Fat provisioning in winter impairs egg production during the following spring: a landscape-scale study of blue tits

Authors


Correspondence author. E-mail: j.d.blount@exeter.ac.uk

Summary

  1. Provisioning of garden birds is a growing phenomenon, particularly during winter, but there is little empirical evidence of its true ecological impacts. One possibility is that winter provisioning could enhance subsequent breeding performance, but this seems likely to depend on the types of nutrients provided. For example, whereas effects of macronutrients such as fat are unlikely to be carried over to influence breeding in small passerines, micronutrients such as dietary vitamin E (an antioxidant) may be stored or have lasting health benefits.
  2. Here, we examine the carry-over effects of winter food supplements on egg production in wild populations of blue tits (Cyanistes caeruleus). Over three consecutive years, birds were provisioned with fat, fat plus vitamin E or remained unfed (controls).
  3. The provision of fat in winter resulted in smaller relative yolk mass in larger eggs and reduced yolk carotenoid concentrations in early breeders. However, these effects were not seen in birds provisioned with fat plus vitamin E. Lay date, clutch size, egg mass and yolk vitamin E concentrations were not significantly affected by winter provisioning treatment.
  4. Our results indicate that winter provisioning can have important downstream consequences, in particular affecting investment in egg production several weeks or months later.
  5. Provisioning is widely applied to support garden bird populations and for the conservation management of endangered species. However, our results challenge the assumption that such practices are always beneficial at the population level and emphasize how the ecological impacts can depend on the specific nutritional profile of provisioned foods.

Introduction

The evolution of life-history traits is constrained by the existence of trade-offs amongst them; at the proximate level, such trade-offs are often modulated by variation in the supply of dietary resources (Stearns 1992). Indeed, food supply plays a crucial role in avian ecology, by influencing community structure, regulating population sizes and through effects on individual behaviours and life histories (Newton 1998). For birds in temperate climates, access to sufficient dietary resources is likely to vary seasonally, leading Lack (1954) to hypothesize that populations would be regulated by food availability during the period of least abundance: the ‘winter food limitation hypothesis’. Consequently, over-winter food supply has been shown to enhance body condition and survival in the short term (Jansson, Ekman & Vonbromssen 1981; Brittingham & Temple 1988; Grubb & Cimprich 1990). This is of particular relevance to the ad hoc supplementation of foods within urban habitats, a hugely popular and growing phenomenon worldwide (Jones & Reynolds 2008). Provisioning of garden birds is most prevalent in winter months, when energetic demands are high and natural food is relatively scarce (Chamberlain et al. 2005). Indeed, the UK and USA together purchase in excess of 500 000 tonnes of commercial bird food each year (O'Leary & Jones 2006), but the ecological consequences of this enormous resource are little studied (Jones & Reynolds 2008; Robb et al. 2008a).

Whilst the capacity for winter food availability to affect survival is well-established, there can be other, downstream consequences that remain poorly understood. Carry-over effects arise when events in one season or year influence an individual's performance in a later season or year (Harrison et al. 2011). The potential for natural food availability on the wintering grounds to influence subsequent productivity in migratory birds is well-known (reviewed by Harrison et al. 2011). However, whilst numerous experimental studies have investigated the effects of food supplementation just prior to or during breeding on productivity in wild birds (e.g. Svensson & Nilsson 1995; Blount et al. 2002b), effects across seasons have rarely been quantified (Park, Lee & Rhim 2004; Robb et al. 2008b). In a landscape-scale investigation, Robb et al. (2008b) found that blue tits provisioned with peanuts during winter had advanced laying dates and increased fledging success compared with unfed controls, even though provisioning had stopped 6 weeks prior to breeding.

The mechanisms that underlie such carry-over effects of winter provisioning on garden birds are poorly understood. Until recently, it has been assumed that carry-over effects are largely driven by energy availability. However, as ‘income breeders’, small passerines cannot store macronutrients such as fat to any great extent and must rely on daily food intake to meet the energetic demands of reproduction (Drent & Daan 1980). It has recently been suggested, however, that any micronutrient with the potential to be stored could produce carry-over effects (Harrison et al. 2011). In particular, large amounts of fat-soluble dietary antioxidants such as vitamin E and carotenoids can accumulate in subcutaneous fat and liver in birds, forming reserves that may be drawn upon during the breeding season when demand is increased (Negro et al. 2001; Surai 2007; Metzger & Bairlein 2011). Antioxidants can have important effects on life histories. For example, vitamin E and carotenoids provide antioxidant defence against reactive oxygen species (ROS), which result as by-products of metabolism (reviewed by Catoni, Peters & Schaefer 2008; Monaghan, Metcalfe & Torres 2009). In doing so, they prevent important biomolecules from being damaged by ROS (i.e. oxidative stress), which might otherwise lead to physiological dysfunction, disease progression and ageing. However, we are not aware of any study that has tested whether winter provisioning with an antioxidant may have carry-over effects on reproduction in any species.

Whilst it is generally assumed that provisioning of food to wild birds is beneficial at the individual and population levels, in fact this is far from clear (Robb et al. 2008a; Harrison et al. 2010). For example, although winter provisioning may enhance short-term survival, it could result in dependency on feeders, reduced dietary diversity and could even have direct deleterious effects on health (Jones & Reynolds 2008). For example, studies of laboratory rodents have shown that increased dietary carbohydrate and fatty acid availability up-regulates mitochondrial respiration (Iossa et al. 2002), which can result in increased susceptibility to oxidative stress (e.g. Igosheva et al. 2010). Therefore, it is conceivable that winter provisioning of birds could have negative downstream effects on reproductive capacity, but this remains to be experimentally tested. The potential for such deleterious effects seems likely to depend on the nutritional composition of foods. For example, winter provisioning with macronutrients such as fats could in theory impair subsequent reproduction, whereas dietary antioxidants may mitigate any such deleterious effects of macronutrient metabolism.

One aspect of reproduction that seems especially likely to be influenced by winter provisioning is egg production. In birds, egg production carries high costs in terms of both energy and nutrient requirements (Perrins 1996). The egg's lipid- and protein-rich yolk provides the main source of energy for developing embryos, and variation in yolk mass can influence offspring survival (Williams 1994). Nevertheless, it appears that energy supply per se is rarely limiting for egg production (e.g. Bolton, Houston & Monaghan 1992). The yolk also contains maternally derived antioxidants including carotenoids and vitamin E, deposition of which is thought to comprise a maternal effect to enhance offspring performance (McGraw, Adkins-Regan & Parker 2005). The availability of diet-derived antioxidants is potentially limiting for egg production because of scarcity of antioxidants in the environment and/or physiological trade-offs in their usage (Blount, Houston & Møller 2000). Increased deposition of carotenoids into yolk, resulting from supplemental feeding just prior to and during laying, has been shown to result in reduced yolk susceptibility to lipid peroxidation (Blount et al. 2002a; McGraw 2005) and enhanced hatching success (Møller, Karadas & Mousseau 2008), immunity (Saino et al. 2003), nestling plumage colouration (Biard, Surai & Møller 2005), and adult survival and sexual signal expression (McGraw, Adkins-Regan & Parker 2005). There is also considerable evidence that yolk-derived vitamin E can function as a potent antioxidant and immunostimulant in vivo (reviewed by Surai 2007). However, few studies have investigated whether limitations in the supply of vitamin E may underlie life-history trade-offs (but see de Ayala, Martinelli & Saino 2006). Remarkably, there have been no studies of the effects of vitamin E availability on egg production in wild birds, and whether a female's capacity to produce high quality eggs is affected by her access to dietary antioxidants in the preceding winter awaits study.

Any effects of winter provisioning seem likely to be context-dependent, being affected by habitat quality and natural food availability (Kallander 1981). Therefore, it is important to study the ecological consequences of dietary provisioning at a landscape scale, across multiple sites and years (Robb et al. 2008a). Here, we use such an approach to examine the effects of winter provisioning of wild blue tits on egg production during the following spring. To investigate the roles of energy and antioxidants as possible mediators of carry-over effects, populations were fed fat, fat plus vitamin E or remained unfed (controls). Our aim was to determine whether winter provisioning altered the timing of laying, clutch size or egg quality in terms of the relative mass of egg components and levels of yolk antioxidants (vitamin E and carotenoids). We hypothesized that compared with unfed controls, winter provisioning with fat would have no beneficial effects on egg production, whereas winter provisioning with fat plus vitamin E would enhance egg production. Alternatively, if winter provisioning with a macronutrient such as fat impairs future egg production compared with unfed controls, we hypothesized that this impairment would not be seen in birds provisioned with fat plus vitamin E.

Materials and methods

Study site and experimental design

The winter provisioning experiment was conducted over three years from 2007 to 2009, and carry-over effects were measured during the subsequent breeding seasons, 2008–2010. The study took place in Cornwall, UK, at nine deciduous woodland sites, where oak (Quercus spp.), beech (Fagus sylvatica), sweet chestnut (Castanea sativa) and sycamore (Acer pseudoplatanus) were the predominant tree species. Sites averaged 10·7 hectares in size and were situated at least 2 km apart to minimize the possibility of between-site movements of birds. There was no evidence from adult ringing records that birds moved sites between winter and spring, or from year to year. Sites were nominally grouped into three triplets, according to similarities in the composition of tree, understorey and ground cover species. Woodland size, proximity to settlements, level of public access and amount of periphery woodland were also taken into account.

In the first year, each site within a triplet was randomly allocated to one of three provisioning groups: (1) no supplement (hereafter ‘control’), (2) fat only (hereafter ‘fat’) and (3) fat plus vitamin E (hereafter ‘fat + VE’). Treatments were rotated within triplets across years, so that every site received all three treatments over the course of the study. As treatments were replicated three times using different triplet groups in a given year, any potential confounding effects of year have been avoided.

Within the six fed sites each year, feeders were hung c. 4 m from the ground at 100 m intervals along parallel transects (100 m apart) and at an average of nine per site. Feeders were custom designed to prevent access by grey squirrels (Sciurus carolinensis) and other mammals and larger bird species. A total of 346 nest boxes, with a 32-mm-diameter entrance hole, were positioned across all sites (mean ± SE per site: 38 ± 1·4). Boxes were erected at 25 m intervals along transects and parallel to feeders, such that each box was no more than 50 m from a feeder. This design produced an equal density and distribution of feeders and boxes across individual woodlands, c. one feeder and four boxes per hectare.

Winter provisioning experiment

Food was provisioned through the winter only (14 December–4 March 2007/2008; 18 November–11 March 2008/2009 and 2009/2010), leaving a gap of at least one month before laying commenced (8 April, 11 April and 15 April, respectively) and thus allowing carry-over effects to be investigated with confidence. All feeding stations were provisioned with a fresh 150 g fat ball every 10 days. Fat balls for the fat + VE treatment group were supplemented with α-tocopherol (T3251; Sigma-Aldrich, Dorset, UK) at a concentration of 100 mg kg−1 fat, a level equivalent to that found in peanuts (Chun, Lee & Eitenmiller 2005), a popular food provisioned to garden birds. α-Tocopherol cannot be provisioned to wild birds without the use of a ‘carrier’ and, as a lipophilic molecule, it is inevitably co-acquired with fat in natural foods (Blount et al. 2002b). Therefore, the fat + VE treatment group provides an ecologically realistic test of the effects of antioxidant provisioning.

All fat balls were produced from solid vegetable fat (Crisp ‘n Dry, Princes Ltd., Liverpool, UK) 1–2 days in advance of provisioning, using standardized methods adapted from the study by Blount et al. (2002b). Fat was heated to 60 °C for c. 1 h until liquefied and then cooled on ice until viscous. When the fat reached 18–20 °C, yellow food colouring (0·125 mL 100g−1 fat; ASDA Natural Food Colouring, Asda Stores Ltd., Leeds) was added, to increase fat ball attractiveness to target species (McGraw et al. 2006) verified by a pilot study (unpublished data). At the same time, α-tocopherol was added to fat (fat + VE treatment) at the concentration specified previously and stirred thoroughly for 4 mins to homogenize. Fat balls (150 g) were hardened at −20 °C overnight before being deployed. Upon collection, fat balls were weighed (±0·01 g) to determine the levels of consumption. Observations at feeders, beak markings on fat balls and winter mist netting confirmed that food use was dominated by Parid sp., with a mean of 3·32 kg (± 0·12 SE) consumed per site per year. Ring recoveries and stable isotope analysis provided evidence that winter provisioned foods had been utilized by breeding birds (Plummer 2011). Food uptake per 10-day feeding period differed between years, but was not significantly different between treatment groups (general linear mixed model [GLMM] with site/feeder random factor; treatment: math formula = 0·36, = 0·55, year: math formula = 336·29, < 0·001, treatment × year: math formula = 0·03, = 0·98).

Breeding parameters

Nest boxes were inspected every 1–3 days from April to June. Lay date of the first egg was back-calculated by assuming one egg was laid per day, if more than one egg was present (Perrins 1996). After the first egg was laid, nests were visited every 1–2 days until clutch completion and new eggs were marked to establish laying order. Total clutch size was recorded upon clutch completion and total clutch mass determined (±0·1 g) using an electronic balance. One egg, typically the last or second-to-last egg laid, was then removed for the measurement of mass and biochemical analysis. Collected eggs were returned to the laboratory, weighed (±0·001 g) and dissected on the day of collection. The yolk was rolled over damp filter paper to remove traces of albumen, weighed (±0·001 g) and stored at −80 °C until analysis.

Biochemical assays

For the extraction of antioxidants, egg yolk (0·040–0·050 g) was vortexed in 0·7 mL 5% NaCl for 5 s and then homogenized with 1 mL EtOH for 20 s. Hexane (1·5 mL) was added, and samples were further homogenized for 10 s, before being centrifuged for 4 min at 8000 × g and the hexane phase containing the antioxidants drawn off. Extraction was repeated and both hexane extracts combined.

Total carotenoid concentrations in egg yolk were determined by spectrophotometry at 450 nm (Nicolet Evolution 500; Thermo Electron Corp., Hemel Hemstead, UK) with total carotenoid concentration calculated using the extinction coefficient of lutein in hexane (2589, Craft & Soares 1992). Hexane (500 μl) was evaporated to dryness and the residue redissolved in 150 μL DCM and 150 μL MeOH. For the determination of α-tocopherol concentrations, samples (20 μL) were injected into a high-performance liquid chromatography system (HPLC; Dionex Corporation, California, USA). Separation utilized a 3-μ C18 reverse-phase column (15 cm × 4·6 mm) (Spherisorb S30DS2; Phase separations, Clwyd, UK), with a mobile phase of MeOH/water (97 : 3 v/v) at a flow rate of 1·1 mL min−1. Fluorescence detection (Dionex RF2000) was performed at 295 nm (excitation) and 330 nm (emission). The α-tocopherol peak was identified and quantified by comparison with a standard solution of α-tocopherol (T3251 Sigma-Aldrich) in methanol. Total carotenoid and α-tocopherol concentrations are reported as μg g−1 yolk.

Statistical analyses

To test the influence of winter provisioning on egg production, general linear mixed models (GLMM) were applied to the following response variables: lay date, clutch size, clutch and egg mass, and yolk α-tocopherol and total carotenoid concentrations. A log10/log10 GLMM of yolk mass on egg mass was used to examine proportionality of yolk investment. Nest box identity nested within woodland site was specified as the random term, to control for temporal and spatial pseudoreplication. An information-theoretical approach based on Akaike's Information Criterion (AIC) was then used for model selection and model averaging (Burnham & Anderson 2002), appropriate for complex large-scale field investigations as reported here (Whittingham et al. 2006).

All first clutches (= 467) were included in lay date analysis, whilst clutch and egg component analyses excluded clutches with laying breaks >2 days (n = 23) and/or abandoned before incubation (n = 32). Eggs showing any sign of incubation upon dissection (i.e. visible early-stage embryo) were excluded from egg component analyses (16% of eggs collected). For each analysis, a candidate set including all possible models given the predictor variables (Table 1), plus a null model fitted with only the intercept, was compared. A quadratic function of lay date (lay date squared) was initially included to test for nonlinear relationships, but it did not improve model fit according to AICc and had little predictive power (main effect parameter estimate (β) <0·0001 in all cases) and was therefore excluded from further analyses. Normality and homoscedasticity of residuals were checked prior to model selection; concentrations of α-tocopherol and total carotenoids were subsequently log-transformed to correct normality.

Table 1. List of parameters included as fixed explanatory variables in GLMMs to investigate the causes of egg production variation
Fixed predictorsDescription/levelsReason for inclusionResponse
  1. Egg production response variables examined were lay date (LD), clutch size (CS), clutch mass (CM), egg mass (EM), and yolk α-tocopherol (YTOC) and total carotenoid (YCAR) concentrations.

Treatment3 level factor – unfed, fat, fat + VEExamine the effect of winter provisioningLD, CS, CM, EM, YTOC, YCAR, YM
Year3 level factor – 2008, 2009, 2010Life-history traits vary annually (Svensson & Nilsson 1995; Lambrechts et al. 2004) LD, CS, CM, EM, YTOC, YCAR, YM
Lay dateContinuous, 1 = 1 AprilClutch size and breeding performance vary seasonally (Norris 1993; Perrins 1996) CS, CM, EM, YTOC, YCAR, YM
Clutch sizeContinuousAccount for variation in female breeding condition and trade-offs between egg number and quality (Slagsvold & Lifjeld 1990; Perrins 1996) CM, EM, YTOC, YCAR
Log10 (Egg mass (g))ContinuousExamine proportionality of yolk investmentYM
Treatment × Year2-way interactionExamine whether effects of annual variation were consistent between treatmentsLD, CS, CM, EM, YTOC, YCAR
Treatment × Lay date2-way interactionExamine whether effects of seasonal variation were consistent between treatmentsCS, CM, EM, YTOC, YCAR,
Treatment × Clutch size2-way interactionExamine whether effects of clutch size were consistent between treatmentsCM, EM, YTOC, YCAR
Lay date × Clutch size2-way interactionImprove model fit by controlling for decline in clutch size through laying period (Perrins 1996) CM, EM, YTOC, YCAR
Treatment × Log10 (Egg mass)2-way interactionExamine whether proportional yolk investment was consistent between treatmentsYM

Models were compared using AICc (i.e. AIC corrected for small sample size), where the best fitting model has the lowest AICc value and all other models are ranked according to their difference in AICc from the top model (ΔAICc). If a single ‘best’ model could not be identified, model averaging was applied across the most strongly supported models (the confidence set), defined by ΔAICc ≤ 2·0 (Burnham & Anderson 2002). Akaike weights (wi) were used to assess the relative support of models within a confidence set, calculate model-averaged parameter estimates (β) and associated standard errors (SE) and estimate relative importance of explanatory variables (w). The predictive power of top ranking models was assessed by calculating a pseudo-R2 value following Nagelkerke (1991), because the coefficient of determination (R2) cannot be generated directly for mixed models. Where treatment (or a specific treatment interaction) was well-supported for inclusion in the best model, further testing was applied to assess between-treatment group differences. Using AICc, the top GLMM model within the confidence set was compared with replicate models, in which two focal treatment groups were paired. In this instance, strong support for a between-treatment group difference was concluded if the model in which the two groups were paired was ΔAICc > 2 from the original GLMM model. All statistical analyses were conducted in R version 2·12·2 (R Development Core Team 2011) using libraries nlme (Pinheiro et al. 2010) and MuMIn (Bartoń 2011).

Results

Timing of laying

Lay dates did not differ between treatment groups, but were strongly predicted by between-year differences (mean ± SE per year: 26·6 ± 0·6; 24·7 ± 0·5; 30·2 ± 0·5, respectively, where 1 = 1 April). The top model, featuring year only (wi = 0·993, n = 467, pseudo-R2 = 0·109), was at least 141 times better supported by the data than all alternatives within the candidate set that also included treatment and the treatment × year interaction (wi > 0·007).

Clutch size and relative mass of egg components

There was strong support for an effect of treatment on the relationship between egg mass and yolk mass; all models within the confidence set for proportionality of yolk investment included the treatment × log10 (egg mass) interaction (Table 2, n = 299, Fig. 1). Post hoc testing revealed that this was driven by differences between the fat and fat + VE treatment groups (ΔAICc = 8·836, for paired model against top model). Whilst fat-fed females had proportionally small yolks compared with controls, fat + VE females produced proportionally larger yolks as egg mass increased (Table 3). But there was no evidence that proportional yolk investment differed between the control group and the fat (ΔAICc = 1·672) or the fat + VE treatment groups (ΔAICc = 0·413). The importance of year, clutch size and lay date in the model was relatively smaller (Table 3).

Table 2. Confidence sets of ranked models for the analyses of maternal investment in egg production, based on Akaike's information criterion corrected for small sample size (AICc)
Rank Model parametersaLog-likelihoodAICcΔAICc w i b Pseudo-R2
  1. a

    Treat, winter feeding treatment; EM, egg mass; LD, lay date; CS, clutch size; ×, interaction term.

  2. b

    Akaike weight for the model within the confidence set.

(a) Proportional yolk mass (log 10 (yolk mass))
1 (treat × log10(EM))581·7−1144·80·0000·2280·486
2 (treat × log10(EM)) + year583·8−1144·70·0270·2250·494
3 (treat × log10(EM)) + LD582·4−1144·10·6360·1660·489
4 (treat × log10(EM)) + CS584·3−1143·80·9930·1390·489
5 (treat × log10(EM)) + year + CS584·3−1143·61·2150·1240·496
6 (treat × log10(EM)) + year + LD584·3−1143·41·3440·1170·496
(b) Clutch size
1 year + LD−756·81527·90·0000·4440·142
2 year + LD + treat−755·01528·40·5300·3400·149
3 year + (treat × LD)−753·31529·31·4430·2160·156
(c) Clutch mass
1 year + LD + CS−439·1894·70·0000·7170·857
2 year + (LD × CS)−439·0896·51·8570·2830·857
(d) Egg mass
1 year + CS294·3−574·20·0000·7050·046
2 year + CS + LD 294·5−572·51·7460·2950·047
(e) α-Tocopherol concentration
1 year + (LD × CS)−210·6439·80·0001·0000·111
(f) Total carotenoid concentration
1 year + (LD × CS) + (treat × LD)−79·0185·20·0001·0000·252
Table 3. Relative variable importance (w), model-averaged parameter estimates (Est.) and standard errors (SE) for variables represented in the confidence sets of maternal egg investment analyses
ParameterProportion yolkClutch sizeClutch massEgg massα-Tocopherol concentrationTotal carotenoid concentration
w Est.SE w Est.SE w Est.SE w Est.SE w Est.SE w Est.SE
Intercept −0·6780·014 10·9310·487 1·4070·491 1·2530·037 4·1890·508 2·6350·364
Clutch size0·2630·0000·001   1·0001·0350·0541·000−0·0080·0031·0000·1250·0571·0000·1100·039
Lay date0·2830·0000·0001·000−0·0960·0151·000−0·0200·0150·2950·0000·0011·0000·0350·0171·0000·0370·012
Log10(egg mass) 1·0000·8700·102               
Treatment1·000  0·556           1·000  
fat 0·0070·010 −0·3690·681          −0·3150·176
fat + VE −0·0090·010 −0·2550·424          0·2030·167
Year0·466  1·000  1·000  1·000  1·000  1·000  
2009 0·0040·006 −0·3320·182 −0·2890·094 −0·0330·013 −0·1340·066 −0·0020·042
2010 0·0040·006 0·3050·183 −0·0690·094 −0·0270·013 0·1180·067 −0·1590·043
Lay date × clutch size      0·2830·0000·002   1·000−0·0060·0021·000−0·0040·001
Treatment × lay date   0·216           1·000  
fat × lay date    0·0100·022          0·0140·006
fat + VE × lay date    0·0020·012          −0·0090·006
Treatment × log10(egg mass)1·000                 
fat × log10(egg mass) −0·1880·138               
fat + VE × log10(egg mass) 0·2250·140               
Figure 1.

Relationship between yolk mass and egg mass. Lines fitted using model-averaged parameter estimates. Testing the difference in the slopes (b) against 1·0 (isometry) reveals that increases in egg mass are accompanied by a proportionate increase in yolk mass for control (= 0·870 ± 0·102 (SE); t103 = 1·23, = 0·200) and fat + VE (= 1·095 ± 0·138; t92 = 0·69, = 0·492) groups. But yolk mass increases proportionately less than egg mass in the fat group (negative allometry; = 0·682 ± 0·135; t98 = 2·35, = 0·021).

Winter provisioning treatment explained less than half the variation in clutch size compared with year or lay date when excluding possible interactions (= 0·435; Table 2, models 1 and 2 only, n = 413) and was poorly estimated as indicated by relatively high standard errors (Table 3).

Variation in total resource deposition was investigated in an analysis of clutch mass, controlling for clutch size. Clutch mass was not influenced by treatment, and similarly neither was individual egg mass (Table 2, n = 388 and 312, respectively; no models contained treatment or treatment interactions within 2 ΔAICc confidence sets). Year was the best predictor of total clutch and individual egg mass variation; females laid clutches of reduced mass in 2009 and had comparatively large eggs in 2008 (= 1·000; Table 3). However, the models within the egg mass confidence set explained only a small proportion of the variation (pseudo-R2 = 0·046 and 0·047; Table 2). Furthermore, a variance components analysis of model 1 (Table 2), using restricted maximum likelihood (REML), revealed that 71·7% of egg mass variation was attributed to interclutch variation, with woodland site accounting for 3·5% and nest box for 24·9%, indicating that inherent differences among females were the greatest predictor of egg mass variation.

Egg yolk antioxidants

Variation in yolk α-tocopherol concentration was not affected by winter provisioning treatment. In this analysis, the best-supported model received a high level of support compared with all others in the candidate set (unadjusted wi = 0·748; Table 2), suggesting that differences were the result of annual and seasonal variation. α-Tocopherol levels were lower in 2009 and decreased as the season progressed, but to a greater extent in larger clutches (Table 3).

By contrast, evidence of a difference in total carotenoid concentration between the treatment groups received a high level of support in an interaction with lay date (Table 2). Post hoc comparisons revealed that yolk total carotenoid concentration in relation to laying date differed between the fat-fed treatment group compared with fat + VE (ΔAICc = 12·346, for paired model against top model) and control groups (ΔAICc = 2·469), whilst the fat + VE and control groups were comparatively similar (ΔAICc = 1·580; Table 3, Fig. 2). Compared with females of the other groups, fat-fed females invested fewer carotenoids into their eggs early in the season, but more carotenoids later in the season. Both year and the lay date by clutch size interaction also received strong selection probabilities, whereby total carotenoid levels were reduced in 2010 and showed a seasonal increase, to a greater extent in smaller clutches ( 0·540, Table 3).

Figure 2.

Relationship between yolk total carotenoid concentration and lay date. Lines fitted using model-averaged parameter estimates. See Tables 1 and 2 for statistical findings and text for details.

Discussion

The results of this study demonstrate pervasive, downstream effects of winter provisioning on egg production during the following spring. However, the effects were strongly influenced by the types of nutrients provided. Previous studies of carry-over effects in birds have considered that macronutrients such as fats are key-limiting resources; an increase in dietary macronutrient supply may either result in storage in body tissues for later use or may result in boosted body condition in one season or year, such that individuals perform better in a subsequent season or year (reviewed by Harrison et al. 2011). Our results are striking because they demonstrate that (1) increased dietary fat availability in winter can in fact impair subsequent egg production in terms of relative yolk mass and egg carotenoid deposition early in the breeding season and (2) macronutrients such as fats are clearly not the only nutritional currency that can invoke carry-over effects; negative effects of winter provisioning were not seen in birds fed fat together with vitamin E.

It is well-established that larger eggs confer benefits to offspring in terms of growth and survival, with these effects attributed to yolk resources (Williams 1994). But furthermore, as well as providing the major energetic requirements for early development, the yolk comprises a cocktail of micronutrients and maternally derived compounds known to influence offspring fitness, such as antioxidants, immunoglobulins and hormones (Gasparini et al. 2001; Blount et al. 2002b; Groothuis et al. 2005). Thus, increasing yolk mass can benefit offspring in terms of enhanced embryonic growth and post-hatching reserves (Peach & Thomas 1986; Bourgault et al. 2007). Egg size is a relatively inflexible trait within females, compared with between-individual differences (Christians 2002). Consistent with this, we found no effects of winter provisioning treatment on egg mass per se. However, variation in the yolk component as a function of total egg mass reflects the absolute difference in the nutrient and energy content of an egg. As such relative yolk mass is an important determinant of egg quality (Williams 1994), which females may modulate adaptively or because of constraint, depending on their physiological condition and access to resources. Typically in altricial and precocial bird species, yolk mass varies in direct proportion to egg mass (i.e. an isometric relationship; Williams 1994). For a small proportion of females constrained to lay small eggs, fat provisioning appears to have been beneficial. But, as egg mass increased, fat provisioning led to a significant decline in relative yolk mass (negative allometry; Fig. 1). However, this deleterious consequence of winter provisioning was not seen in birds that had received fat together with vitamin E.

How could winter provisioning with fat apparently reduce the capacity of females to produce large egg yolks? As income breeders, blue tits are incapable of storing sufficient amounts of endogenous macronutrients to fuel reproduction (Drent & Daan 1980); therefore, fat provisioned in winter is unlikely to have been utilized directly for egg formation several weeks or months later. Instead, yolk mass is a function of food availability in the days leading up to laying in income breeders (Ardia, Wasson & Winkler 2006). It therefore appears that winter provisioning with fat subsequently impaired the capacity of birds to acquire, assimilate and/or mobilize key resources required for yolk formation during egg production. Yet, the addition of vitamin E to fat supplements mitigated this. We think the most likely explanation is that birds may come to rely on readily accessible provisioned foods in winter (Brittingham & Temple 1992). Whilst a high-fat diet provides an abundance of energy, it could mean that birds fail to obtain a balanced, natural diet and are lacking in key nutrients such as antioxidants. Indeed, a high-fat diet should increase the requirement for antioxidant protection, polyunsaturated fatty acids in particular being highly susceptible to oxidative damage (Igosheva et al. 2010). In our study, it was logistically impossible to follow individual birds throughout winter and spring, and thus, we were unable to assess the effects of provisioning on oxidative stress levels during winter and to relate this to breeding performance. However, it seems possible that fat-fed females may have suffered oxidative stress that impaired their ability to invest resources in egg composition. Vitamin E is a potent antioxidant, capable of breaking the chain of lipid peroxidation (Surai 2007), and therefore, acquisition of dietary vitamin E could have mitigated the oxidative burden imposed by a fatty diet.

Yolk α-tocopherol concentrations reduced over the laying period and were unaffected by winter provisioning treatment. It is therefore unlikely that α-tocopherol acquired through winter provisioning was stored for later use during egg production. We found no significant effect of provisioning treatment on laying date. However, compared with both unfed and fat + VE females, individuals provisioned with fat alone produced eggs with relatively low concentrations of carotenoids early in the season, but relatively high concentrations of carotenoids later in the season. There is strong selection for breeding early in blue tits, as in many other bird species, because early breeders afford their offspring higher survival and recruitment prospects (Perrins 1996). It has been shown that supplementing the diet of blue tits with carotenoids just prior to and during laying results in significantly elevated levels of carotenoids in eggs (Biard, Surai & Møller 2005). Increased levels of yolk carotenoids can counter lipid peroxidation (Blount et al. 2002a; McGraw, Adkins-Regan & Parker 2005), to which the lipid-rich yolk and rapidly growing embryo are highly susceptible, and have a range of beneficial effects in nestlings such as increased immunity and survival (Saino et al. 2003; Biard, Surai & Møller 2005; McGraw, Adkins-Regan & Parker 2005). However, females in poor condition or experiencing oxidative stress may deposit fewer carotenoids into their eggs in favour of somatic maintenance (e.g. Hõrak, Surai & Møller 2002; Blount et al. 2004; Isaksson, Johansson & Andersson 2008). It is perhaps more important to lay early than to produce eggs that contain high levels of carotenoids. It seems likely that in our study, fat-fed birds attempted to lay at the optimal time even though it was at the cost of producing eggs that contained relatively low levels of carotenoids. Potentially, this could be amplified if winter provisioning enabled relatively low-quality individuals to enter the breeding population, which would otherwise not have bred at all. In the fat-fed treatment group, where nutrient acquisition appeared to have deleterious consequences for egg production, a change in phenotypic structuring could have resulted in reduced average egg quality at the population level.

It is possible that the influence of winter provisioning reduced as the season progressed, because of the increase in time between food uptake and egg laying. However, we do not know why fat-fed birds that laid late produced eggs with relatively high concentrations of carotenoids, compared with females of the other provisioning treatments. It is well-established that dietary access to carotenoids increases over the laying season in Parids. Caterpillar supply, which provides the main food resource for breeding tits, increases in number and quality across the laying period (Arnold et al. 2010). The increase in carotenoid allocation to eggs later in the season in fat-fed birds was not simply because they produced smaller clutches; although clutch size decreased over the laying season, this decline did not differ significantly amongst provisioning treatments. One possible explanation is that fat-fed birds had low survival prospects, that is no expectation of future reproduction, and therefore invested more in current reproduction as the season neared its end (Royle, Surai & Hartley 2003; Alonso-Alvarez et al. 2004). We do not have the data to assess this possibility, but this would be an interesting direction for further work.

As urban land cover expands, gardens are expected to play an increasingly important role in the conservation of biodiversity (Chamberlain, Cannon & Toms 2004). The provision of food for garden birds has been thought likely to benefit this objective (e.g. Toms & Sterry 2008), although there is limited and indeterminate evidence of its ecological impacts (e.g. Robb et al. 2008b; Harrison et al. 2010). More generally, food provisioning has also been applied as a conservation tool to manage endangered populations (e.g. Armstrong, Castro & Griffiths 2007; Oro et al. 2008). Our study is the first to report deleterious effects of provisioning that were carried over from one season to the next. We emphasize, however, that our study focussed on egg phenotypes; it will be important to see how these effects translate into fitness consequences. The mechanism by which these negative effects were generated is of key importance; the provision of energy-rich fat supplements in winter had negative consequences for female egg investment several weeks after provisioning stopped. Yet at the population level, this was mitigated by the provision of fat together with vitamin E. This is the first direct evidence that the specific nutritional composition of provisioned foods may determine whether carry-over effects on breeding performance are positive or negative at the population level. Therefore, where provisioning is practiced as a conservation tool, careful consideration should be given to the nutritional composition of foods. Whether winter provisioning of garden bird species is considered to be beneficial or deleterious may depend on whether effects are interpreted at the level of individuals or populations. Provisioning may lead to a reduction in average levels of egg quality at the population level. However, if provisioning enables certain low-quality individuals to breed, when they might otherwise have died or survived only as non-breeders, this would clearly enhance their lifetime reproductive success and may in fact boost the overall population size. It is evident that further work at the level of individuals is needed to understand how winter feeding may be used to benefit wild bird populations in the future.

Acknowledgements

We thank the many landowners and fieldworkers who made this research possible, Chris Mitchell for assistance in the laboratory and Nick Carter for logistical support. Egg collection for biochemical analysis was licensed by Natural England (licence numbers: 20081292, 20091379 and 20100947). This work was funded by a Natural Environment Research Council (NERC) CASE studentship (to KEP, JDB, SB and DEC), a Royal Society Research Fellowship (to JDB), the BTO and Gardman Ltd.

Ancillary