By continuing to browse this site you agree to us using cookies as described in About Cookies
Notice: Wiley Online Library will be unavailable on Saturday 7th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 08.00 EDT / 13.00 BST / 17:30 IST / 20.00 SGT and Sunday 8th Oct from 03.00 EDT / 08:00 BST / 12:30 IST / 15.00 SGT to 06.00 EDT / 11.00 BST / 15:30 IST / 18.00 SGT for essential maintenance. Apologies for the inconvenience.
Phillip Gienapp, Department of Biosciences, Ecological Genetics Research Unit, University of Helsinki, Viikinkaari 1, P.O. Box 65, FI – 00014, University of Helsinki, Finland. Tel.: +358 45 2626646; fax: +358 9 191 57694; e-mail: email@example.com
Condition, defined as the amount of ‘internal resources’ an individual can freely allocate, is often assumed to be environmentally determined and to reflect an individual’s health and nutritional status. However, an additive genetic component of condition is possible if it ‘captures’ the genetic variance of many underlying traits as many fitness-related traits appear to do. Yet, the heritability of condition can be low if selection has eroded much of its additive genetic variance, or if the environmental influences are strong. Here, we tested whether feather growth rate – presumably a condition-dependent trait – has a heritable component, and whether variation in feather growth rate is related to variation in fitness. To this end, we utilized data from a long-term population study of Siberian jays (Perisoreus infaustus), and found that feather growth rate, measured as the width of feather growth bars (GB), differed between age-classes and sexes, but was only weakly related to variation in fitness as measured by annual and life-time reproductive success. As revealed by animal model analyses, GB width was significantly heritable (h2 = 0.10 ± 0.05), showing that this measure of condition is not solely environmentally determined, but reflects at least partly inherited genetic differences among individuals. Consequently, variation in feather growth rates as assessed with ptilochronological methods can provide information about heritable genetic differences in condition.
Despite the fact that the expression of condition-dependent traits have often been shown to be influenced by additive genetic effects, a priori predictions about their heritability and genetic variability can be difficult to make. Whereas positive directional selection on condition is expected to erode additive genetic variance in the underlying traits (Fisher, 1958; Merilä & Sheldon, 1999), condition-dependent traits can counteract this by accumulating new genetic variance from many underlying loci (cf. Houle, 1992; Houle et al., 1996). Yet, as they may also accumulate environmental variation from many sources (Price & Schluter, 1991; Merilä & Sheldon, 1999), a trait can harbour substantial additive genetic variation but still have a low heritability (Price & Schluter, 1991). Hence, from practical point-of-view, the use of a trait as an indicator of environmental quality (e.g. Carlson, 1998) would benefit from understanding the relative importance of genetic and environmental influences on trait expression.
The wing and tail feathers of most birds have more or less regular banding patterns known as feather growth bars (Fig. 1). These feather growth bars are commonly used as condition indicators (e.g. Carbonell & Telleria, 1999; Perez-Tris et al., 2000; Stratford & Stouffer, 2001). Feather growth bars are caused by alternating dark and light bands of material produced during day and night, respectively (Wood, 1950; Brodin, 1993). These bands do not seem to result from different amounts of pigment laid down into the feather because they can also be found in pure white feathers (Grubb, 1995). Although this hints at micro-structural differences, the exact mechanism causing these – often quite distinct – patterns is still unknown (Grubb, 1995). Nevertheless, because one dark and one light band constitutes the part of the feather that is produced during 1 day, feather growth bars can be used to quantify the rate of feather growth and this measure has been proposed (Grubb, 1989) and also widely used as an indicator of condition (e.g. Yosef & Grubb, 1992; Carlson, 1998; Takaki et al., 2001; Vangestel & Lens, 2006). Birds in good nutritional condition should be able to allocate more resources to feather growth, and hence, have both faster feather growth rate and wider feather growth bars than birds in poor nutritional condition. This has been verified in experimental and semi-experimental (cf. food supplementation to wild birds) settings (e.g. Grubb & Cimprich, 1990; Jenkins et al., 2001). It has also been experimentally shown that physiologically stressed birds experience reduced feather growth rate (Talloen et al., 2008). However, until to date, no study has addressed the question whether individual variation in feather growth bars could be at least partly heritable, and if so, how much of the variation could be attributed to genetic effects.
The aim of this study was to investigate the relative importance of environmental and genetic factors in determining variation feather growth patterns in a wild bird population, and in particular, whether this variation is predictive of individual differences in fitness as measured by variation in life-time reproductive success and survival probability. To this end, we capitalized on data from a long-term (1974–2006) study of individually marked and pedigreed population of Siberian jays (Perisoreus infaustus) from where individual level data on annual and life-time reproductive success was available.
Field work and data collection
Siberian jays (Perisoreus infaustus) are comparably small (80 g) corvids inhabiting boreal forests from Fennoscandia to Siberia. They are group-living and maintain large year-around territories (up to 1–4 km2 in our study area), which are preferably set up in old-growth coniferous forests (Blomgren, 1964; Eggers et al., 2005). Natal dispersal and territory establishment takes place in late summer or early spring after the young have gained their independence (e.g. Blomgren, 1964; Ekman et al., 1999; Griesser et al., 2006). A typical family group consists of the breeding pair plus retained offspring and sometimes juveniles that are not offspring of the breeding pair. Group sizes can vary from two to six individuals (Ekman et al., 1994, 2002; Lillandt et al., 2003).
The study area is located in Ostrobothnia (SW-Finland), near Kristiinankaupunki (62°22′N, 21°30′E). Annual monitoring of the Siberian jay population started in 1974 in a 120-km2 sized forest area. Subsequently, the study area has been enlarged several times and now covers about 1000 km2 with a maximum N–S and E–W distance of about 75 and 50 km, respectively. Mainly for ‘historical’ and practical reasons, the study area is divided into seven sub-areas. For a more detailed description of the study area, see Lillandt (1993) and Lillandt et al. (2003).
Individuals were observed or caught at feeding stations in July–October. At these occasions, unringed birds were banded with colour- and standard aluminum rings, and the identity of already ringed individuals was recorded. Body weight was measured from nearly all birds, but tarsus length less systematically. Therefore, tarsus length could not be included in our analyses. A feather (since 1976) and/or blood (since 1997) samples were taken from all individuals for subsequent DNA-analyses. One or two outermost tail feathers were collected and preserved in individually labelled paper envelopes. Parentages were established based on nine microsatellite markers (Lillandt et al., 2002; Jaari et al., 2008) by assigning fathers and mothers using Cervus 3.0 (Kalinowski et al., 2007). Additional information comes from observational data based on the fact that parents treat their own offspring nepotistically when compared to unrelated individuals in the group (see Ekman et al., 1994). Hence, the used pedigree was constructed as described in (Lillandt et al., 2001) combining marker information with behavioural observations. Age of the individuals was determined from banding records, or if unringed, they were classified as ‘juveniles’ (1st calendar year) or ‘adults’ (2nd calendar year or older) on the basis of plumage characteristics (Svensson, 1992). This means that the age of individuals captured as adults for the first time is only their minimum age. Sex was determined from sex chromosomal genes CHD1W and CHD1Z (Fridolfsson & Ellegren, 1999). In total, 1041 feathers from 844 individuals could be included in the analyses. Life-time reproductive success (LRS) of an individual was defined as the number of offspring that recruited into the breeding population and 342 individuals with complete LRS data could be included in the analysis. The analysis of annual reproductive success was based on 182 pairs.
Growth bar measurements
All feathers – together with a length standard – were photographed using a Canon EOS 5D and a standard 24–105-mm zoom lens with lighting coming from the side in a shallow angle. Growth bar (GB) width was measured from the digital photographs using ImageJ software (Abramoff et al., 2004). The width of one GB was defined as the distance between the mid-lines of the dark parts of adjacent growth bars because this was the least ambiguous way to define the width of one growth bar (Fig. 1). The width of at least five (mean = 18.1, SD = 4.41, n = 1041) GBs was measured in the distal half of each feather. The average of these measures was defined as GB width of the given feather and used in all analysis to reduce measurement error. To evaluate the reliability of our GB measurements, 33 feathers were blindly re-measured and the repeatability (following Lessells & Boag, 1987) was r =0.66.
We first used linear mixed models (LMM) to analyse variation in GB width as a function of sex, age, and body weight including individual, sub area and year as random effects. We were interested in sex, age and body weight effects because social dominance, which is related to these variables in Siberian jays (Ekman et al., 2002, P. Gienapp & J. Merilä, unpublished), could affect the amount of food an individual is able to obtain and thereby feather growth rate. Individual identity was included as a random effect to account for repeated measurements of the same individual, and sub-area and year to account for effects of common environment. The statistical significance of the random effects was determined with likelihood-ratio tests comparing models including and excluding, respectively, the effect in question. Fixed effects were tested using ‘conditional’ Wald-statistic, divided by the denominator d.f. All ages over seven (calendar) years were lumped into one age class to avoid problems with small sample sizes in older age classes. Hence, age was included in the analyses as a factor with seven levels. Note that when analysing the effect of body size on GB width variation, we corrected it by removing variance because of individual identity, year, area, sex, age and capture date. Had we not done this, the use of raw body weight and including individual identity, year, area, sex, age and capture date as additional independent variables would not have ‘corrected’ body weight for these effects, but rather, estimated their effects on GB width.
Subsequently, an animal model analysis (e.g. Lynch & Walsh, 1998; Kruuk, 2004) was used to estimate variance components in feather growth. Animal models are generally well suited for the quantification of genetic parameters of natural populations, as they use all available information in the pedigree and can accommodate potential problems stemming from inbreeding and selection. All fixed and random effects that were statistically significant in the previous analyses were included in the animal model analyses. In short, sex, age (‘juvenile’ vs. ‘adult’), their interaction and tail feather length were included as fixed effects and year, sub-area, individual and the additive genetic effect as random effects. Feather length was included to account for a correlation of GB width with body size rather than body weight, as a proxy for body size, because the relationship between feather length and GB width, but not between body weight and GB width, was statistically significant. Failing to account for the effect of body size would have inflated the heritability estimate because part of it would then simply reflect the heritability of body size (h2 ≈0.5; P. Gienapp & J. Merilä, unpublished).
The relationship between feather growth (i.e. GB width) and fitness was analysed using Poisson regression with correction for overdispersion. Annual (ARS) and life-time reproductive success (LRS) were modelled as functions of annual GB width and GB width averaged over an individual’s life-time, respectively. Variation in GB width could be also affected by sex, age, feather length, year and sub-area. By regressing ARS and LRS against raw GB width, these effects could have confounded the analysis, and including these effects as explanatory variables would have tested their effect on fitness but not ‘corrected’ GB width for them. Hence, residual GB width corrected for sex, age, feather length, year and sub-area effects was used. ARS and LRS were defined as the number of offspring that recruited into the breeding population produced in any given year or over an individual’s life-time, respectively. ARS was analysed with a Generalised Linear Mixed Model (GLMM) including breeding pair as a random effect to account for repeated breeding events of the same individuals. LRS was analysed using a Generalised Linear Model (GLM), as no random effects needed to be included.
A proportional hazards model (Cox, 1972) was used to analyse the relationship between GB width and annual survival. The proportional hazards model is a robust and versatile tool because it can handle censored data, time-dependent variables and random effects (‘frailties’) without making assumptions about the distribution of the data (Therneau & Grambsch, 2000). Hence, also individuals that were still alive at the end of the study period could be included in the analyses to increase the sample size. To account for annual variation in survival probability, year was included into the model as a time-dependent ‘frailty’ term. Because the model cannot handle missing values and GB width was missing in at least some years for almost all individuals, individual average GB widths were used in the analysis.
ASReml 2.0 (VSN International, Hemel Hempstead, UK) was used for LMM- and the animal model-analyses, and R 2.7.0 (R Development Core Team, 2007) for GLM- and GLMM-analyses and the proportional hazards model.
Sex and age differences
A total of 625 individuals with known exact age could be included in this analysis. Growth bar (GB) width did not correlate with length of the tail feather (F1,739.6 = 0.07, P = 0.79), but differed significantly between sexes (F1,645.1 = 5.84, P = 0.017) and age classes (F6,557.2 = 36.3, P < 0.001) with males and older birds having wider GBs than females and younger birds. The main difference between age classes was between ‘juveniles’ and older individuals (Fig. 2) and when the analysis was restricted to ‘adults’ (2nd calendar year or older) significant differences among age classes disappeared (F5,83.1 = 1.12, P = 0.36). Consequently, only two age classes (‘juvenile’ vs. ‘adult’) were used in the subsequent analyses and also individuals without known exact age were included. This inclusion increased the data set to 844 individuals with 1041 observations. We now tested also for an interaction between sex and age, which was significant (F1,1012.2 = 5.14, P = 0.025; (Fig. 3) revealing that sex difference increased with increasing age. In these data, there was also a positive correlation between tail feather length and GB width (F1,1006.0 = 4.15, P = 0.043).
Relationship with body weight
As body weight depended on individual (Likelihood ratio test (LRT): Chi2 = 335.7, d.f. = 1, P < 0.001), year (LRT: Chi2 = 133.0, d.f. = 1, P < 0.001) and area (LRT: Chi2 = 4.07, d.f. = 1, P = 0.044) as well as age (‘juvenile’ vs. ‘adult’, F1,1203.4 = 15.5, P < 0.001), sex (F1,785.6 = 741.1, P < 0.001) and day of capture (F1,968.0 = 10.4, P = 0.002), we used body weight corrected for these effects when analysing the relationship between body weight and GB width. As in the previous analysis, GB width was related to sex (F1,699.2 = 12.37, P < 0.001) and age (F1,842.0 = 175.79, P < 0.001), but not with residual body weight (F1,814.6 = 0.58, P = 0.45). All interactions were nonsignificant (age × body size: F1,839.9 = 0.25, P = 0.62; sex × body size: F1,805.6 = 0.94, P = 0.34; sex × age: F1,835.1 = 3.18, P = 0.076).
Relationship with reproductive success and survival
There was no correlation between mean residual GB width and life-time reproductive success (GLM with Poisson-distribution and correction for overdispersion: Chi2 = 0.05, P = 0.82). Hence, individuals with on average wide GBs for their sex, age and general body size (as indicated by feather tail length) did not enjoy higher fitness. However, the number of recruits produced by a pair was negatively correlated with residual GB width of the tail feather grown in the same year in females (GLMM with Poisson-distribution and correction for overdispersion, pair and year as random effects: b = −1.01 ± 0.42, z = −2.407, P = 0.016) but not in males (b = 0.73 ± 0.63, z = 1.158, P = 0.25; interaction: z = 1.702, P = 0.09).
Survival probability varied among years (Chi2 = 87.9, P < 0.001) and the hazard was negatively related to residual GB width (b = −0.23 ± 0.079, Chi2 = 8.79, P = 0.003) meaning that individuals with wider GBs survived better. This relationship was identical for males and females (interaction sex × GB width, Chi2 = 0.65, P = 0.42), and there was no strong effect of sex on survival (Chi2 = 3.0, P = 0.08).
Quantitative genetic analysis
The animal model revealed that except for sub-area (Chi2 = 2.07, P = 0.15), all random effects explained a significant amount of variation in GB widths (Table 1). The large effect of individual identity showed that ca. 23% of variation in the GB width was explained by nonheritable individual properties, whereas only ca 5% of variance owed to among year effects (Table 1). The additive genetic effects fell in between these, giving a narrow sense heritability (h2 = VA / VP) of 0.097 (SE = 0.045) after ‘correcting’ for the included fixed effects. The coefficient of additive genetic variance (CVA) for GB width was estimated to be 4.61 (SE = 3.16).
Table 1. Results for the random effects from the animal model analysis of growth bar (GB) width.
Proportion ± SE
Given are explained variance, proportion of variance explained and test statistics (and significance) of the different effects. Significances were tested by comparing a model including the effect with a model without the effect using likelihood ratio tests with a single degree of freedom. The proportions explained for the ‘individual’ (permanent environment) and ‘additive genetic’ effect are the repeatability and heritability of GB width, respectively.
nsP > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.
0.004 ± 0.004
0.053 ± 0.023
0.231 ± 0.064
Additive genetic effect
0.097 ± 0.045
Analysing a large data set of more than 800 individuals, we found a small but significant heritability of feather growth bar (GB) width. Feather growth rate, measured as GB width, is generally viewed as a condition-dependent trait linked to nutritional status during moult and this has been confirmed through experiments (see Grubb, 1995). As pointed out in the introduction, a condition-dependent trait, such as feather GB width, can be heritable if the traits affecting feather growth rate harbour additive genetic variance. For example, the amount of resources an individual has access to during molt could depend on its dominance status, health, food conversion efficiency or other physiological traits. Dominance is related to personality, which has a heritable component (Drent et al., 2003). Similarly, physiological characteristics are heritable (e.g. Ketola & Kotiaho, 2009). Consequently, it may not be surprising that a number of studies have found low to moderate heritabilities of condition-dependent traits, including residual body mass in birds (Merilä, 1996; Meriläet al., 1999, 2001; Jensen et al., 2003; Parker & Garant, 2004; Birkhead et al., 2006), lipid and glycogen content in flies (Blanckenhorn & Hosken, 2003), immunocompetence in sheep, birds and insects (e.g. Coltman et al., 2001; Ryder & Siva-Jothy, 2001; Råberg et al., 2003). Hence, our study adds to growing evidence (see also: Kruuk et al., 2000; Merilä & Sheldon, 2000; McCleery et al., 2004) that traits closely related to fitness – such as life-history or condition-dependent traits – can be heritable despite opposing expectations (Gustafsson, 1986; Jones, 1987).
Genetic variability is expected to be reduced in small and isolated populations because of effects of genetic drift and inbreeding (Lynch & Walsh, 1998). Our study population lies near the Southern limit of the species distribution range, is bordered by the Baltic Sea and genetically distinct from other Finnish Siberian jay populations (Uimaniemi et al., 2000). Furthermore, levels of genetic variability in neutral marker genes in the study population is lower when compared to other Siberian jay populations (Uimaniemi et al., 2000; Jaari et al., 2008), and there is also evidence for inbreeding (Alho et al., 2009). It is therefore possible that the comparably low heritability of feather GB width reflects generally low levels of additive genetic variation in this population. However, heritabilities of other morphological traits in this population range from h2 =0.49 (body weight and tarsus length) to h2 =0.56 (wing and tail length; P. Gienapp & J. Merilä, unpublished). As these estimates correspondence to typical estimates from other avian studies (e.g. Meriläet al., 1998; Jensen et al., 2003; McCleery et al., 2004; Parker & Garant, 2004), it seems unlikely that the low heritability GB width owes to generally low levels of genetic variation in this population.
Although we found a small but significant additive genetic component of feather growth rate (GB width), the fact remains that the trait seems to be mainly environmentally determined. Year and the residual variance accounted for about 70% of the total phenotypic variance in it, and another 20% of the variation was explained by nonheritable individual properties (Table 1). While we can only speculate what these properties are, they could e.g. relate to consistent differences among individual in food abundance in their territories. At any rate, these permanent individual effects reflect environmentally determined individual condition.
As to reliability of the heritability estimate of GB width, we note that the relatively high measurement error (ca. 34%) in this trait may bias heritability estimates downward. On the other hand, unaccounted environmental covariance between relatives can bias heritability estimates upwards, and this is a concern in the studies of wild populations in particular. Because the animal model integrates information from distant relatives, it is thought to suffer less from this bias than the parent–offspring regression (Kruuk, 2004), but it is currently unclear how much pedigree information is necessary to avoid this bias entirely (Kruuk & Hadfield, 2007). We used an animal model and tried to control for all possible confounding factors in our analyses. This may at least partly explain why our estimate of heritability for feather growth rate lies at the lower end of the range condition-dependent traits. Yet, it would be premature to conclude that the heritability of feather growth would be lower than that of other condition-dependent traits as many of the studies have not used animal model approach that may yield lower and less upwards biased estimates than traditional methods (Kruuk & Hadfield, 2007). Furthermore, heritabilities are difficult to compare across studies because they are conditional on the fixed effects structure of the applied models (Wilson, 2008).
We observed that juveniles had significantly narrower growth bars (GBs) and hence a slower growth rates than adults (Fig. 2). The feathers of juveniles are grown for the most part during their growth period in the nest (Blomgren, 1964). This could mean that they are simply not able to divert as much resources to feather development as adults during their post-breeding moult. Although there were no sex differences in feather growth rates among juveniles, adult males had significantly wider GBs than adult females even after taking into account the longer tail feathers of males. As males are generally dominant over females (Piper, 1997), this may be explainable by their better access to food. Furthermore, as the moult occurs after breeding, females, which produce and incubate clutches, may have been energetically more stressed than males and therefore able to allocate less energy to feather growth than males. In fact, we observed that females raising more offspring in a given year experienced lower feather growth rates than females raising less offspring. This indicates a trade-off between reproductive effort and somatic maintenance. This trade-off may be exacerbated by the prolonged parental care in Siberian jays: fledglings can stay in the natal territory and are given access to food by their parents. Such a trade-off is generally expected but seldom observed as a larger variation in resource acquisition can mask the variation in resource allocation (van Noordwijk & DeJong, 1986).
Individuals with wider GBs and hence higher feather growth rate survived on average better. This finding is in line with the general assumption that feather growth rate reflects individual condition. Part of this condition may be because of the quality of the individual’s territory: individuals in territories with high food abundance during the molt may also be able to hoard an ample supply of food, which improves their winter survival probabilities.
In birds, condition-dependent traits, such as residual body mass or feather growth rate have ‘traditionally’ been viewed and used as indicators of environmentally determined ‘condition’ (e.g. Grubb, 1989; Alatalo et al., 1990; Schluter & Gustafsson, 1993). A reliable indicator of environmentally determined condition would be useful for studies in ecology or evolutionary biology. For example, it could be used to test for spatial variation in habitat quality without needing to develop suitable criteria and to carry out a direct assessment of the habitat (e.g. Carlson, 1998). A measure of environmentally determined condition could also be useful when studying ‘selection on the environmental component’ (sensu Price et al., 1988). It would allow to directly test whether fitness is related to the trait of interest or with to environmentally determined condition, which simultaneously affects the trait. However, our results together with earlier similar results from other traits (e.g. Merilä, 1996; Meriläet al., 1999; Parker & Garant, 2004; Birkhead et al., 2006) would caution against this, and call for alternative approaches to test for purely environmental effects.
One possibility to do this could be to separate the trait of interest into a genetic and environmental component using quantitative genetic methodology. This approach has been used to test for ‘selection on environmental component’ in avian breeding time (Sheldon et al., 2003; Gienapp et al., 2006) or antler size in red deer (Kruuk et al., 2002) by predicting individual breeding values and comparing selection on them (genetic selection) and selection on phenotypes (phenotypic selection). There is however a caveat with the use of breeding values (Postma, 2006; Hadfield et al., 2010). Instead of predicting breeding values one should estimate the genetic selection from the genetic covariance between the trait of interest and fitness (Hadfield et al., 2010).
In conclusion, the results show that feather growth rate – a trait traditionally considered to reflect individual condition and determined by environmental factors – harbours additive genetic variance and is heritable. Hence, variation in feather growth rates seems to convey information about inherited differences among individuals. Yet, a large part of the variation in feather growth rates seems to partition into environmental effects related to sex, age and identity of the individuals. Consequently, while variation in feather growth rates can be used as indicator of environmental conditions experienced by individuals, it should be kept in mind that part of the variance in this trait is because of genetic differences among individuals.
This work was funded by grants from the Academy of Finland (to PG and JM), Maj and Tor Nessling foundation (to JM) and Kone foundation (to PG and JM). We thank Bo-Göran Lillandt for putting the feathers (and the other data) at our disposal and Simo Rintakoski for carefully photographing and measuring them. Erik Postma, Maria Delgado, Alistair Wilson, Tadeusz Kawecki and an anonymous reviewer gave useful comments on the manuscript.