SEARCH

SEARCH BY CITATION

Keywords:

  • genotype-by-environment variance;
  • phenotypic plasticity;
  • selection

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Phenotypic plasticity is an important strategy for coping with changing environments. However, environmental change usually results in strong directional selection, and little is known empirically about how this affects plasticity. If genes affecting a trait value also affect its plasticity, selection on the trait should influence plasticity. Synthetic outbred populations of Arabidopsis thaliana were selected for earlier flowering under simulated spring- and winter-annual conditions to investigate the correlated response of flowering time plasticity and its effect on family-by-environment variance (Vg×e) within each selected line. We found that selection affected plasticity in an environmentally dependent manner: under simulated spring-annual conditions, selection increased the magnitude of plastic response but decreased Vg×e; selection under simulated winter-annual conditions reduced the magnitude of plastic response but did not alter Vg×e significantly. As selection may constrain future response to environmental change, the environment for crop breeding and ex situ conservation programmes should be carefully chosen. Models of species persistence under environmental change should also consider the interaction between selection and plasticity.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Phenotypic plasticity occurs when a single genotype expresses alternative phenotypes depending on environmental cues, and it can be an important factor allowing populations to colonize and persist in new and changing environments (Price et al., 2003; Yeh & Price, 2004; Ghalambor et al., 2007; Crispo, 2008). In addition, plasticity may also intensify or attenuate evolved responses in trait value, depending on the proximity of the plastic response to the phenotypic optimum (Ghalambor et al., 2007). Recent concerns about the ability of species to persist and adapt to new environmental conditions have renewed interest in the complex relationship between adaptive changes in trait mean and plastic responses and their consequences for population mean fitness (Charmantier et al., 2008; Gienapp, 2008; Chevin & Lande, 2010). However, it remains unclear whether the plasticity of a trait can evolve independently of the trait mean or whether they constrain each other’s evolution (Pigliucci, 2005; Crispo, 2008; Auld et al., 2010).

The relationship between the value of a trait in one environment and its plasticity has been the subject of considerable debate (Via & Lande, 1985; Schlichting, 1986; Via et al., 1995; Pigliucci, 2005). Via & Lande (1985) proposed that plasticity is a function of the differential expression of the same genes in different environments. Under this model, the plasticity of a trait can have a strongly correlated response to direct selection on that trait (as long as the cross-environment genetic correlation does not equal 1). Alternatively, the plasticity of a trait might be under separate regulatory control from the trait itself (Schlichting & Pigliucci, 1995). Under the second model, direct selection on a trait should have little effect on the plasticity of that trait. Although these two mechanisms are not mutually exclusive, it is important to determine empirically whether correlated response to directional selection involves change in plasticity. This will enable us to ascertain whether the current response of a trait to selection may constrain future evolution by reducing or increasing plasticity.

The most powerful approach to evaluate the evolutionary independence of trait value and its plasticity is the use of selection experiments (See Scheiner, 2002; Callahan, 2005; Garland & Kelly, 2006 for extensive reviews on this topic). This approach allows for the investigation of how plasticity evolves as a correlated trait when selection is controlled to target the trait mean. Although artificial selection experiments can never replicate field conditions exactly, and thus directly inform us about the evolution of traits in nature, they are ideal to understand the genetic relationship between the mean and plasticity. Despite their advantages, selection studies of plasticity in higher plants are relatively rare due to their large size and relatively long generation times (but see Mather & Jinks, 1982 and Falconer, 1990). The use of the model organism Arabidopsis thaliana allows many plants to be grown in a short period of time. Previous studies using this species have mainly performed ‘line sorting’ because A. thaliana reproduces primarily by self-fertilization. Here, we take advantage of a set of lines selected for earlier flowering derived from a synthetic outbred population of A. thaliana and hand-pollinated after selection (Scarcelli & Kover, 2009) to investigate the correlated response in plasticity. These lines have been subject to five generations of artificial selection under two different growth conditions, causing a reduction in mean flowering time (FT) of approximately two standard deviations.

Arabidopsis thaliana is a weedy, colonizing plant species in the Brassicaceae family with a wide geographical distribution (Redei, 1970). Accordingly, extensive natural variation has been observed for life-history traits, such as FT (Napp-Zinnn, 1985; Alonso-Blanco & Koornneef, 2000; Korves et al., 2007). Natural accessions of A. thaliana have long been broadly categorized into ‘fast-cyclers’ and ‘long-flowerers’ (Napp-Zinnn, 1985). Long-flowerers are identified by late flowering in growth chambers and/or the requirement of cold spells (vernalization) before flowering. These have been thought to exhibit a ‘winter-annual’ life history whereby they germinate in the autumn, overwinter as rosettes and flower early in spring. Fast-cyclers flower early in the growth chamber and have been thought to correspond to the spring-annual habit, where both germination and flowering occur in the spring of the same year. Whereas under field conditions, there is much more continuous variation in FT (Wilczek et al., 2009), simulated spring- and winter-annual treatments can elicit very different flowering strategies (Kover et al., 2009).

Flowering time has been shown to be under strong selection under field conditions in A. thaliana (Korves et al., 2007) and related crop species (Franks et al., 2007). Plasticity in FT is thought to be an important determinant of the ability of many species to respond to changing climates (Metcalf et al., 2003; Nicotra et al., 2010). Accordingly, the genetic basis of FT has been extensively studied in many plants and, in particular, in A. thaliana, where more than 70 flowering genes have been characterized in pathways controlled by light, temperature and circadian clock genes (reviewed by Putterill et al., 2004). It seems plausible, given the large number of genes and environmental cues involved in determining FT, that the mean FT could evolve independently from its plastic response, because they may be regulated by different environmental sensing pathways. We have previously shown that selection for earlier flowering under ‘winter-annual’- and ‘spring-annual’-simulated growth conditions is mediated by different changes at the genetic level (Scarcelli & Kover, 2009). However, the genes selected in each environment had a positive correlated effect in the other environment (Kover et al., 2009). Evidence for differential genetic control of FT under different environmental conditions has also been suggested by Ungerer et al. (2003), by showing that different QTLs are identified when A. thaliana plants are grown under long and short day length.

Here, we investigate whether directional selection for early flowering affects the magnitude of and variation in FT plasticity to simulated spring- or winter-annual conditions and whether this effect is environmentally dependent. We also test whether there is evidence for plasticity costs that could constrain response to selection.

Materials and methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Experimental design

The synthetic outbred population used in this study is described in detail by Scarcelli et al. (2007). Briefly, 19 natural accessions of A. thaliana from a wide geographical distribution were randomly intermated for five generations. In the first generation, six replicated founder populations were cultivated in growth chambers running either a simulated ‘winter-annual’ (winter treatment) or a ‘spring-annual’ programme (spring treatment). All six populations were initiated using full siblings from the same 200 families randomly chosen from the outbred population. Data from this first generation provide the estimates of the relationship between mean and plasticity before selection. From each basal population, a selected and a control line were derived, producing six lines selected for early flowering and six control lines. Every generation, plants were checked daily for germination and then for the appearance of floral buds (bolting), and FT was calculated as the number of days between germination and bolting. The selected lines were produced by randomly crossing the 50 earliest flowering plants after assigning them to 25 hand-pollinated crosses. Crosses were performed in floral buds emasculated prior to pollen maturation to avoid any seeds being the result of self-fertilization. Control lines were produced similarly by crossing 50 randomly selected plants per generation. This protocol was repeated for five generations.

The ‘spring treatment’ simulates conditions experienced in spring-annual life histories with transitions in temperature and day length reflecting changes from spring into summer (14 °C:10 °C day:night and 8 h:16 h light:dark, transitioning to 21 °C:18 °C day:night and 16 h:8 h light : dark). The ‘winter treatment’ simulates the light and temperature transitions expected in winter-annual life histories with the programme starting with autumn–winter transitions (16 °C: 10 °C day:night and 8 h:16 h light: dark transitioning to 4 °C:4 °C day: night and 6 h:18 h light:dark), followed by the same programme as in the spring-annual treatment. This temporally varying environment was designed to give a range of environmental cues akin to what A. thaliana may experience in the field. This is preferable to the uniform conditions used in many growth chamber experiments, although it is recognized that plants in natural conditions still experience a far more complex environment.

At the end of the selection protocol, the 12 lines (three control and three selected bred in both spring and winter treatments) were planted in a ‘reciprocal transplant’ fashion: three full sibs from each of the 25 crosses performed in the fifth generation of selection were grown under each treatment (1200 plants in each) (for further details, see Kover et al., 2009). After the plants had senesced, the total number of fruits produced by each plant was counted, providing an estimate of fitness (fruit number and seed set have been shown to be highly correlated in A. thaliana (Mauricio & Rausher, 1997; Westerman & Lawrence, 1970).

Data analysis

This study considers both changes in the magnitude of the populations’ mean plasticity and changes in genotype × environment variation (Vg×e). Although they both tell us something about the response of the populations to different environments, they measure distinct aspects of plasticity. Mean plasticity shows the magnitude of the average response of the population of genotypes to these specific environments at a point in evolutionary time, whereas Vg×e indicates the range of possible plastic responses. Plasticity is most commonly estimated as the differences in mean for a given genotype when grown in different environments and was here calculated as the family mean FT in winter treatment minus mean FT in spring treatment.

To determine the relationship between FT and FT plasticity in the population prior to selection, we calculated the Pearson’s rank correlation coefficient and mean FT for the 200 full-sib families in spring- and winter-annual treatments. To determine the existence of genotype-by-environment interaction prior to selection, we used the linear model: FT = family + environment + family × environment. Family-by-environment interactions can arise from either changes in genetic variance across environments or a change in the rank order of genotypes across environments (crossing of reaction norms). We used the equations from Cockerham (1963, p.88) to identify the percentage of Vg×e that can be attributed to crossing following the method of Johnson (2007) for applying to a mixed-effect model.

To determine whether selection on trait values affected the plasticity in FT, we used the following linear model: plasticity = breeding environment + selection treatment + breeding environment × selection treatment + line (breeding environment × selection treatment). The same model factors were used to determine the effect of selection treatment on cross-environment mean FTs (mean of spring and winter treatments). As a post hoc test, to directly compare the mean plasticity and cross-environment mean for control and selected lines separately for each breeding environment, we used exact permutation tests (Ludbrook & Dudley, 1998). Observed differences between mean plasticity in control and selected treatments were compared with the distribution of differences obtained by randomizing all observed FT values in the six populations within each environmental condition 46 656 times (216 possible bootstrap permutations in first group × 216 possible bootstrap permutations in second group) into two groups of three populations (the maximum number of possible independent permutations) and calculating their plasticity.

To determine the effect of selection for early flowering on the genetic variance for plasticity, variance components for familyxenvironment interactions (Vg×e) and total phenotypic variance (Vp) in each of the 12 populations were calculated using the R package lme4 (Bates & Maechler, 2009). The following model was used: FT = family + environment + family × environment, with environment set as a fixed effect and family as a random effect. Confidence limits (± 95%) for each population’s Vp and Vg×e variance components were constructed from the 2.5 and 97.5 percentiles of 1000 parametric bootstraps. Differences in the variance components Vp and Vg×e for control and selected lines were tested separately in the spring and winter treatments using exact permutation tests. We compared the observed differences in values with a null distribution obtained by randomizing the variance components observed in the six populations into all possible permutations of two groups of three populations.

To determine the costs of FT plasticity, we estimated selection gradients (Lande & Arnold, 1983) as modified by Weinig et al. (2006) where the mean fruit count (fitness) of a family in a given environment is regressed onto the family mean FT and FT plasticity. The fitness analysis was performed only for the control lines, because strong artificial selection for early flowering could have affected the relationship between plasticity and fitness. Costs of plasticity are indicated by negative partial regression coefficients in the selection model (i.e. more plastic genotypes have lower fitness), whereas positive partial regression coefficients indicate that plasticity is adaptive or that homoeostasis (maintaining the same phenotype independent of the environment) is costly, i.e. more plastic genotypes have higher fitness. We estimate both the ‘local costs’ (i.e. detected in only one environment) and the ‘global costs’ (when they are detected in both treatments) (Sultan & Spencer, 2002).

All data analyses were performed using R (R Development Core Team 2009).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Trait mean and plasticity prior to selection

An anova for FT for the 200 full-sib families in the basal outcrossed population revealed a highly significant G×E interaction (F187,653 = 1.8, P = 5.7 × 10−8). A plot of the reaction norms for the 200 families showed both nonparallelism and crossing, despite a much larger effect of treatment (Fig. 1); 48.3% of the G×E can be attributed to crossing of reaction norms, indicating that environmental changes can lead to changes in adaptive value of these genotypes.

image

Figure 1.  Reaction norms for flowering times under the spring and winter treatments for the 200 families in the basal population prior to selection.

Download figure to PowerPoint

We observed a strong negative relationship between family mean FT and plasticity under spring conditions (Fig. 2), indicating that families that flower early in the spring treatment are also more plastic. However, the relationship between FT in winter and plasticity is positive and not statistically significant (Fig. 2). Thus, we expect that selection for early flowering under the spring treatment should increase the magnitude of plasticity, whereas selection under the winter treatment should not affect the magnitude of plasticity.

image

Figure 2.  Relationship between phenotypic plasticity and mean flowering time prior to selection in the spring (r = −0.9, t186 = −28.11, P < 2.2 × 10−16) and winter (r = 0.14, t186 = 1.96, P = 0.0517) treatments. Best fit lines based on linear models are provided for better visualization of the trend.

Download figure to PowerPoint

Correlated responses to selection in plasticity

Plasticity and cross-environment mean FT values for selected and control lines are shown in Table 1. The anova for plasticity indicates that breeding environment significantly affected plasticity, with plants grown in the spring treatment being more plastic than plants bred in the winter treatment (F1,215 = 10.92, P = 0.001). Although plants under selection treatment were overall more plastic than plants in the control treatment, the difference is not significant (F1,215 = 2.96, P = 0.0869). This can be explained by the fact that selection has opposite effects on plasticity depending on the environment in which it was carried out (Fig. 2), as confirmed by the presence of a significant environment by the selection treatment interaction (F1,216 = 56.22, P = 1.5 × 10−12). There is also evidence of heterogeneity among replicated lines, with a significant effect of lines nested within environment and selection treatment (F8,223 = 2.53, P = 0.0118). Exact permutation test shows that among the plants bred under spring conditions, plants in selected lines were significantly more plastic than in the control lines (Table 2). In contrast, plants selected for early flowering in winter showed a significant decrease in plasticity relative to the controls. We also found that plants selected for early flowering in the spring treatment were significantly more plastic than those selected for early flowering in the winter treatment. Although winter control lines were more plastic than spring control lines, the difference was not significant.

Table 1.   Mean and standard error of the mean (SE) for the magnitude of plasticity in flowering time and cross-environment means for each line; estimates for total phenotypic (Vp) and family × environment (Vg×e) variance components are also presented. Confidence intervals for Vp and Vg×e can be found in the Table S1.
Breeding environmentSelection treatmentMean plasticitySECross-environment meanSEVpVg×e
SpringControl33.781.1654.290.9546.369.37
SpringControl33.861.2157.271.0362.537.40
SpringControl31.280.7552.900.5825.452.00
WinterControl37.110.7753.040.7026.923.15
WinterControl35.801.1755.231.0668.141.87
WinterControl32.320.9856.870.8151.278.38
SpringSelected37.940.6545.520.3615.091.75
SpringSelected37.430.4845.170.3510.651.26
SpringSelected38.970.5646.420.2612.411.48
WinterSelected30.191.0747.330.6830.657.42
WinterSelected32.080.4844.570.3011.420.99
WinterSelected32.210.7945.280.4823.946.02
Table 2.   Results of exact permutation tests for differences in magnitude of plasticity and cross-environment mean flowering time between control and selected lines in the spring and winter treatments. P-values < 0.05 are shown in boldface.
 ControlSelectedMean differenceP
Magnitude of plasticity
 Spring32.9838.115.130.0084
 Winter35.0831.49−3.590.0293
 Mean difference−2.106.62  
 P0.09710.0071  
Cross-environment mean
 Spring54.8245.7−9.120.0075
 Winter55.0545.73−9.320.0076
 Mean difference−0.23−0.03  
 P0.43520.4812  

The cross-environment means showed no significant effect of breeding environment (F1,215 = 0.24, P = 0.6227) or of an interaction between breeding environment and selection treatment (F1,215 = 0.002, P = 0.961). However, a significant effect of selection was observed (F1,215 = 405.83, P < 2 × 10−16). Exact randomization tests indicate that selection significantly reduced the cross-environment mean FT by around 9 days, irrespective of whether the selection took place in spring or winter conditions (Table 2).

Effect of selection on G×E variance

Variance components and 95% confidence limits for each line are shown in Tables 1 and S1. As expected, exact permutation tests show that selected lines have reduced total phenotypic variance (Vp) relative to the controls, independent of the environmental conditions under which selection was performed (Table 3). Whereas selection for early flowering in spring conditions significantly reduced Vg×e relative to the control lines (Table 3), plants selected for early flowering in winter conditions showed no significant change in Vg×e.

Table 3.   Results of exact permutation tests for differences in total phenotypic variance (Vp) and family × environment variance (Vg×e) for control and selected lines in the spring and winter treatments. P-values < 0.05 are shown in boldface.
 ControlSelectedMean differenceP
Vp
 Spring44.77912.715−32.060.021
 Winter48.77422.002−26.770.042
 Mean difference−3.9952−9.28695  
 P0.3840.065  
Vg×e
 Spring6.2551.497−4.760.035
 Winter4.4674.8100.340.438
 Mean difference1.7877−3.3128  
 P0.2620.070  

Flowering time plasticity and fruit production (fitness)

Analysis of selection gradients found no significant partial regression coefficients for fitness on the plasticity of the plants in any of the control lines in either spring or winter treatments (Table S2).

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

There has been much interest recently on the importance of phenotypic plasticity as a coping strategy for organisms confronted with changing environments, and the need to include plasticity in evolutionary models of species persistence under climate change has been highlighted (Chevin & Lande, 2010; Chevin et al., 2010; Nicotra et al., 2010; Reed et al., 2010). Maintenance of plasticity is generally considered as positive, and its maintenance has been mainly evaluated against possible contemporary fitness costs (e.g. van Tienderen, 1997; DeWitt et al., 1998; Relyea, 2002; Ernande & Dieckmann, 2004). However, changes in environment usually lead to strong directional selection, and little is known about how selection on trait value affects plasticity. We find that selection for earlier FT caused a correlated response on plasticity and that the direction and magnitude of this response depended on the growth conditions under which selection occurred. The direction of the observed effects is in agreement with trait correlation prior to selection; however, the observed correlated response in winter was not expected to be significant. In addition, changes in the magnitude of plasticity were not indicative of effects on Vg×e, which determines future responses to selection on plasticity. The relationship between trait value and plasticity is therefore complex, making it challenging to incorporate into general models. Nevertheless, our results indicate that response to directional selection can compromise plastic responses in the short term and longer term (direct effect on plasticity or on Vg×e, respectively), depending on environmental conditions. Thus, the maintenance of plasticity in populations might be constrained not only by fitness costs, but also by its intrinsic relationship with the trait value.

Ideally, organisms would benefit from modifying trait values to better fit their environmental optima, without compromising their ability to respond plastically to unpredictable contemporary changes in environmental conditions. Thus, a genetic system where trait value and plasticity were independent will better allow adaptation and flexibility. However, the existence of ‘plasticity genes’ has been much criticized in favour of a model where the same genes affect a given trait under different environments, but to different extents (allelic sensitivity). If the genes that contribute to FT variation were largely nonoverlapping in the two environments, selection in one environment would have no effect on the phenotype in the other environment, resulting in changes in plastic responses and cross-environment means (due to changes in the slope of the norm of reaction). Alternatively, if the same genes equally affect FT in two different environments, changes in mean FT would cause similar changes in the other environment and the evolution of FT would not affect the plastic response. Evidence for both types of genetic architecture for FT exists. For example, Ungerer et al. (2003) found different QTL in plants grown under long and short day lengths, whereas Schwartz et al. (2009) found that a single locus mediates flowering responses to photoperiod and temperature cues.

Figure 3 shows that in our study, much of the change in FT due to selection corresponds to changes in the height of the norm of reaction (as opposed to slope), indicating that a large proportion of the genetic changes involved loci with effects in both environments. Accordingly, the cross-environment mean FT is reduced by the same amount by selection in winter and spring treatments.

image

Figure 3.  Effect of selection on population mean reaction norms for flowering time under the spring and winter treatments after six generations of selection or control breeding. Control populations are represented by solid lines, and populations selected for early flowering are represented by dashed lines. Black lines represent populations kept under the spring treatment; grey lines represent populations bred in simulated winter-annual conditions.

Download figure to PowerPoint

Although selection had the same overall effect on FT in both treatments, the slopes of the reaction norm changed in opposite directions in the two growth conditions. This result is partly consistent with the Jinks–Connolly rule (Jinks & Connolly, 1973; Falconer, 1990), which proposes that selection away from the cross-environment mean (for early flowering in spring treatment) results in an increase in plasticity, whereas selection towards the cross-environment mean (for early flowering in the winter treatment) results in a decrease in plasticity. This rule holds because direct responses should always be greater than correlated responses, because genetic correlations tend to be generally smaller than 1. However, this rule does not explain the difference in magnitude in the correlated response in the two treatments (larger in the spring treatment). Czesak et al. (2006) proposed that such asymmetrical correlated response in plasticity results from selection operating on a combination of genes with effects on both environments and genes with effects specific to one environment. Evidence of environment-specific selection in our selected lines was observed at the FRIGIDA locus, where nonfunctional alleles were significantly selected for in the simulated spring-annual conditions, but not in the winter (Scarcelli & Kover, 2009). It is likely that response to selection includes other genes with similar environment-specific effects.

Considering that fine-tuning of FT with environment cues is thought to be a critical life-history trait, it is surprising that we found no evidence for plasticity to increase fitness in this study. However, similar results have been observed often for FT plasticity, and even when costs are detected, they tend to be small (e.g. Dorn et al., 2000; Weinig et al., 2006). Nevertheless, our results only indicate that plasticity does not significantly affect fitness under the growth conditions in our experiment; it is possible that under natural conditions, these genotypes would experience an advantage or a cost.

Dissection of the molecular pathways that regulate FT in Arabidopsis and other plants suggests a complex network that contains ‘receptor genes’ that perceive environmental cues and ‘floral integrators’ genes that actually promote floral induction (Blazquez et al., 2003; Michaels, 2009; Schwartz et al., 2009). The activity of the integrators is thought to be modulated by a signal cascade that starts with the receptors. Such a network suggests that evolution in FT can occur through changes in allele frequency in receptor genes with different perceptions of the environment. Response through such genes would affect FT only in environments with the appropriate cues and would be perceived as evolution in plasticity (changes in the difference in mean flowering between environments). Alternatively, changes in allele frequency in floral integrator genes, which requires different amount of signalling from receptors, can also cause changes in FT, but those would be perceived as independent of the environment. Our results suggest that much of the response to selection for earlier flowering involves core parts of the network, such as floral integrators, which reduce FT s in both environments. It is also possible that the response could have involved receptors for cues that were present in both environments (such as changes in day length). However, some of the genetic factors selected have to be involved in specific responses to cues present in a single environment, because lines selected in one of the environments always flower earlier than the lines selected in the other environment (Kover et al., 2009). Given that the largest change in plasticity is observed in the spring treatment, it is likely that there are more environment-specific changes that occurred in this environment than in the winter treatment.

According to theory, directional selection in a stable environment should reduce variation for plasticity when only a limited number of genotypes can express the selected phenotype. Accordingly, selection for early flowering under both treatments reduced the variance for plasticity (Vg×e). However, the reduction in variation for plasticity in plants grown in winter conditions is smaller and not significant, suggesting that selection in winter should be less of a constraint to future plastic response to changing environments. Selection appears to have little effect on the amount of crossing of reaction norms in the population, suggesting that there will be little effect on the maintenance of the Vg×e remaining in the population following selection. Although Vg×e is reduced by selection, the proportion of the total phenotypic variance due to plasticity is increased. This suggests that although the variation in plasticity is reduced, it may become a more important avenue for future response to selection. A possible explanation for this fact is that the loci with larger effects on flowering (which are usually selected first) are the ones that affect FT similarly in both environment and that future response to selection will be mainly through the loci of smaller and environment-specific effects that affect the slope of the norm of reaction.

Changes in climate have already caused changes in phenology and, in particular, in FT (Fitter & Fitter, 2002; Menzel et al., 2006; Franks et al., 2007). One of the predictions for anthropogenic climate change is that environments will become more extreme and less predictable (Meehl et al., 2000; IPCC 2001), leading to the conclusion that plasticity would be a beneficial trait to cope with the uncertainty. While this expectation makes much intuitive sense, the definition of plasticity needs to be carefully considered in this context. Plasticity is most commonly defined as the difference in trait means across environments; species and populations with steeper norms of reaction are usually considered more plastic (e.g. Balaguer et al., 2001; Burns & Winn, 2006; Funk, 2008). However, it is unlikely that current norms of reaction would be still adaptive after significant climate changes (Visser, 2008). Thus, a population with more variation among norms of reaction (increased Vg×e) might be in a better position to cope with environmental shifts. In this study, we find that the effect of selection on plasticity will depend on which estimate of plasticity is being considered (estimates of magnitude of plastic response or Vg×e). We argue that the latter definition should be favoured (unless the norms of reaction are for fitness) and conclude that selection under some environmental conditions can be detrimental for populations’ ability to cope with environmental uncertainty.

Finally, our results suggest that breeders should consider carefully the environment in which artificial selection is to be carried out to enable the resultant varieties to be robust to unpredicted environmental variation. Equally, conservation programmes ex situ need to consider carefully which environment to keep organisms in to not adversely affect future re-introduction.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Thanks to Ryan George, Tilly Elldridge and Yoland Savriama for technical help and Robert Platt for programming advice. The manuscript was much improved by helpful discussions with Richard Preziosi, Jason Wolf and two anonymous reviewers. DS was supported by a NERC PhD studentship, and PXK was supported by a NERC research grant.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Materials and methods
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Supporting Information

Table S1 Line means and standard errors for the magnitude of plasticity in flowering time and cross-environment means; REML estimates for total phenotypic (Vp) and family × environment (Vg×e) variance components; Bootstrap confidence limits for variance components; percentage of Vg×e that can be attributed to crossing of reaction norms.

Table S2 Partial regression coefficients for fruit number on magnitude of plasticity in control lines. The slope is indicated by the partial regression beta coefficient (b) and the significance by the probability (P).

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer-reviewed and may be re-organized for online delivery, but are not copy-edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

FilenameFormatSizeDescription
JEB_2360_sm_TableS1-2.xls25KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.