SEARCH

SEARCH BY CITATION

Keywords:

  • adaptive value;
  • constraint;
  • growth;
  • Parus major;
  • stress

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

1. Predation risk is known to influence behaviour and development, especially in breeding animals. Mothers may be selected to transfer information about the intensity of such risk to their offspring through maternal effects and thus influence their development.

2. Here, we test for this maternal effect via manipulation of perceived predation risk by exposing great tit females before and during ovulation to stuffed models and sounds of either a predator bird (sparrowhawk –Accipiter nisus) or of a non-predatory control (song thrush –Turdus philomelos) in their environment. Offspring of exposed mothers were then raised by foster parents subjected to no treatment in order to separate maternal effects from effects during post-hatching parental care.

3. Nestlings of mothers under increased predator density were smaller than those of control mothers yet showed higher growth rates of the wings. Additionally, first-year recruits from the predator treatment had longer wings at maturity.

4. This maternal effect may be a passive consequence of higher circulating stress hormone levels in mothers. The accelerated wing growth during the nestling stage may be a result of compensatory growth.

5. Alternatively, the accelerated wing growth, coupled with the longer wings at maturity, suggest that the maternal response to the environmental risk may be adaptive since lower weight and bigger wings are a selective advantage for predator evasion. In this case the maternal effect probably influences the distribution of resources to different growth functions in offspring.

6. We show for the first time through an environmental rather than a direct hormonal manipulation, that predation risk may elicit adaptive maternal effects in birds.


Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Predation is an important evolutionary force affecting free-living animals. Many anti-predator traits as well as anxiety related behaviours have evolved in response to predation risk (Lima & Dill 1990; Ghalambor & Martin 2000; Eggers et al. 2006; Cresswell 2008). During reproduction, predation can have a particularly strong influence, and parents may, for example, alter current reproductive effort in response to the present environmental predation pressure (e.g. Fontaine & Martin 2006). Stress levels may be higher when predators are present (Scheuerlein, Van’t Hof & Gwinner 2001), and in species where both parents provide parental care, the loss of one parent may increase the work load of the remaining parent (e.g. Griggio & Pilastro 2007; Aho et al. 2009). Due to the smaller size, under-developed senses and responses, and in some species the inability to recognize predators (Kullberg & Lind 2002), offspring may be particularly vulnerable to predation. In addition, parents may reduce the rate of food provisioning in order to reduce their own risk of predation (Scheuerlein & Gwinner 2006), to reduce nest exposure to predators (Eggers, Griesser & Ekman 2005, 2008) or to increase nest defense behaviour (Scheuerlein, Van’t Hof & Gwinner 2001). In a comparative study on birds, hatching and fledging periods were longer under lower predation risk (Bosque & Bosque 1995). Thus, predation risk may affect the evolution of behaviours such as parental guarding, mobbing, or nest site choice, but also important life-history traits such as optimal body mass or size, clutch size or age of maturation (Remes & Martin 2002; Roff, Remes & Martin 2005).

Females, via adaptive maternal effects, may influence offspring development and behaviour as a response to different stimuli such as the presence of parasites in bird nests (Tschirren, Richner & Schwabl 2004; Gallizzi et al. 2008), temperature (Andersen et al. 2005), resource levels (Biard, Surai & Møller 2007; Berthouly, Cassier & Richner 2008) and density of conspecifics (McCormick 2006; Hargitai et al. 2009). Only a few studies have demonstrated both that predation risk can induce maternal effects and that the effect on offspring phenotype is adaptive (Storm & Lima 2010). A previous study by Saino et al. (2005) suggested that female barn swallows respond to predatory environmental cues with a maternal effect, yet this seems to be detrimental to offspring, at least in the short term: egg-laying females that were exposed to a predator model increased the levels of corticosterone (CORT), a stress hormone, in their eggs. In a subsequent study where stress hormone levels were increased by CORT injection into eggs, hatchability was reduced and nestlings grew smaller. The question as to whether these effects are adaptive or a result of a stress-induced developmental constraint is still unclear (Saino et al. 2005; Breuner 2008; Love & Williams 2008a).

Here, we assess the missing direct link between maternal predation risk and some potentially adaptive traits of the offspring, in other words, the direct effect of an altered predator environment on the development of young. Since this requires disentangling the maternal effects from the effects of the rearing environment, we exposed female great tits before and during egg laying to models and calls of a typical predator, the sparrowhawk, but then transferred whole clutches at hatching to nests of females without an experimental predator treatment. As a control, we exposed females to models and calls of a non-predatory bird, the song thrush and also transferred their young into nests without any treatment. We then followed the growth and development of the young and, in the following year, assessed morphological traits of recruiting offspring originating from the two treatments into the breeding population in order to test for the adaptive value of the predation-induced maternal effects.

Materials and methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The study was conducted during spring 2009 in a natural population of great tits (Parus major) in the Bremgartenwald forest near Bern, Switzerland (46º57′N, 7º24′E). The forest was divided into 22 plots comprising either 14 or 15 nest boxes each. Plots were approximately two great tit territories apart (c. 120 m, Gosler 1993) in order to reduce the effect of one plot’s treatment on the neighbouring plot. We visited nest boxes every third day from the beginning of the breeding season in order to determine the start of nest building and egg laying, and every day from the 10th day of incubation to determine hatching date.

In order to prevent any parasite related maternal effects (Tschirren, Richner & Schwabl 2004; Gallizzi, Guenon & Richner 2008; Gallizzi et al. 2008) we heat-treated the nest material of all the nests when it reached 2 cm in height, and additionally cleaned the nest box thoroughly using a hard brush to remove any remaining fleas or larvae from the box, thus eliminating all naturally occurring ectoparasites in the nest at this stage (Richner, Oppliger & Christe 1993). Reproductive cycles of fleas take about 3 weeks (Tripet & Richner 1999), and flea immigration into nests is weak (Heeb et al. 1996).

Experimental treatments

We increased the perceived predator density, and hence perceived predation risk, in six of the plots by displaying taxidermic models of sparrowhawks (Accipiter nisus) and simultaneously playing sparrowhawk calls from a portable loudspeaker. The sparrowhawk is a common predator of great tits that breeds during spring and early summer when the tits fledge (Perrins 1979; Gosler 1993). As a control treatment we performed in six other plots simulations of song thrushes (Turdus philomelos), an open nesting species that neither predates on great tits, nor competes with them for access to nesting sites. Both bird species used in the treatments are common and have been observed in our forest (personal observations; Helfenstein F. pers. comm.). The remaining 10 plots were kept untreated, and were used as foster plots for rearing of the nestlings.

We predefined two thresholds to determine the start of the treatment in a plot: (i) once five nests in the plot reached a stage indicating the territory is used and the nest is likely to be finished; or (ii) once at least one nest in the plot reached a final stage before laying eggs (egg cup clearly visible, often padded with fur). When either of these thresholds was reached, we randomly assigned the plot to one of three treatment groups: ‘predator’, ‘control’ or ‘foster’ plot. Randomization was performed as follows: We created in advance blocks of three types of treatment groups in a random order, with two foster types in each block (e.g. block i = [‘predator’, ‘foster’, ‘control’, ‘foster’]). We then assigned the first plot that had reached one of the above thresholds to the first treatment in the first block, the second plot to reach a threshold to the second treatment in the first block, and so forth. When the first block had been filled with all treatment types we assigned the next plots to the second block and so forth. If more than one plot had reached the threshold on the same day, we rolled a die to determine the assignment. This process ensured that seasonality effects known to be correlated with bird quality and other life-history traits (Verhulst & Nilsson 2008) were randomized between the two treatments, and that nests of the maternal treatments hatched at similar dates as foster nests allowing exchange of broods. The last two plots to cross the threshold were assigned to be foster plots due to the higher need for such plots, and in order to keep a balance between the number of plots under each simulation treatment.

We started performing simulations in a plot once it had been assigned the predator or control treatment. Eight wooden poles, 2·5 m high, were placed in central locations in each plot so that each pole stood in proximity to a few nest boxes. The poles were mounted with a plant saucer turned upside down to serve as rain protection. Every day, either in the morning or in the afternoon (alternated daily), we placed two taxidermic models (for c. 1·5 h), mounted on a small stick, on two of the poles in each plot, following a predetermined sequence. Thus, a simulation was performed in each location in the plot every fourth day. During the simulation we played typical sounds of the corresponding bird species using a portable loudspeaker (FoxPro NX3 game caller, FoxPro, USA, http://www.gofoxpro.com/).

Females of the control and the ‘higher predation risk’ groups were exposed to the treatments for a similar number of days before starting to lay eggs (mean ± SE: 9·2 ± 0·7 and 8·7 ± 0·6 days respectively; Wilcoxon rank-sum test: W = 1880·5, = 0·920).

Egg mass

On the third day of incubation we measured mass of the total clutch (±0·1 g) and counted the number of eggs in order to calculate mean egg mass. The measurement procedure took no longer than 10 min and no nests were deserted as a consequence.

Brood exchange and nestling measurements

We visited nests daily to determine the start of egg laying, incubation and hatching date. We collected the fourth laid egg and replaced it with a dummy egg in each nest in order to analyse egg yolk composition (not reported here).

We measured nestling mass using a portable scale (±0·1 g) when the nestlings were 2 days old (hatching = day 0). We then exchanged entire broods between treatment and foster nests of similar brood size (±1) and hatching date. We marked nestlings for later identification by plucking specific combinations of tuft feathers. During the transfer nestlings were kept in small padded plastic boxes. We placed heating bags underneath the pads to keep nestlings warm. Altogether the transfer process took 1 h at the most, i.e. each nestling spent a maximum of 30 min outside of a nest box. We blocked the entrance hole of the nest boxes from which nestlings were removed with a small piece of cloth to prevent parents from potentially deserting the nest after finding an empty nest box. No nests were deserted as a consequence of the brood exchange procedure.

We measured nestlings again when 8 and 14 days old. On these days, in addition to mass, we measured nestling sternum and tarsus length (±0·05 mm). We also measured nestling wing length on these days as the distance between the end of the radius and the furthermost point of the wing (±0·5 mm). On day 8 we took a small blood sample from each nestling for molecular sexing (Griffiths et al. 1998). Sex ratio did not differ between the two treatment groups (Wilcoxon rank-sum test, = 187, = 0·744).

Parental feeding

When the nestlings were 8 days old, we placed a camcorder in each nest box above the entrance hole to record parental feeding behaviour. All nest boxes were equipped during the whole season with a dummy camera in order to habituate the birds to the camera. To further reduce the effect of placing a camera, we excluded the first half an hour of each video from the analysis. We then counted the number of feeding trips of each parent for 1 h.

Measurements of first-year recruits

In addition to ringing each nestling with a numbered identification ring on the eighth day of its life, we marked nestlings originating from maternally manipulated nests with a white plastic ring on each leg. This allowed us to identify first-year recruits from these nestlings from the start of breeding during spring 2010. Using spring traps we captured the marked birds and measured their mass, tarsus and wing length. One person (MC) performed all the measurements blind with respect to bird origin. Dispersal distance was calculated by calculating the Euclidean distance between GPS points of the nest from which each recruit had fledged and the box in which it was breeding.

Statistical procedures

All statistical analyses were done using R (R Development Core Team, 2009). To test the effect of our treatments on nest desertion we defined four nest-desertion categories: nests that were deserted at an early nesting stage (nest building had started, but not advanced), nests that were deserted at an advanced nesting stage (nest built and almost ready for eggs), nests that were deserted after some eggs had been laid, and nests that were not deserted. We then performed a chi-square test of independence between desertion rate (frequency in each category) and the two treatments.

The time period where females were exposed to the treatment before starting to lay eggs was highly correlated with both the date on which the first egg was laid (= 0·816, < 0·001), and the date of hatching (= 0·697, < 0·001). This prevented us from testing this factor independently within our statistical models. Since laying and hatching dates convey more biological information than the time span between treatment and hatching we chose to use the former in the relevant models.

The difference in mean egg mass between the treatments was analysed using a linear model with laying date of the first egg as a covariate, and its interaction with the treatment.

Nestling morphological traits were analysed separately using linear mixed effects models (Pinheiro et al. 2008) allowing us to estimate treatment effects on each singular trait and possible trade-offs between traits [in contrast to a general measurement for nestling size for which we performed a principal component analysis (PCA); see later]. In all models we included the nest of origin as a random factor to account for the non-independence of nestlings originating from the same genetic family or sharing the same rearing environment (since we exchanged whole broods, nest of origin accounts for the rearing environment as well). Except in the analysis of wing length, we did not include the plot as a random factor since this factor did not explain in the other models a significant proportion of the variation in the data but renders models more complex, and may thereby lead to ill-estimation of model parameters.

In nestling trait models we entered the predator simulation treatment, brood size at hatching, hatching date (centred on the mean hatching date), nestling mass-rank on day 2 (heaviest ranked first), sex and age (repeated measurements models) as fixed factors. The identity of the person measuring (four people performed measurements; each nest and each nestling were always measured by the same person, blind to nestling treatment group of origin) was entered as a fixed effect to control for differences in measurements between performers. Interactions between the treatment and other fixed factors that were suspected to be biologically relevant were added and tested in the analysis. These included the interactions between the treatment type and hatching date, nestling sex and nestling age, together with an interaction between nestling rank and nestling age (repeated measures models). The effect of hatching date on nestling growth rates is often quadratic (Verhulst & Nilsson 2008), and seemed likely when inspecting our data. We tested this relationship in all the models, but found no statistical support for it (models not reported here). When necessary we performed post-hoc tests by running the same models on the single levels tested for the difference (e.g. on different ages for wing analysis).

Nestling mass, wing and sternum lengths were square root transformed to match modelling assumptions of normality of the residuals and homoscedasticity. In addition, possible outliers (i.e. data points with extremely large residual value compared with the majority of model residuals) were singularly inspected to decide whether to keep them in the analysis. Six such outliers were identified, two from the control group and four from the predator treatment. These corresponded to nestlings that were much smaller than their siblings on all three measurement days and on multiple morphological traits. Asynchronous incubation and hatching regularly occur in the great tit (e.g. Hõrak 1995), and these six nestlings have likely hatched at least 1 day after their siblings. Whereas the results did not differ qualitatively if the six nestlings were included or excluded, performing the analyses without these six nestlings allowed both conforming to the modelling assumptions and a better estimation of model coefficients and their associated P-values.

As a general measurement for nestling size we used the first principal component (PC1) of a PCA (singular value decomposition) performed on tarsus, wing and sternum lengths of the nestlings on each measurement day (8 and 14) separately. Four people performed the measurements, inducing inter-observer variance not corrected for when performing the PCA, which could affect the PC1 score of each nestling – in principle a relative measure of size. To overcome this problem we first subtracted, for the measurements of each morphological trait, the mean of each performer’s measurement on that specific trait, thereby centring measurements according to performer. Only then did we perform the PCA on the centred values. PC1 explained 0·711 and 0·473 of the variance on days 8 and 14 respectively. The three size measurements were highly correlated with each other as well as with PC1 (all < 0·001), which thus describes general size where larger individuals have higher PC1 scores. PC1 scores were analysed separately for days 8 and 14 because differences in variances and in loadings between measurements of the 2 days cause scores not to represent the same relation of nestling traits (in other words, a similar change in PC1 score of days 8 and 14 reflects different changes in nestling size). Additionally, due to allometry, the loadings of the morphological traits differ on the two measurement traits, thus influencing the PC1 scores differently.

For the analysis of the combined number of feeding trips of both parents, we used a GLM with quasi-Poisson error distribution given the overdispersion when using Poisson distribution (residual deviance 89·466 on 39 degrees of freedom). The results were qualitatively similar with both error distributions.

Analysis of traits of recruits was done using linear mixed effects models, with nest of origin as a random factor given the family connections between recruits. We tested for the effects of the maternal treatment on traits while controlling for sex and for the interaction between sex and the treatment. However, since the structure of the non-independent data was highly unbalanced (from one to three individuals originating from the same family), we generated 10,000 permutations of the distribution of the recruits’ nests of origin within the treatments, keeping family ties and the structure of the data (i.e. number of sibling pairs/triplets in each treatment), and ran the model on each permutation. The percentage of occasions in which a treatment coefficient of models from permutated data was more extreme than the coefficients of the model gave us a permutated P-value estimating the significance of the treatment effects (Faraway 2005).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We successfully exchanged 46 broods: 21 from control nests and 25 from the ‘higher predation risk’ treatment. Summary statistics for nests and nestlings are presented in Table 1. The probability to fledge did not differ between the treatment groups (0·986 and 0·966 in the control and ‘higher predation risk’ groups respectively; GLM family binomial: χ21,321 = 1·43, = 0·261).

Table 1.   summary of biometrics for offspring of the different treatments
 MeasurementControlHigh predation risk
  1. Data from nests that have been exchanged in the experimental manipulation. Values are mean ± SD. Laying and hatching date refer to 15.3.2009 as day 0. Number of nestlings is number of nestlings on day 2 after hatching. Treatment time is from the beginning of the simulations until start of egg-laying.

  2. PC = principal component.

NestsTreatment time (days)7.57 ± 4.349.00 ± 4.99
Laying date34.95 ± 3.1136.48 ± 4.43
Hatching date57.19 ± 3.4357.60 ± 4.15
Brood size at hatching7.14 ± 1.427.04 ± 1.48
Mean egg mass (g)1.60 ± 0.111.57 ± 0.10
Feeding rate (events/h)19.6 ± 9.3918.83 ± 5.88
 Fledging age (days)18.85 ± 1.5619.59 ± 1.01
  FemalesMalesFemalesMales
NestlingsN68699270
Day 2Mass (g)3.48 ± 0.603.30 ± 0.633.17 ± 0.673.27 ± 0.66
Day 8Mass (g)12.74 ± 0.9513.01 ± 1.0712.32 ± 1.1113.02 ± 1.20
Sternum (mm)7.80 ± 0.757.88 ± 0.707.87 ± 0.857.93 ± 0.88
Wing (mm)22.84 ± 2.4622.68 ± 2.4821.40 ± 2.4621.53 ± 2.17
Tarsus (mm)14.60 ± 0.8214.62 ± 0.9614.15 ± 1.1314.66 ± 1.08
PC10.31 ± 1.320.52 ± 1.44−0.55 ± 1.45−0.10 ± 1.38
Day 14Mass (g)18.00 ± 1.0018.79 ± 1.4317.60 ± 1.0118.48 ± 1.19
Sternum (mm)11.39 ± 1.0011.50 ± 0.9211.54 ± 1.0511.76 ± 0.91
Wing (mm)49.65 ± 1.7250.17 ± 2.4748.77 ± 2.3149.86 ± 2.35
Tarsus (mm)17.71 ± 0.6418.20 ± 0.7317.52 ± 0.6818.25 ± 0.76
PC1−0.03 ± 0.920.64 ± 1.23−0.63 ± 1.110.32 ± 1.07
RecruitsN98138
Dispersal distance (m)1101.8 ± 898.2998.1 ± 600.91407.2 ± 991.41379.9 ± 652.7
Mass (g)17.46 ± 0.9318.20 ± 0.8117.13 ± 0.9617.59 ± 0.64
Wing (mm)72.72 ± 1.2576.25 ± 2.0974.65 ± 1.5476.56 ± 1.08
Tarsus (mm)17.83 ± 0.5418.77 ± 0.5317.63 ± 0.6518.49 ± 0.43

The higher predation risk treatment did not significantly explain the probability of nest desertion for females in any of the four defined categories (χ2 = 0·199, d.f. = 3, = 0·977). Our treatment had no significant effect on clutch size (F1,43 = 0·331, = 0·568), with females laying 8·9 ± 1·3 and 8·7 ± 1·6 eggs (mean ± SD) in the control and the ‘higher predation risk’ treatments respectively. Clutch size depended, however, on the timing of breeding and increased as the season progressed (F1,43 = 5·337, = 0·026). Similarly, mean egg mass did not differ between the treatment groups (F1,109 = 1·004, = 0·319), but eggs laid later in the season were increasingly heavier (coefficient ± SE: 0·007 ± 0·003 g, F1,109 = 1·004, = 0·009). The interaction between laying date and the treatment was not significant (F1,109 = 0·181, = 0·672). Brood size at hatching was unaffected by the treatment (linear model: F1,43 = 0·098, = 0·755).

Offspring of females that laid eggs under increased predation risk were lighter than those of control mothers (Table 2). Additionally, they were smaller than control offspring both on day 8 and on day 14 (day 8: F1,39 = 11·719, = 0·001; day 14: F1,39 = 8·124, = 0·007; Fig. 1). Although on day 8 males were only marginally larger than females (F1,250 = 3·191, = 0·075), on day 14 they were significantly larger than females (F1,248 = 25·473, < 0·001). Additionally, nestling size was affected by nestling hatching order, reflected by mass-rank on day 2, whereby hatching later reduced body size both on day 8 and on day 14 (coefficient ± SE day 8: −0·423 ± 0·022, F1,250 = 380·525, < 0·001; day 14: −0·239 ± 0·023, F1,248 = 109·667, < 0·001).

Table 2. anova table for variables influencing nestling mass0·5
VariableCoefficient (SE)d.f.FP
  1. A repeated measurements model with nestling identification nested within nest of origin as random factors. Age was included as a categorical factor with three levels for mass models and two for sternum, wing, and tarsus models.

  2. pred = predator treatment; m = males.

Brood size0·006 (0·008)1,360·4430·510
Hatching date0·006 (0·005)1,361·5080·228
Maternal treatment (pred.)−0·080 (0·027)1,368·8290·005
Performer 3,364·7650·007
Rank on day 2−0·060 (0·003)1,251498·564<0·001
Sex (m)0·047 (0·003)1,25118·811<0·001
Age 2,5877353·926<0·001
Hatching date × Maternal treatment (pred.)−0·008 (0·007)1,361·4290·240
Rank on day 2 × Age (day 8)  (day 14)0·022 (0·004) 0·047 (0·004)2,58783·532<0·001
Maternal treatment (pred.) × Sex (m)0·023 (0·015)1,2512·3350·130
Maternal treatment (pred.) × Age (day 8)  (day 14)0·015 (0·016) 0·008 (0·016)2,5870·4090·665
image

Figure 1.  Offspring of mothers under increased predation risk were smaller than those of control mothers: the first principal component (PC1; mean ± 95% CI) of a principal component analysis combining nestling tarsus, wing and sternum lengths represents nestling size on days 8 and 14.

Download figure to PowerPoint

Examining each morphological trait separately revealed that the treatment affected morphological development differently for each trait (Table 3). Although sternum growth did not significantly differ between the two groups, wing growth of offspring from females exposed to higher predation risk was accelerated after day 8 compared to that of offspring from control females, as revealed by the significant interaction between nestling age and treatment (Table 3). Indeed, a post-hoc analysis revealed that while on day 8 nestlings of ‘high predation risk’-mothers had shorter wings (coefficient ± SE: −0·127 ± 0·054, F1,10 = 5·648, P = 0·039; Fig. 2), the difference on day 14 was no longer significant, albeit marginally, (coefficient ± SE: −0·073 ± 0·033, F1,10 = 4·821, = 0·053; Fig. 2). There was no sex-specific effect on wing growth (interaction between sex and treatment > 0·1). The effect of the treatment on skeletal size, represented by tarsus length, depended on nestling sex. Our treatment had no significant effect on males on either measurement day, whereas there was a negative main effect of the ‘higher predation risk’ treatment on female nestlings (post-hoc analysis: males –F1,33 = 2·266, = 0·142; females –F1,35 = 15·146, < 0·001; Fig. 3a,b).

Table 3.   Summary of variables influencing nestling morphological measures
MeasurementVariableCoefficient (SE)d.f.FP
  1. Three separate repeated measurements models, all with nestling identity nested within nest of origin as random factors. In wing models these were nested within plot of origin, also a random factor since it was significant for this response. Age was included as a categorical factor with two levels (days 8 and 14) for sternum, wing, and tarsus models. Significant values are marked in bold.

  2. pred = predator treatment; m = males.

Sternum0·5Brood size0·004 (0·005)1,360·7510·392
Hatching date0·000 (0·003)1,360·0210·886
Maternal treatment (pred.)−0·034 (0·018)1,363·3630·075
Performer 3,3650·337<0·001
Rank on day 2−0·021 (0·003)1,25257·800<0·001
Sex (m)0·027 (0·013)1,2524·3690·038
Age0·538 (0·020)1,293694·078<0·001
Hatching date × Maternal treatment (pred.)−0·003 (0·004)1,360·8130·373
Maternal treatment (pred.) × Sex (m)0·004 (0·018)1,2520·0500·822
Rank on day 2 × Age0·011 (0·004)1,2937·7880·006
Maternal treatment (pred.) × Age0·021 (0·017)1,2931·4420·231
Wing0·5Brood size0·042 (0·010)1,2616·725<0·001
Hatching date−0·000 (0·006)1,260·0020·962
Maternal treatment (pred.)−0·143 (0·043)1,1011·0550·008
Performer 3,266·7710·002
Rank on day 2−0·074 (0·004)1,252369·175<0·001
Sex (m)0·042 (0·019)1,2524·6860·031
Age2·124 (0·026)1,2936597·940<0·001
Hatching date × Maternal treatment (pred.)−0·003 (0·008)1,260·1910·665
Maternal treatment (pred.) × Sex (m)0·016 (0·027)1,2520·3800·538
Rank on day 2 × Age0·042 (0·005)1,29372·418<0·001
Maternal treatment (pred.) × Age0·091 (0·022)1,29317·183<0·001
TarsusBrood size−0·006 (0·036)1,360·0320·859
Hatching date−0·20 (0·022)1,360·8370·366
Maternal treatment (pred.)−0·531 (0·122)1,3618·912<0·001
Performer 3,3629·810<0·001
Rank on day 2−0·229 (0·016)1,252211·332<0·001
Sex (m)0·318 (0·074)1,25218·459<0·001
Age2·653 (0·114)1,292539·948<0·001
Hatching date × Maternal treatment (pred.)−0·006 (0·028)1,360·0500·823
Maternal treatment (pred.) × Sex (m)0·257 (0·103)1,2526·2260·013
Rank on day 2 × Age0·169 (0·022)1,29260·473<0·001
Maternal treatment (pred.) × Age0·115 (0·096)1,2921·4240·234
image

Figure 2.  The effect of the treatment on nestling wing length depended on age. Eight days old nestlings of mothers under increased predation risk had significantly shorter wings than the control nestlings. When the nestlings were 14 days old the difference between the groups was smaller and no longer significant. Mean ± 95% CI presented (asterisk represents < 0.05 in a post-hoc test; see text).

Download figure to PowerPoint

image

Figure 3.  Tarsus length of nestlings from the control (black) and predator (grey) groups: (a) males did not differ in their tarsus length between groups; (b) females from the high predation risk group had significantly smaller tarsi on both measurement days (Tukey adjusted post-hoc tests, day 8: z = −3.941 < 0.001; day 14: z = −2.652 P = 0.037). Figures show mean ± 95% CI.

Download figure to PowerPoint

Parental feeding rate was not different for parents hosting nestlings from predator treatment or control nests (Table 1; model summary Table 4). Larger broods received more feedings, and within pairs males fed on average 2·52 as many times as females (paired Wilcoxon signed rank test: V = 71, < 0·001).

Table 4.   Number of feeding events per hour of both parents combined
Model familyVariableCoefficient (SE)d.f.χ2P
  1. GLM with Quasipoisson error distribution. Performing the analysis with a Poisson error distribution resulted in overdispersion due to one possible outlier, though the result was not qualitatively different.

  2. pred = predator treatment. Coefficients presented are back-transformed.

QuasipoissonBrood size1·101 (0·041)15·6380·018
Hatching date0·972 (0·015)13·4580·063
Maternal treatment (pred.)0·968 (0·112)10·0820·774
Residual 39  

We captured 38 recruits during spring 2010. There was no significant difference in the recruitment rate between the two treatments (χ2 = 0·0056, d.f. = 1, = 0·94). The correlation between wing length on the day 14 and as adults was weak and not significant (= 0·184, = 0·269). Wings of recruits originating from the ‘higher predation risk’ treatment were significantly longer than those from the control (Tables 1 and 5). Recruits from the ‘high predation risk’ treatment did not significantly differ from control recruits in body mass or tarsus length, nor did they disperse to a significantly different distance from the nest (Table 5). The interaction between the treatment and sex was not significant in any of the models (for wing length = 0·079; for the rest > 0·16).

Table 5.   Treatment coefficients and permutated P-values for first year recruits
ResponseCoefficient-predator (SE)Permutation P-value
  1. Coefficients stem from mixed effects model. P-value from 10,000 permutations of birds’ maternal treatment keeping family ties and data structure consistent.

  2. m = males.

Dispersal distance (m)445·75 (358·09)0·217
Mass−0·36 (0·38)0·418
Wing1·78 (0·67)0·012
Tarsus−0·25 (0·25)0·342

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

In this study we manipulated the level of predation risk perceived by female great tits before and during egg laying. Nestlings from both groups were raised after hatching in non-manipulated host nests with no additional predatory stress. This design allowed us to assess whether varying levels of predation risk during ovulation may translate into maternal effects leading to different offspring phenotype (Mousseau & Fox 1998).

Our results indicate that females indeed influenced offspring phenotype depending on predation risk, and this could be realized via alteration of egg composition and/or incubation behaviour since we found no difference in mean egg mass between the treatments. Nestlings that hatched from eggs laid by mothers exposed to increased predator density during laying were smaller and lighter a few days before fledging, than those produced by control mothers (at the age of 14 days, where mean fledging age ± SD was 19·24 ± 1·34 days, in the great tit skeletal size and mass on day 14 are close to their final size. Perrins 1979; Gosler 1993). The shorter tarsi found for female offspring of the high predation risk treatment might also indicate an interaction between initial resource limitation caused by the maternal effect and competition from larger male siblings. Body mass and size at fledging have long been suggested as good predictors of nestling future survival, especially during winter (e.g. Perrins 1979; Tinbergen & Boerlijst 1990; Schwagmeyer & Mock 2008). Body size is also a determining factor for social dominance in great tits (Gosler 1993), and thus birds that are larger at fledging may have a higher probability to acquire high status and thereafter a good territory or a good mate, and have higher reproductive success. The results of offspring mass and size could thus be taken to suggest that a predator-rich environment around egg laying has, via maternal effects, detrimental consequences for offspring fitness.

Such an interpretation may imply that our observations are a result of a physiological constraint on females, which when stressed, may lay eggs with higher levels of stress hormones (Saino et al. 2005) and/or may change their foraging or breeding behaviours (e.g. Morosinotto, Thomson & Korpimäki 2010; Lima, 2009, Lima & Dill 1990; Lima 1990; Cresswell 2008). The level of stress hormones in eggs has strong effects on life-history and fitness related traits such as offspring growth, sex ratio and immunity (e.g. Saino et al. 2002; Groothuis et al. 2005; Rubolini et al. 2005; Warner, Radder & Shine 2009) as well as on fear and feeding related behaviours (e.g. Rubolini et al. 2005; Janczak, Braastad & Bakken 2006). We thus cannot entirely exclude that the harmful effects we detected here are a by-product of maternal stress during laying.

In addition to reduced mass and size, we also found that the growth rate of wings differed significantly between the two treatment groups, with offspring of mothers under increased predation risk exhibiting accelerated wing growth. According to the idea that our observations are because of maternal stress transferred to offspring via CORT in the eggs, it may be argued that these offspring had a late start for wing growth due to the effects of the treatment as observed in mass and size. Nestlings may allocate resources to speed up their wing growth in order to fledge at the same time as their siblings, even at the cost of reduced mass (Nilsson & Svensson 1996). Such compensatory growth may incur costs later in the individual’s life (e.g. Metcalfe & Monaghan 2001; Johnsson & Bohlin 2006). This view would be supported by the fact that the wing length of the high predation risk offspring did not differ from that of control nestlings, or might have even still been smaller, a few days before fledging.

On the other hand, the assemblage of all our results suggests that what we observed can be interpreted as an adaptive maternal strategy of females, preparing their offspring to a predator-rich environment by influencing wing growth and thus flight performance. Previous studies investigating the effects of maternal stress on the offspring often took the approach of injecting stress hormones into the eggs and reported negative effects on nestling growth. Unfortunately, most of the studies did not measure/report whether the injection of CORT into the eggs affected wing growth specifically, leaving this unknown. Saino et al. (2005) found no effect of CORT on wing length at fledging, but in the experiment of Love & Williams (2008b) female offspring of one group from CORT injected eggs had longer wings than their controls. Chin et al. (2009) hypothesized that some of the negative effects of in ovo CORT on nestlings may be balanced out by positive effects on flight performance. In an environment rich with post-fledging predators (e.g. sparrowhawks), increased flight performance by decreasing wing loading may have a substantial effect on the probability of survival (Chin et al. 2009). Indeed, mass has been shown to be costly for flying maneuverability (Witter, Cuthill & Bonser 1994) and fat reserves, which increase mass, have been suggested to be costly for escaping predators following the finding that in the presence of sparrowhawks adult great tits reduced their mass (Gosler, Greenwood & Perrins 1995; Gentle & Gosler 2001). Similarly, greenfinches exposed to models of sparrowhawks reduced their body mass even though they were allowed to feed ad libitum (Lilliendahl 1997). Reduced mass at fledging was also observed in blue tits (Parus caeruleus) when growing in the presence but not in the absence of sparrowhawks (Adriaensen et al. 1998). Additionally, in the presence of sparrowhawks heavier fledglings were selected against, suggesting an adaptive function of lower body mass (Adriaensen et al. 1998). In line with Chin et al. (2009) hypothesis, our results show both that wing growth rate was accelerated for offspring of females exposed to higher predation risk during egg laying, and that this effect persisted into adulthood whereby first year recruits from this group had longer wings. We show such an effect for the first time without a direct manipulation of yolk hormones, but rather by manipulation of the maternal environment.

The fact that growth rate of wings and feathers differed significantly between the two treatment groups, with nestlings of the ‘higher predation risk’ group having accelerated wing growth in the second part of their development, i.e. after day 8, may represent a higher investment into wing growth. Although wing length on day 14 did not differ between nestlings of the two groups, final wing length in great tits is only achieved some time later, and as opposed to skeletal size and to mass, wings on day 14 post-hatch are only about 2/3 of their final length (Perrins 1979; Gosler 1993). The longer wings of recruits from the risk treatment group seem to support that accelerated growth of their wings continued into adulthood.

As an alternative explanation, selection might have been stronger on birds from the high predation risk group than on control birds and thus lead to the difference in wing length of the recruits from the two groups. We would then, however, also expect a difference in tarsus length, which was not the case and contrasts with the situation on day 14 when high predation risk offspring had smaller tarsi and wings that did not significantly differ in length. If stronger selection would nevertheless have occurred, it would represent the cost paid by the high predation risk offspring for increasing their wing growth rate to fledge at the typical age, although starting from a worse start due to maternal stress (indicated by the smaller mass/size in younger ages). Accelerated wing growth could then be seen as a costly compensatory growth mechanism. It does not, however, explain the longer wings of the recruits since compensation would suggest wings to grow to an optimal or required length for fledging, and investing into longer wings would then be unnecessary (except for overcompensation which may presumably be costly as well). Also, structures and tissues grown rapidly due to compensation are more prone to development instability and weakness (Metcalfe & Monaghan 2003). Dawson et al. (2000), for example, found that bird flight feathers grown faster during molt were more sensitive to wear. We would thus expect wear to be more severe in wings of high predation risk offspring, whereas our findings indicate the opposite.

The accelerated wing growth of offspring from the ‘high predation risk’ group obviously required increased resource use for the wings. Since resources are limited, investing more into wing growth may exhaust resources from other body and growth functions such as mass, other skeletal growth, and the immune system (Nilsson & Svensson 1996; Lochmiller & Deerenberg 2000; Berthouly, Cassier & Richner 2008). The observed maternal effect may in this case work by changing the distribution of resources in growing offspring, affecting the trade-off between different traits. Differential growth, i.e. preferential allocation of resources to the growth of body systems and structures according to their importance has been previously shown in birds (e.g. Øyan & Anker-Nilssen 1996; Kunz & Ekman 2000; Mainwaring, Dickens & Hartley 2010). However, differential growth has been previously shown as a post-hatching response to resource levels. Further research is needed to test the possibility that differential growth is affected by maternal effects, and to show the mechanism by which this could happen. It may even be that CORT deposited in eggs has a role in regulating the allocation of resources to the growth of different morphological traits in nestlings: when injected to kestrel nestlings in a recent study (Muller, Jenni-Eiermann & Jenni 2009), CORT had a differential effect on the growth of different structures. Such an effect would explain the seemingly deleterious effects of in ovo CORT in studies that did not assess wing growth. Although we collected one egg from each treated nest, we cannot yet examine the role CORT played creating this effect due to several reasons. First, measuring the levels of CORT in the great tit yolks requires pooling together many yolks due to the seemingly low amount of CORT present (T. Groothuis, pers. comm.). This would reduce our sample size too much for any statistical comparison. Secondly, a recent study showed that concentrations of CORT measured in chicken eggs may in reality reflect concentrations of progesterone and its precursors (Rettenbacher, Möstl & Groothuis 2009), thus putting in doubt any result we could obtain. Further measurements on the yolk composition from the eggs we collected are possible (e.g. concentration of other hormones such as testosterone), but still require work.

We found no difference in the feeding rate of foster parents bringing up nestlings of the two experimental groups. In altricial birds, nestling begging is the major factor influencing parental feeding rate (reviewed by Tarwater, Kelley & Brawn 2009). Generally, parents respond to an increase in begging by increasing feeding effort. Several studies have found a change in nestling begging rate or intensity after an artificial increase in the levels of the stress hormone CORT (e.g. Kitaysky et al. 2003; Rubolini et al. 2005). Yet, the fact that we observed no difference in feeding rate suggests that there was no difference in the begging behaviour of the nestlings of the two origins. The quantity of food nestlings actually received could still have been different, however, if parents had changed, as a response to different begging intensity, the food load or type brought to nestlings (e.g. Wright et al. 1998; Eggers, Griesser & Ekman 2008; Tarwater, Kelley & Brawn 2009). However, great tits are generally single prey loaders (Royama 1970) and changing the type of prey parents brought to the nest, for example by selectively choosing larger prey items or specific prey types, would probably require increased foraging time and thus reduced feeding rate, which was not observed. Thus, it seems reasonable to believe that the results observed for nestling growth are a product of physiological differences between the nestlings of the two treatments rather than differences in begging behaviour.

Additionally, it is possible that the adaptive value of the maternal effect would be further revealed in older offspring adopting different behaviours. In a recent study on crickets Storm & Lima (2010) described a trans-generational maternal effect on offspring behaviour resulting from exposure of mothers to predator cues. Offspring of mothers exposed to predatory cues exhibited a more risk aversive behaviour in the presence of predatory spiders, and this behaviour seems to have contributed to their higher survival when facing the risk of predation. It is possible that such behavioural effects would also arise in the offspring in this study, yet these would be visible only when older. It has already been suggested that high levels of CORT in bird eggs may alter the risk taking behaviour of offspring (Groothuis et al. 2005). Such a response may be adaptive if it increases the survival probability of offspring in a predator-rich environment (Storm & Lima 2010).

Finally, exposure to predators during laying may act in mothers on the trade-off between current and future reproduction where mothers favour a reduced investment into the current brood to the benefit of a higher investment in future broods. Hayward & Wingfield (2004) observed that female Japanese quail injected with CORT during egg laying produced nestlings that were smaller in the first weeks of their lives but not in adulthood. They suggested that reduced offspring growth rates may benefit parents by reducing parental effort. As in their study, we found no difference in mass or tarsus length between first-year recruits. However, in order to strongly assess the validity of this hypothesis a much larger sample size of offspring would be needed, and/or following female lifetime reproductive success. We did not find a significant difference in parental feeding rate between the foster parents of nestlings of both groups, which would be one of the predictions of this hypothesis.

To summarize, although it has been previously shown that predation risk increases the stress response in reproducing mothers and that this type of response may be deleterious (e.g. Saino et al. 2005), most studies focus on the stress hormones alone, ignoring other possible mechanisms. Here, we show for the first time via environmental rather than hormonal manipulation that predation risk can induce maternal effects in free-living birds, affecting offspring growth. We suggest that these effects are adaptive by decreasing wing load thereby enhancing survival.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank Alan Haynes, Joshka Kufmann, and Joana Agudo for their assistance with field work. Fabrice Helfenstein gave useful comments on an early version of the manuscript. We are also thankful to three anonymous referees for their helpful remarks. Taxidermic models were kindly provided by Sirpa Kurz of the Naturhistorisches Museum, Bern. The study was financially supported by a Swiss National Science Foundation grant (3100A0-102017) to HR, and performed under animal experimentation permit 117/07 of Canton Bern Animal Experimentation Committee to MC.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References