Selection in a fluctuating environment and the evolution of sexual dimorphism in the seed beetle Callosobruchus maculatus

Authors


Lára R. Hallsson, Animal Ecology/Department of Ecology and Genetics, Evolutionary Biology Centre (EBC), Uppsala University, Norbyvägen 18D, 75236 Uppsala, Sweden.
Tel.: +46 18 471 2672; fax: +46 18 471 6484; e-mail: l.hallsson@gmx.com

Abstract

Temperature changes in the environment, which realistically include environmental fluctuations, can create both plastic and evolutionary responses of traits. Sexes might differ in either or both of these responses for homologous traits, which in turn has consequences for sexual dimorphism and its evolution. Here, we investigate both immediate changes in and the evolution of sexual dimorphism in response to a changing environment (with and without fluctuations) using the seed beetle Callosobruchus maculatus. We investigate sex differences in plasticity and also the genetic architecture of body mass and developmental time dimorphism to test two existing hypotheses on sex differences in plasticity (adaptive canalization hypothesis and condition dependence hypothesis). We found a decreased sexual size dimorphism in higher temperature and that females responded more plastically than males, supporting the condition dependence hypothesis. However, selection in a fluctuating environment altered sex-specific patterns of genetic and environmental variation, indicating support for the adaptive canalization hypothesis. Genetic correlations between sexes (rMF) were affected by fluctuating selection, suggesting facilitated independent evolution of the sexes. Thus, the selective past of a population is highly important for the understanding of the evolutionary dynamics of sexual dimorphism.

Introduction

Sexual dimorphism of homologous male and female traits is widespread among animals, and sexual size dimorphism has received great attention in particular (Fairbairn, 1997, 2007; Blanckenhorn et al., 2007; Stillwell et al., 2010). A difference in size between the sexes is ultimately a result of sex-specific selection acting on the homologous trait resulting in different optima for the two sexes (Price, 1984; Arak, 1988; Schluter et al., 1991; Blanckenhorn, 2000, 2005, 2007; Fairbairn, 2007). However, effects of climate or other ecological and environmental variables, such as temperature, can have different effects on the sexes. This can be due to different fitness consequences of different trait sizes, differences in the underlying genetic architecture or differences in the degree of plasticity between the sexes (Fairbairn, 2005; Teder & Tammaru, 2005; Blanckenhorn et al., 2006; Stillwell & Fox, 2007). Thus, sexual dimorphism evolves and is also a plastic trait, and when studying sexual dimorphism and its evolutionary dynamics, it is important to consider differences in selection, sex-specific genetic variances and covariances, and differences in plasticity between the sexes.

Two hypotheses have been proposed as proximate explanations for sexual dimorphism. In either hypothesis, one sex is thought to be more plastic in response to a change in the environment than the other; thus, sexual dimorphism is explained by differential plasticity of the sexes. Both hypotheses assume that traits that are closely related to fitness are under strong selection, but the hypotheses differ in two crucial ways. First, in their underlying assumption of the type of selection considered, the adaptive canalization hypothesis is mainly considering stabilizing selection (Stillwell et al., 2010), whereas the condition dependence hypothesis is regarding selection to be directional (Bonduriansky, 2007a,b; Stillwell et al., 2010). Second, they differ in their prediction regarding the plasticity of the trait under strong selection: the adaptive canalization hypothesis (Fairbairn, 2005) predicts that traits closely related to fitness will be less plastic because of strong selection acting on them, whereas the condition dependence hypothesis predicts that traits under strong selection are more likely to capture genetic and environmental effects affecting condition and therefore these can be more plastic (Bonduriansky, 2007a,b). Making predictions about sex differences in plasticity or interpreting results according to one or the other hypothesis is often difficult. This is because, for instance, both a decrease in stabilizing selection and an increase in directional selection would lead to an increase in phenotypic plasticity, which then could be explained by either of the hypotheses. Thus, the trait under observation needs to be studied in relation to other traits (Stillwell et al., 2010), or the type of selection acting on the trait has to be known. Alternatively, if both directional and stabilizing selections are acting on the trait of interest, one has to be able to determine their relative importance.

To avoid this potential problem, we studied sex differences in plasticity in a quantitative genetic framework. We used the fact that the hypotheses are based on different types of selection and combine it with knowledge on behaviour of genetic and environmental variation under stabilizing vs. directional selection. Based on that, we can make inferences about effects on genetic and environmental variation under the different hypotheses and finally test the hypotheses empirically. With this indirect approach of testing the hypotheses, we create a link between proximate and ultimate explanations for sexual dimorphism. Whereas differential plasticity serves as a proximate explanation for sexual dimorphism, analysing different kinds of selection and genetic variances/covariances gives insights into the evolution of sexual dimorphism. Based on the knowledge of the type of selection in either hypothesis, we infer that under the adaptive canalization hypothesis, one would see little genetic and environmental variation, because strong stabilizing selection is thought to lead to genetic and environmental canalization (Stearns et al., 1995; Flatt, 2005; Zhang & Hill, 2005). Under the condition dependence hypothesis, on the other hand, one would see increased genetic variation due to strong directional selection (Bürger & Lynch, 1995; Kawecki, 2000).

The environmental temperature is generally known to have a major impact on the life history of insects, and it has been shown to differentially affect sexes (Roff, 2002). Thus, by exposing individuals to within- and across-generation temperature changes and investigating their plastic and evolutionary response for separate sexes, we can gain insight into sexual dimorphism and its evolution. Development time and body size are traits that are known to be strongly dependent on temperature, mainly mediated by developmental processes with higher temperature, resulting in shorter developmental times and smaller size (Sibly & Atkinson, 1994; reviewed by Nylin & Gotthard, 1998). The traits have different fitness consequences for males and females. Female body size is mainly determined by fecundity selection favouring large females (Honek, 1993; Savalli & Fox, 1999). Larger males can have increased mating success and increase female fecundity via nuptial gifts (Fox & Czesak, 2006). Developmental time can be under different selection in the two sexes, for instance in Lycaena butterflies (Fischer & Fiedler, 2000, 2001) or in the seed beetle Callosobruchus maculatus (Savalli & Fox, 1999), through selection for rapid development in males (protandry) and hence a competition advantage over females. Females, on the other hand, might gain from an increased developmental time as this will result in a larger size and hence greater fecundity. Both developmental time and body size are known to respond in a phenotypic plastic way to changes in the environment in both sexes (Atkinson & Sibly, 1997; Angilletta & Dunham, 2003; Stillwell et al., 2007; reviewed by Nylin & Gotthard, 1998). These plastic responses can differ between males and females and induce variation for sexual dimorphism (Fairbairn, 2005; Teder & Tammaru, 2005; Stillwell & Fox, 2007). In the seed beetle C. maculatus body mass at emergence and developmental time are positively correlated traits. Females trade off the advantages of being large with associated benefits of higher fecundity against cost of having to develop longer to achieve larger size (Sibly & Calow, 1986; Moller et al., 1989). Males that emerge earlier have a greater access to virgin females (Savalli & Fox, 1999), but a shorter developmental time results in a smaller emergence mass. Thus, females are larger than males, and males hatch after a shorter developmental time than females (Guntrip et al., 1997). Body mass is thought to be closer related to fitness in females compared with males (Savalli & Fox, 1999), and developmental time is thought to be closer related to fitness in males compared with females (Zonneveld, 1996; Fischer & Fiedler, 2000, 2001).

In this study, we test the two hypotheses by analysing two traits that are known to be important for fitness, but in different ways in males and females, using C. maculatus. For each hypothesis and trait, we can make predictions of the response of the sexes to a rapid (within generation) change in temperature environment regarding phenotypic plasticity, genetic and environmental variances and changes in sexual dimorphism. Under the adaptive canalization hypothesis, we predict that male mass is affected more than female mass, that is, males are more plastic and females are canalized, and that male genetic and environmental variances should be higher. The higher plasticity in males is expected to result in an increased sexual dimorphism in higher temperature. This is predicted to be reversed in developmental time as this trait is of higher importance for male fitness, both compared to females and compared to mass, that is, we expect males to be less plastic than females and to show lower genetic and environmental variances, and thus, sexual dimorphism is expected to be reduced in higher temperature. Under the condition dependence hypo-thesis, we predict that female mass is more plastic because it is thought to be under stronger selection in females compared with males and that females have higher genetic variances than males. A decreased sexual dimorphism in higher temperature is expected. We expect that the opposite prediction holds for developmental time, with males being more plastic than females and showing higher genetic variance, leading to increased sex differences in higher temperature (Table 1). However, not only rapid (within generation) changes in temperature are expected to create changes in sexual dimorphism, but also environmental changes over an extended period of time (across several generations) are of considerable interest when studying sexual dimorphism and its evolution. The environment is unlikely to be constant over time, with fluctuations being the most realistic and biologically relevant way in how the environment is changing (e.g. Lundberg et al., 2000; Boyce et al., 2006; Schreiber, 2010; Hallsson & Björklund, 2012). Sexual dimorphism might change after continued exposure to environmental change. Each of the factors that are known to determine the response within a generation, can potentially change due to selection in the long term. Thus, selection over an extended period of time (across several generations) might affect sex-specific selection, sex-specific amount of genetic variance and/or sex differences in plasticity, resulting in a response that might differ from the response after a rapid shift in the environment (within generation). We therefore subjected our populations to selection for 18 generations, where the temperature was raised successively to the novel temperature. This was done in two major ways, either by letting the temperature increase by the same amount every generation generating a linear trend towards the novel temperature (Trend lines) or by letting the temperature increase but with fluctuations around the mean increase (Fluctuation lines), while a control (Control lines) was kept at constant temperature. Using this approach, we can analyse not only the effect of selection compared to a rapid shift to a new environment but also the role of environmental fluctuations. We evaluate the effect of a rapid (within generation) change in the environment and test the predictions under the different hypotheses explicitly by investigating the response of the Control across temperatures, and we investigate the effect of continued temperature changes (over several generations) including fluctuation by comparing the Fluctuation lines with the Control lines after selection. We found phenotypic plasticity and genetic and environmental variance to be generally affected differently by the different selection regimes (Hallsson & Björklund, 2012). In this study, we concentrate on sex differences (in plasticity and genetic and environmental variances) in response to within- and across-generation changes in the environment.

Table 1.   Expected differences between males (M) and females (F) in phenotypic plasticity (PP), genetic and environmental variation (CVG and CVE) and changes in sexual dimorphism (SD) in higher rearing temperatures under the adaptive canalization hypothesis and the condition dependence hypothesis. Mass at emergence is assumed to be more important for female fitness, whereas developmental time is assumed to be more important for male fitness.
 Adaptive canalization hypothesis (stabilizing selection)Condition dependence hypothesis (directional selection)
PPSDCVGCVEPPSDCVGCVE
  1. ‘↑’ and ‘↓’ indicate an expected increase/decrease in SD, respectively, and ‘?’ indicates that there is (to our knowledge) no literature/studies on this effect.

Mass at emergenceM > FM > FM > FM < FM < F?
Developmental timeM < FM < FM < FM > FM > F?

Materials and methods

To investigate both the within-generation response of sexual dimorphism to a change in temperature and the evolution of sexual dimorphism in response to a changing environment, we conducted a long-term experimental evolution experiment in the laboratory over 18 generations and a split-brood experiment at the end of the long-term experiment. We used the seed beetle C. maculatus as a model organism (stocks were provided by Peter Credland, University of London). We are interested in sex differences in the phenotypic response and its underlying changes in genetic variance and covariances as well as correlation between sexes.

The seed beetles (Nigerian mixed strain) were adapted to 30 °C for approximately 90 generations in the laboratory before the experiment. Beetles were kept on black-eyed beans as substrate. The generation time is approximately 25 days at 30 °C. During the experiment, we separated generations by transferring about 200 adult individuals to 100 g of fresh beans at peak of emergence. Individuals were exposed to three different selection regimes including a trend of successive increase in temperature (from 30 to 36 °C; about 0.3 °C every generation) over 18 generations in total. The first treatment involved only a linear trend from 30 to 36 °C, hereafter called ‘Trend’. In the other two selection regimes, environmental fluctuations (noise) were added on top of the trend of temperature increase. One treatment had high positive temporal autocorrelation in the noise, hereafter called ‘Red’ treatment and one treatment had no temporal autocorrelation in the noise, hereafter called ‘White’ treatment. Both Red and White noise treatments are collectively referred to as ‘Fluctuation’ treatment lines from here on. A control hereafter called ‘Control’ was kept at a constant temperature of 30 °C. Trend, Red and White treatment lines were reared in 36 °C in their 18th and last generation of the selection experiment. The Control and the three different treatments contained four replicates each, resulting in 16 lines in total.

After the 18th and last generation of selection, we conducted a split-brood experiment. With this approach, we aimed to mirror the response of the different treatment lines and sexes to exposed changes in temperature, compare the response between treatments and sexes and show whether the response of population and sex is due to genetic and/or phenotypic plastic changes. We measured mass at emergence and developmental time of all lines (including Control) and separate sexes. The split-brood experiment lasted over two generations. We formed mating pairs of virgin beetles and let them mate and lay eggs on 40 g of substrate (sufficiently high to minimize within-bean larvae competition) for 72 h in room temperature (23 °C). Ten pairs per replicate per line were formed, resulting in 160 pairs in total. Beans with attached eggs were split up into two temperature environments: one environment (30 °C) representing the environment of origin to which all lines have been adapted to at the beginning of the long-term experimental evolution experiment and the other environment (36 °C) representing the environment the treatment lines have been selected towards during the experiment. The split-up resulted in full-siblings of each family being raised across environments. Hatching individuals were collected and counted daily during their emergence period, separating sexes (developmental time data). Mass at emergence was measured on days two, four and seven of the period of emergence; individuals were sexed, frozen and dried in an oven at 50 °C for 48 h, and their mass was obtained to the nearest 0.01 mg (using a Sartorius Genius Microbalance model ME235P).

In total, 1668 individuals were measured for mass at emergence (i.e. on average, 26 observations per treatment, replicate, temperature and sex) and 3457 for developmental time (approximately 54 per treatment, replicate, temperature and sex).

Statistical analysis

The analysis was conducted using the MCMCglmm package (Hadfield, 2010) in r 2.13.0 (R Development Core Team, 2011). We used uninformative proper priors for both fixed and random effects. Fixed effect priors were normally distributed with an expected value (mean) of zero and variance of 108. Random effects priors (one for each random effect and its interaction with fixed effects) were inverse-Wishart-distributed. We allowed the Markov chain a burn-in period of 3000 iterations, after which we ran 15 000 iterations and sampled every 10th iteration from the posterior distribution, resulting 1500 stored values per chain (Hadfield, 2011; http://stat.ethz.ch/CRAN/web/packages/MCMCglmm/index.html, Vignette: MCMCglmm Course Notes). These settings resulted in an appropriate convergence of the chain. Convergence was assessed by checking for potential autocorrelations of consecutive values in the chain and via visual inspection of potential trends in the chain as well as of the shape of the posterior density distribution of fixed and random effects respectively. Autocorrelation between consecutive values was low (< 0.04), there were no trends in the chain, and the posterior distributions were not skewed. We present the posterior mode and 95% credible intervals (95% CI) unless explicitly stated otherwise.

We fitted linear mixed models separately for each trait (developmental time, mass at emergence) and each treatment line (Control, Trend, Red, White). Temperature (30 vs. 36 °C) and sex of the offspring were fitted as fixed effect predictors, and family and replicate identity were included as random effects. As we were interested in the covariances between sexes and between temperature environments, we also fitted the interactions of sex and temperature with the two random effects. For the family identity and the replicate identity random effects, we estimated unstructured variance–covariance matrices, that is, one variance for each temperature and sex (four variances) and all covariances between sexes and temperatures (six covariances). For the residual variance–covariance matrix, we fixed all covariances to zero, because each individual was measured in only one environment, and hence, there is no replication in our data to estimate a residual covariance.

Our models estimate all relevant sex-specific phenotypic and genotypic variance–covariance matrices. We investigated all estimates for separate sexes. We converted variances to standardized coefficients of variation by dividing variance by the phenotypic mean trait value (Houle, 1992). The coefficient of genetic variation (CVG) measures the variation among families in each environment (covariance of full-sibs; Falconer & Mackay, 1996) and thus the amount of genetic variation in a treatment line in a certain environment. CVG typically includes dominance and maternal effects. Thus, we can regard the estimated CVG as an upper bound to VA (Falconer & Mackay, 1996; Conner & Hartl, 2004). The coefficient of environmental (CVE) variation measures the variation among individuals within each family and environment, which represents environmental variability/degree of environmental canalization, and it includes all the variance due to nongenetic origin (Stearns et al., 1995). In our study, we were able to control for external environmental conditions (half of the individuals of a family are raised in a common environment). CVE contains the variance due to internal conditions (e.g. developmental noise), G × E, phenotypic plasticity and epistasis (discussed in Hallsson & Björklund, 2012). CVG and CVE measure the degree of genetic and environmental canalization, respectively (Stearns et al., 1995). The genetic correlations between the sexes (rMF) within each environment were assessed by dividing the covariance between sexes by the square root of the product of variance for each sex and measured the degree of independence between homologous male and female traits.

Within each treatment line, significance of estimates of the fixed effects can be assessed by their posterior distribution not overlapping zero. We tested for significance of variance components using model comparison based on the deviance information criterion (DIC) (Wilson et al., 2010). To explicitly test for effect of sex on each of the parameters (i.e. for treatment, environment, CVG, G × E and CVE) within each treatment line, we assessed whether the credible interval of the distribution of sex differences (females – males) did overlap zero. Distributions of sex differences were calculated and tested for each of the parameters. To evaluate differences between treatment lines, we had to compare estimates between models. Therefore, we calculated a distribution of treatment differences for each sex and tested for a significant difference from zero. The distributions of treatment differences (= delta distribution) were generated from posterior distributions of fixed and random effects, respectively. Thus, we calculated the difference between the posterior distribution of each parameter in one treatment line and the posterior distribution of corresponding parameter in the other treatment. Mean and standard deviation of the resulting delta distribution were calculated, and the estimate was considered significant if the mean (± 2 SD) did not include zero. Estimates of the genetic correlation between the sexes within each treatment and environment were evaluated by their credible intervals.

Results

Body mass at emergence

Sexes differed significantly in their size at both 30 and 36 °C, with females being significantly larger than males in all lines (Fig. 1a, Tables 2 and S2, Appendix S1.1a). Female size decreased more than male size across temperature environments, and hence, sexual size dimorphism was smaller in 36 °C compared to 30 °C (Fig. 1a, Appendix S1.1a). The effect of environment was overall significant for females but not for males (females: mean = 0.18 mg, 95% CI = 0.071, 0.28; males mean = 0.095 mg; 95% CI = −0.042, 0.21). The response of sexes across environments was similar for all treatment lines (Fig. 1a, Appendix S1.1a).

Figure 1.

 (a) Mean phenotype for mass at emergence (in mg) for each sex, treatment line and temperature presented with 95% credible intervals (based on variation among of 40 families) (b) Mean phenotype for developmental time (in days) for each sex, treatment line and temperature presented with 95% credible intervals (based on variation among of 40 families). Control (C), Trend (T), Red (R) and White (W) in 30 and 36 °C are abbreviated by C30, C36, T30, T36, etc. Significant effect of environment within each line and sex is indicated by an asterisk *. Significant differences between selection lines and Control (i.e. effect of selection) are indicated by triangles; significant difference in the Control within each temperature and sex is indicated a triangle pointed up (Δ), and significant difference in Control across temperatures (i.e. treatment-by-environment interaction) is indicated by a triangle pointed down (∇). Estimates of significance are presented in Appendix S1.1.

Table 2.   Summary of results; sex difference in mean phenotype, genetic (CVG) and environmental (CVE) coefficient of variation.
TreatmentMean phenotypeCVGCVE
Mass at emergenceDevelopmental timeMass at emergenceDevelopmental timeMass at emergenceDevelopmental time
  1. *Denotes significant effects found, – denotes no significant effect found. For details on significances see Tables S2 and S3.

Control**
Trend**
Red***
White**

Females but not males tended to show an increase in CVG across temperatures (Fig. 2a, Appendix S1.2a). There was a sex-specific effect on CVG in the White treatment line (Tables 2 and S3). Both males and females differed from the Control in their CVG, resulting in a significant CVG-by-sex-by-treatment interaction [delta distribution: mean (± 2 SD) = 7.39 (4.41)]; females showed a decreased, whereas males showed an increased CVG (Table 3, Fig. 2a, Appendix S1.2a).

Figure 2.

 (a) Genetic coefficient of variation (CVG) (b) Environmental coefficient of variation (CVE) for mass at emergence for separate sexes; presented with 95% credible intervals. Abbreviations as in Fig. 1; *indicates significant effect of environment within each treatment line and sex. Triangles indicate significant differences of each treatment line to Control line for separate sexes; difference within temperature (Δ) and across temperatures (∇). Estimates of significance are presented in Appendix S1.2.

Table 3.   Summary of results; treatment differences of mean phenotype, genetic (CVG) and environmental (CVE) coefficient of variation, tested against Control for separate sexes.
ContrastMean phenotypeCVGCVE
Mass at emergenceDevelopmental timeMass at emergenceDevelopmental timeMass at emergenceDevelopmental time
FMFMFMFMFMFM
  1. F = female, M = male; *Denotes significant effects found, – denotes no significant effect found. For details on significances, see Figs 1–3; Appendix S1.1–S1.3.

Control-Trend******
Control-Red******
Control-White******

Also in CVE, we found a sex-specific response across temperatures; females showed an increased CVE (especially in Control and Trend), whereas males showed an unchanged or even decreased CVE (Fig. 2b, Tables 2 and S3, Appendix S1.2b). Sexes of the Red treatment lines differed from the Control, resulting in a significant CVE-by-sex-by-treatment interaction (delta distribution: mean (± 2 SD) = 2.61 (1.44); Table 3, Fig. 2b, Appendix S1.2b). In the Control, females show an increased CVE in higher temperature, but in the Red line, CVE was decreased. In contrast, the CVE of the Red males was increased compared with the CVE of the Control males (Fig. 2b, Appendix S1.2b).

Genetic correlations between sexes within each environment and treatment were generally high (near one) and significantly different from zero for both Control and most treatment lines (except White treatment line in 30 °C) (Table 4).

Table 4.   Cross-sex genetic correlation (rMF) for each treatment line and temperature. rMF is tested for significance against zero by evaluation of the credible interval (not including zero).
TreatmentMass at emergenceDevelopmental time
rMF in 30 °CrMF in 36 °CrMF in 30 °CrMF in 36 °C
  1. *Denotes significant effect found.

Control0.996 (0.78; 0.99)*0.997 (0.80; 0.99)*0.516 (0.16; 0.77)*0.776 (0.45; 0.91)*
Trend0.998 (0.95; 0.99)*0.997 (0.90; 0.99)*0.553 (0.12; 0.77)*0.826 (0.54; 0.92)*
Red0.996 (0.88; 0.99)*0.993 (0.36; 0.99)*0.289 (−0.09; 0.64)0.459 (−0.13; 0.76)
White0.931 (−0.98; 0.99)0.994 (0.71; 0.99)*0.558 (0.07; 0.75)*0.181 (−0.23; 0.74)

Developmental time

There were no sex differences in developmental time and no effect of environment or treatment (Tables 2 and 3, Fig. 1b,Table S2). We could not detect sex differences in CVG within and across environments for most of the lines (Tables 2, S2 and S3). However, there were significant effects of treatment on CVG (both males and females showed a generally reduced CVG, Table 2, Fig. 3a, Table S3, Appendix S1.3a), and sexes in the Red treatment line responded different across temperatures compared with sexes in the Control, resulting in a significant CVG-by-environment-by-sex-by-treatment interaction [delta distribution: mean (± 2 SD) = 1.98 (1.22)]. Both sexes in the Control show an increased CVG across temperatures, but this could only be found in the females in the Red line, resulting in a significant sex difference in CVG in 36 °C in the Red line (Table 3, Fig. 3a, Table S3, Appendix S1.3a).

Figure 3.

 (a) GVG and (b) CVE for developmental time for separate sexes. Abbreviations as in Fig. 1; * indicates significant effect of environment within each treatment line and sex. Triangles indicate significant differences of each treatment line to Control line for separate sexes; difference within temperature (Δ) and across temperatures (∇). Estimates of significance are presented in Appendix S1.3.

We found a sex-specific response across environments in CVE in the Control but not in the treatment lines. Control females, but not males, showed an increase in CVE across temperatures (significant sex-by-environment interaction for CVE, Table 2, Fig. 3b, Appendix S1.3b). In the selection lines, both males and females showed a significant increase in CVE across environments (Fig. 3b, Appendix S1.3b); this resulted in a significant CVE-by-environment-by-sex-by-treatment interaction for Trend and White lines [delta distribution: mean (± 2 SD) = −1.36 (0.77) and −1.42 (0.91) respectively]. Generally, the treatment lines had a lower CVE compared with the Control line (Table 3, Fig. 3b, Appendix S1.3b).

The genetic correlation between the sexes was significantly different from zero in Control and Trend, but not in the Red and White treatment lines (except White in 36 °C, Table 4). The Control and Trend lines showed a higher genetic correlation compared to the Fluctuation lines, where the genetic correlations between the sexes were low (credible interval including zero; Table 4).

Discussion

We found that sexual size dimorphism decreased in the higher temperature due to females having a greater reduction in body mass. This was accompanied by an increase in CVG and CVE in females across temperatures (Figs 1a and 2a,b; Control line, Appendix S1.1a and S1.2a,b). These findings support the condition dependence hypothesis. After selection in a fluctuating environment (Red and White line), we found lower CVG in females and higher CVG in males (White) and lower CVE in females and higher CVE in males (Red) for body mass (Fig. 2a,b, Appendix S1.2a,b). For developmental time, we found lower CVG in males compared with females (Red; Fig. 3a, Appendix S1.3a). These findings support the adaptive canalization hypothesis.

Generally, individuals hatch earlier and at smaller body mass in 36 °C. We found a sex-specific effect of a rapid (within generation) shift in temperature on body mass at emergence, with females being more plastic than males and having increased CVG and CVE. Having in mind that body mass is thought to be closer related to fitness in females compared with males and thus under stronger selection, this result is in line with the condition dependence hypothesis (Table 1). Females are more plastic (less canalized) in their response to a change in condition, which is correlated with a decrease in sexual size dimorphism. The response of the males is more stable across environments; thus, they are generally more canalized against environmental perturbations. The increase in CVG across environments was more pronounced in females compared with males. We suggest that strong directional selection on female body size leads to a lower degree of genetic canalization (Stearns et al., 1995) and higher genetic variance (Bürger & Lynch, 1995; Kawecki, 2000). Furthermore, a higher CVG after a rapid shift in the environment implies that females have a greater potential to respond to selection than males (Houle, 1992). The CVE increased across environments for females but not for males suggests that the females are the sex that is less environmentally canalized against environmental perturbations (Stearns et al., 1995). All these findings support the expectations under the condition dependence hypothesis.

That females are under stronger directional selection to be large compared with males and therefore more plastic has previously been shown by Teder et al. (2008), Tammaru et al. (2009) and Stillwell et al. (2010). Our findings with regard to differential phenotypic plasticity are in line with previous findings of Stillwell & Fox (2007; reviewed by Stillwell et al., 2010), who tested temperature effects on sexual dimorphism over a range of temperatures in C. maculatus. They found male body size to exhibit more plasticity in response to low rearing temperatures (according to their predictions) but females to be the more plastic sex in response to high rearing temperatures (unexpected). As the temperature exposure in our study falls in to the high temperature range of Stillwell & Fox’s study (2007; reviewed by Stillwell et al., 2010), our results confirm the pattern for C. maculatus. However, as the beetles were adapted to different temperatures prior to the experiment [25 °C in Stillwell & Fox’s study (2007) vs. 30 °C in our study], a direct comparison of results might be less straightforward.

The decrease in sexual dimorphism in higher temp-erature could be generated by sex-specific responses to physiological stress. Fecundity was generally lower, and the number of hatching individuals was reduced in higher temperature (Hallsson & Björklund, 2012). This suggests that the higher temperature is physiologically stressful. Females might be more sensitive to developmental stress compared with males and thus unable to reach their target body size when reared at high temperatures. Alternatively, our findings could be explained by males being under stronger stabilizing selection for body mass and therefore less plastic than females (adaptive canalization hypothesis, Fairbairn, 2005). However, we find this an unlikely explanation as males had higher CVG and CVE in 30 °C compared with females, which would not be the case if males were under stronger stabilizing selection compared with females.

Developmental time is more closely related to fitness in males, as males that hatch early have an advantage in male–male competition over females (Savalli & Fox, 1999). Stronger stabilizing selection on this trait in males should result in low plasticity, lower genetic variance and a more canalized response (adaptive canalization hypo-thesis), whereas strong directional selection should result in higher phenotypic plasticity and higher CVG (condition dependence hypothesis) (Table 1). Against expectations, we found no sex differences in phenotypic plasticity or CVG. However, there was a sex-specific increase in CVE across environments in the Control line, where females but not males showed an elevated CVE. We suggest that because of its high importance for fitness, developmental time is under strong stabilizing selection in males, and therefore, the response of males is more environmentally canalized. For both traits, we found females to be more sensitive to a change in the environment than males. For body size, we found strong evidence that this pattern can be explained by the condition dependence hypothesis, with females being under stronger directional selection and therefore more plastic. For developmental time, there is suggestive evidence that males are under stronger stabilizing selection and therefore more canalized against environmental perturbations.

Sex difference in fluctuating environments

Selection in a fluctuating environment had significant impact on the response of sexes. Although our results for body size in the Control line generally support the condition dependence hypothesis, with females being the more plastic sex, we found contrasting results in the Fluctuation lines. We found a sex difference in CVG with a consistently lower CVG in females and a higher CVG in males (White). We suggest that this might be due to females experiencing stronger stabilizing selection (and relaxed directional or both) due to exposure to a fluctuating environment, whereas males might experience stronger directional or relaxed stabilizing selection (or both). However, as body size is more important for females’ fitness compared to males’ and thus under stronger selection, the sex difference is most likely driven by strong selection on females. Thus, our results support the adaptive canalization hypothesis. Selection in a fluctuating environment leads to a more genetically canalized response of females and a reduced potential to respond to selection. In contrast, it leads to a less canalized response and greater genetic variation and therefore higher potential to respond to further selection in males.

Also, patterns of environmental sensitivity were altered by selection in a fluctuating environment. As discussed, we found the amount of environmental variation to be sex and environment specific for body mass at emergence (particularly in the Control and Trend line); females are the sex being more sensitive/less environmentally canalized to a change in the environment compared with males. However, this relationship was reversed in one of the Fluctuation lines (Red). In contrast to the Control, males show a decreased degree of environmental canalization (i.e. higher CVE), whereas females show increased degree of environmental canalization (lower CVE). This result supports the adaptive canalization hypothesis and suggests that females have been under strong stabilizing selection due to exposure to a fluctuating environment.

For developmental time, we were unable to detect a sex difference in CVG. However, an interesting exception to this overall pattern is the finding of a sex difference in CVG after selection in a fluctuating environment (Red), where males but not females show a decreased CVG in higher temperature and therefore a stable response across environments. Under the assumption that developmental time has a greater influence on male compared with female fitness, this result supports the adaptive canalization hypothesis. We suggest that selection in a fluctuating environment leads to stronger stabilizing selection on males compared with females, resulting in a higher degree of genetic canalization for the trait in question. Thus, as developmental time is highly important for male fitness and the predictability of a fluctuating environment is low, holding trait expression constant and buffer it against environmental perturbations is beneficial in males. Moreover, a sex-specific response in the Fluctuation lines is facilitated because the genetic correlation between the sexes was low after selection in a fluctuating environment (Table 4). This implies that males and females can respond independently. Furthermore, it suggests that a change in environmental conditions over time can change the genetic variance–covariance matrix in a way that rMF is weakened. This suggests that genetic correlations are environment dependent (Lyons et al., 1994; Vieira et al., 2000; Sgrò & Hoffmann, 2004; Robinson et al., 2009) and respond to direct selection (Wilkinson et al., 1990; Shaw et al., 1995; Paulsen, 1996; reviewed by Steppan et al., 2002; Björklund, 2004; Arnold et al., 2008).

Our results highlight that fluctuations in the selective past have differential impact on the sexes. We suggest that the frequent change in the direction of selection in a fluctuating environment might result in net stabilizing selection on traits that are most important for fitness in the respective sex (i.e. body mass for females and developmental time for males). Thus, although we generally found strong support for the condition dependence hypothesis, selection in a fluctuating environment can alter sex-specific patterns of genetic and environmental canalization, indicating support for the adaptive canalization hypothesis.

It is therefore highly important to consider the selective past of a population when studying the evolution of sexual dimorphism in general and testing the hypothesis in particular. Knowledge of the selective past, the predominant type of selection acting on the trait under observation and its importance for fitness in each of the sexes are essential to draw meaningful conclusions. And only if all these parameters are known, one can compare results between studies, populations and/or species.

Conclusion

Our results highlight that environmental variation in terms of temperature differences can create variation in sexual size dimorphism with sexes differing in the degree of canalization against environmental changes. Although we found support for the condition dependence hypothesis for body mass at emergence, our results suggest that selection in a fluctuating environment can lead to increased stabilizing selection on traits that are most important for fitness in the respective sex, supporting the adaptive canalization hypothesis. We emphasize that the selective past of a population and especially fluctuations in the environment are of general importance when studying sexual dimorphism. In this study, we show that fluctuating environments can have major consequences for sex-specific responses to selection, environmental sensitivities and plasticity. Fluctuating environments can lead to a sex-specific higher potential to respond to selection and trait-dependent changes in patterns of sex-specific environmental sensitivity, and finally they facilitate an independent response of the sexes to selection, due to low genetic correlation between the sexes. Thus, future investigations should consider the selective past of a population when studying sex-specific responses to environmental variation.

Data deposited at Dryad: doi: 10.5061/dryad.j3459

Ancillary