Can evolution of sexual dimorphism be triggered by developmental temperatures?

Authors


Tarmo Ketola, Centre of Excellence in Evolutionary Research, Department of Biological and Environmental Science, University of Jyväskylä, P. O. Box 35, FI-40014, Finland.
Tel.: +358 40 721 4852; fax: +358 14 2602321; e-mail: tarmo.t.ketola@jyu.fi

Abstract

Genetic prerequisites for the evolution of sexual dimorphism, sex-specific heritabilities and low or negative genetic correlations between homologous traits in males and females are rarely found. However, sexual dimorphism is evolving rapidly following environmental change, suggesting that sexual dimorphism and its genetic background could be environmentally sensitive. Yet few studies have explored the sensitivity of the genetic background of sexual dimorphism on environmental variation. In this study, on Drosophila melanogaster, we used a large nested full-sib–half-sib breeding design where families were split into four different developmental temperatures: two constant temperature treatments of 25 and 30 °C and two cycling temperatures with means of 25 and 30 °C, respectively. After emergence, we tested heat shock tolerance of adult flies. We found that sexual dimorphism was strongly affected by temperature during development. Moreover, we found that female heritability was significantly lower in flies developing at hot temperature and more so under hot and cycling temperatures. Interestingly, most of the genetic variation for heat shock tolerance was orthogonal (i.e. noncorrelated) between sexes, allowing independent evolution of heat shock tolerance in males and females. These findings give support to the hypothesis that the evolution of sexual dimorphism can be influenced by the environments experienced during development.

Introduction

Males and females may have different fitness optima, and opposing selection on homologous traits in each sex is expected to lead to sexual dimorphism. The evolution of sexual dimorphism may occur if heritabilities are sex specific and genetic correlations are lower than one. The lower the genetic correlation, the easier it is to develop sexual dimorphism (Lande, 1980; Rice, 1984; Cowley et al., 1986; Lynch & Walsh, 1998). Empirical data rarely reveal evidence that the above assumptions are met (Roff, 1997; Poissant et al., 2009). Regardless, sexual dimorphism appears to evolve rapidly in response to environmental changes (Badyaev, 2002), suggesting that experienced environments influence expression of sexual dimorphism and thereby its potential to evolve. Understanding the mechanisms that lead to sex differences and sexual conflict, where traits and genes favoured by one sex are costly to the other, is important for resolving how genetic variation in fitness is maintained and for explaining patterns of sexual selection; for example, why females do not seem to get any benefits from males for compensating their lower resource allocation on reproduction (Chapman et al., 2003; Delcourt et al., 2009; Mokkonen et al., 2011).

Females and males are considered to maximize their fitness differently, and even homologous traits in males and females can be under antagonistic selection (i.e. intralocus sexual conflict). However, males and females should share a common gene pool causing conflicting selection pressures for alleles expressed in both males and females (Fairbairn et al., 2007). However, this hereditary dilemma might be overcome if trait expression evolves sex specific. Although sexual dimorphism is often observed, evidence from natural populations is somewhat perplexing as sex-specific heritabilities appear not to be common and large positive genetic correlations are frequently found between homologous traits in males and females (Roff, 1997; Poissant et al., 2009). What makes these findings paradoxical is the fact that in several cases where there seem to be no genetic potential for the evolution of sexual dimorphism, sexual dimorphism changes in response to the environment (Badyaev, 2002). To resolve this paradox, Badyaev (2002) suggested that sexual dimorphism and genetic determinants of sexual dimorphism must be environmentally sensitive during development and ontogeny. While sexual dimorphism is often found to be sensitive to environments, little is known about environmental sensitivity of the genetic background of sexual dimorphism (LeGalliard et al., 2005; Poissant et al., 2009). Given that heritabilities and genetic correlations, in general, have been found to be environmentally sensitive, it is likely that the evolution of sexual dimorphism is also environment dependent (Hoffmann & Merilä, 1999; Sgró & Hoffmann, 2004; Charmantier & Garant, 2005). For example, unfavourable environmental conditions may impact levels of genetic variation through the expression of stress sensitive mutations and changes in environmental variance (Rutherford & Lindquist, 1998; Bubliy and Loeschcke 2001). The full genetic potential of individuals may also be limited in some environments resulting in changes in heritabilities and genetic correlations across environments (Hoffmann & Merilä 1999; Sgró & Hoffmann, 2004).

In Drosophila, temperature acts as a strong selective force particularly during development when larvae and pupae are exposed to daily temperature fluctuations in their natural habitat (Hoffmann et al., 2003). Although differences in mean temperatures are of great importance, experiments examining traits under constant temperatures may not have direct relevance to natural highly variable conditions. Studies on the adult heat shock response in D. melanogaster have found that different mechanisms underlie adaptation to variable environments and stable environments further highlighting the need to examine trait responses also in variable temperature environments (Krebs & Loeschcke, 1995a,b; Sørensen et al., 2001, 2005; Sarup et al., 2006, Bozinovic et al. 2011). Even fewer studies have explored the sensitivity of sexual dimorphism of fitness-related traits to environmental manipulation, and in most of the cases traits cannot be explicitly linked to the selective environments or to fitness (Poissant et al., 2009, but see: Delcourt et al. 2009). Moreover, adaptation to fluctuating environments will influence the evolution of sexual dimorphism, as cross-sex genetic correlations and sexual dimorphism could vary unpredictably when environments vary (Poissant et al., 2009). This is partly due to genotype by environment (G by E) interactions that are common for most traits (DeWitt & Scheiner, 2004).

Here, we experimentally determined how differences in developmental temperatures influence sexual dimorphism in adult heat shock tolerance, sex-specific heritability of heat shock tolerance, and cross-sex correlations of heat shock tolerance in D. melanogaster. We utilize a nested full-sib–half-sib breeding design where families were split into four different developmental temperatures: two constant temperature treatments of 25 and 30 °C and two cycling temperatures with means of 25 and 30 °C, respectively. Drosophila species are suitable for such studies as thermal resistance differs in males and females (Krebs & Loeschcke, 1994; Andersen et al., 2010) suggesting sex-specific genetic architecture of thermal tolerance (Morgan & Mackay, 2006; Norry et al., 2007). We hypothesize that if the thermal environment during development contributes to the evolution of sexual dimorphism, we will find that developmental temperature impacts on sexual dimorphism, sex-specific heritable variation and/or genetic correlations between sexes (Lyons et al., 1994; Badyaev, 2002).

Materials and methods

Origin and maintenance of flies

The mass-bred D. melanogaster population used in the experiment was collected in Southern Tasmania, Australia (43.15°S, 147°E). Offspring from twenty field inseminated females were combined to establish the population. Prior to the experiment, flies were reared for 20 generations on an oat-sugar-yeast-agar medium and maintained at a minimum population size of 1000 individuals under a 12 : 12 h light/dark cycle at 25 °C. Larval density of the parental generation (generating parents used in the experiment) was controlled by collecting 30 eggs into each of 100 vials, after which vials were incubated at 25 °C until emergence. Virgin flies were sexed and transferred to vials in groups of ca. 30.

Quantitative genetic experiment

A standard nested full-sib–half-sib design was implemented (Lynch & Walsh, 1998). To begin the experiment, ca. 150 vials with one sire and five dams in each were collected at 3–4 days of age. Flies were allowed to mate for 30 h, after which females were transferred into separate vials with medium filled spoons, for egg laying. Where possible, we collected eggs from each female into 12 vials with ten eggs per vial. Eggs were collected over a 3-day period with vials randomized across four temperature treatments. The experimental vials were, on each collection day, transferred to climate chambers at 8.30 p.m. Vials were labelled such that day of egg collection could be distinguished.

Treatments were as follows: constant 25 and 30 °C and cycling 25 and 30 °C. In the fluctuating temperature regimes, mean temperature in the cabinets was equal to the mean temperatures in the cabinets with constant temperatures. The temperature cycles for the cabinets with fluctuations were 18 h of lower temperatures followed by 6 h of hotter temperatures; 23 and 31 °C for fluctuating 25, and 28 and 36 °C for fluctuating 30 °C. The ‘day temperatures’ were initiated at 2 p.m. and ended at 8 p.m. (CET). The temperature time series were chosen so that within the extreme treatments, a clearly lower egg-to-adult viability was found, but not so stressful that all offspring died. Moreover, although temperature changes were abrupt, the realization of temperature changes in vials is slower as ca. 1 h is required for temperatures to converge to the desired temperatures inside the vials. Using this method, we did not wish to test or mimic any particular environment, but to test whether different thermal environments have different developmental outcomes.

Three days after egg collection, paper was added to the medium to allow larvae to crawl upon and pupate. During the experiment, the location of the trays and racks in the chambers were randomized daily. Following 3 days from the mass hatching event, individuals were transferred to empty vials with food spoons (as a source of moisture and nutrition). All individuals were then reared at 25 °C for 40–44 h. This time period at 25 °C was chosen to reduce the short-term, immediate, physiological and nutritional responses of rearing treatments on heat shock tolerance. Sex ratio of emerged flies did not differ from equal sex ratio at any of the temperatures (see Table 2, statistics not shown).

Heat shock tolerance was assessed by placing vials with flies from the four temperature treatments into a water bath for 1 h at 38.3 °C. The water bath temperature was set so that we obtained live and comatose individuals from all environments and both sexes (tested in preliminary experiments). Following the heat shock, flies were allowed to recover at 25 °C for 2 h where after individuals were counted, sexed and classified, according to their ability to stand on their legs. An individual that was unable to stand was considered heat intolerant whereas standing individuals were considered heat tolerant. To account for the potential effects of different water baths on the heat shock tolerance of individuals, the identity of water bath was recorded and included as a factor in the statistical analysis.

Data analysis

The impact of developmental temperature, sex and their interaction on heat shock tolerance were estimated with spss (v. 18; SPSS Inc., Chicago, IL, USA) using generalized linear mixed models with logit link. Day of egg collection and water bath identity were included as fixed factors. A sexual dimorphism index (SDI) was calculated as in Poissant et al. (2009). Quantitative genetic analyses (variance component and heritability estimation, and factor analytic modelling) were performed using ASReml 3.0 (VSN International Ltd, Hemel Hempstead, UK). We studied heat shock tolerance of males and females developed at all four temperature regimes separately with univariate models including sire, dam nested within sire and vial as random factors. Additionally, we included effects of 3 different days of egg collection and water bath identity as fixed effects. As our data was in binary format, we used bin function with logit link in all analyses. Estimation at the underlying (liability) scale assumes a residual variation of π2/3 for the logit link, and thus residual variation is practically fixed in all analyses. Therefore, this method does not allow residual variation to be different in different environments, preventing testing if the heritability changes could be due to environmentally sensitive residual variation (Hoffmann & Merilä, 1999). Heritabilities were calculated as in Falconer & Mackay (1996) and tested by z-tests (estimate/standard error), as likelihood ratio tests are not recommended for comparing complex models, especially if traits are binary (Gilmour et al., 2002). Tests were one tailed as theoretically variances can only be positive.

Next, we utilized factor analytic modelling in resolving the genetic correlation structure between heat shock tolerance in both sexes and at all developmental temperatures:

image(1)

where inline image is the reduced rank additive genetic covariance matrix of traits, Γ is the matrix of factor loadings and Ψ is a diagonal matrix of variances specific to developmental environments and sexes (Meyer, 2009). This method allows estimation of multivariate axes of genetic variation directly from the data, after which genetic correlations can be retrieved from the estimated matrix of factor loadings Γ. In addition to general genetic variation that is described by loadings of environments on eigenvectors, one can allow also existence of environment-specific genetic variation Ψ (Meyer, 2009). As likelihood comparisons of models using binary analysis can yield misleading results (Gilmour et al., 2002; Wilson et al., 2011), we compared the models with more traditional methods by observing the regular coefficient of determination adjusted by the number of parameters fitted (R2adj). This statistic was calculated from the residuals and predicted values according to Zar (1999; equation: 20.23, p: 423). Differences between the two best models were minute (see below), and the best model (R2adj = 0.216) consisted of two factors with specific genetic variation for sire, as well as one factor for dam and simple effects of vial over all environments. However, due to problems with respect to convergence of some of the parameters of this model, we chose the next best model (R2adj = 0.215), which consisted of two factors for sire and simple effects of dam and vial. Regardless of model choice, the sire effects remained the same; in both cases, genetic effects of sexes were associated with different factors. If sire was fitted as a simple effect across environments, in addition to dam and vial effects, the coefficient of determination was clearly lower (R2adj = 0.191). More complex models failed to converge altogether.

Significances of factor loadings were assessed by comparing diagonals of the inverse of the average information matrix to the estimated loadings as in Cudeck & O’Dell (1994). Approximation of sampling variances for genetic correlations was performed by utilizing the method outlined by Krinsky & Robb (1990; see also Hole, 2007). Shortly, we created 10 000 random factors, drawn at random from a multivariate normal distribution, with means of factor loadings and covariation structure of factor loadings determined by the inverse of the average information matrix (.vvp – file in ASReml). G-matrix (based on randomized matrix of factor loadings ΓΓ’) and genetic correlations based on G-matrix (e.g. Falconer & Mackay 1996) were resolved for each matrix of randomized factor loadings (eqn 1). The significances of correlations were determined by calculating the proportion of draws where the correlation overlapped zero. Similarly, differences between pair-wise correlations were determined by observing whether correlation estimates in draws overlapped. Because testing was two tailed, P < 0.05 was determined if not more than 2.5% of the draws overlapped zero. Differences between pair-wise correlations were calculated analogously. This method has proven reliable for approximating standard errors in economic applications (Krinsky & Robb, 1990; Hole, 2007).

Results

Sensitivity of sexual dimorphism to developmental conditions

We found that development in hotter and more variable environments increased heat shock resistance of emerging adults in both sexes (Fig. 1, Table 1). Males had higher heat shock resistance compared to females. The difference between sexes in heat tolerance became larger in hotter temperature environments (Fig. 1, Table 1). Based on the within temperature test on sex differences and on the SDI, the largest deviation between the sexes was found at constant 30 °C (Wald χ2 = 215.17, P < 0.001, SDI = 0.213), followed by cycling 25 °C (Wald χ2 = 99.62, P < 0.001, SDI = 0.169) and cycling 30 °C (Wald χ2 = 100.97, P < 0.001, SDI = 0.153). The smallest difference was found for flies developed at constant 25 °C (Wald χ2 = 8.81, P < 0.003, SDI = 0.076). Rack identity in water baths (Wald χ2 = 578.13, d.f. = 12, P < 0.001) and the day of egg collection (Wald χ2 = 14.95, d.f. = 2, P < 0.001) also influenced heat shock tolerance. The model indicated also significant effects of sire (Wald χ2 = 2128.67, d.f. = 12, P < 0.001) and dam (nested within sire: Wald χ2 = 2323.28, d.f. = 487, P < 0.001) on heat shock tolerance.

Figure 1.

 Heat shock tolerance of male (black bars) and female (grey bars) Drosophila melanogaster after development (egg-to-adult) at different temperatures. Sexual dimorphism indices (SDI) were calculated as in Poissant et al. (2009).

Table 1.   Results from a generalized linear model (with logit link) testing effects of sex, mean temperature (Temperature), cycling temperature (Cycles) and their interaction on heat shock tolerance. The model also contained the effects of identity of water bath, day of egg collection, and identities of sire and dam (nested within sire). These effects were all highly significant (P < 0.001; data not shown).
VariableWaldd.f.P
Sex381.291< 0.001
Temperature1037.151< 0.001
Cycles217.391< 0.001
Temperature × Cycles30.8521< 0.001
Sex × Temperature48.761< 0.001
Sex × Cycles3.4910.062
Sex × Temperature × Cycles16.791< 0.001

Sensitivity of genetic variation and heritability to developmental temperature conditions

Males were found to have low heritabilities for heat shock tolerance ranging from 0.030 to 0.092 (Table 2). These estimates did not differ between developmental temperatures (P > 0.39, in all). In contrast, female heritability estimates for heat shock tolerance were clearly higher, ranging from 0.544 to 1.367 (Table 2; see the discussion for an explanation of the > 1 heritabilities). In females we found that higher average temperature resulted in lower heritabilities (pair-wise test of pooled h2 estimates from 25 and 30 °C, Table 2a; z = 2.925, P = 0.004), and this was due to larger sire variation (z = 2.675, P = 0.008) and lower vial variation (z = 2.171, P = 0.030) at colder temperatures. Temperature fluctuations did not affect the significance of heritability estimates (z = 1.692, P = 0.091) or other variance components in females (all z-tests nonsignificant). When temperatures were tested pair-wise between treatments, estimates of heritability differed significantly between females that developed at cycling 30 °C (h2 = 0.544) and females that developed at all other temperatures (pair-wise z-tests, P < 0.006, in all). This was due to the differences in sire variation. For flies developed at cycling 30 °C, the sire variance was clearly lower than for flies developed at constant 25 °C (z-test = 2.84, P = 0.004), cycling 25 °C (z-test = 4.20, P < 0.001) or constant 30 °C (z-test = 3.71, P < 0.001). There was also a tentative heritability difference between flies that developed at stable 25 °C (h2 = 1.016) and cycling 25 °C (h2 = 1.366; z-test = 1.85, P = 0.065) (Table 2).

Table 2.   Effects of developmental temperatures on quantitative genetic estimates of male and female heat shock tolerance, pooled according to mean temperatures (Panel a), pooled according to temperature fluctuations (Panel b) or estimating all treatment levels separately (Panel c). Variance components of sire (Vsire), dam (Vdam) and vial (Vvial) and heritabilities (h2) with their associated standard errors that correspond to the estimates at the underlying scale. Statistical significance (P) is based on z-tests. Estimation at the underlying (liability) scale assumes a residual variation of π2/3 for the logit link, and thus residual variation is omitted from the table. When pooled, we included a fixed factor for temperatures in the models, to control for the average effect of temperature on heat shock survival.
 nSiresDamsVSireSEVDamSEVVialSEh2SEP
(a)
 Males
  25 °C10 9581385690.0520.0310.3250.0530.7190.0590.0480.0280.044
  30 °C88601385690.0680.0400.3890.0690.7760.0750.0600.0350.044
 Females
  25 °C11 1711385732.0660.2990.3190.0580.7650.0661.2840.127< 0.001
  30 °C88511355571.1380.1770.2090.0610.9980.0850.8080.101< 0.001
(b)
 Males
  Constant10 3931385690.0670.0350.3810.0580.6060.0590.0620.0320.026
  Cycling94251385680.0880.0420.4090.0660.7230.0680.0780.0370.017
 Females
  Constant10 4491385691.7860.2600.2920.5840.7900.7061.1600.121< 0.001
  Cycling95731365731.2560.1890.2460.0580.8640.0770.8880.106< 0.001
(c)
 Males
  Constant 25 °C54771375690.0600.0430.3460.0840.5410.0860.0570.0410.082
  Cycling 25 °C54811385680.0490.0480.4620.0970.6530.0940.0440.0430.153
  Constant 30 °C49161385690.1090.0580.4240.1050.5930.1040.0980.0520.030
  Cycling 30 °C39441335550.0270.0540.2550.1210.8200.1410.0250.0490.305
 Females
  Constant 25 °C55511385691.4490.2400.2610.0900.7190.1041.0140.126< 0.001
  Cycling 25 °C56201365732.2120.3420.3580.0970.6570.1021.3580.142< 0.001
  Constant 30 °C48981355571.8250.2920.2980.1060.7750.1061.1800.136< 0.001
  Cycling 30 °C39531345470.6750.1350.0450.1070.9230.1410.5480.097< 0.001

Sensitivity of cross-sex genetic correlations to developmental conditions

Cross-sex genetic correlations between males and females were sensitive to developmental temperature. For flies developed at constant 25 °C, the genetic correlation was significantly negative (P = 0.0078). Other cross-sex genetic correlations were positive but did not deviate significantly from zero (P = 0.20, 0.55, 0.55 for flies developed at cycling 25 °C, constant 30 °C and cycling 30 °C, respectively). We did not find evidence for differences in cross-sex genetic correlations between flies developed at constant and cycling 25 °C (P = 0.079), nor in other pair-wise comparisons (P > 0.2, in all). All other cross-environment genetic correlations were highly positive (Table 3a). This can also be seen from factor loadings where males and females were associated with different factors (Table 3b). The first factor explained by far the majority of genetic variation with 93.7% of the genetic variation attributable to variation in females. The second factor characterizing genetic variation of males explained only additional 4.8%. Specific genetic variances explained 1.4% of the genetic variation, but were not statistically significant. Factor loadings and loading patterns resembled closely what was found for genetic variances in the univariate analyses.

Table 3.    (a) Genetic variance covariance matrix of male and female Drosophila melanogaster heat shock tolerance after development in four different thermal environments. In diagonal are genetic variances (italics), below the diagonal are the covariances, and above the diagonal the genetic correlations. Statistical significances are marked with asterisks for genetic correlations. Cross-sex genetic correlations are highlighted with bold. (b) Loadings of genetically determined heat shock tolerance in different thermal environments on genetic factors (Γ1,Γ2) and the amount of genetic variation that was found to be specific to environments and sexes (Ψ).
 MalesFemales
Constant 25 °CCycling 25 °CConstant 30 °CCycling 30 °CConstant 25 °CCycling 25 °CConstant 30 °CCycling 30 °C
  1. Asterisks denote statistical significance of estimates; *, ** and *** equal P < 0.05, P < 0.01 and P < 0.001, respectively.

(a)
 Males
  Constant 25 °C0.0700.9300.975*0.988*−0.130**0.0640.0760.004
  Cycling 25 °C0.0750.0940.989*0.976**0.2420.3060.2940.363
  Constant 30 °C0.0840.0980.1060.998***0.0930.1590.1470.218
  Cycling 30 °C0.0880.1010.1090.1140.0250.0920.0790.151
 Females
  Constant 25 °C−0.0470.1010.0410.0111.8500.998***0.999***0.992***
  Cycling 25 °C−0.0280.1570.0870.0522.2742.8090.999***0.998***
  Constant 30 °C−0.0290.1290.0680.0381.9472.4022.0550.997***
  Cycling 30 °C−0.0010.0850.0540.0391.0321.2801.0940.585
(b)
 Ψ0.01630.0280.065
 Γ1−0.0190.0910.0490.0281.358***1.676***1.433***0.763***
 Γ20.264***0.292***0.322***0.336***−0.0790.019−0.0050.053

Discussion

The evolution of sexual dimorphism is expected if genetic variances are sex specific, genetic correlations between sexes are below 1 and if selection pressures differ between males and females. However, sex-specific genetic variation is rarely observed (Roff, 1997; Poissant et al., 2009). Nevertheless, sexual dimorphism appears to evolve rapidly in response to environmental perturbations, suggesting that sexual dimorphism and its genetic background is sensitive to environments experienced during development and ontogeny (Badyaev, 2002). Our experimental results are in agreement with the observations made by Badyaev (2002).

Sensitivity of sexual dimorphism to developmental conditions

Developmental temperatures were found to affect sexual dimorphism. This was revealed by a significant sex-by-treatment effect on the level of heat shock tolerance. Males were much more resistant than females when developing in hotter environments. The difference between sexes decreased in cooler environments. Temperature cycles did not clearly affect sexual dimorphism but effects of cycles were found to be sensitive to mean temperatures. The largest difference was found between males and females developing at constant 30 °C. In contrast, a small difference was found between sexes for flies developed at constant 25 °C. The prevalence of environmentally sensitive sexual dimorphism is unclear. A number of studies find rather little evidence (Hare et al., 2011; Delph & Bell, 2008; LeGalliard et al., 2005; Fox et al., 2004), whereas other studies support the presence of environmental effects on sexual dimorphism (DeBlock & Stocks, 2003; Rosivall et al., 2010; Miller & Emlen, 2010; Kleinteich & Schneider, 2010; Dubiec et al., 2006). Developmental temperature also had an effect on average heat shock tolerance as we found that hot and cycling conditions during development resulted in a higher tolerance to heat shock at the adult stage. This result could be due to selection – only the fittest individuals survive at stressful temperatures or due to developmental acclimation (Levins, 1968; Kristensen et al., 2008).

Sensitivity of genetic variation and heritability to developmental temperature conditions

The largest difference between estimates of heritability for heat shock tolerance was found between the sexes rather than between developmental temperatures. Heritability estimates for males ranged between 0.03 and 0.09 and for females between 0.55 and 1.38. There are several potential reasons for observing heritability estimates higher than one (Cowley et al., 1986; Cowley & Atchley, 1988; de Boer & van Arendonk, 1992; Lynch & Walsh, 1998; Wilson, 2008) but as there was a strong sex bias in heritability estimates, the most likely cause is that the X chromosome contained genetic variation for heat shock tolerance (Cowley et al., 1986; Cowley & Atchley, 1988). The X chromosomes are known to carry QTL with strong effects on heat resistance (Morgan & Mackay, 2006; Norry et al., 2007). Such effects can easily cause h2 estimates to be larger than one, although this should not be theoretically possible (Puurtinen et al., 2009). For example, expected sire to daughter similarity due to additive genetic effects is 0.25 but expected sire to daughter similarity due to X chromosome effects is 0.5. Thus, if the sire variation contains large amount of X-chromosomal variation, and sire variation is multiplied by 4 to obtain the heritability (h2 = (Vsire*4)/Vp; Falconer & Mackay, 1996), the heritability estimate can easily exceed one in females (Cowley et al., 1986; Cowley & Atchley, 1988). Unfortunately, with the design used here, we cannot separate the additive genetic variance due to Y and X chromosomes and autosomes (Lynch & Walsh, 1998; Fairbairn & Roff, 2006). Therefore, the causal reasons for inflated female heritabilities warrant further studies, especially as X-chromosomal inheritance is expected and found to play a major role in sexually antagonistic expression of many traits (Rice, 1984; Gibson et al., 2001; Innocenti & Morrow, 2010).

Our results showing higher heritabilities in females concur with the findings of Drobniak et al. (2010), where cell-mediated immune response in blue tits was heritable in females but not in males. Moreover, Blanckenhorn (2002) found higher broad-sense heritability estimates of developmental time in female dung flies (however this pattern was somewhat reversed in hind tibia length). Fairbairn (2007) reports slightly higher broad-sense heritabilities in female water striders’ traits. Interestingly, there seems to be no good review available for sex specificity of heritabilities. Such a review is needed because even in the case when cross-sex genetic correlations are high, sexual dimorphism can rapidly evolve if heritabilities differ between sexes (Lynch & Walsh, 1998). Nevertheless, our study shows that the evolution of heat shock tolerance occurs more profoundly via selection on females, although this is potentially confounded by X-chromosomal inheritance.

If heritability and variance components are sensitive to temperatures during development, this allows for efficient selection in environments where the heritabilities are inflated. Furthermore, if the environment affects sex-specific heritabilities disproportionally, this could contribute to the rapid evolution of sexual dimorphism (Lynch & Walsh, 1998). We found that the most significant effect of the environment was observed in female heritability estimates. Here, we observed a significantly lower heritability in flies developing at hotter temperatures and particularly at cycling 30 °C. The changes in female heritability estimates were clearly due to altered sire variation, suggesting allelic sensitivity to the environment or differential selection on genetic variation during development. Heritability estimates for male heat shock tolerance were also lowest under the same conditions than in females albeit not significantly different from the other treatments.

As heritability estimates of adult heat shock tolerance were found to depend on temperature during development, this indicates that the expression of genetic variation differs across environments. There is however no clear trends with respect to whether heritabilities are generally higher or lower in stressful environments (reviewed in Hoffmann & Merilä, 1999; Charmantier & Garant, 2005). Recent quantitative genetic studies, for example, suggest that the genetic variances and heritabilities of traits may not be sensitive to environmental variation (Messina & Fry, 2003; Pakkasmaa et al., 2003; Fox et al., 2004) whereas others have found the opposite pattern (Bubliy & Loeschcke, 2001; Hendrickx et al., 2008).

Sensitivity of cross-sex genetic correlations to developmental conditions

Males and females share a common gene pool and thus the evolution of sexual dimorphism can be hindered by genetic correlations between homologous traits in males and females. The evolution of sexual dimorphism is facilitated if genetic variation in males and females is uncorrelated, as was evident from results from our factor analysis with two factors clearly explaining either female or male variation (Table 2). This result suggests a strong opportunity for independent evolution of heat shock tolerance in both sexes. Furthermore, cross-sex genetic correlations showed most of the genetic correlations were not different from zero, allowing rather independent evolution of heat shock tolerance in males and females. Evidence for environmental effects on cross-sex genetic correlations is scarce and nonconsistent (Simons & Roff, 1996; Lyons et al., 1994; Fox et al., 2004; Delcourt et al., 2009), and thus, much more emphasis must be given to this issue.

Conclusions

We found evidence that for adult heat shock survival, sexual dimorphism and its genetic background are dependent on developmental temperatures. First, the temperature during development influenced the degree of sexual dimorphism. Second, the expression of genetic variation was sensitive to the environment especially in females, and finally, genetic variation was mostly sex specific. Together, these results suggest that the developmental environment can strongly affect sexual dimorphism and alter the evolutionary trajectories by alleviating expression of genetic material for selection to act upon. Our results also show that irrespective of the environment, male and female heat shock tolerances are free to evolve independently from each other. Thus, environmentally based sensitivity of sexual dimorphism and its evolutionary potential can provide a viable explanation for previous findings of fast changes in sexual dimorphism following environmental perturbations (Badyaev, 2002).

Acknowledgments

We thank D. Andersen, L. van Koll, M. Thybring, P. Sarup, C. Vermeulen, J. Witt, L. Andersen, N. Le, K.S. Pedersen and J. Overgaard for excellent help in the laboratory, A.A. Hoffmann for providing the fly stocks, the Academy of Finland for funding for TK, the Danish research council for frame grants to VL and a Steno stipend to TNK and the Aarhus Stress Network for supporting the visit of TK to Aarhus University.

Ancillary