SEARCH

SEARCH BY CITATION

Keywords:

  • animal model;
  • comb;
  • condition;
  • genetic correlation;
  • heritability;
  • residual mass

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

We studied the quantitative genetics of sexually selected traits in a captive population of red junglefowl (Gallus gallus L.) using a multi-generational ‘animal model’ approach. We found significant heritability of mass, tarsus length (both strongly sexually dimorphic), residual mass, and male comb (a fleshy head ornament) length. Residual mass has a genetic correlation between the sexes smaller than unity and so could show partially independent responses to selection in the two sexes. In males, tarsus length and mass were not genetically correlated, and this produced a negative genetic correlation between tarsus length and residual mass. The male red junglefowl's comb, an ornament influencing female choice, is highly condition dependent. We show that expression of this ornament is heritable, however, and shows strong genetic correlation with a condition index, residual mass. Because residual mass is partly influenced by various aspects of condition, it appears that comb size has ‘captured’ genetic variability in condition.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Much of our understanding of the evolution of sexually selected traits in vertebrates has been based on tests of the functions of these traits within a population or the distributions of traits among species (for examples, see Andersson, 1994). However, sexually selected traits, like all traits, develop as the result of interactions within the genotype and between genes and the environment (Kotiaho et al., 2001; Garant et al., 2004; but see Griffith et al., 1999; Merilä & Sheldon, 1999). To understand the evolution of sexually selected traits, it is thus important to determine the extent to which genetic determination of these traits differs between males and females, the pattern of genetic variance and covariance influencing the development of different traits within each sex, and the relative role of environment and genetics in determining expression of these traits (Price, 1996; Meriläet al., 1998; Badyaev, 2002). For instance, if two traits are strongly positively genetically correlated in males and females, distinct evolutionary pathways in the two sexes (sexual dimorphism) will be limited (Meriläet al., 1998; Jensen et al., 2003). Similarly, given a strong positive genetic correlation within either males or females, conflicting selection pressures on the two traits will limit the evolutionary response, while on the other hand a low genetic correlation will facilitate independent evolution of the two traits (Lande & Arnold, 1983).

The expression of some sexual signals appears to be dependent on individual condition (reviewed in Andersson, 1994). It may be that such traits have evolved developmental sensitivity to a number of aspects of condition. If condition is influenced by heritable aspects of the phenotype (for instance, parasite resistance), expression of the condition-dependent signal should ‘capture’ the genetic variability in condition and become heritable itself (Rowe & Houle, 1996). Thus, if a condition-dependent sexual signal appears heritable, we would expect this signal to show significant genetic correlations with heritable aspects of condition. This mechanism could explain why, despite strong directional selection, sexually selected traits appear to maintain their genetic variability (Rowe & Houle, 1996; see Kotiaho et al., 2001 for an example).

The aim of this study was first to partition phenotypic variation of sexual dimorphic characters within and among sexes to estimate genetic and environmental influences on these characters. To achieve this, we conducted multi-generational quantitative genetic analyses using ‘animal model’ techniques developed by domestic animal breeders. These methods have recently been applied successfully to the study of behavioural ecology and evolution to take advantage of the increased statistical power derived from multi-generational pedigree data (Kruuk, 2004). Based on data from three generations of captive red junglefowl (Gallus gallus L.), we examined residual mass (often referred to as the body condition index) and the two sexually dimorphic traits (mass and tarsus length) from which this index is derived. We asked to what extent the patterns of genetic variance and covariance influencing the development of these traits were similar among traits within sexes as well as between males and females. Sexual selection has presumably acted on these traits to create the dimorphism in the past, and current genetic correlations among traits and between the sexes will influence their continued evolution. We also considered the male junglefowl's comb, a highly condition-dependent sexual ornament. We examined its heritability and genetic correlations with other traits to establish potential constraints on its evolution and look for evidence of capture of genetic variance as predicted by Rowe & Houle (1996).

Study species

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Red junglefowl are the wild ancestors of the domestic chicken. The red junglefowl used in this study were descended from two populations acquired by the San Diego Zoo in the 1940’s. These zoo birds, like nearly all red junglefowl wild or captive, were probably introgressed to some extent with domestic chickens (Peterson & Brisbin, 1999). Despite some introgression, the zoo birds are quite similar morphologically to wild-type junglefowl and they differ dramatically in morphology from commercial chicken breeds subject to intense artificial selection. The zoo maintained a large population with free range of the zoo grounds for several decades before a large founding population (150 birds, sex ratio approximately unity) was acquired by researchers at the University of New Mexico (UNM; Albuquerque, NM, USA) in the mid-1980s.

Reproductive behaviour in this population of red junglefowl has been studied extensively. Male red junglefowl possess a fleshy comb atop their head that varies in size with male health (Zuk et al., 1990a; Zuk et al., 1990b), residual mass (or mass with tarsus as a covariate vs. comb length; Parker, 2003) and dominance status (Ligon et al., 1990; Zuk & Johnsen, 2000; Parker et al., 2002). Male dominance is an important determinant of reproductive access to females (Collias & Collias, 1996). However, female junglefowl also show active mate preferences (Collias & Collias, 1996) and prefer males with larger combs (Parker & Ligon, 2003). The sons of large-combed males inherit large-combs and improved residual mass compared with the sons of small-combed males (Parker, 2003). Males are much larger than females (Table 1). We have little direct evidence of the selective forces shaping red junglefowl sexual dimorphism. However, in males, residual mass, mass, and tarsus length can all correlate positively with dominance rank (T. Parker, unpublished data; Parker et al., 2002), probably because they bestow an advantage in male conflict. There may be selection for heavier females (Parker, 2002), although presumably this selection has been more strongly counter-balanced than that on males.

Table 1.  Sample sizes, phenotypic means, and estimates of variance components and heritability (h2) of morphological traits of male and female red junglefowl. Quantitative genetic estimates were generated with univariate animal models for male and female traits separately. Variance components include residual variance (VR), additive genetic variance (VA), and variance attributable to brood effects (VB) (comparable with maternal effects in this case).
TraitsnMean (SD)VRVAVBh2 (SE)
  1. Heritability **P < 0.01, ***P < 0.001.

Females
 Tarsus length (mm)26759.4 (3.3)3.305.031.880.54 (0.085)***
 Mass (g)267711.9 (85.0)999.984156.091549.500.62 (0.085)***
 Residual mass (g)2670.00 (75.0)1223.001924.50905.370.48 (0.094)***
Males
 Tarsus length (mm)22571.7 (3.4)5.944.700.400.43 (0.105)***
 Mass (g)2251025.8 (135.3)6110.484032.942409.080.32 (0.106)**
 Residual mass (g)2250.00 (114.4)1440.495740.431308.260.68 (0.120)***
 Comb length (mm)22155.4 (8.8)15.9423.1619.610.39 (0.117)**

Previous quantitative genetic analysis of morphological traits in this population was limited to comparing trait values of 18 fathers with measurements of their sons’ traits (Johnson et al., 1993). Tarsus length showed moderate (0.41 ± 0.20 SE), although marginally significant, heritability and comb length showed moderate (0.24 ± 0.16 SE) but nonsignificant heritability. Neither mass nor relative mass were examined.

Data

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Most of the birds used in this study were those included in a previous analysis designed to separate paternal genetic effects related to comb size from differential maternal investment in response to male comb size (Parker, 2003). A founding generation of 17 males and 24 females of unknown relatedness hatched in 1998 were bred to produce a second generation in 1999. Of the 201 chicks from the 1999 generation reaching adulthood, 24 males and 56 females were bred to produce a third generation in 2000. Of the birds hatched in 2000, 256 survived to be measured at adulthood. Thus 498 individuals are included in our pedigree. In 1999, most breeders were allowed to produce a second brood after the spring experimental breeding season. These birds were not included in previously published analyses of experimental treatment effects limited to first broods, but are included here where adding individuals with morphological data to the pedigree is a statistical asset. All birds produced in 1999 and 2000 were sired by artificial insemination. Females incubated their own eggs and reared chicks until three to five weeks old, at which point chicks were moved to standard density, single-age flocks containing up to 60 individuals. All birds were housed in large aviaries out of doors, were fed identical moderate protein poultry diets, and had ad libitum access to food and water. See Parker (2003) for addition details on the breeding and rearing conditions.

We took mass and tarsus length measurements for birds in the founding population after the birds reached adult size, but not at a consistent age because age was not known. However, we measured male comb size for most individuals in this founding group at approximately 6–8 months of age. This imprecision (limited to only 8% of the individuals in our pedigree) is likely to introduce some random variance into our analyses. In the two subsequent generations, we took mass, tarsus length, and comb length measurements at exactly 26 weeks (6 months) of age, 4 weeks after adult skeletal size had been reached (T. Parker, unpublished data). Mass gain is decreasing at this point, but individuals continue to become heavier for at least 4 weeks, and combs can continue to grow for at least another year. We measured mass to the nearest gram on an electronic balance and lengths with dial callipers to the nearest 10th of a millimetre. For males and females separately, we calculated residual mass (a condition index) as the residuals from a regression of mass on tarsus length.

Phenotypic coefficient of variation

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

We compared the phenotypic coefficient of variation (CV) between male comb length and male mass and tarsus length to determine if comb length, a demonstrated sexual signal, showed greater proportional variability than nonsignal traits. In order to compare the CV of mass, a three dimensional trait, on the same scale as the CV of comb length, a one dimensional measurement, we took the cube root of mass before calculating the CV (Lande, 1977).

Identifying environmental correlates of morphology

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Prior to conducting quantitative genetics analyses, we conducted mixed-models analyses of variance (SAS 8.2, proc mixed) to identify environmental variables correlated with the phenotypic traits of interest (tarsus length, mass, residual mass, comb length). Identifying correlated environmental variables was important so that we could then include these variables as fixed effects in our quantitative genetics analyses to account for heterogeneity in environmental effects on the phenotype. We included brood identity as a random variable in these mixed models to control for nonindependence of individuals within broods. Brood identity was not known for birds hatched in 1998, although these birds did come from two reproductively distinct groups. Thus, brood identity for 1998 birds was actually one of two groups of origin rather than true brood identity. Three variables were found to be important influences on the morphological traits. Environment of rearing, a categorical variable, was most important. We defined five alternative environments of rearing: (i) at UNM hatched in spring 2000, (ii) at UNM hatched in spring 1999, (iii) at UNM hatched in summer 1999, (iv) away from UNM location-A hatched spring/summer 1998, and (v) away from UNM location-B hatched spring/summer 1998. The two non-UNM locations were private properties within 40 km of UNM. We found that this variable influenced all traits except female tarsus length, and thus we included it as a fixed effect for all traits except female tarsus length in the quantitative genetics analyses. The second variable found to be important was the environment of rearing for the individual's mother. The categories for this variable were similar to those for the previous variable. This variable was a significant predictor of male mass and condition, and so we included it as a fixed effect in the quantitative genetic analyses of these traits. Lastly, hatch year was a significant predictor of female tarsus length, and so we included hatch year as a fixed effect in quantitative genetic analysis of this trait.

Heritability and components of variance

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

We estimated heritability of phenotypic traits through a mixed model restricted maximum likelihood (REML) estimation procedure using the software package VCE4 (Neumaier & Groeneveld, 1998). We thus used the pedigree information (available for 498 individuals) to fit an individual ‘animal model’ (Lynch & Walsh, 1998), which partitions phenotypic variance in a quantitative character into its additive genetic and other fixed and random components such as common environment (Meyer, 1989). As described above, three categorical variables, environment of rearing, environment of mother's rearing, and year, were included, where appropriate, as fixed effects to account for temporal and spatial heterogeneity in environmental effects on the phenotype. The brood identity (combination of parental identities) was always fitted as a random effect to account for common-environment effects specific to the individual brood, as well as maternal effects (Kruuk, 2004). Paternal effects are also accounted for by brood identity, but they are likely to be quite weak because we used artificial insemination, and so the only environmental sire effects could be those mediated through chemicals in the ejaculate. No evidence exists for such a sire effect in birds, although it has almost never been investigated (Parker, 2003). Total phenotypic variance (VP) of each trait was therefore partitioned into additive genetic variance (VA), common environmental variance because of brood (VB), and residual variance (VR). The narrow-sense heritability (h2) was estimated as the ratio of the additive genetic variance (VA) to the total phenotypic variance (VP): h2 = VA/VP.

Genetic covariance

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

We used a multivariate animal model to calculate the genetic correlation among traits both within and among sexes. The same fixed and random effects were used in these analyses. In the case of the genetic correlation between the sexes, any deviation from unity might also potentially constrain evolutionary response, particularly if selection acted differently on the two sexes. Standard errors for heritabilities and genetic correlations were computed using the program VCE4 (Neumaier & Groeneveld, 1998), and significance of estimates was tested using two-tailed t-tests.

Conditional heritability

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

For each trait in males and females we determined the amount of additive genetic variance independent of the additive genetic variance in that trait for the other sex (Hansen et al., 2003; Jensen et al., 2003). Based on the additive genetic variance for a trait in sex x [VA(x)] and y [VA(y)], and the additive genetic covariance for this trait between the sexes [COVA(x|y)], we calculated the conditional additive genetic variance [VA(x|y)] for each sex as VA(x|y) = VA(y) − [COVA(x|y)]2/VA(x). We could then calculate, for each trait in both sexes, the conditional heritability: inline image (Jensen et al., 2003). This is the heritability of the trait in one sex that is independent of the genetic variance for this trait in the other sex.

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Red junglefowl are highly sexually dimorphic. Based on our measurements, male mass was 44% higher than female mass (Table 1). Also, males’ tarsi were 20% longer than females’ (Table 1). Because combs on these young birds were still growing, the mean comb length (Table 1) was well below the values for most adults (T. Parker, unpublished data). By the end of their first breeding season, this mean would have increased by approximately 20–30 mm (T. Parker, unpublished data). The phenotypic coefficient of variation (CV) for the cube root of male mass [4.44 ± 0.21 (SE)], for raw male mass [13.19 ± 0.63 (SE)], and for male tarsus [4.74 ± 0.22 (SE)], were all significantly lower than the phenotypic CV for male comb length [15.88 ± 0.77 (SE)].

Among sexes genetic estimates

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

All the traits we examined showed significant heritability (Table 1). Females showed a higher heritability of mass than did males (t490 = 2.21, P = 0.028; Table 1) because residual variance was much higher in males (males: VR/VP = 0.49 ± 0.10 SE; females; VR/VP = 0.15 ± 0.07 SE; see also Table 1). Heritability estimates for residual mass did not differ significantly between males and females (Table 1). Males and females also did not show significantly different heritability of tarsus length (Table 1). Genetic correlations among males and females for each trait were highly significantly different from 0 (all P < 0.001) (Table 2). Genetic correlation between male and female tarsus was not different from 1 (Table 2). For male and female mass, genetic correlation was marginally nonsignificantly different from unity (P = 0.099; Table 2). However, male and female residual mass genetic correlation was significantly <1 (P < 0.001; Table 2). Therefore, the pattern of genetic covariance for this trait is not entirely the same for males and female. When we calculated conditional heritability (the potential for the traits to evolve separately in each sex) for mass and residual mass, we found rather high values (0.15–0.51) for males and females (Table 2). In other words, it seems that mass or residual mass could both potentially respond to selection independently in males and females.

Table 2.  Brood effect (rB) and genetic (rG) correlations between females and males for three morphological traits. The tarsus length model did not converge and thus we were unable to generate standard errors for the genetic correlation between female tarsus and male tarsus. We also present the conditional additive genetic variance (VA(x|y)) and the conditional heritability (inline image) for each trait in males and females. These values represent the potential for each trait to evolve separately in each sex.
 rB (SE)rG (SE)FemaleMale
VA(x|y)inline imageVA(x|y)inline image
Tarsus length  (mm)0.522 (.)1.00 (.)0.000.000.000.00
Mass (g)0.81 (0.164)0.76 (0.145)1730.3070.261455.2070.15
Residual  mass (g)0.92 (0.147)0.56 (0.113)1572.1130.383424.4070.51

Within sex genetic estimates

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

When we examined males and females separately, we found similar patterns of phenotypic correlations among traits for both sexes. In both sexes mass and tarsus length and mass and residual mass show highly significant positive phenotypic correlations (Tables 3 and 4). However, patterns of genetic correlations among traits differed for males and females (Tables 3 and 4). In females, mass and tarsus length were strongly positively genetically correlated (P < 0.001), as were residual mass and tarsus length (P < 0.001; Table 3). However, in males, mass and tarsus length were not significantly genetically correlated (P = 0.338), and residual mass and tarsus were significantly negatively correlated (P = 0.004; Table 4). These different patterns between males and females are not the result of the use of different covariates for tarsus in the two sexes. When the female data were re-analysed (data not shown) using the same covariate for tarsus as was used for male tarsus (environment of rearing), results did not differ from those we present here.

Table 3.  For female junglefowl, phenotypic (r2; above the diagonal) and genetic (plus SE; below the diagonal) correlations among morphological traits. Because mass and residual mass strongly correlate our model did not converge, and we were unable to generate standard errors for genetic correlations when both variables were included in the model. Therefore the genetic correlation for tarsus length and mass were generated in a multivariate animal model without residual mass and the genetic correlations between tarsus length and residual mass were generated with a multivariate animal model not including mass (Jensen et al., 2003).
TraitTarsus length (mm)Mass (g)Residual mass (g)
  1. *Phenotypic correlations P < 0.001.

Tarsus length (mm) 0.22*0.00
Mass (g)0.842 (0.053) 0.80*
Residual mass (g)0.605 (0.123)  
Table 4.  For male junglefowl, phenotypic (r2; above the diagonal) and genetic (plus SE; below the diagonal) correlations among morphological traits. Because mass and residual mass strongly correlate our model did not converge, and we were unable to generate standard errors for genetic correlations when both variables were included in the model. Therefore genetic correlations among all variables except residual mass were generated in one multivariate animal model. Correlations with residual mass were generated in a separate multivariate animal model not including mass (Jensen et al., 2003).
TraitTarsus length (mm)Mass (g)Residual mass (g)Comb length (mm)
  1. *Phenotypic correlations P < 0.001.

Tarsus length (mm) 0.28*0.000.05*
Mass (g)0.170 (0.177) 0.72*0.55*
Residual mass (g)−0.488 (0.166)  0.53*
Comb length (mm)−0.477 (0.125)0.643 (0.085)0.861 (0.068) 

In both sexes, a significant proportion of variance in mass (males 0.19 ± 0.06 SE; females: 0.23 ± 0.05 SE) and residual mass (males: 0.15 ± 0.07 SE; females: 0.22 ± 0.05 SE) phenotype (VB/VP) was explained by brood effects, but neither females (0.07 ± 0.05 SE) nor males (0.04 ± 0.05 SE) showed significant brood effect on tarsus. These brood effects were strongly correlated between males and females (Table 2) suggesting maternal effects had similar influences on both sexes.

We found significant heritability of comb length in males (Table 1). Comb length was significantly positively genetically correlated with mass (P < 0.001), but negatively genetically correlated with tarsus length (P < 0.001; Table 4). Most interestingly, comb length and residual mass were strongly genetically correlated (Table 4), consistent with capture of genetic variance in condition by a sexually selected ornament. This correlation was significantly >0 (P < 0.001) and <1 (P = 0.042). Also consistent with its demonstrated condition dependence, the amount of variance attributable to brood effect for comb was nearly as much as that attributable to additive genetic effects (Table 1).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

Although red junglefowl are strongly sexually dimorphic, there currently exist some constraints on the further evolution of dimorphism. The genetic correlation between males and females for tarsus length is 1.0, indicating that neither male nor female tarsus length can respond to selection without a concurrent response in the other sex. However, both mass and residual mass show signs of partially independent genetic covariance in males and females (genetic correlations <1.0 and conditional heritability >0). This explains previous results from this data set in which sons of large-combed sires had significantly higher residual mass than sons of small-combed sires, but daughters showed a weaker, nonsignificant trend in this direction (Parker, 2003). Because mass and residual mass should be able to respond independently to selection in both males and females, further increases or decreases in sexual dimorphism could be facilitated. However, divergence between the sexes would be concurrently hindered by genetic nonindependence of male and female tarsus length. A recent study of house sparrows (Passer domesticus) found no between-sex genetic correlations significantly <1.0 (Jensen et al., 2003). This different pattern may partly relate to the fact that house sparrows are much less sexually size-dimorphic than junglefowl.

Genetic correlations among traits within a sex can also constrain evolutionary responses. We observed positive genetic correlations among female traits, indicating that a positive response to selection in one trait will produce positive responses in other traits. Positive genetic correlations are common among morphological traits (Roff, 1996). In male junglefowl, however, the situation appears more complex than in females. Because of a lack of genetic correlation, mass and tarsus length should be able to evolve independently, but an evolutionary increase in residual mass and comb length would appear to require evolutionary decreases in tarsus length. The negative genetic correlation between tarsus and residual mass in males appears to be an inevitable outcome of two factors: (i) in males, tarsus length and mass are not significantly genetically correlated, so birds inheriting longer tarsi will not necessarily inherit higher mass, and (ii) for a bird of a given mass, residual mass will always be smaller if tarsus length is longer. This is not seen in females because there is a positive genetic correlation between mass and tarsus length, so individuals with longer tarsi are likely to also have greater mass. Interestingly, an almost identical pattern was found in male house sparrows by Jensen et al. (2003), although they did not interpret it as a potential statistical artefact. We will discuss the biological relevance of this finding, but first, we explore the statistical and biological implications of using residual mass as an index of condition.

There are potential statistical pitfalls associated with using residuals as an index of condition (Darlington & Smulders, 2001; Garcia-Berthou, 2001; Freckleton, 2002). Authors critical of residuals have suggested as alternatives using mass with tarsus length as a covariate in analyses of covariance (Garcia-Berthou, 2001) or multiple regression (Freckleton, 2002). Accordingly, we conducted our quantitative genetics analyses a second time using mass with tarsus as covariate as a condition index instead of residual mass. Our estimates, including the negative genetic correlation, were similar with this method, but matrices would not converge to produce confidence intervals. Therefore, we chose to report results based on residuals instead.

Regardless of the potential statistical pitfalls, we must also consider the biological utility of using residual mass as a proxy for condition in analyses such as these. Residual mass has a long history of successful use as an index of condition in birds and other taxa (e.g. Hochachka & Smith, 1991; Linden et al., 1992; Boltnev et al., 1998; Sorci & Clobert, 1999; Kotiaho et al., 2001; Meriläet al., 2001). Much of this work has been quite fruitful, and very interesting evolutionary patterns have been uncovered through its study (e.g. Meriläet al., 2001). However, variation in mass is not always a reliable index of condition (e.g. Gosler & Harper, 2000). Appropriateness of residual mass should thus be assessed on a case-by-case basis.

Our results demonstrate that genetic variation influences phenotypic variation in residual mass, a finding common in birds (e.g. Meriläet al., 2001; Jensen et al., 2003). The question central to interpretation of our results is: does heritability of residual mass represent heritability of condition? In other words, does residual mass ‘capture’ genetic variation in condition in the sense of Rowe & Houle (1996)? The non-exclusive alternative is that there is a heritable range of residual mass in the population that is unrelated to individual condition. The most intuitive way for a trait to capture genetic variation in condition would be for the trait to be sensitive to phenotypic components of condition. As such, because phenotypic condition is influenced by the environment and by genes (e.g. both exposure to and susceptibility to pathogens), under this scenario, capture of genetic variation in condition will come at the price of capture of environmental variability in condition. Expression of a condition-dependent trait will then reflect genetic quality to the extent that variation in condition is due to variation in genes (Rowe & Houle, 1996).

In captive red junglefowl, much evidence demonstrates that residual mass reflects condition. For instance, diseased red junglefowl are often lighter than healthy individuals. This has been documented through experimental infection with intestinal nematodes (Chappell et al., 1997); a treatment which decreased mass more consistently than tarsus during development (Zuk et al., 1990b). Observation from years of handling junglefowl confirm that ill adult junglefowl tend to loose mass (T. Parker, personal observation, J. D. Ligon, personal communication). The birds in our study apparently were exposed to pathogens (possibly from wild birds outside, or occasionally entering, the aviaries), and some individuals temporarily stopped gaining mass during development, presumably because of illness. In our quantitative genetics analyses, mass and residual mass both showed stronger brood effects than tarsus length, indicating greater sensitivity to common environmental influences. Several patterns observed with dominance interactions also suggest that residual mass is related to condition. Residual mass and mass controlling for tarsus in a multiple regression can be strong predictors of male dominance rank in captivity (r = 0.56 − 0.72; T. Parker, unpublished, Parker, 2003). Residual mass or mass may not be influenced by dominance status (Zuk & Johnsen, 2000; Parker et al., 2002). Instead, it may directly influence competitive ability or correlate with other factors influencing competitive ability, such as aerobic capacity, a plausible indication of condition, and a known correlate of dominance rank in male junglefowl (Hammond et al., 2000). Finally, dominance interactions can cause adult male junglefowl to loose mass (Zuk & Johnsen, 2000), possibly because these interactions are stressful (Parker et al., 2002). Thus, although our birds were provided with ample food and sheltered from some challenges of wild living, evidence is strong that residual mass did reflect condition.

Because variation in residual mass in red junglefowl appears to represent variation in measures of individual condition, it is an excellent trait with which to compare a sexually selected ornament for a test of the ‘capture of genetic variance’ hypothesis (Rowe & Houle, 1996) to explain heritability of condition dependent ornaments. Extensive empirical evidence demonstrates that the red junglefowl comb is condition dependent. Comb size is altered by manipulating individuals’ social status [r = 0.35 − 0.43; if not reported in citation, we estimated correlation coefficients from reported statistics using the program MetaWin (version 1.0) (Sinauer Associates, Sutherland, MA, USA) to generate a standard measure of effect size] (Zuk & Johnsen, 2000; Parker et al., 2002) or health (r = 0.30 − 0.43) (Zuk et al., 1990b; Chappell et al., 1997). Comb size also correlates with unmanipulated variation in health (r < 0.61) (Zuk et al., 1990a), immune function (Zuk et al., 1995; Zuk & Johnsen, 1998), and dominance status (r = 0.62) (Parker et al., 2002). Residual mass, of course, also correlates with comb size (r = 0.53) (Parker, 2003) and variation in mass explains statistically much of the difference in comb size between experimentally infected and healthy birds (Chappell et al., 1997). Further, we observed a significantly stronger phenotypic coefficient of variation for comb length than for mass. In other words, comb shows greater proportional variation than does mass. This is consistent with the role of comb length as a signal selected to amplify information about variation in male quality. We found that residual mass and comb size are strongly genetically correlated, indicating that comb size has ‘captured’ genetic variation in condition (see Rowe & Houle, 1996; Kotiaho et al., 2001). In other words, the genetic variance influencing differences in residual mass is, to a large extent, the same genetic variance influencing differences in comb size. If, as we propose, some of the genetic variance influencing residual mass is the genetic variance influencing various aspects of condition (such as disease susceptibility or competitive ability) then comb size will represent this genetic variance as well. Our results are similar to those of some other recent studies finding evidences of heritable condition correlated with heritable ornament expression (e.g. David et al., 2000; Kotiaho et al., 2001). In particular, Kotiaho and co-workers (2001) demonstrated that in the dung beetle (Onthophagus Taurus) that heritable differences in residual mass were strongly genetically correlated with heritable variation in male display rate (a condition-dependent trait preferred by females).

Now that we have explored the biological relevance of residual mass and its relationship with comb size, we can return to the question of the negative genetic correlation between tarsus length and these other two traits. A negative genetic correlation between two traits should mean that parallel evolutionary change will be constrained (Lande & Arnold, 1983). Would evolutionary elongation of tarsi be hindered by positive selection for birds with higher residual mass? If residual mass itself were only an index of other aspects of condition, then there might not be strong selection on residual mass, per se, but only on the aspects of condition underpinning it. Thus the negative genetic correlation between residual mass and tarsus length would not necessarily hinder evolution of tarsus length. However, some heritable aspect of mass and/or residual mass that may be independent of condition may itself be under selection, possibly because it influences male competition or comb size. That both residual mass and comb size show a similar negative genetic correlation with tarsus length indicates that residual mass does have a direct influence on comb growth. Therefore evolutionary increase in relative mass or comb size would be partly constrained if selection favoured stable or increased tarsus length. Of course, this genetic correlation is only partial, so such divergence would not be entirely prevented. If residual mass were just a correlate of comb size and they were both influenced by underlying aspects of condition, the phenotypic and genetic correlation between the two would weaken during an evolutionary change in tarsus length, but would return to its early strength under stabilizing selection on tarsus length.

Because we have conducted these analyses with the ancestor to the domestic chicken, we can compare our results to those from domestic birds for which animal models methods are routinely used to assess quantitative genetics parameters. Domestic chickens, especially those typically studied by poultry scientists, are heavily artificially selected. Males in commercial lines can weigh between 3 and 4 kg (Hagger, 1994), three times more than junglefowl, and these domestic birds have radically altered sexual ornaments, morphology, and reproductive patterns (Pizzari et al., 2004). Despite these great differences, the quantitative genetics of the morphology of these groups appears similar. In domestic lines, as in our population of junglefowl, male and female body mass show a strong positive genetic correlation (Hagger, 1994; Le Bihan-Duval et al., 1998; Mignon-Grasteau et al., 1998; Mignon-Grasteau et al., 1999), heritability of mass is moderate (Mignon-Grasteau et al., 2001) to high (Hagger, 1994) and tarsus length is strongly heritable as well (Tixier-Boichard et al., 1995) (Table 5). Heritability of female body mass in chickens is sometimes (Le Bihan-Duval et al., 1998; Mignon-Grasteau et al., 1999), but not always (Tixier-Boichard et al., 1995; Mignon-Grasteau et al., 2001) higher than that for male body mass (Table 5). Unfortunately, neither residual mass nor comb length was included in any of the animal model analyses we located for domestic chickens. Comb length is rarely reported in poultry studies (Johnson et al., 1993), but it does appear heritable in chickens, and this heritability seems to be partially mediated through heritability of a trait important to condition; immune response (Verhulst et al., 1999). Thus there is little evidence for divergence in quantitative genetic parameters between artificially selected lines and wild-type birds. However, at this point we can only draw comparisons based on a few traits.

Table 5.  Some estimates of heritability (h2) of body mass in male and female domestic chickens based on the animal model method. Where available, we also show genetic correlations (rG) between female and male body mass.
 AgeMale h2Female h2rG
Hagger (1994)Adult0.790.730.84
Le Bihan-Duval et al. (1998)Week 70.390.460.94
Mignon-Grasteau et al. (1998)Week 80.280.430.84
Mignon-Grasteau et al. (1999)Week 80.390.450.97
Mignon-Grasteau et al. (1999)Week 360.610.640.51
Mignon-Grasteau et al. (2001)Week 200.490.43 
Mignon-Grasteau et al. (2001)Week 360.610.64 
Tixier-Boichard et al. (1995)Adult0.610.560.71

In conclusion, our analyses suggest that much of the pattern of genetic variance and covariance of sexually dimorphic traits is the same in both sexes but that there are still some differences that could potentially lead to independent responses to selection in the two sexes. Length of the male's comb is heritable and shows strong genetic correlation with residual mass. Because residual mass is partly influenced by various aspects of condition, it appears that comb size has ‘captured’ genetic variability in condition as predicted by theory explaining the persistence of genetic variability in sexual signals under strong directional selection (Rowe & Houle, 1996; see also Kotiaho et al., 2001).

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References

J.D. Ligon, A. Kodric-Brown, and R. Thornhill provided advice throughout data gathering. N. Parker, C. Fellows, J. Rosenfield, and J. Hill were helpful in many ways. J. Abbott, M. Edgar, J. Mayfield, S. Mercen, E. O'Keefe, G. Quintana, and Y. Romero assisted with animal care, husbandry, and data collection. R. Ricci and F. Gurule were consistently helpful in facilitating animal care. Two anonymous referees provided excellent suggestions for the manuscript. While conducting these analyses, THP was supported by a U.S. National Science Foundation (NSF) International Research Fellowship (INT-0202704) and DG was supported by a Natural Sciences and Engineering Research Council of Canada (NSERC) Postdoctoral Research Fellowship, and by a Biotechnology and Biological Sciences Research Council (BBSRC) grant to B.C. Sheldon and L.E.B. Kruuk. Funding for the research came from a U.S. National Science Foundation Grant (IBN-0072995), a Sigma Xi Grant-in-Aid of Research, the American Ornithologists’ Union, and University of New Mexico sources including the Department of Biology's Graduate Research Allocation Committee and Grove Scholarship Committee, the Graduate and Professional Student Association's Student Research Allocation Committee, and the Office of Graduate Studies’ Research, Planning, and Travel fund. The research presented here was described in Animal Research Protocol No. 9817-B approved on 25 September 1998 and by an addendum approved 13 December 1999 by the Institutional Animal Care and Use Committee of the Main Campus of the University of New Mexico.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Study species
  6. Data
  7. Data analysis
  8. Phenotypic coefficient of variation
  9. Identifying environmental correlates of morphology
  10. Heritability and components of variance
  11. Genetic covariance
  12. Conditional heritability
  13. Results
  14. Among sexes genetic estimates
  15. Within sex genetic estimates
  16. Discussion
  17. Acknowledgments
  18. References