SEARCH

SEARCH BY CITATION

Keywords:

  • additive genetic variance;
  • body mass;
  • habitat quality;
  • natural selection;
  • Parus caeruleus;
  • tarsus length

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Quantifying the genetic variation and selection acting on phenotypes is a prerequisite for understanding microevolutionary processes. Surprisingly, long-term comparisons across conspecific populations exposed to different environments are still lacking, hampering evolutionary studies of population differentiation in natural conditions. Here, we present analyses of additive genetic variation and selection using two body-size traits in three blue tit (Parus caeruleus) populations from distinct habitats. Chick tarsus length and body mass at fledging showed substantial levels of genetic variation in the three populations. Estimated heritabilities of body mass increased with habitat quality. The poorer habitats showed weak positive selection on tarsus length, and strong positive selection on body mass, but there was no significant selection on either trait in the good habitat. However, there was no evidence of any microevolutionary response to selection in any population during the study periods. Potential explanations for this absence of a response to selection are discussed, including the effects of spatial heterogeneity associated with gene flow between habitats.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Understanding the evolution of life-history and morphological traits in natural populations requires analysing the genetic basis of these traits as well as the selection acting on them. Studies combining a quantitative genetics approach with selection analyses on long-term monitored populations have recently grown in number (e.g. Merilä & Fry, 1998; Kruuk et al., 2002; Sheldon et al., 2003). However, generalizing results from single populations may be misleading, and in particular, very little is known of the variation in the genetic basis of and selection on quantitative traits across space, between different populations of the same species occupying different habitats (but see e.g. Ebert et al., 1991; Bennington & McGraw, 1996).

There has been evidence that the variance components of phenotypic traits can vary with environmental conditions (review in Hoffmann & Merilä, 1999). However, tests of environmental effects on genetic variation in laboratory experiments or natural populations gave equivocal results (review in Hoffmann & Parsons, 1997). Most studies on Drosophila or other invertebrates suggest that stressful conditions tend to increase the heritability of morphological traits such as thorax length (e.g. Ebert et al., 1991; Imasheva et al., 1998; Bubliy et al., 2003). However, investigations on size-related traits in birds (Gebhardt-Henrich & van Noordwijk, 1991, 1994; Merilä, 1996; Larsson et al., 1997; Charmantier et al., 2004), fish (Garant et al., 2003) or mammals (Réale et al., 1999) have shown a decrease of heritability in unfavourable conditions. Data collected on birds usually concern variation of conditions as a response to year-to-year variation (van Noordwijk et al., 1988; Larsson, 1993; Larsson et al., 1997) or due to brood size manipulation (Merilä, 1997; Merilä & Fry, 1998). In these studies, lower heritability in unfavourable conditions results from either increasing environmental variance (VE) or from decreasing additive genetic variance (VA). A decrease in VA in the harsh environments is explained by genotype × environment interactions (GEI) (Hoffmann & Merilä, 1999), i.e. when the expression of genetic variance depends on the environmental conditions. For example, unfavourable conditions can constrain the expression of growth-promoting loci during food-shortage periods (Hoffmann & Merilä, 1999), especially when there is a limited time window in which growth occurs (van Noordwijk, 1988; van Noordwijk & Marks, 1998). Overall, investigations of variance components in different natural habitats remain scarce (but see Garant et al., 2003, 2004).

When comparing different environmental conditions, we also expect variation in selection pressures. We refer here to directional selection because stabilizing selection will not generate an evolutionary response of a change in the mean; it is also less commonly reported (Kingsolver et al., 2001). Directional selection on size-related traits should be more intensive under poor conditions where constraints on growth and development are higher than in favourable environments (e.g. Endler, 1986; Nager et al., 1998).

In this paper, we compare components of phenotypic variation and selection pressures in three populations of a passerine bird species occupying environments of different habitat quality. Having quantified the heritability and directional selection on two body size traits, we can then compare observed temporal trends with theoretical predictions of the magnitude of microevolutionary response. However, in doing so, two limitations of selection analyses need to be considered. First, measurements of selection can be heavily biased in situations in which the focal trait and fitness are both positively affected by the same environmental factors (Rausher, 1992; Stinchcombe et al., 2002). The appearance of positive selection on the trait is then generated by the confounding factors of environmental conditions, but such environmental covariance will not result in any evolutionary response (Kruuk et al., 2003). A recent method avoids this bias by calculating selection gradients using estimates of individual breeding values (i.e. the sum of the additive effects of an individual's genes on a given trait) rather than phenotypic values (Rausher, 1992; Mauricio & Mojonnier, 1997), thus eliminating bias resulting from environmental covariances.

A second consideration is that evolutionary change in a phenotypic trait depends not only on its heritability but also on its genetic correlations with other traits (Falconer & Mackay, 1996; Lynch & Walsh, 1998). When a phenotypic character is correlated with other characters through pleiotropy or linkage disequilibrium, selection on this character occurs both because of direct selection and also because of indirect selection on correlated traits (Lande & Arnold, 1983). Avian morphometric traits such as tarsus length and body mass are well known to be correlated (e.g. Boag & van Noordwijk, 1987; Green, 2001), hence much of the variation they display will stem from the same set of pleiotropic ‘body size genes’ (Boag & van Noordwijk, 1987). However, multivariate statistics have rarely been used in studies of wild populations (but see Kruuk et al., 2001, 2002; Potti et al., 2002) because they require substantially larger sample sizes than univariate analyses (Falconer & Mackay, 1996).

In an attempt to investigate evolutionary responses in organisms occupying habitats which differ in quality, we compare here the components of phenotypic variance and the selection on fledgling morphometric traits in three populations of blue tits (Parus caeruleus), and test for corresponding patterns of microevolution. Previous studies have shown that the three study habitats constitute a gradient of food abundance and parasite loads, and thus a gradient in resource availability for the raising of chicks (Blondel et al., 1993, 1999; Hurtrez-Boussès et al., 1998; Tremblay et al., 2003). Through long-term monitoring of these populations, data are available on both morphological traits and pedigree information, thus enabling the partitioning of phenotypic variance.

The aim of this study was two-fold. First, we investigated the extent to which differences in environmental conditions affect chick body mass and tarsus length, their quantitative genetics and the selection on these traits. We predicted (i) increasing heritability with increasing habitat quality due to increasing VA and possibly decreasing VE (Hoffmann & Merilä, 1999), and (ii) weaker selection on size-related traits in the higher-quality habitat. Secondly, we tested whether predictions of evolution derived from selection analyses and quantitative genetics match observed changes over time.

Study sites and data collection

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

The study was carried out in three populations of blue tits P. caeruleus (see e.g. Blondel et al., 2001 for a description of the study sites). Two of the populations were located on the French island of Corsica; one is an evergreen habitat dominated by the Holm oak Quercus ilex (Pirio, hereafter called P) where blue tits have been monitored using nest boxes since 1976, and the other is a deciduous forest dominated by the Downy oak Q. humilis (Muro, hereafter called M) monitored since 1994. These two study sites are separated by only 25 km, yet gene flow between the two populations is partly limited by a mountain barrier culminating at 1700 m (with the lowest mountain pass at 800 m) and by adaptive differences to local habitats in life-history traits such as laying date (Lambrechts et al., 1997; Blondel et al., 1999) or behavioural traits such as social dominance (Braillet et al., 2002). Corsican blue tits belong to the subspecies P. c. ogliastrae, which is c. 15% smaller than the mainland counterpart (Dias & Blondel, 1996). The mainland study site is a deciduous forest of downy oaks in the south of France (La Rouvière, hereafter called R) where nest boxes were erected in 1990. In the three study sites, nest box density was c. 2.5 nest boxes ha−1.

The evergreen holm oak Q. ilex renews only 30% of its foliage yearly; a critical point is that foliage development occurs 1 month later in this species than in the deciduous downy oak Q. humilis which renews at once the totality of its foliage (Blondel et al., 1993). As a consequence, the leaf-eating caterpillars representing the main prey for tits occur 1 month later and in consistently fewer number in evergreen than in deciduous habitats (Zandt et al., 1990). The low food supply in evergreen habitats has repeatedly been shown to be critical for blue tits so that study site P is considered as a low quality habitat where blue tits are significantly smaller and in worse condition than those living in deciduous habitats (Blondel et al., 1993, 1999). The routine collection of caterpillar frass for more than 10 years (details in Zandt et al., 1990) provided evidence for (i) a low food abundance in site P, (ii) a c. 10-fold higher abundance of food at site M, which is a 150-year-old forest, and (iii) an intermediate abundance of caterpillars in the 70-year-old coppice in the mainland site R. Moreover, studies of the dipteran blowflies of the genus Protocalliphora have shown a very high ectoparasite load in site P, which negatively affects chick growth and breeding success (Hurtrez-Boussès et al., 1997a,b), and a much lower ectoparasite abundance in M and R. Thus, habitat quality in terms of food resources and parasite load can be ranked as P < R < M.

Blue tit monitoring consisted of routine inspection of nest boxes to collect breeding and morphometric parameters from all birds. Chicks were ringed during the nestling period with individually numbered rings provided by the CRBPO, France. At an age of 14 or 15 days, chick body mass was measured to the nearest 1/10 g with a Pesola® spring balance (PESOLA, Baar, Switzerland) and their tarsus length to the 1/100 mm with a calliper. Parents were captured in nest boxes when chicks were 8–15 days old and were identified or ringed. Standard measures on body mass and tarsus length have been recorded continuously for chicks since 1989, 1991 and 1994 for P, R and M, respectively. Data from broods used in various cross-fostering or other experiments performed at the three study sites were excluded. The last breeding season included in the analyses was 2002.

Offspring recruitment to the breeding population, i.e. survival from chick to adulthood, was estimated by the capture of breeding adults in nest boxes in succeeding years. This is a classical ecological approach to estimate contributions to following generations in long-term individually based vertebrate studies (Clutton-Brock, 1988; Newton, 1989).

Quantitative genetic analysis and estimation of heritability

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Sample sizes of the quantitative genetic analyses are shown in Table 1. For each population, variance components of chick tarsus length and body mass were estimated using a restricted maximum likelihood (REML) procedure of estimation using an ‘animal model’ (Lynch & Walsh, 1998; Kruuk, 2004). This model was fitted using the software package VCE versions 3 and 4 (Variance Component Estimation, available at http://w3.tzv.fal.de/eg/vce4/vce4.html, Neumaier & Groeneveld, 1998), as described and discussed for collared flycatchers (Ficedula albicollis) and red deer (Cervus elaphus) in Meriläet al. (2001b) and Kruuk et al. (2002). VCE uses the pedigree information to fit an individual animal model, a form of mixed model which partitions the phenotypic value of individual traits into a sum of fixed and random effects, including a random effect of individual genetic merit, for which the variance–covariance structure is determined by the additive genetic relatedness matrix (Lynch & Walsh, 1998). Initially developed by animal breeders, the animal model approach is more flexible and yet more powerful than the conventional models used in estimating quantitative genetic parameters (e.g. parent–offspring regression, full and half-sib analysis) because it can accommodate unbalanced data sets and because it exploits the relatedness between all pairs of individuals in the pedigree (see recent review in Kruuk, 2004). We firstly used standard linear models (fitted in SAS, proc GLM) to determine the fixed effects that should be included in our final analysis. Year of measure, identity of measurer and sex had a significant effect on tarsus length and body mass (all P < 0.0001) and were thus fitted as fixed effects in the variance components analysis. For 19% of chicks, sex was known from molecular data. We combined this information with the visual sexing in the field by the three best measurers, whose mean visual accuracy was 88.0% (estimated by comparison with the genetic sexing), amounting to 44% of chicks with attributed sexes. Genetic sexing represented 42.6% of these 44%, and the combined accuracy was 94.3%. Hence, sex was included in the quantitative genetic analyses as a three-level (male/female/unknown) fixed effect.

Table 1.  Population means and standard errors (SE), components of phenotypic variances (VA: additive genetic variance; VB: variance attributable to the common brood environment; VR: residual variance), heritability (h2), coefficient of additive genetic variation (CVA) and correlations between chick tarsus length (mm) and body mass (g).
 Tarsus lengthBody mass
PRMPRM
  1. Estimated from animal models with year, measurer and sex as fixed effects and brood identity as a random effect, in three populations of blue tits (P < R < M for habitat quality).

  2. **P < 0.01; ***P < 0.001, after sequential Bonferroni corrections.

Trait mean (SE)15.90 (0.01)16.61 (0.01)16.32 (0.01)9.54 (0.02)10.65 (0.02)10.28 (0.02)
VA0.145***0.159***0.106***0.227***0.508***0.292***
VB0.076***0.083***0.028***0.401***0.815***0.065**
VR0.088***0.087***0.084***0.221***0.132**0.101**
h2 (SE)0.469 (0.055)0.483 (0.053)0.485 (0.079)0.267 (0.050)0.349 (0.051)0.638 (0.113)
CVA2.392.4021.9954.9486.7185.264
Number of chicks263738281602326538991597
Number of broods541517221661528220
Phenotypic correlations between both traits0.465***0.527***0.323***   
Genetic correlations between both traits0.195***0.243***0.091**   

Our models included an individual's additive genetic value and brood identification number as random effects. The total phenotypic variance of each morphometric trait (VP) was therefore partitioned into the following components:

  • image

where VA was the additive genetic variance, VB the environmental variance attributable to the common brood environment (and any dominance variance between full siblings, Kruuk et al., 2001) and VR the residual variance (including environmental effects other than those specific to the brood, any further nonadditive genetic effects and error variance, Falconer & Mackay, 1996).

The narrow-sense heritability defined as h2 = VA/VP describes the resemblance between parents and offspring. Statistical significance of variance components and of differences between heritabilities was assessed by t-tests. To control for the high number of tests, we performed sequential Bonferroni corrections for each type of analysis (Rice, 1989). We also provide the coefficient of additive genetic variation defined as CVA = 100√(VA)/trait mean (Houle, 1992), as a measure of additive genetic variance that is unaffected by other components of variance.

Finally, we predicted individual breeding values, i.e. the sum of the additive effects for a given trait for each individual (Falconer & Mackay, 1996). The breeding value is a measure of individual genetic merit, defined as the expected phenotypic value of its offspring. Using the pedigree information and the estimated variance components, we calculated best linear unbiased predictors (BLUPs) of the breeding values for each individual with known phenotypic value (Lynch & Walsh, 1998) running animal models with the software PEST (Neumaier & Groeneveld, 1998).

Phenotypic and genetic correlations

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We estimated pairwise Pearson's correlation coefficients between tarsus length and body mass. To control for the fixed effects of year, measurer and sex, the Pearson's coefficients were estimated on the residuals of the traits’ regression over these three factors.

For each population, pairwise genetic correlations were estimated using multivariate animal models with year, measurer and sex as fixed effects generating genetic covariances between the pairs of traits. The REML procedure of estimation partitions the phenotypic variance into the same sources of variation as those used in the univariate model and was run with VCE (Groeneveld, 1995; Neumaier & Groeneveld, 1998).

Selection analysis

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Selection on each trait in each population was estimated from its association with survival, specifically recruitment to the breeding population. Observed local recruitment rates in our populations were 4.9, 7.6 and 4.6% for P, R and M, respectively, ranging from 1 to 13% depending on year. Annual local recruitment rates were not significantly correlated among the three populations (pairwise correlation coefficients: rP/R = −0.19, n = 11 years; rR/M = −0.41, n = 8 years; rP/M = −0.46, n = 8; all n.s.). In 1997, a very strong rainstorm (see Blondel et al., 1999 for details) occurred on June 5 and 6 in Corsica resulting in high mortality of chicks and low population mean body mass (8.7 g) of surviving fledglings at site P. Only one of the 1997 P fledglings was recruited in later years, so data from P in 1997 were excluded from the selection analyses because they would give excessive statistical weight to this unique recruited chick.

For the three study populations, we conducted year-specific analyses using yearly standardized phenotypic values (0 mean and unit variance) and yearly relative fitness measures, that is 0 (not recruited) or 1 (recruited) values divided by the yearly mean local recruitment. For comparison, we also conducted overall analyses using phenotypic and fitness values standardized using the overall mean and variance and overall relative fitness, thus averaging over temporal variation.

Selection differentials

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We measured total directional and stabilizing (or disruptive) phenotypic natural selection on tarsus length and body mass using univariate analyses with local recruitment as a measure of fitness (Arnold & Wade, 1984a,b). We used least squares regression of relative recruitment on the standardized trait values (proc GLM, SAS Institute, Inc., 1992). The correlation between fledgling recruitment and the phenotypic measure of a trait is the selection differential, with linear models giving an estimation of the directional selection differential (S) and quadratic models estimating the stabilizing (if negative) or disruptive (if positive) selection differential (c). Statistical significance of the selection differentials was tested with a generalized linear mixed model (proc MIXED, SAS Institute, Inc., 1992) with brood identification as a random effect to account for the nonindependence of chicks measured in the same brood (Meriläet al., 1997; Kruuk et al., 2001).

Selection gradients

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

The previous univariate analyses were followed by comparable multivariate regressions to disentangle the direct and indirect effects of selection following Lande & Arnold's (1983) regression technique for correlated characters. Direct selection was thus expressed in directional standardized selection gradients (β) estimated from linear models and stabilizing (or disruptive) standardized selection gradients (γ) estimated from quadratic models.

Selection on breeding values

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We repeated selection analyses using estimates of individual genetic merit instead of phenotype, to avoid the potential problem of biases in selection estimates resulting from environmental covariances (Rausher, 1992; Mauricio & Mojonnier, 1997; Stinchcombe et al., 2002). We used the BLUP breeding values estimated in the quantitative genetic analysis to test for directional or stabilizing/disruptive selection on genotypes, using year-specific and overall analyses equivalent to those described above for phenotypic values.

The overall estimations of selection on phenotypic values as well as breeding values were run again standardizing the data within cohorts instead of within populations. The results for selection differentials, selection gradients and response to selection were similar for the two types of standardization (data not shown).

Response to selection

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We tested for changes over time in phenotypic values of tarsus length and body mass as well as in estimated breeding values. When a trait was under significant selection, we computed the predicted response to selection using heritability estimates and selection differentials (R = h2S, where h2 is the heritability of the trait based on the univariate analyses and S is the nonstandardized directional selection differential, Falconer & Mackay, 1996). Generation times estimated following Charlesworth (1994) were 2.64 for P, 2.00 for R and 1.91 for M. We then tested whether the slope of the linear regressions of annual means over time (proc REG, SAS Institute, Inc., 1992) differed significantly from the predicted value using t-tests.

To investigate the influence of parental origin (native or immigrant to each study site) on the evolution of chick morphology in response to selection, we repeated the above analyses splitting broods in two categories: chicks whose father or mother was a resident of the study site, i.e. was born on the same site as its chicks, vs. chicks for which neither parent was ringed as a chick on the same site.

Phenotypic variation between populations

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We found significant phenotypic differences in chick tarsus length and body mass among the three study sites (means in Table 1, for tarsus length: F2,6510 = 732, P < 0.0001, for body mass: F2,6508 = 573, P < 0.0001; all pairwise comparisons also significant). Birds were smaller in the island populations (P and M) than on the mainland (R), and smaller in the poor evergreen habitat P than in the richer M habitat (Table 1).

Quantitative genetic analysis and estimation of heritability

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Heritability of tarsus length and body mass was significantly greater than zero in all three populations (all P < 0.0001; Table 1). For body mass, there was a trend of increased heritability with increasing habitat quality for body mass with heritability ranging from 0.267 ± 0.05 in P to 0.638 ± 0.113 in M (Table 1, t-tests, P/R: t1187 = 1.1, n.s.; R/M: t746 = 2.3, P < 0.05; P/M: t879 = 3.0, P < 0.01). The increase in heritability was attributable to an increase in VA for R (P/R: t1187 = 3.3, P < 0.001) and a decrease of VB for M (R/M: t746 = 17.65, P < 0.0001). However, when standardizing the VA with the trait mean, CVA did not show the same trend of increase with increasing habitat quality (Table 1). On the contrary, for tarsus length the levels of variance components VA, VB and VR were of similar magnitude in the three populations, and heritability of the trait did not differ with habitat quality (Table 1, pairwise t-tests, all P n.s.).

Selection on phenotypes

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

Univariate analyses over the whole study period showed significant positive directional selection on standardized tarsus length and body mass in the evergreen Corsican site P and the mainland site R (Table 2). In the year-by-year analyses in P, 10 of 12 annual selection differentials on tarsus length were positive but only two significantly differed from zero (1990, 1995), and 11 of 13 were positive for body mass and only one significant (1989, presumably reflecting reductions in statistical power in the annual estimates). In R, all 11 selection differentials were positive for both traits, but only one was significant (1997) for tarsus length and five (1994, 1996, 1997, 2000, 2001) for body mass. There was no evidence for stabilizing selection on the morphometric traits in any population. There was also no significant directional selection in the rich habitat (M), either for the overall analysis or for individual years.

Table 2.  Selection on tarsus length and body mass of blue tit chicks for recruitment: overall standardized selection differentials (linear models for directional selection: S and quadratic models for stabilizing/disruptive selection: c) and selection gradients (linear: β and quadratic: γ) from univariate and multiple regression models, respectively.
nS (SE)c (SE)β (SE)γ (SE)
  1. n is the number of chicks measured.

  2. **P < 0.01; ***P < 0.001.

Tarsus length
 P15980.274 (0.111)**0.037 (0.069)0.186 (0.124)0.035 (0.069)
 R30460.212 (0.067)**−0.046 (0.040)−0.001 (0.077)−0.031 (0.042)
 M1277−0.000 (0.134)0.000 (0.092)−0.022 (0.147)0.008 (0.093)
Body mass
 P20200.314 (0.099)***−0.052 (0.063)0.189 (0.132)−0.032 (0.035)
 R31180.421 (0.065)***0.002 (0.044)0.419 (0.077)***0.013 (0.048)
 M12720.044 (0.133)−0.019 (0.026)0.049 (0.146)−0.021 (0.026)

When including both traits simultaneously in the analyses, the selection gradients did not confirm the positive selection on tarsus length either in P or in R, or the selection on body mass in P (all tests n.s.). The positive selection on body mass in R was confirmed in the multivariate analysis (P < 0.0001; Table 2). Ten of 11 year-specific selection gradients were positive of which four were significantly greater than zero (1994, 1996, 1997, 2000). The multivariate analyses on the deciduous Corsican site M showed no selection at all on either trait.

Annual selection gradients were not significantly correlated among the three populations, except for a surprising negative correlation of body mass selection between R and M (pairwise correlation coefficient: rR/M = −0.90; n = 8; P < 0.01).

Within populations, directional selection gradients on tarsus length and on body mass were not significantly related to the recruitment rate (all regressions n.s.; n = 12, 11 and 8 years for P, R and M, respectively).

Selection on breeding values

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

The univariate and multivariate analyses on breeding values of tarsus length and body mass showed a positive directional selection on breeding values of tarsus length in P, which was not significant after the sequential Bonferroni correction (overall selection differential: P = 0.035; overall selection gradient: P = 0.026; Pcritic = 0.008, Table 3). The directional selection on body mass in R was confirmed on breeding values in both univariate and bivariate analyses. In M, there was weak stabilizing selection on tarsus length which was not significant after the Bonferroni correction (overall selection differential: P = 0.040; overall selection gradient: P = 0.047; Pcritic = 0.01, Table 3).

Table 3.  Selection on breeding values of chick tarsus length and body mass for recruitment as first-year breeders (see Table 2 for legend).
nS (SE)c (SE)β (SE)γ (SE)
  1. *P < 0.05; **P < 0.01.

  2. After sequential Bonferroni corrections, all P < 0.05 were not significant.

Tarsus length
 P15940.950 (0.433)*−0.152 (1.047)1.179 (0.484)*−0.146 (1.042)
 R30430.379 (0.246)−0.958 (0.554)0.588 (0.277)−0.767 (0.579)
 M1269−0.326 (0.690)−4.807 (2.406)*−0.584 (0.741)−4.656 (2.437)*
Body mass
 P20200.159 (0.381)−0.197 (0.357)−0.532 (0.478)−0.078 (0.371)
 R31180.401 (0.134)**−0.212 (0.158)0.385 (0.155)**−0.140 (0.168)
 M12720.252 (0.347)−0.181 (0.165)0.343 (0.370)−0.196 (0.168)

Annual selection gradients on breeding values were not significantly correlated among the three populations (all pairwise correlation coefficients n.s.).

Phenotypic and genetic response to selection

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

The linear regression of annual mean phenotypic values over time showed no evidence of a statistically significant change over time in any population (Fig. 1a,b; all tests n.s.). Using the equation R = h2S, the expected response to selection on body mass in R is 0.176 g per generation or 0.088 g year−1. Despite the positive selection on body mass in R, chick body mass did not increase significantly over the course of the study period (b = 0.023 ± 0.035 g year−1, t11 = 0.67, n.s.) and the observed rate of change was significantly less than the predicted rate of change (t11 = 1.85, P < 0.05). Similarly, the observed rate of change of chick tarsus length in R (b = −0.021 ± 0.011 mm year−1) was significantly less than the predicted rates of change (R = 0.031 mm year−1; t11 = 4.69, P < 0.001). Finally in P, the observed rates of change of chick tarsus length and body mass (b = 0.008 ± 0.011 mm year−1 and −0.045 g year−1) differed significantly from the predicted rates of change (R = 0.028 mm year−1 and 0.030 g year−1; t12 = 1.76 and 2.77, P < 0.05 and <0.01).

image

Figure 1. Variation over time and between populations P, M and R for (a) mean fledging tarsus length, (b) mean fledging body mass, (c) mean estimated breeding values of tarsus length and (d) mean estimated breeding values of body mass. P: squares, M: diamonds and R: triangles.

Download figure to PowerPoint

At the genotypic level linear regressions of annual means did not show any changes in breeding values over time (Fig. 1c,d; all P > 0.1) and the observed rate of change of body mass breeding values in R (b = 0.009 ± 0.006 g year−1) was significantly less than the predicted rate of change (R = 0.070 g year−1; t11 = 10.53, P < 0.0001).

Finally, we compared patterns of change in chicks of immigrant vs. resident birds. In the mainland site R, 58.2% of breeding males and 34.7% of breeding females were ringed as chicks in the same study site. The observed rate of change of body mass for offspring of resident males (b = 0.038 ± 0.049 g year−1) did not differ significantly from the predicted rate of change of 0.088 g year−1 (t11 = 1.03, n.s.). However, the rate of change was significantly less for chicks of nonresident males (b = −0.014 ± 0.049 g year−1, t11 = 2.07, P < 0.05). The results showed a similar pattern when considering chicks of resident females (Fig. 2; b = 0.141 ± 0.068 g year−1; test of difference from predicted change: t11 = 0.776, n.s.) and chicks of nonresident females (b = 0.012 ± 0.040 g year−1, t11 = 1.89, P < 0.05). Nevertheless, changes in body mass for chicks of resident or nonresident parents were still never significantly different from zero (all t-tests, n.s.). Body mass in the mainland site R was the only trait with significant selection gradients; however, the other traits under weak selection (body mass in P and tarsus length in P and R) showed the same trend for a higher response to selection for chicks of residents than for chicks of other parents.

image

Figure 2. Evolution over time of phenotypic body mass in R for chicks of nonresident females (white triangles) and resident females (black triangles). In both cases, regressions did not show any significant change over time (see text for equations and tests).

Download figure to PowerPoint

Variance components and habitat quality

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

This study confirms earlier conclusions from other vertebrate species (e.g. Larsson et al., 1998; Réale et al., 1999; Kruuk et al., 2001, 2002) of substantial levels of additive genetic variation underlying morphometric traits, in spite of directional selection acting on these traits. Bird tarsus length and body mass showed significant heritabilities in all three populations (Table 1), of similar order of magnitude as those found before in this or closely related species (Dhondt, 1982; van Noordwijk et al., 1988; Merilä & Fry, 1998; Merilä & Sheldon, 2001).

We also showed that the variance components of morphometric traits can vary between populations of the same species occupying different habitats. The results for body mass agree with the trend found in earlier studies of vertebrate morphological traits that heritability decreases as environmental conditions become more difficult (see review of 19 studies of birds in Merilä & Sheldon, 2001). In this case, the decrease in heritability in the harsh quality environment was due to a decrease in the additive genetic variance along with an increase in the common brood-environment effect. The increase in variance between broods can be interpreted as relatively higher impact of parents and territory quality in unfavourable habitats. When food is scarce, parental care and availability of local resources will be an important determinant of brood fitness, increasing the between-brood variance in morphometric traits when compared with the within-brood variance. We have previously shown that in the evergreen Corsican site P, infestation by the parasite Protocalliphora (Hurtrez-Boussès et al., 1997b) increases the between-brood variance in chick morphology (Charmantier et al., 2004).

However, these results should be treated with caution for two reasons. First, increasing heritability with habitat quality is not fully supported by our study because the heritability of tarsus length (in contrast to body mass) did not differ between the populations, and because no replication is available for habitat quality. Secondly, higher heritability does not necessarily mean higher evolvability. Houle (1992) has shown that in some cases, CVA gives a more direct estimate of the evolvability of the trait, or its ability to respond to selection, as it is standardized by the trait mean and thus does not depend on the magnitude of the total trait variance (Houle, 1992; Roff, 1997). In our case, when comparing body mass evolution in the three blue tit populations, higher heritability in M than in P is due to high VA but also to low variation between broods. Standardizing by the trait mean instead of VP shows that relative evolvabilities do not follow the predicted pattern of M > R > P (Table 1). This emphasizes the necessity to consider genetic variance through different measures (Houle, 1992).

Response to selection

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

The results in this study failed to reveal any evolutionary response to selection. Using theoretical expectations of evolutionary responses to selection, our analyses of chick body mass in R predict an increase of 0.09 g each year. Yet there was no evidence for a significant change in chick body mass in the course of the study period. In the same line of reasoning, the slight selection on body mass in P and tarsus length in P and R did not result in any increase in those traits over the study period (Fig. 1). Such discrepancies have been frequently reported in natural populations (e.g. Larsson et al., 1998; Kruuk et al., 2001, 2002; Meriläet al., 2001a; Sheldon et al., 2003). Meriläet al. (2001c) have reviewed the possible explanations for this paradox and we will now discuss seven potential explanations for the lack of correspondence between predicted and observed responses to selection in our study: (i) overestimation of heritability due to maternal/common environment effects; (ii) biased estimation of heritability due to genotype × environment interactions (GEI); (iii) biased estimation of heritability due to misassigned paternities; (iv) overestimates of the magnitude of selection due to environmental covariance; (v) selection constrained by genetic correlations; (vi) selection fluctuating in time or space; (vii) response to selection affected by indirect genetic effects.

The first risk of bias in heritability estimates comes from maternal and common environment effects of offspring belonging to the same brood (Meriläet al., 2001c), which will upwardly bias estimates of additive genetic variance if not taken account of. One advantage of our analyses was that we could include brood identity as an additional random effect in the animal model and thereby overcome the problem of inflation of heritability due to common nest environment effect that occurs in classical full sib analyses (Boag & van Noordwijk, 1987). The common brood environment effect explained as much as 47.3% of the variation in chick body mass in P and 56.0% in R, and would have resulted in considerable overestimation of the heritability of body mass if not included in the model (see example in Kruuk, 2004).

Genotype × environment interactions can also lead to a biased estimation of heritability. GEI can occur when parents and offspring are reared in different environments. In their review on avian quantitative genetics, Merilä & Sheldon (2001, pp. 227–228) discuss a potential bias arising when two environments are unbalanced in size or frequency across the landscape, leading to higher environmental correlation between parents and offspring in central than in marginal habitats. In Corsica, the most common habitat type is the evergreen one. If there was higher environmental correlation between parents and offspring in these habitats because most of the breeding parents were recruits from evergreen habitats, then heritability of traits would be artificially increased. However, as heritabilities were lower in the evergreen habitat in Corsica, our results are robust concerning this potential bias.

A third potential source of error in heritability estimation comes from the occurrence of extra-pair paternity. Levels of extra-pair paternity have been found to be relatively high in these study sites, i.e. between 14 and 25% of chicks studied depending on the population (Charmantier & Blondel, 2003). The incorrect paternal links in the pedigree can result in an underestimation of the additive genetic variance of the studied traits. We did not have enough known genotypes in our pedigrees to test how much the quantitative parameters would change after removal of the extra-pair young. However, analyses on morphometric traits in collared flycatchers have shown that the bias due to similar levels of extra-pair paternities may be negligible (Meriläet al., 1998); note that these levels of error are also similar to those in mammalian studies in which paternities were assigned using genetic data (Kruuk et al., 2002). In any case, misassigned paternity would result in a downward in the predicted response to selection, and hence could not explain the discrepancies in this study between predicted and observed changes. Thus, none of the three situations relating to heritability estimation provide an adequate explanation for the lack of correspondence between observed and predicted trends.

A fourth possible source of error is the incorrect estimation of selection. One explanation discussed by Meriläet al. (2001c), based on the reasoning of Price et al. (1988), implies a stronger association of fitness with phenotypic values than with breeding values, due to an environmental covariance between the trait and the fitness measure (see also Kruuk et al., 2002; Stinchcombe et al., 2002). Selection gradients on the focal trait are then overestimated because of confounding environmental factors influencing both the trait and the fitness measure. One of the great advantages of the animal model approach is that we can estimate individual breeding values and therefore test the selection acting on them (Rausher, 1992; Mauricio & Mojonnier, 1997; Kruuk, 2004). Surprisingly, in our study, selection gradients estimated on breeding values were either of similar magnitude or even higher than those estimated on phenotypes (Tables 2 and 3). Therefore the absence of response to selection cannot be due to selection being associated predominantly with the nonheritable component of the phenotype. Another potential source of error comes from the use of local recruitment, which is used in many bird studies to reflect the total recruitment of breeding offspring. However, a review of over 22 long-term studies has shown that on average two-thirds of the offspring emigrate to breed outside study plots (Lambrechts et al., 1999). Investigations as to whether this dispersal is random with regards to the morphometric traits in this study system would be necessary to fully validate our estimations of selection differentials, but studies of collared flycatchers found no relationship between dispersal and size (Pärt, 1990).

A response to selection may also be constrained by genetic correlations with other unmeasured traits under opposing selection. Although we have used bivariate statistics including tarsus length and body mass to control for the correlation between these traits, there is still a possibility that the absence of evolutionary response to selection on body size is due to genetic correlations with other targets of selection (Lande & Arnold, 1983). Ideally, a multivariate analysis should include all other traits with which body mass is genetically correlated, yet such analysis is necessarily constrained to traits that have been measured (Lande & Arnold, 1983; Meriläet al., 2001c). Another genetic correlation which might constrain an evolutionary response is that which occurs between sexes (Roff, 1997), if it is associated with opposing selection pressures in the two sexes. However, we ran sex-specific analyses and did not find any evidence for sexually antagonistic selection (data not shown); hence, this does not explain the absence of evolutionary response.

As a sixth possible explanation for an absence of response, theoretical models have shown that temporal and spatial heterogeneity can maintain genetic variation (Ewing, 1979). Selection can fluctuate in direction and/or intensity between years, for example, when contrasting climatic conditions induce changes in food supply, as has been shown for a population of medium ground finch Geospiza fortis on the island of Daphne Major in the Galápagos (Gibbs & Grant, 1987, 1995). In our study, the analyses showed some evidence for spatial heterogeneity in selection in the three populations of blue tit occupying different habitats. Two of these populations (P and M) are only separated by 25 km, yet the strength of annual selection on tarsus length and body mass as well as on their estimated breeding values did not show any correlation. For example, directional selection on tarsus length was null or negative (not significantly, see Tables 2 and 3) in the rich habitat (M) but positive in the harsh environment (P). Although there are probably some geographical and ethological barriers to migration between populations P and M (Lambrechts et al., 1997; Braillet et al., 2002), they are presumably connected by gene flow (Dias et al., 1996; Blondel et al., 1999, 2001): in M, only 22.8% males and 15.6% females were ringed as chicks in the same study site in which they subsequently bred, and in P natal philopatry has similar levels: 37.0% of males and 16.1% of females. It is hence possible that because of gene flow between the two populations, immigrants from M which have not been under positive directional selection are recruited in P, constraining the evolution of body size traits in response to the positive selection occurring in P. A mosaic of habitat patches dominated by either deciduous or evergreen oaks also characterizes the mainland landscape occupied by blue tits. However, a preliminary analysis comparing chicks from resident parents to chicks from immigrants shows that neither the magnitude nor the temporal trend in immigrant phenotypes are consistent with the hypothesis that gene flow alone is responsible for the observed stasis (Fig. 2). Finally, we found that all significant directional selection gradients were positive. Interestingly, a previous selection analysis on reproductive success of blue tit adults has shown that small adult males have a higher breeding success than large males in P whereas the opposite is true in R (Blondel et al., 2002). This results in stronger sexual dimorphism in adults in R than P. It also suggests that in P, natural selection acts in different directions on males depending on life stages, with higher recruitment for larger male chicks but higher reproductive success for smaller adult males. Such a trade-off between survival and reproduction has already been reported in the study of female body size in the song sparrow Melospiza melodia (Schluter & Smith, 1986). This type of variation of selection depending on life stages can be an important source of maintenance of additive genetic variance as well as a potential explanation for the absence of response to selection. Hence, there is some evidence for fluctuating selection pressures through different life-history components.

A final potential source of bias in the estimation of response to selection is the presence of indirect genetic effects. Our analyses show large common environment effects, which may be due in part to maternal effects. These maternal effects may in part be due to the mother's genotype, i.e. indirect genetic effects. Such effects can alter the response to selection, and may constrain or even reverse an evolutionary response, depending on the sign of the covariance between the direct and the indirect genetic effects and the respective selection pressures (Kirkpatrick & Lande, 1989; Lynch & Walsh, 1998). Further complex quantitative genetics models, which we will explore in future work, will be necessary to quantify the relevant components of variance and covariance.

In conclusion, our analyses illustrate high levels of additive genetic variation underlying morphometric traits, as well as substantial common brood environment effects. Directional selection on heritable morphometric traits in blue tits differed between environments occupied by different populations. The spatial heterogeneity associated with gene flow between habitats as well as potential indirect genetic effects are two of the possible explanations for the absence of response to selection, and call for further investigations. Finally, our comparison of two traits in three different populations produced different impressions of the factors determining evolutionary processes in each, suggesting the need for a degree of caution when generalizing from results on a single population or trait.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References

We warmly thank all the people who have participated in data collection over the years. Thanks also to Erik Postma for valuable discussions and to Philippe Jarne, Edmund D. Brodie III, Arie van Noordwijk and two anonymous referees whose comments greatly improved the manuscript. A. C. was supported by a grant from the ‘Ministère Français de l'Education Nationale, de l'Enseignement et de la Recherche’ and a Marie Curie Host Fellowship (QLK5-1999-50768). LK is supported by the Royal Society, London.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study sites and data collection
  6. Quantitative genetic analysis and estimation of heritability
  7. Phenotypic and genetic correlations
  8. Selection analysis
  9. Selection differentials
  10. Selection gradients
  11. Selection on breeding values
  12. Response to selection
  13. Results
  14. Phenotypic variation between populations
  15. Quantitative genetic analysis and estimation of heritability
  16. Phenotypic and genetic correlations
  17. Patterns of selection
  18. Selection on phenotypes
  19. Selection on breeding values
  20. Phenotypic and genetic response to selection
  21. Discussion
  22. Variance components and habitat quality
  23. Response to selection
  24. Acknowledgments
  25. References
  • Arnold, S.J. & Wade, M.J. 1984a. On the measurement of natural and sexual selection: applications. Evolution 0: 720734.
  • Arnold, S.J. & Wade, M.J. 1984b. On the measurement of natural and sexual selection: theory. Evolution 38: 709719.
  • Bennington, C.C. & McGraw, J.B. 1996. Environment-dependence of quantitative genetic parameters in Impatiens pallida. Evolution 50: 10831097.
  • Blondel, J., Dias, P.C., Maistre, M. & Perret, P. 1993. Habitat heterogeneity and life history variation of Mediterranean blue tits. Auk 110: 511520.
  • Blondel, J., Dias, P.C., Perret, P., Maistre, M. & Lambrechts, M.M. 1999. Selection-based biodiversity at a small spatial scale in a low-dispersing insular bird. Science 285: 13991402.
  • Blondel, J., Perret, P., Dias, P.C. & Lambrechts, M.M. 2001. Is phenotypic variation of blue tits (Parus caeruleus L.) in mediterranean mainland and insular landscapes adaptive? Genet. Sel. Evol. 33: S121-S139.
  • Blondel, J., Perret, P., Anstett, M.-C. & Thébaud, C. 2002. Evolution of sexual size dimorphism in birds: test of hypotheses using blue tits in constrated Mediterranean habitats. J. Evol. Biol. 15: 440450.
  • Boag, P.T. & van Noordwijk, A.J. 1987. Quantitative genetics. In: Avian Genetics (F.Cooke & P. A.Buckley, eds), pp. 4578. Academic Press, New York.
  • Braillet, C., Charmantier, A., Archaux, F., Dos Santos, A., Perret, P. & Lambrechts, M.M. 2002. Two blue tit Parus caeruleus populations from Corsica differ in social dominance. J. Avian Biol. 33: 446450.
  • Bubliy, O., Loeschcke, V. & Imasheva, A. 2003. Genetic variation of morphological traits in Drosophila melanogaster under poor nutrition: isofemale lines and offspring-parent regression. Heredity 86: 363369.
  • Charlesworth, B. 1994. Evolution in Age-structured Populations, 2nd edn. Cambridge University Press, Cambridge.
  • Charmantier, A. & Blondel, J. 2003. A contrast in extra-pair paternity levels on mainland and island populations of Mediterranean blue tits. Ethology 109: 351364.
  • Charmantier, A., Kruuk, L.E.B. & Lambrechts, M.M. 2004. Parasitism reduces the potential for evolution in a wild bird population. Evolution 58: 203206.
  • Clutton-Brock, T. 1988. Reproductive Success. University of Cichago press, Chicago.
  • Dhondt, A.A. 1982. Heritability of blue tit tarsus length from normal and cross-fostered broods. Evolution 36: 418419.
  • Dias, P.C. & Blondel, J. 1996. Local specialization and maladaptation in the Mediterranean blue tit (Parus caeruleus). Oecologia 107: 7986.
  • Dias, P.C., Verheyen, G.R. & Raymond, M. 1996. Source-sink populations in Mediterranean blue tits: evidence using single-locus minisatellite probes. J. Evol. Biol. 9: 965978.
  • Ebert, D., Yampolsky, L. & van Noordwijk, A.J. 1991. Genetics of life history in Daphnia magna. II. Phenotypic plasticity. Heredity 70: 344352.
  • Endler, J.A. 1986. Natural Selection in the Wild. Princeton University Press, Princeton, NJ.
  • Ewing, E.P. 1979. Genetic-variation in a heterogeneous environment. 7. Temporal and spatial heterogeneity in infinite populations. Am. Nat. 114: 197212.
  • Falconer, D.S. & Mackay, T.F.C. 1996. Introduction to Quantitative Genetics. Longman, Essex, UK.
  • Garant, D., Dodson, J.J. & Bernatchez, L. 2003. Differential reproductive success and heritability of alternative reproductive tactics in wild Atlantic salmon (Salmo salar L.). Evolution 57: 11331141.
  • Garant, D., Sheldon, B.C. & Gustafsson, L. 2004. Climatic and temporal effects on the expression of secondary sexual characters: genetic and environmental components. Evolution, in press.
  • Gebhardt-Henrich, S.G. & van Noordwijk, A.J. 1991. Nestling growth in the great tit I. Heritability estimates under different environmental conditions. J. Evol. Biol. 3: 341362.
  • Gebhardt-Henrich, S.G. & van Noordwijk, A.J. 1994. The genetical ecology of nestling growth in the great tit. Environmental influences on the expression of genetic variances during growth. Funct. Ecol. 8: 469476.
  • Gibbs, H.L. & Grant, P.R. 1987. Oscillating selection on Darwin's finches. Nature 327: 511513.
  • Grant, P.R. & Grant, R.B. 1995. Predicting microevolutionary responses to directional selection on heritable variation. Evolution 49: 241251.
  • Green, A.J. 2001. Mass/length residuals: measures of body condition or generators of spurious results? Ecology 82: 14731483.
  • Groeneveld, E. 1995. REML VCE, a Multivariate Multi-Model Restricted Maximum Likelihood (Co) Variance Component Estimation. Package, Version 4.2.5 User's Guide. Institute of Animal Husbrandry and Animal Behaviour, Federal Research Center of Agriculture (FAL), Mariensee, Germany.
  • Hoffmann, A.A. & Merilä, J. 1999. Heritable variation and evolution under favourable and unfavourable conditions. Trends Ecol. Evol. 14: 96101.
  • Hoffmann, A.A. & Parsons, P.A. 1997. Consistent heritability changes under poor growth condition. Trends Ecol. Evol. 12: 460461.
  • Houle, D. 1992. Comparing evolvability and variability of quantitative traits. Genetics 130: 195204.
  • Hurtrez-Boussès, S., Blondel, J., Perret, P. & Renaud, F. 1997a. Relationship between intensity of blowfly infestation and reproductive success in a Corsican population of blue tits. J. Avian Biol. 28: 267270.
  • Hurtrez-Boussès, S., Perret, P., Renaud, F. & Blondel, J. 1997b. High blowfly parasitic loads affect breeding success in a Mediterranean population of blue tits. Oecologia 112: 514517.
  • Hurtrez-Boussès, S., Blondel, J., Perret, P., Fabreguettes, J. & Renaud, F. 1998. Chick parasitism by blowflies affect feeding rates in a Mediterranean population of blue tits. Ecol. Lett. 1: 1720.
  • Imasheva, A.G., Loeschcke, V., Zhivotovsky, L.A. & Lazebny, O.E. 1998. Stress temperatures and quantitative variation in Drosophila melanogaster. Heredity 81: 246253.
  • Kingsolver, J.G., Hoekstra, H.E., Hoekstra, J.M., Berrigan, D., Vignieri, S.N., Hill, C.E., Hoang, A., Gibert, P. & Beerli, P. 2001. The strength of phenotypic selection in natural populations. Am. Nat. 157: 245261.
  • Kirkpatrick, M. & Lande, R. 1989. The evolution of maternal characters. Evolution 43: 485503.
  • Kruuk, L.E.B. 2004. Estimating genetic parameters in wild populations using the ‘animal model’. Phil. Trans. R. Soc Lond. B, in press.
  • Kruuk, L.E.B., Merilä, J. & Sheldon, B.C. 2001. Phenotypic selection on a heritable size trait revisited. Am. Nat. 158: 557571.
  • Kruuk, L.E.B., Slate, J., Pemberton, J.M., Brotherstone, S., Guinness, F. & Clutton-Brock, T. 2002. Antler size in red deer: heritability and selection but no evolution. Evolution 56: 16831695.
  • Kruuk, L.E.B., Merila, J. & Sheldon, B.C. 2003. When environmental variation short-circuits natural selection. Trends Ecol. Evol. 18: 207209.
  • Lambrechts, M.M., Blondel, J., Hurtrez-Bousses, S., Maistre, M. & Perret, P. 1997. Adaptive inter-population differences in blue tit life-history traits on Corsica. Evol. Ecol. 11: 599612.
  • Lambrechts, M.M., Blondel, J., Caizergues, A., Dias, P.C., Pradel, R. & Thomas, D.W. 1999. Will estimates of lifetime recruitment of breeding offspring on small-scale study plots help us to quantify processes underlying adaptation? Oikos 86: 147151.
  • Lande, R. & Arnold, S.J. 1983. The measurement of selection on correlated characters. Evolution 37: 12101226.
  • Larsson, K. 1993. Inheritance of body size in the barnacle goose under different environmental conditions. J. Evol. Biol. 6: 195208.
  • Larsson, K., Rattiste, K. & Lilleleht, V. 1997. Heritability of head size in the common gull Larus canus in relation to environmental conditions during offspring growth. Heredity 79: 201207.
  • Larsson, K., van der Jeugd, H.P., van der Veen, I.T. & Forslund, P. 1998. Body size declines despite positive directional selection on heritable size traits in a barnacle goose population. Evolution 52: 11691184.
  • Lynch, M. & Walsh, B. 1998. Genetics and Analysis of Quantitative Traits. Sinauer Associates Inc., Sunderland, Mass.
  • Mauricio, R. & Mojonnier, L.E. 1997. Reducing bias in the measurement of selection. Trends Ecol. Evol. 12: 433436.
  • Merilä, J. 1996. Genetic variation in offspring condition: an experiment. Funct. Ecol. 10: 465474.
  • Merilä, J. 1997. Expression of genetic variation in body size of the collared flycatcher under different environmental conditions. Evolution 51: 526536.
  • Merilä, J. & Fry, J.D. 1998. Genetic variation and causes of genotype-environment interaction in the body size of blue tit (Parus caeruleus). Genetics 148: 12331244.
  • Merilä, J. & Sheldon, B.C. 2001. Avian quantitative genetics. Curr. Ornithol. 16: 179255.
  • Merilä, J., Sheldon, B.C. & Ellegren, H. 1997. Antagonistic natural selection revealed by molecular sex identification of nestling collared flycatchers. Mol. Ecol. 6: 11671175.
  • Merilä, J., Sheldon, B.C. & Ellegren, H. 1998. Quantitative genetics of sexual size dimorphism in the collared flycatcher, Ficedula albicollis. Evolution 52: 870876.
  • Merilä, J., Kruuk, L.E.B. & Sheldon, B.C. 2001a. Cryptic evolution in a wild bird population. Nature 412: 7679.
  • Merilä, J., Kruuk, L.E.B. & Sheldon, B.C. 2001b. Natural selection on the genetical component of variance in body condition in a wild bird population. J. Evol. Biol. 14: 918929.
  • Merilä, J., Sheldon, B.C. & Kruuk, L.E.B. 2001c. Explaining stasis: microevolutionary studies in natural populations. Genetica 112/113: 199222.
  • Nager, R.G., Keller, L.F. & Van Noordwijk, A.J. 1998. Understanding natural selection on traits that are influenced by environmental conditions. In: Adaptive Genetic Variation in the Wild (T. A.Mousseau, B.Sinervo & J. A.Endler, eds), pp. 95115. Oxford University Press, New York.
  • Neumaier, A. & Groeneveld, E. 1998. Restricted maximum likelihood estimation of covariances in sparse linear models. Genet. Sel. Evol. 30: 326.
  • Newton, I. 1989. Lifetime Reproduction in Birds. Academic Press, London.
  • van Noordwijk, A.J. 1988. Two stage selection for body size in the great tit. Proc. Int. Orn. Congr. Ottawa 1986: 14081415.
  • van Noordwijk, A.J. & Marks, H.L. 1998. Genetic aspects of growth. In: Avian Growth and Development (J. M.Starck & R. E.Ricklefs, eds), pp. 305323. Oxford University Press, Oxford.
  • van Noordwijk, A.J., van Balen, J.H. & Scharloo, W. 1988. Heritability of body size in a natural population of the great tit (Parus major) and its relation to age and environmental conditions during growth. Genet. Res. 51: 149162.
  • Pärt, T. 1990. Natal dispersal in the collared flycatcher: possible causes and reproductive consequences. Orn. Scand. 21: 8388.
  • Potti, J., Davila, J.A., Tella, J.L., Frias, O. & Villar, S. 2002. Gender and viability selection on morphology in fledgling pied flycatchers. Mol. Ecol. 11: 13171326.
  • Price, T.D., Kirkpatrick, M. & Arnold, S.J. 1988. Directional selection and the evolution of breeding date in birds. Science 240: 798799.
  • Rausher, M.D. 1992. The measurement of selection on quantitative traits: biases due to environmental covariances between traits and fitness. Evolution 46: 616626.
  • Réale, D., Festa-Bianchet, M. & Jorgenson, J.T. 1999. Heritability of body mass varies with age and season in wild bighorn sheep. Heredity 83: 526532.
  • Rice, W.R. 1989. Analysing tables for statistical tests. Evolution 43: 223225.
  • Roff, D.A. 1997. Evolutionary Quantitative Genetics. Chapman & Hall, New York.
  • SAS Institute, Inc. 1992. SAS User's Guide, Statistics. SAS Institute, Inc., Cary, NC.
  • Schluter, D. & Smith, J.N.M. 1986. Natural selection on beak and body size in the song sparrow. Evolution 40: 221231.
  • Sheldon, B.C., Kruuk, L.E.B. & Merilä, J. 2003. Natural selection and inheritance of breeding time and clutch size in the collared flycatcher. Evolution 57: 406420.
  • Stinchcombe, J.R., Rutter, M.T., Burdick, D.S., Tiffin, P., Rausher, M.D. & Mauricio, R. 2002. Testing for environmentally induced bias in phenotypic estimates of natural selection: theory and practice. Am. Nat. 160: 511523.
  • Tremblay, I., Thomas, D.W., Lambrechts, M.M., Blondel, J. & Perret, P. 2003. Variation in blue tit breeding performance across gradients in habitat richness. Ecology 84: 30333043.
  • Zandt, H., Strijkstra, A., Blondel, J. & van Balen, H. 1990. Food in two mediterranean blue tit populations: do differences in caterpillar availability explain differences in timing of the breeding season. In: Population Biology of Passerine Birds. An Integrated Approach (J.Blondel, A.Gosler, J. D.Lebreton & R.McCleery, eds), pp. 145155. Springer-Verlag, Berlin, Heidelberg.