trap type or day of sampling and the time between blood sampling and freezer storage (to the nearest hour) influenced assays of plasma MDA concentration by fitting day (three-level factor), trap (five-level factor) and time to freezer (covariate) to a linear model of MDA, and assessing their significance by dropping them from the model and testing for significant changes in model explanatory power using F-statistics.
We examined how MDA assays varied with age across our sample of 105 female Soay sheep. We did this by fitting linear models of MDA with different age functions and groupings and comparing their explanatory power. Previous analysis of life history, parasitological and morphological data suggests important differences between lamb, yearling and adult age classes (Catchpole et al. 2000; Clutton-Brock & Pemberton 2004; Coulson et al. 2001). A recent study has also shown that females in this population show senescent declines in survival probability and reproductive performance from around 6 years of age, with performance substantially reduced amongst females aged 9 years or more (Wilson et al. 2007). We compared models incorporating a linear and quadratic function of age with a series of models with different age groupings, including lambs, yearlings, adults and older adults aged greater than 6 or 8 years of age. The groupings used are listed in Table 1, and all age functions or groups were also compared to a model in which no age term was fitted.
Table 1. A comparison of models of age variation in a marker of plasma phospholipid oxidative damage (malondialdehyde; MDA) in female Soay sheep. Models are presented in order of ascending AIC with the best fitting model (lowest AIC) highlighted in bold. The difference in AIC between each model and the best fitting model (ΔAIC), the number of parameters in the model and the model weight are also presented. All models included terms associated with trapping and sampling conditions: trap used, day of trapping and time in hours between taking the sample in the field and freezing the sample. The model with the lowest AIC suggested that the main age differences in MDA were between lambs and other age classes
Number of parameters
Age not fitted
We then ran separate linear models for specific age groups of sheep to test life history correlates of MDA assays. Our model of MDA in lambs included 84 male and female lambs with growth rate, birth weight, FEC, twin status and sex as main effects along with all first-order interactions with sex to test for differences in the factors influencing oxidative damage in early life between males and females. We also ran a separate model for yearling females only, including FEC, body mass and whether or not the yearling had bred or not the preceding spring. Although seven of the 20 yearlings sampled produced a singleton the preceding spring, only one of their lambs survived the neonatal period and so we were unable to test for effects of reproductive effort through lactation. A final model was run for the remaining 46 adult females including FEC, body mass, recent reproductive effort, total reproductive effort and age at first reproduction. Age was also included as a linear covariate to ensure that any effects of total reproductive effort were not simply the result of older females having had more opportunities to breed across their lifetimes.
In general, we used a model simplification approach, sequentially dropping terms with the lowest marginal F-statistic until only significant terms remained in the model (following Crawley 2002). However, for the comparison of different models of age-specific variation in MDA in females, several models had exactly the same number of fitted terms (e.g. quadratic fit and lamb/other groupings, see Table 1) and statistical evaluation of the difference in the models based on F-statistics was not possible. We therefore compared the different age models using Akaike information criterion (AIC), and the model with the lowest AIC value was selected. However, we note that application of an AIC-based model selection approach to select the best model of MDA for lambs, yearlings and adult females yielded the same final model as our model simplification approach (data not shown).
There was no suggestion that day of trapping, trap or the time it took to process a sample from the field to the freezer were correlated with measured plasma MDA concentrations in Soay sheep. Dropping the three variables associated with sampling and storage of blood (day, trap and time from sampling to freezer) from a linear model of plasma MDA concentration including all sampled individual sheep did not significantly change model explanatory power (F(6,97) = 0·66, P = 0·68). Furthermore, addition of these three terms to the lamb, yearling female and adult female models described below did not result in a significant increase in model explanatory power. We therefore excluded these three variables from the following analyses.
age effects in females
Analysis of females (N = 105) revealed that lambs had significantly higher plasma levels of oxidative damage than other age classes, but there were no apparent differences in MDA concentration amongst older age classes (Fig. 2; Table 1). The age model with the lowest AIC grouped females into lambs and all other age classes (Table 1), and this two-group age term was a significant predictor of plasma MDA variation (F(1,97) = 5·42, P = 0·02). Restricting analysis to only samples from adult females, there was no evidence for either a quadratic (F(1,37) = 0·65, P = 0·43) or linear (F(1,38) = 0·17, P = 0·68) difference in plasma MDA with age (Fig. 3).
male and female lambs
There was no evidence for any difference in plasma MDA between the male and female lambs (Table 2A, Fig. 4). However, in both sexes of lamb we found that increased growth rates over the first 4–5 months of life were associated with raised plasma MDA concentrations (Fig. 4; F(1,82) = 5·27, P = 0·02; b = 3·55 ± 1·55 SE). Only growth rate remained as a significant term in linear models of lamb plasma MDA; all other main effects and all interactions with lamb sex were nonsignificant and dropped from the model (Table 2A). The interaction between birth weight and August mass, which would assess relative change in lamb mass without correcting for the time period between measures, was not significant in a model of MDA including main effects for birth mass and August mass (birth mass × August mass interaction: F1,82 = 0·20, P = 0·66). This suggests that it is the absolute and not relative mass increase early in life that influences MDA level amongst Soay lambs.
Table 2. Linear models of plasma malondialdehyde (MDA) concentrations for different age groups of Soay sheep. All models show terms retained as significant following model simplification along with terms dropped from the model during simplification, in reverse order of their elimination from the model. Models shown for (A) lambs of both sexes, (B) yearling females, (C) adults females
Sex × Twin
Sex × FEC
Sex × Growth
Sex × Birth weight
Bred or not
Recent reproductive effort
Age at first reproduction
Total reproductive effort
It is worth noting that, whilst model selection identified growth rate as the key predictor of MDA levels in lambs, birth weight and twin status were closely related to growth rate amongst these individuals. Twins had reduced growth rates compared to singletons (Fig. 5a; estimated mean difference: –0·019 ± 0·005 SE, F(1,82) = 15·17, P < 0·001). Heavy born lambs tended to grow faster across the first 4–5 months of life (Fig. 5b; b = 3·16 ± 0·34 SE, F(1,82) = 84·25, P < 0·001). Heavy born lambs also tended to be born later in the spring than light-born lambs, suggesting they actually had reduced time for post-natal growth (regression of birth date on standardized birth weight: F2,82 = 9·28, P < 0·01, b = 0·024 ± 0·008 SE). The slope of the relationship between lamb August mass and birth weight was significantly greater than unity (t(82) = 6·27, P < 0·001), which demonstrates that early growth rate was more rapid in lambs with high birth weight (Fig. 5b). Fitting these variables on their own within models of lamb MDA showed a significant positive association between birth weight and MDA (b = 0·135 ± 0·060 SE, F(1,82) = 4·98, P = 0·03) and marginally nonsignificant reduction in MDA amongst twins (difference = –0·131 ± 0·078 SE, F(1,82) = 2·80, P = 0·098). The lack of significant effects of these two terms in models including growth rate suggests that the variation in oxidative damage is driven largely by absolute growth rate, but that light born lambs or lambs born as twins are likely to show reduced growth rates and thus reduced levels of plasma MDA.
There was no evidence that any of the fitted terms significantly predicted plasma MDA levels amongst yearling females. August FEC, body mass and whether the yearling had bred or not during the preceding spring were all nonsignificant and dropped from the model of yearling plasma MDA (Table 2B).
None of the fitted terms remained as significant in the final model of plasma MDA amongst adult females (Table 2C). August FEC, recent and total reproductive effort and age at first reproduction did not explain variation in adult MDA, in contrast to predictions based on the assumption that parasite burden and reproductive investment should result in costly increases in ROS production and/or reduced resources to invest in antioxidant defences. It is worth noting that when body mass and recent reproductive effort remained in the model, following sequential deletion of all other terms, dropping recent reproductive effort caused a marginally nonsignificant reduction in model explanatory power (F(1,42) = 2·54, P = 0·09). However, the estimated effect of this term was in the opposite direction to that predicted by life history theory, as females that produced two surviving offspring the previous spring had lower MDA levels than females with one or no surviving offspring.
Levels of oxidative damage to phospholipids in Soay sheep plasma samples were raised in lambs relative to all other age classes (Fig. 2). In mammals, birth is likely to represent a major oxidative challenge, whilst the processes of development and growth demand a consistently raised metabolism that would be expected to increase production of ROS (Gaál et al. 1996; Przybylska et al. 2007). These factors, coupled with poorly developed antioxidant defence and damage repair systems, might lead us to expect raised levels of oxidative damage in very young individuals. However, work on domestic ungulates suggests that milk, and particularly colostrum provided by the mother in the first few days after birth, is rich in non-enzymatic antioxidants, which may help compensate for deficiencies in the antioxidant defence system post-partum and prevent oxidative damage (Przybylska et al. 2007). In two detailed longitudinal studies of the oxidative profiles of cattle calves, Gaál and colleagues concluded that calves coped adequately with changes in oxidative state immediately post-partum but demonstrated subsequent increases in ROS production and oxidative damage in the weeks following birth. They argued that this was due to raised metabolism associated with rapid growth and a shift in diet from milk to grass (Gaál et al. 1996; Gaál et al. 2006). The same factors may be responsible for the raised oxidative damage levels observed in the Soay sheep lambs, as measurements were taken following a period of both rapid growth and development and of adjustment to feeding independently and rumination. However, further studies monitoring the antioxidant content of milk and both levels of antioxidants and oxidative damage across the early lives of Soay sheep would be required to understand the relative roles of diet and growth underlying raised plasma MDA amongst lambs.
Lambs that grew more rapidly across the first 4 months of life had increased levels of oxidative damage to phospholipids in plasma samples (Fig. 4). This represents correlational evidence in support of an ‘oxidative cost’ to early growth as would be predicted by life history theory. Laboratory studies of birds and insects in which growth rates were manipulated have found similar costs, either in terms of raised oxidative damage or reduced antioxidant capacity (e.g. Blount et al. 2003; Alonso-Alvarez et al. 2007; de Block & Stoks 2008). Variation in neonatal traits such as birth weight associated with birth cohort and maternal effects are well documented in both the Soay sheep on St. Kilda and in wild ungulate populations more generally (e.g. cohort and maternal effects: Albon et al. 1987; Festa-Bianchet et al. 2000; Forchhammer et al. 2001; Wilson et al. 2005). Our analyses present a possible link between conditions in very early life – as reflected by birth mass, competition for resources both during gestation and lactation with a twin and growth rates in the first 4 months of life – and levels of plasma oxidative damage. Previous studies of wild bird and reptile populations have found evidence for variation between nests or parents in the oxidative state of nestlings (Costantini et al. 2006; Costantini & Dell’Omo 2006b; Rubolini et al. 2006; Bizé et al. 2008; Olsson et al. 2008). Our samples from Soay lambs raise the possibility that increased growth rates associated with high maternal investment and/or food availability in very early life may be an important driver of differences in the oxidative state of juveniles.
Despite the link between early life conditions and oxidative stress apparent here and in other studies of wild vertebrates, the ecological and evolutionary significance of the observed increase in plasma MDA levels amongst faster growing Soay lambs remains unclear. Previous research clearly shows that heavy born lambs have an improved chance of surviving the neonatal period and, amongst those that do survive, the heaviest lambs at around 4 months of age are the most likely to survive their first winter (Clutton-Brock et al. 1992). It seems unlikely that the observed oxidative cost associated with increased growth rates would offset the benefits to survival prospects of being heavier going into the first winter of life. However, we know that male sheep must invest more than females in growth across the first few years of life and selection on adult body mass varies between sexes and age classes in this population (Clutton-Brock et al. 1992; Pelletier et al. 2007). Although we saw no evidence of sex-specific associations between MDA and body mass or growth rates in our 2007 cohort of lambs, Clutton-Brock et al. (1992) showed that yearling mortality rates declined with increasing body mass at around 16 months of age in females, but in males the relationship was curvilinear with the lightest and heaviest yearling males suffering increased risk of mortality. This work suggests that the costs of growth may be sex-specific and may not become apparent until individuals are close to adulthood. Whilst our results provide support for the idea that investment in growth has costs in terms of raised oxidative damage in Soay sheep, a more detailed, longitudinal study is required to determine the future fitness consequences of variation in growth rates and oxidative state during early life in males and females in this population.
Contrary to predictions of the free radical theory of ageing, we found no evidence for any systematic change in oxidative damage levels with age amongst a cross-sectional sample of adult Soay sheep. Our sample did include unusually old females (nine females aged 9 years or over) and several cross-sectional studies of domestic and laboratory mammals have found evidence for raised oxidative damage amongst older individuals (Sohal et al. 1994a; Sohal et al. 1994b; Sohal et al. 1995; Gaál et al. 1996). Similarly, we found no evidence for any associations between oxidative damage and reproductive investment or parasite burden, as would be predicted under life history theory treatments of oxidative stress (Costantini 2008; Monaghan et al. 2009). This is in contrast to experimental work on birds, which has demonstrated changes in oxidative damage or antioxidant levels following manipulation of reproductive costs or immune challenge (Alonso-Alvarez et al. 2004; Wiersma et al. 2004; Bertrand et al. 2006; Costantini & Dell’Omo 2006a; Torres & Velando 2007). However, several limitations of our data set, which are likely to be typical of studies in natural settings, mean that it would be unwise to interpret our analyses as evidence against the free radical theory of ageing or a role for oxidative stress in life history trade-offs.
First, our data is cross-sectional and we therefore cannot reliably separate within-individual trade-offs or ageing patterns from between-individual heterogeneity. For example, if a strong antioxidant or damage repair capacity was associated with female survival and longevity, then any within-individual increases in oxidative damage with age could be masked as these low oxidative stress females came to predominate in older age classes (van de Pol & Verhulst 2006). Similarly, if individual sheep vary their investment in reproduction in accordance with their current capacity to deal with the costs of such investment (i.e. individual optimization), tests for trade-offs using cross-sectional data will typically yield correlations in the opposite direction to those predicted by life history theory (Moyes et al. 2006; Hamel et al. 2009). Individual heterogeneity may be greatly reduced in benign laboratory or domestic settings and this may, in part at least, explain the disparity between the present results and those from the laboratory, which are also typically cross-sectional. Second, reproductive trade-offs and ageing rates may depend on past environmental experiences of adult females in our data set and controlling for variation in environmental conditions experienced across an individual's lifetime is always likely to present a major challenge to studies in natural settings. Finally, we should point out that laboratory research suggests oxidative damage can vary widely between tissue types (Sohal et al. 1995; Gaál et al. 1996; Costantini 2008) and our oxidative damage assays in plasma may reflect fairly recent oxidative state because turnover of plasma constituents is high. We cannot determine how well assays of plasma MDA reflect oxidative damage levels across a wider range of tissue types, and our results for adult sheep may well reflect a limited ability to detect cumulative and long-term effects of life history decisions on oxidative state using plasma samples alone.
In conclusion, we have shown raised levels of oxidative damage in plasma amongst lambs relative to other age classes and a positive correlation between lamb growth rate and oxidative damage. The lack of a correlation between oxidative damage and either age, reproductive effort or parasite burden serve to highlight the potential limitations of cross-sectional data and plasma samples for addressing these issues in both laboratory and natural settings (Costantini 2008; Monaghan et al. 2009). Ongoing individual-based studies of birds and mammals, such as the study of Soay sheep on St. Kilda, do provide an opportunity to collect and assay longitudinal samples and link these to existing measures of past conditions, growth rates and reproductive investment. Previous work reveals that where longitudinal data is available, mixed-effects models can be used to good effect to detect trade-offs and model senescence at the within-individual level (Nussey et al. 2006; Nussey et al. 2008). This argues for in depth, longitudinal study of variation in markers of oxidative stress in wild populations that are already subject to long-term, individual-based study. We very much hope this study will stimulate further, longitudinal studies investigating the association between oxidative state, ageing and life history in wild mammal populations.
We are grateful to Colin Selman, Andrea Graham, Lea Harrington, Adam Hayward and three anonymous referees for comments earlier drafts of the manuscript and discussion. We thank all members of the spring and summer 2007 field teams on St. Kilda. We also thank the National Trust for Scotland for permission to work on St. Kilda and QinetiQ, Amey and HSS staff on the island for logistic support. The Soay Sheep Project is supported by the Natural Environment Research Council (NERC). DHN was supported by a NERC postdoctoral fellowship and JDB by a Royal Society University Research Fellowship.