Maternal effects on offspring Igs and egg size in relation. to natural and experimentally improved food supply

Functional Ecology 008,, 68 690 doi: 10.1111/j.1365-435.008.0145.x Maternal effects on offspring Igs and egg size in relation Blackwell Publishing Ltd to natural and experimentally improved food supply P. Karell*,1, P. Kontiainen 1, H. Pietiäinen 1, H. Siitari and J. E. Brommer 1 1 Department of Biological and Environmental Sciences, Bird Ecology Unit, University of Helsinki, P. O. Box 65 (Viikinkaari 1), FI 00014, Finland; and Department of Biological and Environmental Science, Evolutionary Research Unit, University of Jyväskylä, P. O. Box 35, FI 40014, Finland Summary 1. Maternal effects have been suggested to function as a mechanism for transgenerational plasticity, in which the environment experienced by the mother is translated into the phenotype of the offspring. In birds and other oviparous vertebrates where early development is within the egg, mothers may be able to improve the viability prospects of their offspring at hatching by priming eggs with immunological and nutritional components.. We studied how resource availability affects maternal investment in offspring by feeding Ural owl (Strix uralensis, Pall.) females prior to egg-laying in 3 years of dramatically different natural food conditions. 3. Supplementary feeding prior to laying increased body mass and the level of Igs of females measured at clutch completion. Supplementary fed Ural owl females laid larger eggs than control females, and had offspring with higher levels of Igs at hatching compared to offspring of control females. 4. We found variation in maternal allocation of resources to the eggs in response to environmental conditions: during a year of rapidly declining food abundance, maternal Igs in hatchlings were higher, whereas egg size was smaller compared to years with a more stable food supply. 5. Egg size had a positive effect on offspring body mass at fledging, whereas Igs at hatching did not affect Igs at fledging. 6. We conclude that maternal body condition and maternal Igs, as well as hatchling Igs and egg size are limited by food resources during egg production. Hatchlings rely on maternally derived Igs and, hence, our results suggest that mothers with high levels of Igs passively transfer higher Igs levels to their eggs instead of active manipulation of Igs levels in eggs. Ural owl egg size appears to be highly sensitive to short-term changes in food abundance, with important consequences for nestling growth. Key-words: egg size plasticity, ELISA, transgenerational immunity, vole cycle Functional Ecology (008) xx, 000 000 Introduction Resource limitation constrains reproductive decisions and effort (Martin 1987). In addition to the number of offspring to be produced in a given breeding attempt, also the phenotypic quality of the offspring can be modified by the mother (Mousseau & Fox 1998). One determinant of phenotypic quality is propagule size, which is simultaneously both a maternal and offspring character (Bernardo 1996). Propagule *Correspondence author. E-mail: patrik.karell@helsinki.fi size has been found to determine growth and survival in various invertebrates (e.g. Sinervo & McEdward 1988; Bridges & Heppell 1996; Steigenga et al. 005) and vertebrates (e.g. Semlitsch & Gibbons 1990; Chambers & Leggett 1996; Räsänen, Laurila, & Merilä 005). In the past decade many avian studies have focused on the variation in egg size. Increased egg size through food supplementation has been linked to improved hatchability of the egg, increased size at hatching, early growth and survival during the nestling stage (reviewed in Williams 1994; Christians 00, see also Grindstaff, Demas & Ketterson 005). Despite the large number of studies, the relevance of variation in avian egg size is still equivocal (Christians 00, see also Kontiainen et al. 008). 008 The Authors. Journal compilation 008 British Ecological Society

Resource limitation and maternal effects 683 Therefore, ecologists have recently focused on egg composition as an estimate of propagule quality since maternal transfer of components to the offspring via the egg may be an important maternal adaptation that can enhance early development and survival of the progeny (Mousseau & Fox 1998). In birds, females have been shown to be able to modify the contents of eggs in terms of immunoglobulins (Igs) (Grindstaff, Brodie & Ketterson 003), androgens (Verboven et al. 003; Groothuis et al. 005), and antioxidants (e.g. Royle, Surai & Hartley 001, see also Groothuis et al. 006). When offsprings are exposed to novel pathogens and parasites, and since the offsprings cannot produce their own Igs during early ontogeny, they are entirely dependent on innate immunity and maternal immune components at hatching (Klasing & Leschchinsky 1998). Furthermore, there is evidence that maternal immunocompetence is associated with maternal immune transmission to the eggs (Smith et al. 1994). Consequently, maternal allocation of immunological components to the offspring to improve its viability and fitness prospects has received a lot of attention (cf. Saino et al. 00; Grindstaff et al. 003). In general, maternally derived immune components in offspring are induced by the disease environment experienced by the mother (Gasparini et al. 001; Buechler et al. 00; Staszewski et al. 007). In the wild, such components have been found to be positively correlated with maternal body mass (Hargitai, Prechl & Török 006) and improved food conditions prior to egg-laying (Pihlaja, Siitari & Alatalo 006, but see Grindstaff et al. 005), and negatively correlated with increased maternal workload prior to egglaying (Kilpimaa, Alatalo & Siitari 007). Thus, maternal transmission of immunological components to the eggs may be limited by the environmental conditions the mother experiences. During ontogeny increased levels of maternally derived immune components may be beneficial, as they may enable offspring to invest in other crucial elements of growth. In offspring, higher maternal Igs have been positively associated with an offspring s own Ig production (Grindstaff et al. 006; Pihlaja et al. 006) and, furthermore, a trade-off between growth and immunity during the nestling period has been documented for both blue tits (Cyanistes caeruleus, Brommer 004) and magpies (Pica pica, Soler et al. 003; Pihlaja et al. 006). In this paper, we study the effects of supplementary food prior to egg laying on Ural owl (Strix uralensis) female Ig level, egg size and maternal transmission of Igs. Ural owls mainly prey on voles, which show dramatically cyclic population densities in Fennoscandia (Norrdahl 1995). Ural owls are site-tenacious and monogamous (Saurola 1987). Vole abundance fluctuates in a 3 year cycle (low, increase and decrease phase) in southern Fennoscandia (Sundell et al. 004), which strongly determines Ural owl reproductive output (Pietiäinen 1989; Brommer, Pietiäinen & Kolunen 00a). With an average breeding life span of 3 5 years (Brommer, Pietiäinen & Kolunen 1998), Ural owl females will face widely different environmental conditions during their lives. In good vole years female Ural owls gain ample fat reserves and lay the largest clutches (Pietiäinen & Kolunen 1993), and the largest eggs (Kontiainen et al. 008). Offspring born in the increase vole phase, when food conditions improve during the first year, have higher probability of recruitment than offspring born in the decrease phase, when food conditions rapidly deteriorate (Brommer, Pietiäinen, & Kokko 00b). Hence, the vole cycle generates differences in offspring values between vole cycle phases and selects for higher reproductive investment in the increase vole phase (Brommer, Kokko & Pietiäinen 000). By supplementary feeding nests prior to laying in 3 years with different natural food supply (one vole cycle), we investigate whether Ural owl maternal investment in offspring, in terms of Ig levels at hatching and egg size, is food limited under all environmental conditions, and whether maternal investment into egg size and hatchling Igs vary in different natural food conditions. Material and methods The experiments were conducted in 004 006 in a study area of c. 1500 km in Päijät-Häme, Finland, where there are around 180 Ural owl nest boxes available. The territories that where included in the experiment were assigned in advance on the basis of territory occupancy, which in turn was based on either breeding activity in the previous year or on signs of scrapings in the nest box material in autumn. Half of these active territories were assigned to the fed group and the nearest active territory was assigned to the control group, thus forming a treatment pair (see also Statistical Analyses ). By feeding in a nest box in an occupied territory we can be confident that no other than the territorial owls eat the food, since Ural owls defend their territory from intruding con-specifics. Each year new territories, and hence new individual owls, were used in the experiment. In birds of prey males deliver food items to the nest during the pre-laying period to nourish the female into breeding condition (Newton 1979; Meijer, Daan & Hall 1990). In Ural owls, males deliver food for the female to the nest. Thus, by adding supplementary food to the nest boxes we simulated improved territory quality and/or male hunting success. Feeding and visiting of control nests started in mid-february, which was more than 1 month before estimated egg-laying (median laying date 004 006: 1 March, range 11 March 6 April). Supplementary food, which consisted of newly hatched dead rooster chicken Gallus gallus, was delivered to the nests at 5 6-day intervals and control nest boxes were checked with a similar interval to standardise disturbance in both groups. Each feeding consisted of 500 g of rooster chicken, food remaining from the previous feeding event was weighed and removed. We stopped the supplementary feeding in a nest at clutch completion. Supplementary food was eaten prior to laying in all nests belonging to the feeding treatment in this study. On average 3095 ± 38 standard error (SE) g (range 1190 4465 g) of supplementary food was eaten per nest prior to laying. Each experimental year was classified according to the natural vole density. To estimate vole density we trapped voles bi-annually in mid-june when owlets have fledged and early October around the time when owl young become independent. The voles were caught by snap trapping in 33 localities throughout the whole study area. Each trapping locality had three quadrates where one quadrate was placed in an open habitat (clear-cut or re-planted forest), one at the interface between open habitat and forest, and one in the forest. Snap traps were set in quadrates (15 15 m) with three traps in each corner as described in Myllymäki et al. (1971). All traps were triggered for two consecutive nights (544 trap nights in total). Vole 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

684 P. Karell et al. abundance was expressed as a percentage, the number of field- and bank voles trapped per 100 trap nights. For more details on the study area and methods, see Pietiäinen (1989) and Brommer et al. (00a). DATA COLLECTED Eggs were individually marked with a pencil, and their width and length were measured with a calliper (accuracy of 0 05 mm). Egg volume was obtained from the measurements using a speciesspecific formula provided by Pietiäinen, Saurola & Väisänen (1986). The laying interval of Ural owls is c. days, which enabled us to accurately determine the laying sequence of most eggs. Ural owl eggs have a white egg shell, and in case two new eggs had been laid between consecutive nest visits, their laying order was determined on the basis of darkening of the egg shell caused by fine particle in the sawdust getting attached to it. The presumed laying order was later checked at hatching, and the hatching order corresponded with estimated laying order in all cases. Incubating females were caught as soon as possible after clutch completion (median 13 days after laying the first egg). Only 35 of a total of 4 females could be caught within the incubation period, and hence, these 35 females are used in the analyses of female traits. The number of days the female was caught after laying the first egg was not significant in the models of female traits. Females were weighed with a 1500 g Pesola spring balance (5 g accuracy). Body size was estimated from the length of radius-ulna, measured with a ruler (1 mm accuracy) from the elbow to the carpal joint (Pietiäinen & Kolunen 1993). From the females c. 75 µl of blood was drawn into capillary tubes after brachial venipuncture. None of the female Ural owls involved in the experiment deserted the nest due to handling. Nests were checked with a 1 3 day interval around the estimated time of hatching to accurately determine which offspring hatched from which egg. Ural owl offspring hatch asynchronously with - days interval, except in clutches of three or more eggs in which the first two or three eggs are more synchronous. In most cases we could determine exactly from which egg the offspring hatched by direct observation. In a few cases (n = 10), the two first chicks in a brood had both hatched since the previous visit. Hatching takes 1 3 days, and in these ten cases we assigned the largest offspring to the egg where hatching was most advanced on the previous visit. Hatchlings were individually marked with a felt-tip pen, weighed and measured. From the hatchlings c. 60 µl of blood was drawn in a capillary tube after puncturing the brachial vein as soon as possible after hatching (0 3 days of age). Some hatchlings could not be blood sampled due to natural mortality immediately after hatching, because the female feeds any dead young to the remaining siblings. Prior to fledging, when the oldest nestling was c. 4 days old (hatching day = 0) all chicks in a brood were ringed and measured, and a blood sample was taken (mean age of nestlings, 3 9 days ± 0 SE). The blood capillaries were kept in the cold and were centrifuged within 1 h after sampling at 375 g for 5 min in order to separate the blood cells from the plasma. The plasma was separated from the blood and stored at 18 C for an analysis of Ig concentration. Some capillaries were destroyed in the centrifuge or blood ran out of the capillaries when centrifuged, which explains the variation in sample sizes. GENERAL ANTIBODY CONCENTRATION Total Ig concentrations in the plasma of female parents and each hatchling (days 0 and 4) were determined with the ELISA method. Briefly, 96-well microplates (Immuno Plate Maxisorp, Nunc Co., Rochester, NY) were first coated overnight at +4 C with commercial anti-chicken Ig antibody (10 µg/ml, C-6409, Sigma Chemical Co., St Louis, MO). After emptying, the wells were saturated for 1 h with 1% bovine serum albumin (BSA, Roche Diagnostics, Basel, Switzerland) prepared in phosphate buffered saline (PBS, ph 7 4), and then washed three times with PBS-Tween 0 (0 5%). Samples were diluted with 1% BSA PBS and each sample incubated in duplicates (50 µl/well) for 3 h at room temperature. In addition to a plasma sample, each plate received a series of standard solutions. Pooled plasma samples from immunized adults (with diphtheria-tetanus vaccination, P. Karell, H. Pietiäinen, H. Siitari, P. Kontiainen, T. Pihlaja & J. E. Brommer, unpublished data) served as standards. Ig values in samples are presented relative to this standard; arbitrary value of ten equals the mean level of individuals used for the pooled sample. After washing, alkaline phosphatase conjugated anti-chicken Ig antibody (A-9171, Sigma Chemical Co.) was added and the plates were incubated overnight at +4 C (dilution of 1 : 0 000). Finally after last washing, p-nitrophenyl phosphate (1 mg/ml, Sigma 104 Phosphatase Substrate) in a diethanol amine buffer (1 mol/l, ph 9 8) was applied. The optical density was read at 405 nm with a plate reader (Multiskan Plus, Flow Laboratories, Helsinki, Finland). SEX DETERMINATION OF OFFSPRING Ural owl nestlings can not be reliably sexed by their morphology. We extracted DNA from the blood samples using salt extraction. We amplified fragments of the sex-chromosome linked CHD gene in the offspring using the protocol and primers developed by Fridolfsson & Ellegren (1999). Fragments were separated on a % agarose gel stained with ethidium bromide and scored under UV light. For more details, see Brommer et al. (003). STATISTICAL ANALYSES Each year different territories (different parental birds) were used for feeding treatment and control treatment. Thus, in the analyses all data was pooled with year as a co-factor. All biologically meaningful interactions were tested, but for brevity we report only the main effects and the treatment by year interaction. Ig concentrations were log-transformed in all analyses. All analyses were done in R 6 1. Female traits were analysed with linear models with emphasis on treatment and year effects and their interaction. We used a stepwise backwards modelling approach as described in Crawley (00) to achieve the minimal adequate model (MAM) for the data. Variables were dropped from the model if P > 0 05 in the F-test to test between the models. We report F-test statistics of both retained variables and tests between models as described in Crawley (00). Analyses of offspring traits (egg size, hatchling Ig concentration, fledgling body mass and fledgling Ig concentration) were analysed with stepwise backwards linear mixed models (LMM) to achieve the MAM for the data. Only hatchlings less than 70 grams, which are < 3 days old (H. Pietiäinen, unpublished data), were included in the hatchling Ig analyses in order to exclude the possibility that the hatchlings own Ig production would affect the results. We first ran the models with brood identity nested in treatment pair as a random effect to account for variance due to treatment pair. The variance of treatment pair was, however, virtually zero in all models and we therefore omitted the nested design, and here we report results from stepwise backwards LMMs, where the variance due to the random effect brood identity has been accounted for. The LMM was 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

Resource limitation and maternal effects 685 solved using Maximum Likelihood during the stepwise backward procedure (Pinheiro & Bates 000). The significance of the random effect was based on a likelihood ratio test (LRT), where -two times the difference in likelihood of a model with and without this random effect was tested as a χ value with one degree of freedom (Pinheiro & Bates 000). The results of the stepwise backwards LMMs are reported as the variables retained in the MAMs followed by the dropped variables from the full models in the order they were dropped. Test statistics of the dropped variables refer to a LRT (χ ) between the model retaining and the model excluding the dropped variable, where a non-significant (P > 0 05) likelihood ratio indicates a better fit for the model without the dropped variable (Crawley 00). Therefore, also marginally significant fixed effects (P ~ 0 05) may be retained in the MAM if that model has a better fit according to the LRT than the model without the variable. For more details on the stepwise backwards procedure in LMMs with normal errors, see Crawley (00). Offspring survival was analysed as a Generalised LMM (GLMM) with binomial errors, and brood identity as a random effect. In order to achieve the MAM by model simplification we used Laplace approximation of likelihood for the binomial GLMM, implemented with the lmer function (Matrix package, by D. Bates and M. Maehler). The laplacian method is a better approximation of maximum likelihood than the penalized quasi likelihood (PQL) method (Breslow & Clayton 1993; Wolfinger & O Connell 1993) and allows for a comparison of models with different fixed effects. We tested between the models with likelihood ratio tests between the model retaining and the model excluding the variable, and used AIC to compare between models. Both the AIC and the likelihood ratio tests produced the same MAM for the data. This GLMM approach does not allow F-tests of fixed effects. We therefore report the χ statistics of the likelihood ratio tests between models in the results. The direction of the effect of a variable is based on evaluation of its (significant) coefficient. Results NATURAL FOOD SUPPLY DURING THE FEEDING EXPERIMENTS The 3 years during which the food supplementation experiments were conducted varied drastically in terms of natural food supply (Fig. 1). The breeding season of 004 was a low phase of the vole cycle (Fig. 1). Vole numbers increased to 18 85 % in autumn 004 and stayed high through the owl breeding season in the vole cycle increase phase of 005 (10 % in June). The vole density further increased and reached its peak in autumn 005 at 36 4%, after which it rapidly declined during the owl breeding season of 006 to 0 1% in June, which was a vole cycle decrease phase (Fig. 1). The vole cycle reflected Ural owl breeding activity in the study population as only 9% (18 of 61) of the active territories produced clutches in 004, whereas in 005 76% (5 of 68) and in 006 58% (49 of 85) of the active territories produced clutches. CLUTCH SIZE, EGG SIZE AND TIMING OF EGG LAYING Clutch size and timing of egg laying varied between years (clutch size, mean ± SE: 004, ± 0 15 (n = 9), 005, 5 33 ± 0 18 (n = 4); 006, 4 75 ± 0 31 (n = 8); laying date, Fig. 1. Vole abundance in autumn (filled circles) and spring (open circles) during the years of the supplementary feeding experiment (004 006). Grey bars indicate the phases in which the Ural owls breed. Values are mean ± SE (see Methods for further details). mean ± SE: 004, 14 April ± 3 (n = 10); 005, March ± 0 7 (n = 4); 006, 0 March ± 1 9 (n = 8)). Supplementary food did not advance timing of laying (year F,39 = 81 38, P < 0 0001; dropped variables: treatment year: F,36 = 61, P = 0 09; treatment F 1,38 = 0 98, P = 0 33,) and did not increase clutch size (year: F,38 = 49 71, P < 0 0001; dropped variables: treatment year F,35 = 0 86, P = 0 43; treatment F 1,37 = 0 01, P = 0 9). However, in all years fed females laid larger eggs than control females (Fig. c, MAM: treatment: F 1,38 = 15 49, P = 0 0003; year: F,38 = 4 57, P = 0 0; random effect: brood ID 69% of variance, LRT: = 119 37, P < 0 0001; dropped variable: treatment year: = 1 87, P = 0 39). From all nests included in the experiments in 004 006, 86% (161 of 187) of the eggs hatched, but there was no difference in hatchability between eggs in control and fed territories (fed 8 6% (76 of 9), control 89 5% (85 of 95), χ = 1 84, d.f. = 1, P = 0 17). FEMALE CONDITION, IMMUNITY AND TRANSMISSION OF IGS TO THE OFFSPRING Supplementary food prior to laying increased the body mass of incubating females, and there were also marked differences across the years of varying natural food supply (Fig. a, MAM: treatment: F 1,31 = 6 5, P = 0 0; year: F,31 = 5 4, P = 0 01; dropped variables: treatment year: F,8 = 0 86, P = 0 43; radius-ulna: F 1,30 = 1 78, P = 0 19). Supplementary food also increased the circulating Ig concentration as fed females had higher levels of Igs than controls (Fig. b, MAM: treatment: F 1,33 = 6 3, P = 0 0; dropped variables: treatment year: F,9 = 43, P = 0 11; year: F,31 = 0 58, P = 0 56). There was no direct relationship between a female s Ig level at clutch completion and her offspring s Ig level at hatching, as Ig in incubating females did not directly explain the mean Ig levels in her brood (LM: year: F,3 = 5 78, P = 0 007; dropped variables: maternal Ig year. F,9 = 0 6, P = 0 77; 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

686 P. Karell et al. Fig.. Effects of supplementary food on female size-corrected weight (a) and female Ig levels (b) at clutch completion, egg size (c) and hatchling Ig (d). Individuals from the fed group are denoted by filled squares and controls by open circles and sample sizes are given above bars. Not all females could be caught, which explains the variation in sample size between treatment groups in (a) and (b). Sample sizes differ in (c) and (d) because all eggs did not hatch and some hatchlings could not be sampled within 3 days after hatching. Ig level is the log transformed value of Ig concentration from the ELISA analysis. Egg size is the volume of individual eggs (see Methods for details). Values are mean ± SE. maternal Ig: F 1,31 = 0 61, P = 0 44,). However, the Ig levels of offspring at hatching were food-limited in all years, as hatchlings of fed females had higher Ig levels than the offspring of control females in all 3 years (Fig. d, MAM: maternal treatment: F 1,34 = 5 71, P = 0 0; year: F,34 = 18 44, P < 0 0001; hatching order: F 6,81 = 10, P = 0 06; sex: F 1,81 = 4 00, P = 0 05; random effect: brood ID: 40% of variance, LRT: = 1 15, P < 0 0001; dropped variables: maternal treatment year: = 0 49, P = 0 78). The MAM retained the effects of sex and hatching order and showed that Ig levels in hatchlings tended to be higher in female offspring and lower in later hatched offspring. Furthermore, in 006 Ig levels in hatchlings were substantially higher than in the two other years (Fig. d, MAM: year 006, t 34 = 3 87, P = 0 0005). CONSEQUENCES OF MATERNAL TREATMENT ON OFFSPRING SURVIVAL, IGS AND SIZE AT FLEDGING We analysed survival of offspring during the nestling period with a GLMM with binomial errors and brood ID as random effect. Survival of offspring decreased with hatching order in the brood (MAM: hatching order: = 1 5, P = 0 00, coefficients of hatching order 4 6: all P < 0 04), but was not improved in offspring of fed females compared to controls or by any other covariable (dropped variables: treatment year: = 3 67, P = 0 16; Ig at hatching: = 0, P = 1; egg size: = 0 3, P = 0 64; treatment: = 1 69, P = 0 19; year: = 4 65, P = 0 10). Egg size had an effect on growth, as offspring that hatched from larger eggs reached larger body mass at fledging (Table 1, egg size). In 004, fledglings weighed less than in 005 or 006 (Table 1, year). Later hatched offspring had a lower body mass at fledging (Table 1). Investigation of the coefficients further showed that in particular offspring in 4th position were smaller than the first hatched offspring, and that relatively few offspring hatched in a later hatching position (Table 1). Ig levels at the fledging stage (around day 4) were on average 33 times (4 9 10 6 ± 1 1 10 6 SE/1 8 10 5 ± 4 99 10 4 SE U ml 1 ) higher than at hatching, and the difference in Ig at hatching between fed and control group was levelled out by day 4 (Table, maternal treatment). Ig levels of the fledglings varied markedly across years and were highest in 005 (Table, year). Maternal treatment had a different effect on the Ig levels of her fledglings in different years, revealed by a significant interaction between maternal treatment and year (Table, feeding year). In particular, in 006 fledglings of fed females had higher Ig levels than fledglings of control females (Table, fed year 006). In both fed and control nests male offspring had higher Ig levels than female offspring (Table, sex). Discussion Supplementary food prior to laying increases a Ural owl female s body mass and Ig level in the incubation period. These fed females do not lay larger clutches, but instead lay larger eggs and have hatchlings with higher Ig levels than control females. Strikingly, these effects of supplementary food persist regardless of the yearly natural food supply, although in particular the vole cycle s increase phase (005 in this study) creates highly favourable natural conditions (cf. Brommer et al. 000). Our results show that egg size and Igs in hatchlings are maternal effects that are strongly resource limited, suggesting that such investments are costly to a Ural owl female, and can only be undertaken when increased food resources are at a female s disposal. Furthermore, increased 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

Resource limitation and maternal effects 687 Table 1. Stepwise backwards linear mixed model (normal errors, brood ID as random effect) of offspring body mass on day 4 Fixed effects Variable Coefficient ± SE Test P Minimal adequate model Constant 484 9 ± 15 3*** Egg size 3 3 ± 1 5** F 1,80 = 11 38 0 001 Year 005 15 4 ± 16 0 F,37 = 7 5 0 00 006 8 8 ± 17 8 Sex 36 ± 7 4*** F 1,80 = 13 3 0 0005 Radius-ulna length 3 8 ± 0 7*** F 1,80 = 58 46 < 0 0001 Hatching order (38) 8 0 ± 9 3 F 5,80 = 67 0 03 3 (7) 9 4 ± 10 3 4 (19) 4 0 ± 1 1*** 5 (7) 18 6 ± 17 4 6 (3) 9 61 ± 5 1 Dropped variables Maternal treatment year = 1 0 0 55 Maternal treatment = 0 37 0 54 Random effect Variance (95% CI) % Brood ID 397 (138 8 1136 0) = 5 7 0 0 Egg size and radius-ulna length were standardized (difference from mean). Sex was coded as 0 for female and 1 for male and the reported coefficients for the factor year are in comparison to year 004. The non-significant variables maternal food treatment (control = 0, fed = 1) and Maternal treatment year are dropped in the minimal adequate model. The test statistics of the dropped variables refer to a log-likelihood ratio test between the model in which the variable is retained and in which it is dropped. Stars indicate the significance of the coefficients in the model (t-test, *** < 0 001, ** < 0 05). Sample sizes for different hatching order are given in brackets. The significance of the random effect was tested with a log-likelihood ratio test (χ ). Table. Stepwise backwards linear mixed model (normal errors, brood ID as random effect) of Ig levels in offspring on day 4 Fixed effects Variable Coefficient ± SE Test P Minimal adequate model Constant 5 71 ± 0 18*** Year 005 0 70 ± 0 19** F,3 = 1 96 < 0 0001 006 0 39 ± 0 3 Feeding 0 33 ± 0 8 F 1,3 = 0 0 0 89 Sex 0 5 ± 0 06*** F 1,73 = 15 46 0 000 Feeding year fed 005 0 7 ± 0 31 F,3 = 3 53 0 04 fed 006 0 80 ± 0 34* Dropped variables Hatching order = 79 0 73 Ig at hatching = 0 68 0 41 Egg size = 1 43 0 3 Random effect Variance (95% CI) % Brood ID 0 053 (0 04 0 115) 38 = 15 4 < 0 0001 The coefficients of year are in comparison to year 004, sex was coded as 0 for female and 1 for male, and maternal treatment ( feeding ) was coded as 0 for offspring of control females and 1 for offspring of fed females. Test statistics are equal to tests in Table 1. Stars indicate the significance of the coefficient in the model (*** < 0 001, ** < 0 01, * < 0 0, < 0 10). egg size has a strong positive effect on offspring body mass at fledging, indicating that there are clear benefits from developing in a large egg (although there is an upper limit, since the largest Ural owl eggs have reduced fitness, see Kontiainen et al. 008). Female body reserves respond both to natural and artificially increased food supply, in accordance with the long-term descriptive pattern (Pietiäinen & Kolunen 1993). Female Ig levels, on the other hand, mainly depend on feeding, although our power to detect different effects in the three phases of the vole cycle may be limited by small sample size in the low phase. Nevertheless, different components in the diet may differentially affect immune function and general body condition. Carotenoids and antioxidants can function as immunostimulants (Chew 1996, but see Blount et al. 00) and farmed chicken (and their hatchlings) are generally rich in 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

688 P. Karell et al. carotenoids and other antioxidants (Karadas et al. 005; McGraw & Klasing 006). Therefore, it is possible that the supplemented rooster chicken hatchlings in this study were rich in carotenoids and antioxidants. Carotenoids in the maternal diet can indeed enhance immune function in hatchling broiler chicken (McWhinney, Bailey & Panigrahy 1989; Haq, Bailey & Chinnah 1996), which may explain the clear pattern of increased levels of Igs in both female and hatchling Ural owls. Alternatively, increased Ig levels could result from recent infections. However, we find it unlikely that the supplementary food would be a major source of infection compared to the natural prey, voles. Instead, we find it most likely that supplementary fed Ural owl females improve both their body mass and their immune function. There is laboratory evidence that a mother s immunity can be transmitted to the offspring via the egg (Smith et al. 1994). The effect of maternal resource limitation on Ig levels in hatchlings has also been found in studies of other altricial birds. Hatchlings of supplementary fed magpie (P. pica) females have increased levels of Igs (Pihlaja et al. 006), and pied flycatcher (Ficedula hypoleuca) females that were forced to work harder prior to laying produced offspring with lower Ig levels (Kilpimaa et al. 007), but, contrastingly, female kittiwakes Rissa tridactyla appear to allocate more Ig to the eggs under adverse than favourable food conditions (Gasparini et al. 007). Nevertheless, only few other studies of maternal effects in wild animals have investigated Ig levels in both mothers and their progeny (Gasparini et al. 00; Grindstaff, et al. 006; Hargitai et al. 006). Our results are to our knowledge the first to experimentally test the role of food resources on mother and offspring Ig levels in different natural food conditions. We find that after experimentally improving the food conditions prior to laying, female Ural owls have increased concentrations of circulating Igs and that the hatchlings of these mothers also have higher concentrations of circulating Igs. Increased Ig levels in hatchlings correspond well to Ig levels in eggs (Gasparini et al. 00; Pihlaja 006) and, therefore, hatchling Ig level is most probably a direct consequence of passive transmission from a mother with higher Ig levels (see also Smith et al. 1994). Nevertheless, mothers may actively allocate differently to eggs of different hatching order (Pihlaja et al. 006) or differently to first clutches and replacement clutches (Gasparini et al. 007), but we only find a weak tendency for within-brood (sex and hatching order) variation in Ig at hatching. We do not find a positive relationship between female Ig levels at clutch completion and Igs at hatching. However, such a relationship between a female s and her offspring s Igs may be hard to discover because of the time-lag between sampling of the mother (after egg-laying) and the time of her Ig transfer to the eggs (days prior to egg-laying). Furthermore, female immune function has been found to be decreasing and highly variable around the time of egg-laying (Oppliger, Christe & Richner 1996; Saino et al. 00). Egg size is considered to be a maternal trait that is rather inflexible on the individual level, and relatively little affected by food conditions (reviewed in Christians 00). We find that Ural owl females are able to respond to improved food conditions by a remarkable increase (up to 16%) in egg size (see also Kontiainen et al. 008). In an earlier review, Williams (1994) concluded that an increase in egg size may be a small additional cost to the female that may be beneficial for the offspring during early development. Our results clearly demonstrate that egg size is a highly plastic trait that is limited by food conditions prior to laying. We also find that offspring hatched from larger eggs grow larger by the end of the nestling period. However, to separately evaluate pre- and post-hatching effects of egg size, hatchling Ig level and maternal food treatment on growth and immunological development, one needs to tease apart origin and rearing effects by designing a crossfostering experiment (e.g. Lynch & Walsh 1998). Since, for logistical reasons we could not perform such a cross-fostering experiment, we are not able to distinguish between pre-hatching maternal effects and post-hatching rearing effects on offspring growth and immune function. Avian cross-fostering studies have reported that egg size improves early growth (e.g. Bize, Roulin, & Richner 00), but not survival of the offspring during rearing (e.g. Bize et al. 00; van de Pol et al. 006, but see Bolton 1991; Pelayo & Clark 003). Food-induced maternal effects on offspring Ig production in a cross-fostering design have to our knowledge not been documented, although experimentally increased maternal antioxidant consumption have been found to improve cell-mediated immunity (Biard, Surai & Møller 005, 007; Rubolini et al. 006). Our study on Ural owls suggests positive effects of egg size on nestling growth and no effects of hatchling Igs on Ig production at fledging. Further studies are, however, needed to explicitly evaluate the pre-hatching maternal effects, such as egg size and immune function, on offspring growth and Ig production. We find variation in maternal effects across years: egg size and maternal Ig levels in hatchlings vary differently depending on natural variations in food supply. In 006, when natural vole abundance was decreasing, hatchlings of both fed and control females had substantially higher Ig levels than in 004 or 005 when vole abundance was increasing. Contrastingly, egg size was small in 006 compared to 004 and 005, and supplementary food did not increase the egg size in 006 to the same level as the fed females eggs in 004 or 005. Increased maternal Ig transfer to the offspring may be especially beneficial if food conditions rapidly deteriorate after laying and parental feeding is demanding (cf. Gasparini et al. 007). It is, however, unclear why egg size is smaller under such deteriorating circumstances. Potentially, females could differently allocate resources into Ig transfer and egg size depending on yearly variations in food availability, but our study lacks the statistical power to evaluate such a phenomenon. Further studies are needed to evaluate potential differential maternal effects on different components under variable natural food conditions. We further find in this study that in the reversed sizedimorphic Ural owl (males are smaller than females) Ig levels at the end of the nestling period are higher in male offspring than in female offspring. Since higher levels of Igs have been suggested to estimate immunological condition (Apanius & 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

Resource limitation and maternal effects 689 Nisbet 006), our result suggests that male offspring would invest more in immune function than female offspring during the nestling period. Interestingly, we find that female offspring tend to have higher Ig levels at hatching, but that this sex-specific effect is reversed at the time of fledging. Development of immunity is considered to be costly (cf. Klasing & Leschchinsky 1998; Lochmiller & Deerenberg 000) and hence, male offspring potentially use more resources to development of immune function as they do not grow as big as female offspring. Such sex-specific investment in growth and immunity would indicate that the smaller sex (males) would not necessarily be cheaper to produce and raise compared to the larger sex (females). Indeed, a previous study of this population has found parental feeding investment in Ural owls to be equal for offspring of both sexes (Brommer et al. 003). In conclusion, we find evidence that food induces increased maternal effects on hatchling Igs and egg size. Furthermore, our results indicate that egg size, but not maternal treatment or Igs at hatching, has positive effects on offspring performance. Further studies are needed to evaluate the consequences of pre- and post-hatching maternal effects on offspring growth and immune function. Acknowledgements All experiments described in this paper were approved by the ethical board for animal experiments. We would like to thank Elina Virtanen for the ELISA work, Jaana Kekkonen and Paula Lehtonen for help with sexing the offspring, and Kalle Huttunen and Heikki Kolunen for assistance in the field. This study was funded by the Academy of Finland (H.P., J.E.B.), the Finnish Cultural Foundation (P.K.), the Swedish Cultural Foundation (P.K.) and Oskar Öflund Foundation (P.K.). References Apanius, V. & Nisbet, I.C.T. (006) Serum immunoglobulin G levels are positively related to reproductive performance in a long-lived seabird, the common tern (Sterna hirundo). Oecologia, 147, 1 3. Bernardo, J. (1996) The particular maternal effect of propagule size, especially egg size: patterns, models, quality of evidence and interpretations. American Zoologist, 36, 16 36. Biard, C., Surai, P.F. & Møller, A.P. (005) Effects of carotenoid availability during laying on reproduction in the blue tit. Oecologia, 144, 3 44. Biard, C., Surai, P.F. & Møller, A.P. (007) An analysis of pre- and posthatching maternal effects mediated by carotenoids in the blue tit. Journal of Evolutionary Biology, 0, 36 339. Bize, P., Roulin, A. & Richner, H. (00) Covariation between egg size and rearing condition determines offspring quality: an experiment with the alpine swift. Oecologia, 13, 31 34. Blount, J.D., Surai, P.F., Nager, R.G., Houston, D.C., Møller, A.P., Trewby, M.L. & Kennedy, M.W. (00) Carotenoids and egg quality in lesser blackbacked gull Larus fuscus: a supplemental feeding study of maternal effects. Proceedings of the Royal Society of London, Series B, 69, 9 36. Bolton, M. (1991) Determinants of chick survival in the lesser black-backed gull relative contributions of egg size and parental quality. Journal of Animal Ecology, 60, 949 960. Breslow, N.E. & Clayton, D.G. (1993) Approximate inference in generalized linear mixed models. Journal of the American Statistical Association, 88, 9 5. Bridges, T.S. & Heppell, S. (1996) Fitness consequences of maternal effects in Streblospio benedicti (Annelida: Polychaeta), American Zoologist, 36, 13 146. Brommer, J.E. (004) Immunocompetence and its costs during development: an experimental study in blue tit nestlings. Proceedings of the Royal Society of London, Series B, (Supplements) 71, 110 113. Brommer, J.E., Karell, P., Pihlaja, T., Painter, J.N., Primmer, C.R. & Pietiäinen H. (003) Ural owl sex allocation and parental investment under poor food conditions. Oecologia, 137, 140 147. Brommer, J.E., Kokko, H. & Pietiäinen, H. (000) Reproductive effort and reproductive values in periodic environments. American Naturalist, 155, 454 47. Brommer, J.E., Pietiäinen, H. & Kokko, H. (00b) Cyclic variation in seasonal recruitment and the evolution of the seasonal decline in Ural owl clutch size. Proceedings of the Royal Society of London, Series B, 69, 647 654. Brommer, J.E., Pietiäinen, H. & Kolunen, H. (1998) The effect of age at first breeding on Ural owl lifetime reproductive success and fitness under cyclic food conditions. Journal of Animal Ecology, 67, 359 369. Brommer, J.E., Pietiäinen, H. & Kolunen, H. (00a) Reproduction and survival in a variable environment: the Ural owl and the three-year vole cycle. Auk, 119, 194 01. Buechler, K., Fitze, P.S., Gottstein, B. Jacot, A. & Richner, H. (00) Parasiteinduced maternal response in a natural bird population. Journal of Animal Ecology, 71, 47 5. Chambers, R.C. & Leggett, W.C. (1996) Maternal influences on variation in egg sizes in temperate marine fishes. American Zoologist, 36, 180 196. Chew, B.P. (1996) Importance of antioxidant vitamins in immunity and health in animals. Animal Feed Sciences and Technology, 59, 103 114. Christians, J.K. (00) Avian egg size: variation within species and inflexibility within individuals. Biological Reviews, 77, 1 6. Crawley, M.J. (00) Statistical Computing. An Introduction to Data Analysis using S-Plus. John Wiley and Sons, New York. Fridolfsson, A.-K. & Ellegren, H. (1999) A simple and universal method for molecular sexing of non-ratite birds. Journal of Avian Biology, 30, 116 11. Gasparini, J., Boulinier, T., Gill, V.A., Gil, D., Hatch, S.A. & Roulin, A. (007) Food availability affects the maternal transfer of androgens and antibodies into eggs of a colonial seabird. Journal of Evolutionary Biology, 0, 874 880. Gasparini, J., McCoy, K.D., Haussy, C., Tveraa, T. & Boulinier, T. (001) Induced maternal response to the Lyme disease spirochaete Borrelia burgdorferi sensu lato in a colonial seabird, the Kittiwake Rissa tridactyla. Proceedings of the Royal Society of London, Series B, Biology Sciences, 68, 647 650. Gasparini, J., McCoy, K.D., Tveraa, T. & Boulinier, T. (00) Related concentrations of specific immunoglobulins against the Lyme disease agent Borrelia burgdorferi sensu lato in eggs, young and adults of the kittiwake (Rissa tridactyla). Ecology Letters, 5, 519 54. Grindstaff, J.L., Brodie, E.D. III & Ketterson, E.D. (003) Immune function across generations: integrating mechanism and evolutionary process in maternal antibody transmission. Proceedings of the Royal Society of London, Series B, Biology Sciences, 70, 309 319. Grindstaff, J.L., Demas, G.E. & Ketterson, E.D. (005) Diet quality affects egg size and number but does not reduce maternal antibody transmission in Japanese quail Coturnix japonica. Journal of Animal Ecology, 74, 1051 1058. Grindstaff, J.L., Hasselquist, D., Nilsson, J-Å., Sandell, M., Smith, H.G. & Stjernman, M. (006) Transgenerational priming of immunity: maternal exposure to a bacterial antigen enhances offspring humoral immunity. Proceedings of the Royal Society of London, Series B, Biology Sciences, 73, 551 557. Groothuis, T.G.G., Eising, C.M., Blount, J.D., Surai, P., Apanius, V., Dijkstra, C. & Müller, W. (006) Multiple pathways of maternal effects in blackheaded gull eggs: constraint and adaptive compensatory adjustment. Journal of Evolutionary Biology, 19, 1304 1313. Groothuis, T.G.G., Eising, C.M., Dijkstra, C. & Müller, W. (005) Balancing between costs and benefits of maternal hormone deposition in avian eggs. Biology Letters, 1, 78 81. Haq, A.-U., Bailey, C.A. & Chinnah, A. (1996) Effect of ß-carotene, canthaxanthin, lutein and vitamin E on neonatal immunity of chicks when supplemented in the broiler breeder diets. Poultry Science, 75, 109 1097. Hargitai, R., Prechl, J. & Török, J. (006) Maternal immunoglobulin concentration in Collared flycatcher (Ficedula albicollis) eggs in relation to parental quality and laying order. Functional Ecology, 0, 89 838. Karadas, F., Pappas, A.C., Surai, P.F. & Speake, B.K. (005) Emryonic development within carotenoid-enriched eggs influences the post-hatch carotenoid status of the chicken. Comparative Biochemistry and Physiology B Biochemistry and Molecular Biology, 141, 44 51. Kilpimaa, J., Alatalo, R.V. & Siitari, H. (007) Pre-hatching maternal investment and offspring immunity in the pied flycatcher (Ficedula hypoleuca). Journal of Evolutionary Biology, 0, 717 74. Klasing, K.C. & Leschchinsky, T.V. (1998) Functions, costs, and benefits of the immune system during development and growth. Proceedings of the Twentysecond International Ornithology Congress, Durban (eds N.J. Adams & R.H. Slotow), pp. 817 835. BirdLife South Africa, Johannesburg. 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690

690 P. Karell et al. Kontiainen, P., Brommer, J.E., Karell, P. & Pietiäinen, H. (008) Heritability, plasticity and canalisation of Ural owl egg size in a cyclic environment. Journal of Evolutionary Biology, 1, 88 96. Lochmiller, R.L. & Deerenberg, C. (000) Trade-offs in evolutionary immunology: just what are the costs of immunity? Oikos, 88, 87 89. Lynch, M. & Walsh, B. (1998) Genetics and Analysis of Quantitative Traits. Sinauer Associates, Massachusetts. Martin, T.E. (1987) Food as a limit on breeding birds: a life-history perspective. Annual Review of Ecology and Systematics, 18, 453 487. McGraw, K.J. & Klasing, K.C. (006) Carotenoids, immunity, and integumentary coloration in red junglefowl (Gallus gallus). Auk, 13, 1161 1171. McWhinney, S.L.R., Bailey, C.A. & Panigrahy, B. (1989) Immunoenhancing effect of ß-carotene in chicks. Poultry Science, 68 (Supplement 1), 94. Meijer, T., Daan, S. & Hall, M. (1990) Family planning in the kestrel (Falco tinnunculus): the proximate control of covariation of laying date and clutch size. Behaviour, 114, 117 136. Mousseau, T.A. & Fox, C.W. (1998) The adaptive significance of maternal effects. Trends in Ecology and Evolution, 13, 403 407. Myllymäki, A., Paasikallio, A., Pankakoski, E. & Kanervo, V. (1971) Removal experiments on small quadrates as a means of rapid assessment of the abundance of small mammals. Annales Zoologici Fennici, 8, 177 185. Newton, I. (1979) Population Ecology of Raptors. Poyser, Berkhamstead. Norrdahl, K. (1995) Population cycles in northern small mammals. Biological Reviews, 70, 61 637. Oppliger, A., Christe, P. & Richner, H. (1996) Clutch size and malaria resistance. Nature, 381, 565. Pelayo, J.T. & Clark, R.G (003) Consequences of egg size for offspring survival: a cross-fostering experiment in ruddy ducks (Oxyura jamaicensis). Auk, 10, 384 393. Pietiäinen, H. (1989) Seasonal and individual variation in the production of offspring in the Ural owl Strix uralensis. Journal of Animal Ecology, 58, 905 90. Pietiäinen, H. & Kolunen, H. (1993) Female body condition and breeding of the Ural owl Strix uralensis. Functional Ecology, 7, 76 735. Pietiäinen, H., Saurola, P. & Väisänen, R.A. (1986) Parental investment in clutch size and egg size in Ural owl Strix uralensis. Ornis Scandinavica, 17, 309 35. Pihlaja, M. (006) Maternal Effects in the Magpie. PhD thesis, University of Jyväskylä, Finland. Pihlaja, M., Siitari, H. & Alatalo, R.V. (006) Maternal antibodies in a wild altricial bird: effects on offspring immunity, growth and survival. Journal of Animal Ecology, 75, 1154 1164. Pinheiro, J.C. & Bates, D.M. (000) Mixed-Effects Models in S and S-PLUS. Springer, New York Berlin Heidelberg. Räsänen, K., Laurila, A. & Merilä, J. (005) Maternal investment in egg size: environment- and population-specific effects on offspring performance. Oecologia, 14, 546 553. Royle, N.J., Surai, P.F. & Hartley, I.R. (001) Maternally derived androgens and antioxidants in bird eggs: complementary but opposing effects? Behavioral Ecology, 1, 381 385. Rubolini, D., Romano, M., Bonisoli Aquati, A. & Saino, N. (006) Early maternal, genetic and environmental components of antioxidant protection, morphology and immunity of yellow-legged gull (Larus michahellis) chicks. Journal of Evolutionary Biology, 19, 1571 1584. Saino, N., Dallára, P., Martinelli, R. & Møller, A.P. (00) Early maternal effects and antibacterial immune factors in the eggs, nestlings and adults of the barn swallow. Journal of Evolutionary Biology, 15, 735 743. Saurola, P. (1987) Mate and nest-site fidelity in Ural and tawny owls. Biology and Conservation of Northern Forest Owls (eds R.W. Nero, R.J. Clark, R.J. Knapton & R.H. Hamre), pp. 81 86. Symposium Proceedings of the USDA Forest Service, General Technical Report RM-14, Winnipeg, Manitoba. Semlitsch, R.D. & Gibbons, J.W. (1990) Effects of egg size on success of larval salamanders in complex aquatic environments. Ecology, 71, 1789 1795. Sinervo, B. & McEdward, L.R. (1988) Developmental consequences of an evolutionary change in egg size: an experimental test. Evolution, 4, 885 899. Smith, N.C., Wallach, M., Petracca, M., Braun, R. & Eckert, J. (1994) Maternal transfer of antibodies induced by infection with Eimeria maxima partially protects chickens against challenge with Eimeria tenella. Parasitology, 109, 551 557. Soler, J.J., de Neve, L., Perez-Contreras, T., Soler, M. & Sorci, G. (003) Tradeoff between immunocompetence and growth in magpies: an experimental study. Proceedings of the Royal Society of London, Series B, Biology Sciences, 70, 41 48. Staszewski, V., Gasparini, J., McCoy, K.D., Tveraa, T. & Boulinier, T. (007) Evidence of an interannual effect of maternal immunization on the immune response of juveniles in a long-lived colonial bird. Journal of Animal Ecology, 76, 115 13. Steigenga, M.J., Zwaan, B.J., Brakefield, P.M. & Fischer, K. (005) The evolutionary genetics of egg size plasticity in a butterfly. Journal of Evolutionary Biology, 18, 81 89. Sundell, J., Huitu, O., Henttonen, H., Kaikusalo, A., Korpimäki, E., Pietiäinen, H., Saurola, P. & Hanski, I. (004) Large-scale spatial dynamics of vole populations in Finland revealed by the breeding success of vole-eating predators. Journal of Animal Ecology, 73, 167 178. Van de Pol, M., Bakker, T., Saaltink, D.J. & Verhulst, S. (006) Rearing conditions determine offspring survival independent of egg quality: a crossfoster experiment with oystercatchers. IBIS, 148, 03 10. Verboven, N., Monaghan, P., Evans, D.M., Schwabl, H., Evans, N., Whitelaw, C. & Nager, R.G. (003) Maternal condition, yolk androgens and offspring performance: a supplemental feeding experiment in the lesser black-backed gull (Larus fuscus). Proceedings of the Royal Society of London, Series B, Biology Sciences, 70, 3 3. Williams, T.D. (1994) Intraspecific variation in egg size and egg composition in birds: effects on offspring fitness. Biological Reviews, 68, 35 59. Wolfinger, R. & O Connell, M. (1993) Generalized linear mixed models A pseudo-likelihood approach. Journal of Statistical Computation and Simulation, 48, 33 43. Received 1 October 007; accepted 5 February 008 Handling Editor: Jonathan Blount 008 The Authors. Journal compilation 008 British Ecological Society, Functional Ecology,, 68 690