"Estimation of Heritability for Fluctuating Asymmetry in Chickens"
Estimation of Heritability for Fluctuating Asymmetry in Chickens by Restricted Maximum Likelihood. Effects of Age and Sex J. L. Campo,1 M. G. Gil, S. G. Davila, and I. Munoz ´ ˜ ´ ´ Departamento de Mejora Genetica Animal, Instituto Nacional de Investigacion Agraria y Alimentaria, Apartado 8111, 28080 Madrid, Spain ABSTRACT The purposes of the present study were to tween sides was not signiﬁcantly different from one, indi- estimate the heritability of the ﬂuctuating asymmetry in cating that differences between sides were purely envi- chickens, using the restricted maximum likelihood proce- ronmental in origin. Different traits rarely showed much dure, and to evaluate the effects of age and sex on the correlation in their level of ﬂuctuating asymmetry, indi- ﬂuctuating asymmetry. Leg, wing, and feather lengths cating that the level of ﬂuctuating asymmetry in all traits and ear-lobe and wattle areas were measured. In experi- did not reﬂect equally the quality of animals. In experi- ment 1, 1,073 birds were used from 2 generations with ment 2, ﬂuctuating asymmetry differences among ages complete pedigree of the Quail Castellana breed to esti- and sexes were investigated at 8, 12, 16, 20, 24, 28, 32, 36, mate the heritability for the ﬂuctuating asymmetry at 36 and 40 wk in 360 birds from the same breed. Signiﬁcant wk of age. The estimated heritability of absolute ﬂuctuat- variation with age was observed in leg length, wing ing asymmetry was not signiﬁcantly different from zero length, feather length (females), and wattle area (females), for all 5 traits, and similar estimates were obtained for which was mainly related to onset of sexual maturity relative ﬂuctuating asymmetry, directional asymmetry, transformed absolute and relative ﬂuctuating asymmetry, and adult stage. Females showed signiﬁcantly greater and 3 alternative indexes of ﬂuctuating asymmetry. The ﬂuctuating asymmetry for ear-lobe area than males. There heritability of the combined absolute or relative ﬂuctuat- were signiﬁcant differences in ﬂuctuating asymmetry for ing asymmetry was still very low, indicating that ﬂuctu- wing length, feather length, and wattle area near the onset ating asymmetry was determined solely by environmen- of sexual maturity, with males having signiﬁcantly tal sources of variation and that ﬂuctuating asymmetry greater ﬂuctuating asymmetry than females for wing estimates should not be confounded by appreciable addi- length and the opposite being true for feather length and tive genetic contributions. The genetic correlation be- wattle area. (Key words: ﬂuctuating asymmetry, heritability, age, sex effect, chicken) 2005 Poultry Science 84:1689–1697 INTRODUCTION of welfare and ﬁtness at the individual level following development (Jones, 1987; Clarke, 1995; Moller et al., 1995; Fluctuating asymmetry (FA) is the most recommended Moller, 1997, 1999; Tuyttens, 2003), although the detection index of developmental instability (Palmer and Strobeck, of asymmetry can be difﬁcult because the left-right differ- 1992). It is considered to be a valid indicator of genetic ences are frequently weak, about 1 to 5% of trait size and environmental stress during development (Clarke et (Merila and Bjorklund, 1995; Gangestad and Thornhill, al., 1986; Leary and Allendorf, 1989; Parsons, 1990, 1992), 1999). Additionally, there are 4 important confounding especially if the analyses are based on the composite FA factors that complicate the analysis of asymmetry: differ- of multiple traits (Leung et al., 2000). Fluctuating asym- ent types of bilateral asymmetry, measurement error, FA- metry is characterized by a normal frequency distribution size relation, and genetic predisposition toward asymme- of left minus right side differences with a mean of zero try. In contrast to FA, directional asymmetry (DA) and (van Valen, 1962; Palmer and Strobeck, 1986, 1992; antisymmetry (AS) have not been generally associated Palmer, 1994; Polak and Trivers, 1994; Swaddle et al., with stress and are characterized by normal distribution 1994). Fluctuating asymmetry is used as a measurement with a mean of not zero, and nonnormal distribution with a mean of zero, respectively. Any 2 or all 3 types of 2005 Poultry Science Association, Inc. Received for publication April 1, 2005. Abbreviation Key: AS = antisymmetry; DA = directional asymmetry; Accepted for publication July 11, 2005. FA = ﬂuctuating asymmetry; L = left value; R = right value; REML = 1 To whom correspondence should be addressed: firstname.lastname@example.org. restricted maximum likelihood. 1689 1690 CAMPO ET AL. asymmetry can occur together in the same trait, particu- age and sex on the FA. The current study relative to the larly the DA-FA combination (van Valen, 1962). As FA, existing literature analyzes for the ﬁrst time the heritabil- measurement errors are normally distributed about a ity of FA in chickens, using the REML procedure under mean of zero, and it is essential to obtain estimates of a general mixed model. FA that are unbiased with respect to measurement error (Leamy, 1984; Palmer and Strobeck, 1986; van Dongen et MATERIALS AND METHODS al., 1999a; van Dongen, 2000). The FA-size relation has commonly been removed by dividing the asymmetry A synthetic breed (Quail Castellana) that originated score by the trait mean (Palmer and Strobeck, 1986), al- from an F2 cross between the Castellana and Buff Prat though it can be undesirable if there is anisometry (Gra- Spanish breeds of chickens (Campo and Orozco, 1986; ham et al., 1998; Leung, 1998; Nachman and Heller, 1999). Campo, 1991) was used in experiments 1 and 2. In experi- Overall most workers agree that FA often has a herita- ment 1, a total of 1,073 birds (358 males and 715 females) ble component, but this is typically small. The heritability from 2 consecutive generations (236 and 837 birds, respec- of FA found in the literature generally ranges from 0.0 tively) with complete pedigree was used to estimate the to 0.1. Heritability has usually been estimated from con- heritability for the FA at 36 wk of age. To study the effect ventional parent-offspring or sib analyses by using differ- of age, 180 females were sampled in experiment 2 and ent types of organisms. In the fruit ﬂy, Reeve (1960), equally divided into 9 age groups, each group consisting Scheiner et al. (1991), Woods et al. (1998), Polak and of 20 birds that were tested for FA at 8, 12, 16, 20, 24, 28, Starmer (2001), and Polak and Stillabower (2004) did not 32, 36, and 40 wk of age. One hundred eighty males were ﬁnd signiﬁcant heritability for bristle number, wing tested similarly. length, wing width, and cross vein length. For the house Five bilateral traits (3 morphological and 2 secondary ﬂy, Chapman and Goulson (2000) reported that wing sexual) were measured. The recorded morphological length FA had no detectable heritable component. For traits were right (R) and left (L) shank (metatarsus), wing butterﬂies, Windig (1998) and Windig and Nylin (2002) (radius), and second primary feather lengths. Only un- indicated that heritabilities were not different from zero damaged birds with intact second primary feathers were for 2 pupal and 2 adult traits, and for moths van Dongen analyzed for feather length (948 instead of 1,073 birds in et al. (1999b) found that the heritability was low for tibia experiment 1). The recorded secondary sexual traits were length. Eggert and Sakaluk (1994) and Roff and Reale right and left ear-lobe and wattle areas. Right and left (2004) found no signiﬁcant heritability in crickets for 3 values of a bird were taken during the same session. All tegminal characters, hind femur, mid femur, fore femur, 5 lengths and ear-lobe and wattle widths were measured and hind tibia. Leary et al. (1985, 1992), Bryden and Heath in millimeters using a digital calliper; ear-lobe and wattle (2000), and Hermida et al. (2002) found no signiﬁcant areas were calculated by multiplying lengths by widths heritability for several meristic and metric traits in rain- (indicated values are measured in cm2). Trait size was bow trout, chinook salmon, or threespine stickleback, re- the mean of the right and left traits [(R+L)/2]. All traits spectively. For the swallow, Moller (1994) and Cadee showed normal frequency distributions. (2000) indicated no signiﬁcant heritability for tarsus, The FA for a trait was deﬁned by the absolute difference wing, or tail lengths. Finally, no signiﬁcant heritability between sides [|R−L|]. The unsigned value (equivalent to was reported by Leamy (1984, 1986, 1997, 1999) for several the mean deviation) is more appropriate than the variance osteometric, dental, skeletal, and mandibular characteris- of right minus left sides, because this index of FA cannot tics in mice; Kruuk et al. (2003) for antler size in deer; be used to measure heritability. Regardless, the mean of and Corruccini and Potter (1981) for dental traits in |R−L| is an estimator of the standard deviation of (R−L), Homo sapiens. the ratio between them being equal to 0.8 (Windig and However, heritability estimates of FA can be large un- Nylin, 2000). A series of steps (Palmer, 1994) were fol- derestimates of the heritability of developmental stability lowed before identifying exhibited asymmetry as FA. (Whitlock, 1996, 1998; van Dongen, 1998a), because the First, the presence of DA (normal distribution with a heritability of FA is equivalent to the repeatability of mean of not zero) and AS (nonnormal distribution with the asymmetry estimate multiplied by the heritability of a mean of zero) was tested by inspection of the distribu- developmental stability. In contrast to FA, both DA and tion of signed right minus left differences (R−L). The AS can have an unknown genetic component (Palmer et presence of DA was tested for using t-tests. Departures al., 1994; Leamy, 1999). However, individual and family from normality (e.g., AS) were assessed using skewness selection for DA in Drosophila was not successful (Coyne, and kurtosis measures; antisymmetry is characterized by 1987; Tuinstra et al., 1990). Heritability of DA is a function bimodal (or broad-peaked) distributions, tending to be of the trait heritability and the genetic and phenotypic platykurtic. If differences in (R−L) exist, asymmetry correlation between trait sides (Roff and Reale, 2004). should be leptokurtically distributed: the greater the mag- The purposes of the present study were to estimate the nitude of differences, the greater the leptokurtosis. Sec- narrow sense heritability of the ﬂuctuating asymmetry in ond, FA and measurement error are normally distributed chickens, using the restricted maximum likelihood about a mean of zero. Thus, it is essential to show that (REML) procedure (Patterson and Thompson, 1971) un- the variance in asymmetry observed between individuals der a general mixed model, and to evaluate the effects of is greater than the variance due to measurement error. HERITABILITY FOR FLUCTUATING ASYMMETRY 1691 Table 1. Mean values and heritabilities ± SE for bilateral asymmetry of various morphological and secondary 1 sexual traits at 36 wk of age, experiment 1 Trait2 Mean Heritability Leg length (mm) (R+L)/2 103.76 ± 0.09 0.56 ± 0.07 (R−L) 0.75 ± 0.06 DA 0.01 ± 0.03 |R−L| 1.54 ± 0.03 0.01 ± 0.03 2|R−L|/(R+L) × 100 1.5 ± 1.2 0.02 ± 0.03 Wing length (mm) (R+L)/2 91.85 ± 0.08 0.39 ± 0.05 (R−L) −0.09 ± 0.09 FA 0.05 ± 0.03 |R−L| 1.8 ± 0.04 0.01 ± 0.03 2|R−L|/(R+L) × 100 1.9 ± 1.4 0.09 ± 0.05 Feather length (mm) (R+L)/2 163.37 ± 0.12 0.02 ± 0.02 (R−L) −0.41 ± 0.27 FA 0.00 ± 0.05 |R−L| 3.79 ± 0.06 0.00 ± 0.02 2|R−L|/(R+L) × 100 0.2 ± 0.1 0.00 ± 0.04 Ear-lobe area (cm2) (R+L)/2 4.28 ± 0.04 0.17 ± 0.04 (R−L) 0.11 ± 0.02 DA 0.02 ± 0.02 |R−L| 0.47 ± 0.02 0.01 ± 0.02 2|R−L|/(R+L) × 100 11.5 ± 0.9 0.00 ± 0.07 Wattle area (cm2) (R+L)/2 17.57 ± 0.10 0.15 ± 0.03 |R−L| 0.02 ± 0.09 FA 0.00 ± 0.00 |R−L| 1.72 ± 0.05 0.03 ± 0.03 2|R−L|/(R+L) × 100 9.9 ± 1.0 0.02 ± 0.03 1 FA = ﬂuctuating asymmetry; DA = directional asymmetry. 2 R = right side; L = left side. The FA is often small and sometimes of the same magni- of the relative asymmetries of the different traits; the tude as measurement error. The determination of mea- summation of n|R−L|/Σ|R−L| across traits only is good if surement error is important because increasing measure- there is low kurtosis (Leung et al., 2000). If no size correc- ment error causes an increase in the degree of departure tion was needed, each |R−L| value would be standardized of the observed FA from the true FA. Large sample sizes and averaged across traits to give a total score. are needed to estimate the heritability of FA, and repeat Narrow sense heritability must be estimated by using a measurements of traits may not be practical. The effect breeding design that is able to distinguish among additive of measurement error should be calculated from a sub- and residual variances. The REML estimator takes into sample of at least 20 birds. The amount of error variation account all the relationships in the data set and can be can then be extracted from the estimates of FA calculated used to estimate additive variance components. The fol- for a larger group of birds for which repeated measures lowing additive genetic mixed model (Henderson, 1984) are not available. was used in experiment 1: y = Xb + Za + e, where y is an Twenty males were measured 3 times at 3 different N × 1 vector of observations for FA (n = 1,073), b is a p sessions (days). Measures were analyzed using a 2-way × 1 vector of ﬁxed sex and generation effects (p = 4), a is ANOVA (Leamy, 1984) with side (ﬁxed) and bird (ran- an N × 1 vector of additive genetic effects, e is an N × 1 dom) as main factors (1 and 19 df), their interaction (19 df), vector of residuals, and X and Z are incidence matrices. and the measurement error (80 df). Signiﬁcant variation Solutions for a and b can be obtained from the Hender- between sides indicates variation in DA, whereas a sig- son’s mixed model equations (Henderson, 1973). The in- niﬁcant interaction indicates signiﬁcant FA (in the ab- direct approach introduced by Misztal and Gianola (1987) sence of AS). The ANOVA gives separate estimates of solves the mixed model equations more efﬁciently. The the variance between sides (DA), the interaction variance REML estimator maximizes the portion of the likelihood (FA), and the error variance (measurement error). Finally, that does not depend on the ﬁxed effects. Graser et al. the product-moment correlation between FA and trait (1987) and Meyer (1989) used the derivative-free algo- size was used to determine if they were independent. If rithm to write the logarithm of the REML. The software a positive relationship was found between the mean value packages PEST (Groeneveld et al., 1990) and VCE and asymmetry of a trait, this effect would be removed (Groeneveld, 1994) were used. by dividing the absolute asymmetry score by the trait To study the effects of age and sex in experiment 2, a mean, deﬁned as the relative FA: [2|R−L|/(R+L)]. Relative 2-way ANOVA (Sokal and Rohlf, 1981) was used follow- FA for all traits had distributions that were not normal ing the model yijk = m + Ai + Sj + ASij + eijk, where yijk is and were transformed to arc-sin square root prior to anal- the FA of age i, sex j, and bird k; m is the overall mean; ysis. Mean relative asymmetry was deﬁned as the mean Ai is the effect of age (i = 1…9); Sj is the effect of sex (j = 1692 CAMPO ET AL. Table 2. Mean squares for ANOVA on ﬂuctuating asymmetry of various morphological and secondary sexual traits measured for 20 birds in 3 sessions, experiment 1 Source of Leg Wing Feather Ear-lobe Wattle variation length length length area area Sides (1) 0.4889 7.5250 69.1764 0.3694 119.3300* Birds (19) 955.8480*** 526.3054*** 717.4350*** 11.9828*** 300.2861*** Interaction (19) 5.6343 3.5272 24.8223** 1.8062*** 17.9803*** Error (80) 4.7753 3.0580 9.4509 0.1556 4.7889 *P < 0.05. **P < 0.01. ***P < 0.001. 1…2); ASij is the interaction; and eijk is the residual (k = There was a signiﬁcant positive correlation coefﬁcient 1…20). If the interaction was signiﬁcant the experiment between the value of absolute asymmetry and characteris- would be reanalyzed to study the effect of age in each tic size for ear lobe (0.32, P ≤ 0.001) and wattle areas (0.17, sex, using a 1-way ANOVA following the model yij = m P ≤ 0.001), whereas the relationship for leg, wing, and + Ai + eij, where yij is the FA of age i and bird j, m is the feather lengths was not signiﬁcant. Relative asymmetry overall mean, Ai is the effect of age (i = 1…9), and eij is could, therefore, be used to control the scaling effect in the residual (j = 1…20). Signiﬁcant differences among both traits. The relative asymmetry for the secondary ages were estimated using the Student-Newman-Keuls’ sexual traits was much higher than that for the morpho- multiple range test (Snedecor and Cochran, 1980). The logical traits. It was less than 2% of trait size for the software package SAS Institute (1998) was used. morphological traits, and about 10% of trait size for the secondary sexual traits. The coefﬁcients of variation of RESULTS each trait were 4, 4, 6, 27, and 32%, respectively. Heritability of absolute and relative FA was zero or Leg length and ear lobe area exhibited DA (t = 11.73, very low and not signiﬁcantly different from zero for all P < 0.001, and t = 5.39, P < 0.001, respectively), and the 5 traits (Table 1). The arc-sin square root transformation right side was consistently greater than the left side (Table of the relative FA yielded similar heritability estimates ± 1). Signed right minus left differences were right skewed SE (0.02 ± 0.03, 0.06 ± 0.04, 0.00 ± 0.00, 0.01 ± 0.03, and for feather length and wattle area (g1 = 4.10, P < 0.001, 0.00 ± 0.02, respectively). The heritability of (R−L), a mea- and g1 = 1.09, P < 0.001, respectively) and leptokurtic for sure of DA, was approximately equal to that of |R−L|. The wing length, feather length, and wattle area (g2 = 63.08, heritability of the mean relative asymmetry, deﬁned as P < 0.001, g2 = 3.47, P < 0.001, and g2 = 19.84, P < 0.001, respectively). Thus, there was no evidence of antisymme- the mean of the relative asymmetries of the different traits, try because bimodal (or broad-peaked) distributions tend was very low (0.01 ± 0.03). A value of heritability equal to be platykurtic. to zero was obtained using the standardized |R−L| value Mean square error for leg length, wing length, feather for each trait and averaging across traits to give a total length, ear-lobe area, and wattle area (Table 2) repre- score. The heritability of each trait was generally greater sented 84.75, 86.69, 38.07, 8.6, and 26.6% of the interaction than that of the asymmetry. Leg length was a highly mean square. Additionally, the interaction was highly heritable trait, with heritability estimate being greater signiﬁcant for feather length, ear-lobe area, and wattle than 0.5, whereas wing length had an intermediate value area. Thus, measurement error was not confounded with of heritability and the very low heritability for feather FA, and the non-DA identiﬁed in this analysis was catego- length was not signiﬁcant different from zero. Ear lobe rized as true FA. Measurement error accounted for 1.47, and wattle areas were lowly heritable traits, with herita- 1.72, 3.83, 3.79, and 4.56% of the total variation, respec- bility estimates less than 0.2 in both cases but signiﬁcantly tively. Between sides mean square was not signiﬁcant for different from zero. leg length, wing length, feather length, and ear-lobe area The phenotypic correlation between sides was less than and provided no evidence of DA. the genetic correlation (Table 3), and heritabilities of each Table 3. Genetic and phenotypic correlation estimates between right and left sides measures, and heritability ± SE of each side, for various morphological and secondary sexual traits at 36 wk of age, experiment 1 Genetic Phenotypic Right side Left side Trait correlation correlation heritability heritability Leg length 1.00 0.90 0.50 ± 0.06 0.52 ± 0.07 Wing length 0.97 0.78 0.28 ± 0.04 0.44 ± 0.05 Feather length 1.00 0.86 0.01 ± 0.02 0.02 ± 0.02 Ear-lobe area 0.99 0.85 0.15 ± 0.03 0.13 ± 0.03 Wattle area 1.00 0.86 0.12 ± 0.02 0.16 ± 0.03 HERITABILITY FOR FLUCTUATING ASYMMETRY 1693 Table 4. Heritability estimates ± SE for other bilateral asymmetry measures of various morphological and secondary sexual traits at 36 wk of age, experiment 1 Fluctuating asymmetry1 Trait |R−L|0.5 log|R−L| (R2+L2)−¹⁄₂(R+L)2 |lnR−lnL| logR−logL Leg length 0.01 ± 0.02 0.00 ± 0.02 0.01 ± 0.08 0.04 ± 0.01 0.01 ± 0.03 Wing length 0.08 ± 0.02 0.05 ± 0.04 0.05 ± 0.08 0.09 ± 0.01 0.11 ± 0.05 Feather length 0.00 ± 0.03 0.00 ± 0.00 0.00 ± 0.18 0.00 ± 0.01 0.00 ± 0.00 Ear-lobe area 0.01 ± 0.02 0.00 ± 0.00 0.01 ± 0.02 0.00 ± 0.01 0.00 ± 0.00 Wattle area 0.06 ± 0.03 0.08 ± 0.04 0.00 ± 0.14 0.00 ± 0.01 0.03 ± 0.04 1 R = right side; L = left side. side were similar for all traits. Correlations between sides was reached at 40 wk, and this value was signiﬁcantly were not signiﬁcantly different from one. greater than that at 24 wk. The effect of sex was signiﬁcant In addition to relative and absolute FA, the heritability for ear lobe area; females had greater relative FA than of 2 transformed values, |R−L|0.5 and log|R−L|, and 3 alter- males. native indexes of FA, (R2+L2)−¹⁄₂(R+L)2, |lnR−lnL|, and Age-by-sex interaction was signiﬁcant for wing length, logR−logL were also calculated. Although |R−L| has a half feather length, and wattle area. Table 7 summarizes the normal distribution, the square root and the logarithmic effect of age for females and males separately. For wing transformed values of |R−L| produced a normal distribu- length, variation associated with age was signiﬁcant in tion. The heritability of these transformed values was both sexes. In females, there were 2 maximum values at similar to that of |R−L| for all traits (Table 4). The same 20 and 40 wk, and these values were signiﬁcantly greater was true for all the 3 alternative indexes of FA. than those at 16, 24, and 32 wk. In males, the relative FA The 3 morphological traits showed high levels of phe- was at a maximum at 16 wk of age, which was signiﬁ- notypic and genetic correlations (Table 5) between their cantly greater than values at 24, 28, 32, and 36 wk. Differ- means (from 0.19 to 0.75 and from 0.64 to 0.95, respec- ence between sexes was signiﬁcant at 16 wk; males had tively), and the same was true for the secondary sexual greater relative FA than females. For feather length and traits (0.52 and 0.59, respectively). Phenotypic and genetic wattle area, variation associated with age was signiﬁcant correlations between morphological and secondary sex- only in females. A maximum value was reached at 20 wk ual traits means were generally low. All 5 traits showed for feather length, and the relative FA at this age was low phenotypic correlations between their levels of FA signiﬁcantly greater than those at 8, 12, 16, 24, 32, and 36 (from −0.06 to 0.19), the genetic correlation being high wk. For wattle area, the relative FA also was at a maxi- between the FA of the secondary sexual traits (0.99) and mum at 20 wk of age with a decrease after that age; ranging from −1 to +1 in all the remaining cases. the maximum value was signiﬁcantly different. For both Mean squares and mean values indicating age and sex traits, difference between sexes was signiﬁcant at 20 wk, effects (experiment 2) on relative FA are presented in and females had greater relative FA than males. Table 6. Ear-lobe area (100 females and 100 males sam- pled) and wattle area (120 females and 120 males sam- DISCUSSION pled) were tested from 24 and from 20 wk of age. Age- by-sex interaction was not signiﬁcant for leg length and The current study fulﬁlled the criteria to estimate the ear lobe area. The effect of age was signiﬁcant for leg narrow sense heritability of FA conﬁdently (Palmer, 1994) length. A minimum value for relative FA was reached at without the effect of confounding factors. Heritability of 24 wk. The mean value increased until a maximum value absolute FA was not signiﬁcantly different from zero for Table 5. Genetic (above diagonal) and phenotypic (below diagonal) correlation estimates between the absolute ﬂuctuating asymmetry, |R−L|, of various morphological and secondary sexual traits at 36 wk of age. Genetic and phenotypic correlation between traits in brackets; experiment 1 Leg Wing Feather Ear-lobe Wattle Trait length length length area area Leg length — 1.00 −1.00 −0.99 0.36 (0.95) (0.64) (−0.07) (0.02) Wing length 0.18 — −1.00 −0.43 0.99 (0.75) (0.74) (−0.02) (−0.07) Feather length −0.06 −0.02 — −1.00 −1.00 (0.21) (0.19) (0.43) (0.07) Ear-lobe area 0.03 −0.01 −0.01 — 0.99 (0.12) (0.13) (0.15) (0.59) Wattle area 0.00 −0.04 0.02 0.19 — (0.16) (0.15) (0.32) (0.52) 1694 CAMPO ET AL. Table 6. Mean squares and mean values indicating age and sex effects on ﬂuctuating asymmetry of various morphological and secondary sexual traits, experiment 2 Source of Leg Wing Feather Ear-lobe Wattle variation1 length length length area area Mean squares Age (8) 0.0038** 0.0069*** 0.1028** 0.0486 0.0702** Sex (1) 0.0012 0.0001 0.0152 0.2518*** 0.0846** Interaction (8) 0.0012 0.0052** 0.0091** 0.0460 0.1042*** Error (342) 0.0015 0.0021 0.0036 0.0244 0.0193 Mean values Leg length Ear-lobe area Females Males Age 8 1.03ab 13.39x 8.61Y 12 1.12ab 16 1.19ab 20 0.89ab 24 0.72b 28 0.93ab 32 0.78ab 36 1.00ab 40 1.34a a,b Means within a trait and column with no common superscript differ signiﬁcantly (P < 0.05). x,y Means within a trait and row with no common superscript differ signiﬁcantly (P < 0.05). 1 Degrees of freedom in brackets for leg, wing, and feather lengths. Degrees of freedom for ear-lobe area were 4, 1, 4, and 190 and for wattle area were 5, 1, 5, and 228, respectively. **P < 0.01. ***P < 0.001. all 5 traits, and similar estimates were obtained for relative not small (Leung, 1998). Transformation of FA or, alterna- FA, DA, transformed absolute FA (square root or log), tively, nonparametric statistics are usually unnecessary transformed relative FA (arc-sin square root), and 3 alter- (Gangestad and Thornhill, 1998) when sample size is large native indexes of FA, with all conclusions remaining the and untransformed or parametric tests are preferred. The same. Use of relative FA could be unnecessary because alternative index log R−log L should be recommended when the FA-size relation was signiﬁcant for ear-lobe and here because there is leptokurtosis (Graham et al., 2003). wattle areas, the coefﬁcient of variation of the trait was The heritability of the combined absolute or relative FA Table 7. Mean squares and mean values indicating age effect for each sex on ﬂuctuating asymmetry of wing length, feather length, and wattle area, experiment 2 Wing length Feather length Wattle area Source of variation1 Females Males Females Males Females Males Mean squares Age (8) 0.0054** 0.0067** 0.0116*** 0.0078 0.1551*** 0.0157 Error (171) 0.0018 0.0024 0.0030 0.0042 0.0194 0.0176 Mean values Age 8 1.60abx 2.06abx 1.52bx 1.53ax 12 1.44abx 1.34abx 0.99bx 1.04ax 16 1.11by 2.32ax 1.62bx 2.12ax 20 2.04ax 1.56abx 3.12ax 1.30ay 21.94ax 6.39ay 24 0.96bx 1.07bx 1.86bx 1.05ax 8.21by 8.22ay 28 1.33abx 1.16bx 2.10abx 1.77ax 6.39by 5.07ay 32 1.07bx 1.25bx 1.83bx 1.86ax 5.85by 9.72ay 36 1.55abx 1.16bx 1.42bx 2.13ax 8.35by 8.11ay 40 2.09ax 1.65abx 2.30abx 2.28ax 8.75by 8.73ay a,b Means within a trait and column with no common superscript differ signiﬁcantly (P < 0.05). x,y Means within a trait and row with no common superscript differ signiﬁcantly (P < 0.05). 1 Degrees of freedom in brackets for wing and feather lengths. Degrees of freedom for wattle area were 5 and 114. **P < 0.01. ***P < 0.001. HERITABILITY FOR FLUCTUATING ASYMMETRY 1695 was still low, as were those for the single traits. This very is equal to h2(1−rA12)/(1−rP12), where rA12 and rP12 are the low heritability estimate indicates that FA for all the 5 genetic and phenotypic correlation between the 2 sides, traits is determined solely by environmental sources of and h2 is the heritability of each side, which is assumed variation and that FA estimates will not be confounded to be the same, and was 0.00, 0.05, 0.00, 0.01, and 0.00, by appreciable additive genetic contributions. Low herita- respectively. The heritability of each side was similar, bility of FA agrees with the high heterosis found by indicating that there was not very strong DA. The pheno- Campo et al. (2000) in the FA of leg length, wing length, typic correlation (rP12) was less than the genetic correla- and feather length. Similarly, Yang et al. (1999) found tion (rA12), and thus the heritability of (R−L) was less than signiﬁcant heterosis for shank weight relative asymmetry the heritability of the trait. Genetic correlation between in 2 crosses of broiler breeders. sides, which is a measure of the effects of pleiotropy Although the heritability of FA has been the subject of or linkage disequilibrium, was not signiﬁcantly different considerable controversy, the most reliable studies from one, indicating that differences between sides were showed very low values in agreement with those in the purely environmental in origin. The lack of a negative current study. Moller and Thornhill (1997a) found a mean correlation between sides agrees with the no evidence of heritability of 0.27, using a meta-analysis of published AS showed by all the traits, because a negative correlation information, even though this analysis was criticized for is a necessary and nearly sufﬁcient cause of AS (van Va- the inclusion of data from studies that did not properly len, 1962). The high heritabilities for leg and wing lengths estimate heritability of FA because of their lack of control are in agreement with other literature on birds in general of potential confounding effects. After excluding studies (Boag and van Noordwijk, 1987) and with the low hetero- that they found to have confounding effects, Whitlock sis found by Campo et al. (2000), whereas the low herita- and Fowler (1997) concluded that the heritability of FA bility for feather length disagrees with the low heterosis ranged from 0.00 to 0.08, and the reevaluation of the found by Campo et al. (2000). original meta-analysis of Moller and Thornhill (1997a) The dominance effect using mixed model methodology yielded recalculated mean heritability of 0.16 (Moller and and REML procedure was predicted (Mrode, 1996), domi- Thornhill, 1997b). The heritability of FA found in the nance variance being generally very low both in bilateral literature for species representing different types of poi- asymmetries and traits. Percentage dominance variances kilotherms and homeotherms organisms generally ranges were only apparent for (R−L) of wattle area (0.03), for from 0.0 to 0.1. A total of 17 species have been used, and |R−L| of leg length and wattle area (0.13 and 0.09), and among them only one was a bird (swallow). For the barn for (R2+L2)−¹⁄₂ (R+L)2 of leg, wing, and feather lengths swallow (Hirundo rustica), Moller (1994) found no signiﬁ- (0.15, 0.17, and 0.02, respectively). Morphological traits cant heritability for tarsus length, wing length, and central typically show low levels of dominance variance (Crnok- tail length, whereas FA of outer tail length had a statisti- rak and Roff, 1995), and hence it can be predicted that cally signiﬁcant heritability, although this estimate seems low levels will be characteristic of FA too. Therefore, FA to be unrealistic (0.80 by father-son resemblance and 1.88 was almost entirely environmental in origin and had nei- by mother-daughter resemblance). Cadee (2000) esti- ther signiﬁcant additive nor dominance genetic compo- mated the heritability of FA for tarsus, tail, and wing nents. Because FA is considered to be the result of micro- sizes by full-sib analysis in the barn swallow under favor- environmental effects during development there could able and unfavorable environments. There was no sig- be a maternal effect that would be detected most easily niﬁcant heritability for tail and wing size, whereas tarsus by difference between inbred lines. asymmetry was signiﬁcantly heritable (0.53 ± 0.26) in the In spite of the fact that the traits could show high levels year with benign weather conditions with heritability be- of phenotypic and genetic correlation in their means, es- ing zero under unfavorable conditions. pecially morphological traits between them and second- The low levels of heritability observed for FA reﬂect ary sexual traits between them, different traits rarely low levels of genetic control over developmental stability. showed much correlation in their level of FA. The lack The heritability of developmental stability is the heritabil- of a strong correlation between the FA of different traits ity of FA divided by the hypothetical repeatability; this indicated that the level of FA in all traits did not reﬂect is (Whitlock, 1998): 0.6366 – 0.3634/(CV|R−L|)2, where CV|R− equally the quality of birds. This result agrees with that of L| is the coefﬁcient of variation of |R−L|. The hypothetical Kellner and Alford (2003), who analyzed several bilateral repeatabilities for the 5 analyzed traits were 0.29, 0.39, traits (tarsometatarsus length, tarsometatarsus width, and 0.27, 0.21, and 0.42, respectively. The heritabilities of de- naris-mandibles distance) in domestic fowl. In their velopmental stability, therefore, were 0.02, 0.03, 0.00, 0.01, study, FA levels were not highly correlated, coefﬁcients and 0.07, respectively, and heritability estimates of FA of correlation ranging from a minimum of 0.07 to a maxi- did not underestimate the heritability of developmental mum of 0.19, showing that asymmetry levels on the differ- stability. The coefﬁcient of variation of |R−L| was very ent traits varied almost completely independently. high, indicating that (R−L) had a leptokurtic distribution Although signiﬁcant variation with age for FA was (van Dongen, 1998b); it was 103, 121, 100, 93, and observed in leg length, wing length, feather length (fe- 131%, respectively. males), and wattle area (females), the only signiﬁcant The estimated heritability of (R−L) agreed very well differences were between near the adult age and near the with its expected value. The expected heritability of (R−L) onset of sexual maturity. No age effect was observed for 1696 CAMPO ET AL. FA in ear-lobe area. There were 3 different patterns of Clarke, G. M., G. W. Brand, and M. J. Whitten. 1986. Fluctuating asymmetry, suggesting different mechanisms by which asymmetry: A technique for measuring developmental stress caused by inbreeding. Aust. J. Biol. Sci. 39:145–153. FA arises during growth (Swaddle and Witter, 1997). Corruccini, R. S., and R. H. Y. Potter. 1981. Developmental corre- First, relative levels of asymmetry for leg length decreased lates of crown component asymmetry and occlusal discrep- early and then increased throughout most of time, ancy. Am. J. Phys. Anthropol. 55:21–31. whereas for wing length in males increased early and Coyne, J. A. 1987. Lack of response to selection for directional asymmetry in Drosophila melanogaster. J. Hered. 78:119. then decreased throughout most of time. The pattern ob- Crnokrak, P., and D. A. Roff. 1995. Dominance variance: associa- served for leg length (metatarsus) disagrees with that tions with selection and ﬁtness. Heredity 75:530–540. indicated by Kellner and Alford (2003) for the tarsometa- Eggert, A. K., and S. S. K. Sakaluk. 1994. Fluctuating asymmetry tarsus length in Lohman Brown pullets until 6 wk of and variation in the size of courtship food gifts in decorated age; levels of asymmetry in that paper increased early in crickets. Am. Nat. 144:708–716. Gangestad, S. W., and R. Thornhill. 1998. The analysis of ﬂuctu- growth and then remained constant. Second, levels of ating asymmetry redux: The robustness of parametric statis- asymmetry for wing length and feather length in females tics. Anim. Behav. 55:497–501. followed a random trend through most of time. This pat- Gangestad, S. W., and R. Thornhill. 1999. Individual differences tern disagrees with that observed by Swaddle and Witter in developmental precision and ﬂuctuating asymmetry: a (1997) who examined the FA in the primary feathers of model and its implications. J. Evol. Biol. 12:402–416. Graham, J. H., J. M. Emlen, D. C. Freeman, L. J. Leamy, and J. starlings, observing that both absolute and relative asym- A. Kieser. 1998. Directional asymmetry and the measurement metry decreased as the feathers develop. Finally, de- of developmental instability. Biol. J. Linnean Soc. 64:1–16. creases of relative asymmetry for wattle area in females Graham, J. H., K. Shimizu, J. M. Emlen, D. C. Freeman, and J. were observed early and persisted over time. Merkel. 2003. Growth models and the expected distribution Females and males showed similar degrees of FA in of ﬂuctuating asymmetry. Biol. J. Linn. Soc. 80:57–65. Graser, H. U., S. P. Smith, and B. Tier. 1987. A derivative-free leg length, whereas the 2 sexes differed for their degree approach for estimating variance components in animal of FA in ear-lobe area; females had a signiﬁcantly greater models by restricted maximum likelihood. J. Anim. Sci. degree of FA than males. Their FA in wing length, feather 64:1362–1370. length, and wattle area were also similar, although near Groeneveld, E. 1994. VCE, a multivariate multimodel REML (co)variance component estimation package (22). Pages 47– the onset of sexual maturity (at 16 or 20 wk of age) there 48 in: Proceedings of the 5th World Congress on Genetics was a signiﬁcant difference. Males had a signiﬁcantly Applied to Livestock Production, Guelph, Ontario, Canada. greater FA than females in wing length at 16 wk of age, Groeneveld, E., M. Kovac, and T. Wang. 1990. PEST, a general and the opposite was true for feather length and wattle purpose BLUP package for multivariate prediction and esti- area at 20 wk of age. In agreement with these results, mation (13). Pages 488–491 in Proc. 4th World Congr. Genet. Appl. Livest. Prod., Edinburgh, Scotland. Moller (1994) found that tarsus length, wing length, and Henderson, C. R. 1973. Sire evaluation and genetic trends. Pages central tail length showed a similar degree of FA in the 10–41 in Proc. Anim. Breeding Genet. Symp. Honor of J. L. 2 sexes of the barn swallow, whereas males differed from Lush. Am. Soc. Anim. Sci., Savoy, IL. females with respect to FA in outer tail length before Henderson, C. R. 1984. Applications of Linear Models in Animal sexual maturity. Breeding. Univ. Guelph, Guelph, Ontario, Canada. Hermida, M., C. Fernandez, R. Amaro, and E. San Miguel. 2002. Heritability and evolvability of meristic traits in a natural REFERENCES population of Gasterosteus aculeatus. Can. J. Zool. 80:532–541. Jones, J. S. 1987. An asymmetrical view of ﬁtness. Nature Boag, P. T., and A. J. van Noordwijk. 1987. Quantitative genetics. 325:298–299. Pages 45–78 in Avian Genetics. F. Cooke, and P. A. Buckley, Kellner, J. R., and R. A. Alford. 2003. The ontogeny of ﬂuctuating ed. Acad. Press, London. asymmetry. Am. Nat. 161:931–947. Bryden, C. A., and D. D. Heath. 2000. Heritability of ﬂuctuating Kruuk, L. E. B., J. Slate, J. M. Pemberton, and T. H. Clutton- asymmetry for multiple traits in Chinook salmon (Oncorhyn- Brock. 2003. Fluctuating asymmetry in a secondary sexual trait: no associations with individual ﬁtness, environmental chus tshawytscha). Can. J. Fish. Aquat. Sci. 57:2186–2192. stress or inbreeding, and no heritability. J. Evol. Biol. Cadee, N. 2000. Genetic and environmental effects on morphol- 16:101–113. ogy and ﬂuctuating asymmetry in nestling barn swallows. Leamy, L. 1984. Morphometric studies in inbred and hybrid J. Evol. Biol. 13:359–370. house mice. V. Directional and ﬂuctuating asymmetry. Am. Campo, J. L. 1991. Use of the sex-linked barring (B) gene for chick Nat. 123:579–593. sexing on an eumelanotic columbian background. Poult. Sci. Leamy, L. 1986. Directional selection and developmental stabil- 70:1469–1473. ity: Evidence from ﬂuctuating asymmetry of dental charac- ˜ Campo, J. L., M. G. Gil, I. Munoz, and M. Alonso. 2000. Relation- ters in mice. Heredity 57:381–388. ships between bilateral asymmetry and tonic immobility re- Leamy, L. 1997. Genetic analysis of ﬂuctuating asymmetry for action or heterophil to lymphocyte ratio in ﬁve breeds of skeletal characters in mice. J. Hered. 88:85–92. chickens. Poult. Sci. 79:453–459. Leamy, L. 1999. Heritability of directional and ﬂuctuating asym- Campo, J. L., and F. Orozco. 1986. Genetic basis of the Melanotic metry for mandibular characters in random-bred mice. J. Prat phenotype. Br. Poult. Sci. 27:361–367. Evol. Biol. 12:146–155. Chapman, J. W., and D. Goulson. 2000. Environmental versus Leary, R. F., and F. W. Allendorf. 1989. Fluctuating asymmetry as genetic inﬂuences on ﬂuctuating asymmetry in the house ﬂy, an indicator of stress. Implications for conservation biology. Musca domestica. Biol. J. Linn. Soc. 70:403–413. Trends Ecol. Evol. 4:214–217. Clarke, G. M. 1995. Relationships between developmental stabil- Leary, R. F., F. W. Allendorf, and R. L. Knudson. 1985. Inheri- ity and ﬁtness. Application for conservation biology. Con- tance of meristic variation and the evolution of develop- serv. Biol. 9:18–24. mental stability in rainbow trout. Evolution 39:308–314. HERITABILITY FOR FLUCTUATING ASYMMETRY 1697 Leary, R. F., F. W. Allendorf, and R. L. Knudson. 1992. Genetic, Polak, M., and R. Trivers. 1994. The science of symmetry in environmental, and developmental causes of meristic varia- biology. Trends Ecol. Evol. 9:122–124. tion in rainbow trout. Acta. Zool. Fenn. 191:79–95. Reeve, E. C. R. 1960. Some genetic tests on asymmetry of stern- Leung, B. 1998. Correcting for allometry in studies of ﬂuctuating opleural chaeta number in Drosophila. Genet. Res. 1:151–172. asymmetry and quality within samples. Proc. R. Soc. Lond. Roff, D., and D. Reale. 2004. The quantitative genetics of ﬂuctu- Biol. Sci. 265:1623–1629. ating asymmetry: a comparison of two models. Evolution Leung, B., M. R. Forbes, and D. Houle. 2000. Fluctuating asym- 58:47–58. metry as a bioindicator of stress: comparing efﬁcacy of analy- SAS Institute. 1998. User’s Guide. Version 6.12. SAS Institute ses involving multiple traits. Am. Nat. 155:101–115. Inc., Cary, NC. Merila, J., and M. Bjorklund. 1995. Fluctuating asymmetry and Scheiner, S. M., R. L. Caplan, and R. F. Lyman. 1991. The genetics measurement error. Syst. Biol. 44:97–101. of phenotypic plasticity. III. Genetic correlations and ﬂuctu- Meyer, K. 1989. Restricted maximum likelihood to estimate vari- ating asymmetries. J. Evol. Biol. 4:51–68. ance components for animal models with several random Snedecor, G. W., and W. G. Cochran. 1980. Statistical Methods. effects using a drivative-free algorithm. Genet. Sel. Evol. 7th ed. Iowa State Univ. Press, Ames, IA. 21:317–340. Sokal, R. R., and F. J. Rohlf. 1981. Biometry. Freeman and Misztal, I., and D. Gianola. 1987. Indirect solution of mixed- Co., London. model equations. J. Dairy Sci. 70:716–723. Swaddle, J. P., and M. S. Witter. 1997. On the ontogeny of Moller, A. P. 1994. Sexual selection in the barn swallow (Hirundo developmental stability in a stabilized trait. Proc. R. Soc. rustica). IV. Patterns of ﬂuctuating asymmetry and selection Lond. Biol. Sci. 264:329–334. against asymmetry. Evolution 48:658–670. Swaddle, J. P., M. S. Witter, and I. C. Cuthill. 1994. The analysis Moller, A. P. 1997. Developmental stability and ﬁtness: a review. of ﬂuctuating asymmetry. Anim. Behav. 48:986–989. Am. Nat. 149:916–932. Tuinstra, E. J., G. de Jong, and W. Scharloo. 1990. Lack of re- Moller, A. P. 1999. Developmental stability is related to ﬁtness. sponse to family selection for directional asymmetry in Dro- Am. Nat. 153:556–560. sophila melanogaster: Left and right sides are not distinguished Moller, A. P., G. S. Sanotra, and K. S. Vestergaard. 1995. Develop- in development. Proc. R. Soc. Lond. Biol. Sci. 241:146–152. mental stability in relation to population density and breed Tuyttens, F. A. M. 2003. Measures of developmental instability of chickens Gallus gallus. Poult. Sci. 74:1761–1771. as integrated, a posteriori indicators of farm animal welfare: Moller, A. P., and R. Thornhill. 1997a. A meta-analysis of the A review. Anim. Welf. 12:535–540. heritability of developmental stability. J. Evol. Biol. 10:1–16. van Dongen, S. 1998a. How repeatable is the estimation of devel- Moller, A. P., and R. Thornhill. 1997b. Developmental stability opmental stability by ﬂuctuating asymmetry? Proc R. Soc. is heritable. J. Evol. Biol. 10:69–76. Lond. Biol. Sci. 265:1423–1427. Mrode, R. A. 1996. Linear Models for the Prediction of Animal van Dongen, S. 1998b. The distribution of individual ﬂuctuating Breeding Values. Commonwealth Agric. Bureaux Int., Wall- asymmetry: Why are the coefﬁcient of variation of the un- ingford, UK. signed FA so high? Ann. Zool. Fenn. 35:79–85. Nachman, G., and K. E. Heller. 1999. Fluctuating asymmetry as van Dongen, S. 2000. Unbiased estimation of individual asym- an index of ﬁtness: Causality or statistical artifact? Oikos metry. J. Evol. Biol. 13:107–112. 86:357–365. van Dongen, S., G. Molenberghs, and E. Matthysen. 1999a. The Palmer, A. R. 1994. Fluctuating asymmetry analyses: A primer. statistical analysis of ﬂuctuating asymmetry: REML estima- Pages 335–364 in Developmental Instabilty: Its Origins and tion of a mixed regression model. J. Evol. Biol. 12:94–102. Evolutionary Implications. T. A. Markow, ed. Kluwer Acad., van Dongen, S., E. Sprengers, C. Lofstedt, and E. Matthyssen. Dordrecht, The Netherlands. 1999b. Heritability of tibia ﬂuctuating asymmetry and devel- Palmer, A. R., and C. Strobeck. 1986. Fluctuating asymmetry: opmental instability in the winter moth (Operophtera bru- Measurement, analysis, patterns. Annu. Rev. Ecol. Syst. mata). Heredity 82:535–542. 17:391–421. van Valen, L. 1962. A study of ﬂuctuating asymmetry. Evolution Palmer, A. R., and C. Strobeck. 1992. Fluctuating asymmetry as 16:125–142. a measure of developmental stability: implications of non- Whitlock, M. 1996. The heritability of ﬂuctuating asymmetry normal distributions and power of statistical tests. Acta Zool. and the genetic control of developmental stability. Proc. R. Fenn. 191:57–72. Soc. Lond. Biol. Sci. 263:849–853. Palmer, A. R., C. Strobeck, and A. K. Chippindale. 1994. Bilateral Whitlock, M. 1998. The repeatability of ﬂuctuating asymmetry: variation and the evolutionary origin of macroscopic asym- A revision and extension. Proc. R. Soc. Lond. Biol. Sci. metries. Pages 203–220 in Developmental Instabilty: Its Ori- 265:1429–1431. gins and Evolutionary Implications. T. A. Markow, ed. Whitlock, M., and K. Fowler. 1997. The instability of studies of Kluwer Acad., Dordrecht, The Netherlands. instability. J. Evol. Biol. 10:63–67. Parsons, P. A. 1990. Fluctuating asymmetry. An epigenetic mea- Windig, J. J. 1998. Evolutionary genetics of ﬂuctuating asymme- sure of stress. Biol. Rev. Camb. Philos. Soc. 65:131–145. try in the peacock butterﬂy, Inachis io. Heredity 80:382–392. Parsons, P. A. 1992. Fluctuating asymmetry. A biological moni- Windig, J. J., and S. Nylin. 2000. How to compare ﬂuctuating tor of environmental and genomic stress. Heredity 68:361– asymmetry of different traits. J. Evol. Biol. 13:29–37. 364. Windig, J. J., and S. Nylin. 2002. Genetics of ﬂuctuating asymme- Patterson, H. D., and R. Thompson. 1971. Recovery of inter- try in pupal traits of the speckled wood butterﬂy, Pararge block information when block sizes are unequal. Biometrika aegeria. Heredity 89:225–234. 58:545–554. Woods, R. E., M. J. Hercus, and A. A. Hoffmann. 1998. Estimat- Polak, M., and W. T. Starmer. 2001. The quantitative genetics ing the heritability of ﬂuctuating asymmetry in ﬁeld Drosoph- of ﬂuctuating asymmetry. Evolution 55:498–511. ila. Evolution 52:816–824. Polak, M., and E. M. Stillabower. 2004. The relationship between Yang, A., D. A. Emmerson, E. A. Dunnington, and P. B. Siegel. genotype, developmental stability and mating performance: 1999. Heterosis and developmental stability of body and Disentangling the epigenetic causes. Proc. R. Soc. Lond. Biol. organ weights at hatch for parental line broiler breeders and Sci. 271:1815–1821. speciﬁc crosses among them. Poult. Sci. 78:942–948.