SEARCH

SEARCH BY CITATION

Keywords:

  • Brassica rapa (wild turnip);
  • competition;
  • costs of plasticity;
  • phenotypic plasticity;
  • recombinant inbred lines

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • • 
    Phenotypic plasticity, the ability of a genotype to express different phenotypes across environments, is an adaptive strategy expected to evolve in heterogeneous environments. One widely held hypothesis is that the evolutionary benefits of plasticity are reduced by its costs, but when compared with the number of traits tested, the evidence for costs is limited.
  • • 
    Selection gradients were calculated for traits and trait plasticities to test for costs of plasticity to density in a field study using recombinant inbred lines (RILs) of Brassica rapa.
  • • 
    Significant costs of putatively adaptive plasticity were found in three out of six measured traits. For one trait, petiole length, a cost of plasticity was detected in both environments tested; such global costs are expected to more strongly constrain the evolution of plasticity than local costs expressed in a single environment.
  • • 
    These results, in combination with evidence from studies in segregating progenies of Arabidopsis thaliana, suggest that the potential for genetic costs of plasticity exists in natural populations. Detection of costs in previous studies may have been limited because historical selection has purged genotypes with costly plasticity, and experimental conditions often lack environmental stresses.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Phenotypic plasticity, where a single genotype expresses different phenotypes across microsites, is commonly viewed as an adaptive strategy in heterogeneous environments. Many studies have suggested that the evolutionary benefit of plasticity is reduced by its costs (reviewed in DeWitt et al., 1998; van Kleunen & Fischer, 2005), that is, the fitness of a plastic relative to a canalized genotype is lower even though the two genotypes express the same trait value in a given environment (Van Tienderen, 1991; Scheiner & Berrigan, 1998; Dorn et al., 2000). However, the evidence for costs is limited when compared with the number of traits measured (Van Tienderen, 1991; Scheiner & Berrigan, 1998; Donohue et al., 2000; Dorn et al., 2000; van Kleunen et al., 2000; Agrawal et al., 2002; Steinger et al., 2003; Stinchcombe et al., 2004; van Kleunen & Fischer, 2005; Weijschede et al., 2006). One obvious explanation for the empirical rarity of costs is that plasticity has no intrinsic costs, for example, if the same sensory equipment and response mechanisms are used in multiple environments. Alternatively, plasticity may only be costly in specific genetic settings, or in stressful (Stinchcombe et al., 2004) or resource-limited environments (van Kleunen et al., 2000; Steinger et al., 2003). Although plants in natural populations would likely experience stressful or resource-limited conditions during their lifetime, most cost studies have been carried out in controlled environments. Finally, genotypes with costly plasticity may have been culled from natural populations by selection (Agrawal et al., 2002; Weinig et al., 2006).

Most studies testing for plasticity costs have used arrays of genotypes from natural populations, from which selection has had time to decrease or eliminate genetic costs of plasticity (Donohue et al., 2000; Dorn et al., 2000; van Kleunen et al., 2000; Agrawal et al., 2002; Relyea, 2002; Steinger et al., 2003; Weijschede et al., 2006; but see Callahan et al., 2005; Weinig et al., 2006). In particular, selection in natural populations would be expected to favor genotypes in coupling phase between plasticity and vigor loci, that is, genotypes in which alleles for plasticity are associated with neutral or favorable alleles at vigor loci, and to favor the evolution of modifiers that reduce plasticity costs resulting from epistasis and/or pleiotropy (DeWitt et al., 1998; Agrawal, 2001). At least in selfing species, where selection effectively acts among clonal variants, genotypes with plasticity costs might well be purged from the population.

In contrast to natural populations, the creation of experimental segregating progenies, such as recombinant inbred lines (RILs), breaks up genetic associations and amplifies variation among RILs, therefore increasing the likelihood of detecting genetic costs (Callahan et al., 2005; Weinig et al., 2006). In one approach to creating RILs, two inbred homozygous parents are crossed to produce a heterozygous F1 individual, which is in turn mated to itself, and the resulting F2s are propagated by single seed descent for six or more generations (Poelman & Sleper, 1995). The generations of selfing (and attendant recombination) used to create RILs may regenerate genotypes with both coupling and repulsion phase associations between plasticity and vigor loci, therefore increasing the potential to detect costs resulting from linkage. The creation of RILs can also generate novel combinations of alleles at interacting loci, which increases the likelihood of detecting costs resulting from epistasis. In comparison to natural genotypes, plasticity costs resulting from pleiotropy are only more likely in RILs if a modifier that decreases pleiotropy has arisen in one but not both of the parental genotypes, or if different modifiers exist in each genotype. A recent study examining plasticity to competition in RILs of Arabidopsis thaliana found significant costs of putatively adaptive plasticity in three out of six measured traits (Weinig et al., 2006), even though a previous study in the same species detected very few costs of plasticity for similar traits in an array of wild genotypes grown under similar conditions (Dorn et al., 2000). These observations are consistent with the hypothesis that genetic costs of plasticity exist, but may be greatly reduced by selection in natural populations.

Competitive environments lend themselves to plasticity studies, because cues of neighbor proximity are well defined and perception of these cues induces multiple phenotypic responses. Low red to far-red light ratios (R:FR) are highly predictive of above-ground competition, because chlorophyll in neighboring plants selectively absorbs light in the red region of the spectrum while transmitting far-red light (Smith, 1982; Casal & Smith, 1989; Neff et al., 2000). Plants respond to low R:FR by elongating stems and petioles, reducing branching, and accelerating development (Smith, 1982). These responses can increase lifetime light interception in crowded settings, and thereby increase fitness (Schmitt et al., 1995; Dudley & Schmitt, 1996; Weinig, 2000a). In addition to adaptive elongation responses, plants may also exhibit maladaptive or nonadaptive reductions in overall growth in crowded stands, as a result of the lower resource levels and overall environmental quality of competitive relative to noncompetitive settings. Costs are more likely to exist when plasticity is adaptive, because genotypes expressing maladaptive plasticity should be removed by selection, as should genotypes with nonadaptive plasticity, when even small costs to plasticity exist.

In this study, we examine selection on architectural and life-history traits and their plastic responses to density in RILs of Brassica rapa to address the questions: is there selection for plasticity to density in Brapa; and is there evidence for costs of plasticity?

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Study system

Brassica rapa L. is an annual (occasionally biennial) plant native to Europe and Asia (Prakash & Hinata, 1980) that was introduced into the USA and Canada as an oilseed cultivar. Over time, B. rapa plants have escaped cultivation and are frequently found in disturbed habitats as well as in agricultural fields (Holm et al., 1997). Seeds often drop around the maternal plant (J. M. Dechaine, pers. obs.) where, depending on germination timing, they likely experience a range of competitive environments.

A total of 150 RILs of Brapa were used to evaluate potential costs of plasticity to density. Brassica rapa typically occurs as an obligate outcrosser, but genotypes capable of selfing are occasionally obtained. To create the RILs, an inbred rapid cycling B. rapa (IMB211) and inbred annual yellow Sarson seed oil genotype (R500) were crossed. The F1 generation was selfed, and the resulting F2s were advanced by single seed descent to the S5 generation; each RIL is expected to be > 94% homozygous (Poelman & Sleper, 1995). The RILs are self-compatible, and show little evidence of inbreeding depression (F. Luy, pers. comm.). RIL seeds from the S5 generation were obtained from Thomas Osborne at the University of Wisconsin, Madison. To obtain enough seed for the experiment, four replicates of every S5 RIL were matured over two plantings in winter of 2003–2004. RILs were self-pollinated, siliques were collected into paper bags upon maturity, and seeds (S6) from each RIL were pooled for planting in the field.

Planting design

On 29–31 April 2004, the S6 RIL seeds were planted into an agricultural field at the University of Minnesota in Saint Paul, MN, USA, into noncompetitive and intraspecific competition treatments (Fig. 1). In a nested plot design, three plots were planted, each with three noncompetitive and three competitive subplots. Fifty RILs were randomly assigned to each subplot within a treatment, and all 150 lines were included within each plot. In the noncompetitive treatments, four replicates of each RIL were spaced at 20-cm intervals. The noncompetitive conditions were maintained by removing emerging weeds as needed throughout the summer. The competition treatment consisted of three short rows of six plants of a RIL each at 5-cm intervals. The four central plants in the middle row were treated as focal plants, and the remaining plants as neighbors. Each focal plant began the experiment with eight interacting neighbors of the same RIL. This planting design was chosen because some RILs had low seed production and having the same plant serve as both a focal plant and a neighbor provided the most efficient use of seeds. Each RIL × treatment combination was planted once in each plot, for a total of 12 replicate plants in each RIL × treatment combination.

image

Figure 1. The planting design. Three plots identical to the one shown were planted. Each plot was subdivided into three subplots of each treatment. Fifty Brassica rapa recombinant inbred lines (RILs) were randomly assigned to each subplot within a treatment. In the noncompetitive treatment, four replicate plants of a RIL (circles) were planted at 20-cm intervals in a row 20 cm from the adjacent RIL. Noncompetitive conditions were maintained throughout the season by hand-weeding. In the intraspecific competition treatment, three columns of six replicate plants of a RIL were planted at 5-cm intervals and 20 cm away from the adjacent RIL.

Download figure to PowerPoint

Trait measurements

Bolting date was recorded as the day on which flower buds were first differentiated from the apical meristem. At bolting, hypocotyl length and the length of the longest leaf were recorded. One week after bolting, petiole length was measured on the longest leaf on each individual. Flowering date was recorded at the first open flower. By early August, the RILs were beginning to senesce. When there were no flowers remaining on a plant and over 50% of the leaves had senesced, the above-ground portion was collected from the field. Several traits were measured following collection from the field: the height of the ‘apical’ inflorescence (differentiated from the apical meristem), the branch number (number of branches produced from axillary meristems in leaves on the apical inflorescence), and the length of the longest branch were measured upon senescence. Fruit production was counted for an estimate of fitness. Only fruits that were swollen and appeared to contain seeds were counted.

Data analysis

proc glm (SAS, 1999) was used to test the fixed effects of treatment, and the random effects of subplot nested within treatment (subplot(treat)), RIL, and RIL × treatment on phenotypic traits. A significant treatment effect indicates that plants respond plastically to density, while a significant RIL × treatment interaction demonstrates that RILs differ in plasticity. The subplot(treat) term effectively controls for large-scale environmental variation in the field site. To control for microenvironmental variation that might jointly affect the expression of plant traits in the group of four replicate focal plants of a RIL that were planted in a row, a model was used that included group nested within the RIL × treatment interaction (group(RIL × treatment)) in the place of subplot(treat). Using this latter analysis, the RIL × treatment interaction was found to be significant for only one trait, presumably because the analysis is very conservative and greatly reduces power. Almost all genotypes had replicates present in each plot, suggesting that reduced power rather than microsite variation accounted for the reduced significance of the RIL × treatment interaction, that is, most microsites were covered by most RILs. Because we were interested in increasing the likelihood of detecting biologically meaningful differences in plasticity, we present results of the analysis using subplot(treat), with the caveat that microenvironmental variation could somewhat inflate estimates of genetic variation in plasticity. Neighbor survivorship was low for some focal plants in the competition treatment, and focal plants with fewer than three living neighbors were removed from the analyses. Final sample size for most traits was between 1550 and 1800 (out of a possible 3600), with the exception that n = 570 for petiole length, because many leaves were wilting or senescing at the time of the petiole census.

None of the phenotypic traits met the ANOVA assumption of equal variances. Hypocotyl length, apical inflorescence height, number of branches, length of the longest branch, and fruit production were square-root transformed, bolting date and petiole length were log-transformed, and flowering date was transformed by taking the inverse square. All transformations improved homoscedasticity. Petiole length was the only trait for which the results of the significance tests differed between the raw and transformed data; the means and ANOVA results for the untransformed data are therefore presented for all of the traits except for petiole length, where ANOVA results for the log-transformed data are presented. Because competitive responses are often expressed as shifts in the scaling of traits (i.e. size of individual organs relative to the whole plant), length of longest leaf at bolting was included as a covariate in the model for length traits (hypocotyl length, apical inflorescence length, branch length, and petiole length) to control for differences among RILs or treatments that are solely attributable to overall differences in plant size. Although biomass was not assessed in this study, leaf length strongly correlates with plant biomass in a number of species, including A. thaliana (Weinig et al., 2006) and Impatiens capensis (Dudley & Schmitt, 1996).

Costs of plasticity were tested for in all traits that showed significant variation in plasticity, as indicated by a significant RIL × treatment interaction effect. Plasticities for each trait were calculated as the difference in the genotypic means between the two density treatments. It was expected that greater length would be adaptive in the competition treatment but not in the noncompetitive treatment. Therefore, plasticity of hypocotyl length, petiole length, apical inflorescence height, and branch length were calculated by subtracting the genotypic means in the noncompetitive treatment from the genotypic means in the competition treatment. Plasticities for days to flowering and branch number were calculated by subtracting the genotypic means in the competition treatment from those in the noncompetitive treatment, because plants were expected to flower earlier and produce fewer branches in the competition treatment. This approach has the advantage of yielding positive values for putatively adaptive plasticity, facilitating the interpretation of negative selection gradients for plasticity as costs of plasticity (see following paragraph).

To test for plasticity costs, a modified Lande & Arnold (1983) partial regression model was used: inline image where the relative fitness (W) of a RIL was regressed on the trait mean in that environment (X1), the plasticity of that trait (inline image), and the mean of a second trait in that environment (X2). The second trait (X2), leaf length at bolting, was included (for length traits) to explicitly test selection on trait scaling. For these analyses, relative fitness was calculated as mean fruit production of a RIL within a treatment divided by grand mean fruit production. For ease of interpretation, trait means and plasticities were standardized to a mean of 0 and a standard deviation of 1 for selection analyses, but unstandardized values are shown in all figures. Signed plasticity values (as described under the plasticity calculations above) were used, in order to test for costs of adaptive plasticity. In addition to signed plasticity values, the use of high replication and an annual plant (fitness is easy to measure accurately) may increase our ability to detect costs of plasticity (van Kleunen & Fischer, 2005). Selection gradients indicate the direction and magnitude of selection on a trait or trait plasticity within a given environment, and a negative selection gradient for the plasticity estimate can be interpreted as a cost of adaptive plasticity that exists in a given environment independent of the average trait value. Costs of plasticity were also evaluated using unsigned, or absolute, values of plasticity. The same number of traits had costs of plasticity using signed vs unsigned values, and we therefore present analyses of the signed plasticities for ease of interpretation. Regression analyses are invalidated by high trait collinearity, as is possible in the current analysis because of the fact that trait plasticities are mathematical functions of the trait means. However, examination of the covariances showed that no trait was significantly correlated with its plasticity.

The above model can be used not only to evaluate plasticity costs, but also to examine selection on each trait. If the direction or magnitude of selection on the trait mean differs across environments, then plasticity is expected to be advantageous and evolve in preference to canalization. To test if the linear selection gradients on a trait differed significantly between treatments, we performed heterogeneity of slopes tests on traits with significant selection gradients in one or more environments. Relative fitness was regressed on treatment, the trait mean, and the trait mean × treatment interaction. Significant trait × treatment interaction terms signify that the direction or magnitude of selection on a given trait differs across treatments. In sum, this analysis provides one means to test the adaptive significance of plasticity. Alternatively, the adaptive significance can be tested by examining the relationship between plasticity and grand mean fitness over two or more environments. This relationship was evaluated using the regression model inline image, where W is relative fitness over both environments for each RIL, inline image is the plasticity of a trait, and inline image is the plasticity of the trait squared. This second approach has the advantage that quadratic in addition to linear selection can be evaluated.

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Genetic variation for plasticity

Average branch number, branch length, and fruit production differed significantly across the two density treatments, such that plants had fewer and shorter branches and produced fewer fruit in the competitive relative to the noncompetitive treatment (Tables 1, 2). The RIL × treatment interaction was significant for all traits tested except days to bolting (Table 2, Fig. 2), indicating that genetic variation for plasticity exists.

Table 1.  Phenotypic responses to density of Brassica rapa
Main effectsTrait means ± 1 SD
CompetitiveNoncompetitive
  1. Trait means of length traits are in millimeters. Back-transformed means and 95% confidence limits (CLs) are included for petiole length.

Hypocotyl length  8.725 ± 5.030  8.844 ± 4.950
Days to bolting 40.524 ± 1.962 40.817 ± 2.480
Days to flowering 48.971 ± 5.552 49.942 ± 5.948
Petiole length 24.547; CL 23.66–26.12 23.988; CL 22.39–25.12
Length of longest leaf 47.767 ± 19.559 41.005 ± 21.165
Apical inflorescence height249.027 ± 128.200237.756 ± 130.479
Branch length 203.61 ± 115.599227.927 ± 120.719
Number of branches  4.136 ± 1.783  4.883 ± 1.926
Fruit production 19.788 ± 29.069 29.277 ± 48.906
Table 2.  ANOVA summary tables (a, b) for recombinant inbred line (RIL) and treatment effects on phenotypic responses to density of Brassica rapa
(a)
SourceHypocotyl length at boltingDays to boltingDays to floweringPetiole length
d.f.MSFd.f.MSFd.f.MSFd.f.MSF
Subplot(treat)MS1783.1 4.40***1720.45.09***17106.94.27***160.12 4.01***
Error140118.9 14924.0 125525.0 3480.03 
RILMS14462.6 2.51***14412.92.75***14289.52.89***1220.05 0.98
Error14224.9 1434.7 13831.0 940.05 
TreatMS14.1 0.1216.20.82132.70.7110.26 2.59
Error5734.4 517.7 5246.1 240.10 
RIL × treatMS14225.0 1.32******1434.71.1713731.01.24*810.05 1.84***
Error140118.9 14924.0 125525.0 3480.03 
Length of longest leafMS1628.633.28***12.896.6***
Error140118.9       3480.03 
(b)             
SourceApical inflorescence heightBranch lengthNumber of branchesFruit production
d.f.MSFd.f.MSFd.f.MSFd.f.MSF
  • Degrees of freedom of the mean squares and error (d.f.), mean squares (MS), and error mean square (error) are shown for subplot (treat), recombinant inbred line (RIL), treatment (treat), RIL ×  treatment, and length of longest leaf (for length traits). Error mean square and d.f. differ because of slight variation in sample size.

  • *

    , P < 0.05:

  • **

    , P < 0.01:

  • ***

    , P < 0.001.

Subplot(treat)MS173.0 × 104  5.03***173.2 × 1045.28***1711.54.91***175748.25.12***
Error13025920.1 13016052.2 13282.3 14921123.5 
RILMS1443.4 × 104  4.17***1442.9 × 1043.83***14411.53.85***1445610.72.97***
Error1418036.4 1427617.5 1413.0 1431890.7 
TreatMS11.8 × 104  1.5019.3 × 1047.60**133.47.45**11.1 × 1045.00*
Error501.2 × 104 471.2 × 104 514.5 552238.5 
RIL × treatMS1408048.4  1.36**1407628.61.26*1403.01.28*1431891.31.68***
Error13025920.1 13016052.2 13282.3 14921123.5 
Length of longest leafMS12.3 × 106386.2***12.4 × 106399.6***
Error13025920.1 13016052.2       
image

Figure 2. Norms of reaction for traits in noncompetitive and competitive treatments. Each line represents the mean trait values of one Brassica rapa recombinant inbred line (RIL) in each environment. A RIL is considered phenotypically plastic for a trait if it displays a significantly different trait mean across environments. Means of 30 (out of 150) randomly chosen RILs are shown for clarity.

Download figure to PowerPoint

Selection analysis

Selection favored decreased hypocotyl length in the noncompetitive treatment (Table 3), as expected under the functional hypothesis that allocating resources to hypocotyl elongation in a noncompetitive setting would be maladaptive. Selection also favored earlier flowering, increased apical inflorescence height and branch length, and increased branch number in both treatments (Table 3). Heterogeneity of slopes tests indicated that selection to increase apical inflorescence height (P < 0.001) and branch length (P < 0.001) was significantly greater in the competitive than in the noncompetitive treatment, while selection to increase branch number was greater in the noncompetitive treatment (P < 0.001). Quadratic selection was significantly negative for branch number and flowering day plasticities, but the distribution of the data was external to the maximum of the fitness function (data not shown) for branch number, indicating that the significance of the quadratic term derived from decelerating fitness gains rather than true stabilizing selection (Mitchell-Olds & Shaw, 1987). Quadratic selection on plasticity for flowering day appeared to have an internal maximum (γ = –0.005, P < 0.009) (Fig. 3).

Table 3.  Analyses of selection on trait means and plasticities for Brassica rapa plants in both noncompetitive and competitive treatments
Main effectsCoefficient (β′)
NoncompetitiveCompetitive
TraitPlasticityLength of longest leafTraitPlasticityLength of longest leaf
  • Columns 2–4 show partial regression coefficients, or selection gradients, standardized to a mean of 0 and standard deviation of 1 for the noncompetitive treatment. Columns 3–6 show the same for the competitive treatment. Significant selection gradients are highlighted in bold. Negative selection gradients for plasticity indicate a cost of plasticity.

  • *

    , P < 0.05;

  • **

    , P < 0.01;

  • ***

    , P < 0.001.

Hypocotyl length–0.211*0.0481.23***–0.0840.0000.436***
Days to flowering–0.461**0.063–0.241***–0.121
Petiole length0.003–0.283*1.15***0.181–0.262**0.363***
Apical inflorescence height0.404**–0.1190.838***0.493***–0.135**0.085
Branch length0.313*–0.0280.957***0.412***–0.104*0.182**
Number of branches0.954***–0.0710.331***0.159*
image

Figure 3. Nonlinear selection on plasticity for flowering time in Brassica rapa. Means of flowering day plasticity for each recombinant inbred line (RIL) are shown on the x-axis. The y-axis shows model residuals of relative fruit production over both the competitive and noncompetitive environments for each RIL. A best-fit second-order quadratic regression is included.

Download figure to PowerPoint

Costs of plasticity

Costs of adaptive plasticity were found in three out of the six traits analyzed (Table 3). Costs of plasticity were detected for apical inflorescence height and branch length in the competitive environment, and for petiole length in both the noncompetitive and the competitive treatments (Table 3, Fig. 4). In addition, it was observed that plasticity for branch number was favored in the competitive treatment.

image

Figure 4. Cost of plasticity for petiole length of Brassica rapa in the competitive environment. Means of petiole length plasticity for each recombinant inbred line (RIL) are shown on the x-axis. The y-axis shows model residuals of relative fruit production in the competitive environment for each RIL. A best-fit linear regression line is included.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Costs of plasticity

Recombinant inbred lines of B. rapa were grown under noncompetitive and competitive conditions to test for the presence of costs of adaptive plasticity in phenotypic responses to density. Costs were detected in three out of six measured traits; these results are particularly relevant given the paucity of evidence for costs of plasticity. Costs of plasticity to density were observed for apical inflorescence height and branch length in the competitive environment. Moderate (β =–0.262 to –0.283) costs of plasticity for petiole length were observed in both treatments. The costs of plasticity for petiole length represent global costs that are expressed in multiple environments, and such costs are much more likely to constrain the evolution of plasticity (and favor canalization) than those expressed in a single environment (Sultan & Spencer, 2002).

Only a limited number of plasticity costs have been detected in previous studies (Agrawal et al., 2002; Relyea, 2002; van Kleunen & Fischer, 2005), even in competitive environments similar to the ones tested in this study (Donohue et al., 2000; Dorn et al., 2000; van Kleunen et al., 2000; Steinger et al., 2003; Weijschede et al., 2006). The lack of observed plasticity costs may arise from the fact that most previous studies have examined arrays of genotypes from natural populations, whereas we used an experimental segregating population of RILs. A recent study in which RILs of A. thaliana were exposed to variable vernalization conditions found significant costs of plasticity for bolting time (Callahan et al., 2005), while a second study using RILs of Athaliana found significant costs of adaptive plasticity to density in three out of six traits tested (Weinig et al., 2006).

The detection of several costs of plasticity in RILs of both B. rapa and A. thaliana suggests that the potential for plasticity costs is present in natural populations, in that genetic costs removed from natural populations by selection can be experimentally regenerated in RILs (Weinig et al., 2006). It should be noted that costs uncovered during studies with RILs are a minimum estimate of the costs that could occur in natural populations, because only alleles from the two parental genotypes are segregating in a RIL population. However, it is also worth noting that the likelihood of detecting plasticity costs in segregating progenies may be greatest in selfing or highly clonal species, where favorable linkage associations existing in natural genotypes or favorable allelic combinations at interacting loci can be experimentally disrupted. In outcrossing species, genotypes in a maladaptive linkage phase between plasticity and vigor loci would likely persist in natural populations, because they would be regenerated in each generation by recombination unless selection against these genotypes was very strong.

The costs of plasticity observed in the current study may also be attributable to the stressful conditions that the B. rapa RILs experienced in the field. Most studies evaluating costs of plasticity are completed in glasshouse or growth-chamber environments where variables outside of the experimental treatment are close to ideal (e.g. plentiful water and sunlight, and no predators). Plasticity costs are expected to be more easily detected in stressful environments, where plants are likely resource-limited and unable to offset costs (van Kleunen et al., 2000; Steinger et al., 2003; Stinchcombe et al., 2004). In the current study, the experimental plants of B. rapa were grown in a field environment where they experienced cool spring weather, summer temperatures up to 35°C, and insect herbivores. For example, although damage was not quantitatively assessed, many plants were damaged by flea beetles while at the cotyledon and four-leaf stage. Furthermore, the only other study to find significant costs in half the traits tested was completed in a field environment similar to our own (Weinig et al., 2006). These results suggest that more studies of plasticity costs are needed under field conditions.

Selection analyses

Selection is expected to purge genotypes with maladaptive plasticity (i.e. plastic responses in the opposite direction of selection), as well as genotypes with nonadaptive plasticity (i.e. plasticity for a trait not under direct selection) if costs of plasticity exist. Costs of plasticity are therefore most likely when plasticity is adaptive (Levins, 1963; Lively, 1986; Van Tienderen, 1991; Weinig et al., 2006). Several studies, including those that incorporated phenotypic manipulations in B. rapa (Schmitt et al., 1995), have demonstrated that increased stem length is adaptive in crowded stands as a result of the favorable effects on light interception (Schmitt & Wulff, 1993; Dudley & Schmitt, 1996; Ballare & Scopel, 1997; Donohue et al., 2000; Weinig, 2000a; Huber et al., 2004). Stem elongation is maladaptive in uncrowded settings as a result of the carbon cost of allocation to stems, the detrimental effects on structural stability (Schmitt et al., 1995), and increased susceptibility to UV damage (Weinig et al., 2004). Adaptive plasticity, as evidenced by reversals in selection across environments (e.g., Schmitt et al., 1995; Dudley & Schmitt, 1996), is rarely documented outside of studies using phenotypic manipulations, because almost all wild-type genotypes respond plastically to local conditions, thereby reducing the range of expressed phenotypes and the opportunity for selection in any single environment. Although the direction of selection was not found to reverse across environments in the current study, selection to decrease hypocotyl length was observed in the noncompetitive environment. Selection also acted more strongly to increase inflorescence and axillary branch lengths in the competitive relative to the noncompetitive treatment, and acted more strongly to increase branch number in the noncompetitive treatment. The observed patterns of selection are therefore largely consistent with previous studies and functional hypotheses regarding adaptive morphologies in crowded vs uncrowded settings; large plants may do well overall, but height is more important to light capture when plants are crowded, whereas increased branching is a more significant fitness determinant in uncrowded settings. Interestingly, we found stabilizing selection on flowering day plasticity. This result is most likely attributable to stabilizing selection on flowering time, in that RILs with extreme plasticities flowered either very early or very late in the season (Fig. 3).

In addition to the potential effects on the evolution of adaptive strategies, plasticity costs may aid in the canalization of novel traits that ultimately lead to species differences (West-Eberhard, 2005), that is, costs of plasticity have implications for micro- and macro-evolutionary processes. If a population experiences an environmental shift, novel phenotypic variation may be induced in a genotype harboring the genetic potential for plasticity. If the new environment occurs consistently across growing seasons, the induced phenotype may continue to be expressed (and possibly experience positive selection) for many generations without any genetic changes within the population (West-Eberhard, 2005). The induced phenotype may continue as an alternative phenotype that is expressed in some populations within the range of a species, or the phenotype may eventually become canalized in some populations after the occurrence of a genetic change, such as mutation or recombination (West-Eberhard, 2005). Plasticity costs increase the likelihood that the phenotype expressed in the novel environment will become canalized instead of remaining plastic (DeWitt et al., 1998; Agrawal, 2001). The higher the costs of plasticity and the stronger the selection on the new trait, the faster a trait will become canalized and contribute to species differentiation.

The large number and magnitude of plasticity costs detected in studies of RILs in B. rapa and A. thaliana suggest that the potential for costs in natural populations may be more prevalent than previously thought. Costs that were not detectable in natural populations because they have been culled by selection may resurface in only a few generations if a population experiences a rare outcrossing event followed by selfing or if selection pressures on plasticity change. Alternatively, costs of plasticity may be prevalent in natural populations, but may go undetected because most experimental environments are not stressful enough for costs to be expressed. Costs of plasticity, particularly costs that are found in multiple microsites, may speed the canalization of traits; for example, costs observed here may reduce the range of settings that favor the evolution of plastic shade-avoidance responses over canalization. Interestingly, populations of Abutilon theophrasti and Impatiens capensis derived from sites that favor greater height exhibit not only plasticity of stem length, but also greater fixed internode lengths (Dudley & Schmitt, 1995; Weinig, 2000b). Although plasticity of stem length to crowding contributes more strongly to overall height than do fixed differences in length, plasticity costs may help to account for the partial canalization of this trait.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The authors thank M. McClellan of the UMN Agricultural Experiment Station for assistance in preparing the field, as well as L. Demink, Z. German, C. Willis, A. Hansen, and B. Meyer for extensive help in managing the field and processing plants on collection. We are also grateful to the National Science Foundation for grants (0227103 and 20091702 to CW) that support work in our laboratory.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • Agrawal AA. 2001. Phenotypic plasticity in the interactions and evolution of species. Ecology 294: 321326.
  • Agrawal AA, Conner JK, Johnson MTJ, Wallsgrove R. 2002. Ecological genetics of an induced plant defense against herbivores: Additive genetic variance and costs of phenotypic plasticity. Evolution 56: 22062213.
  • Ballare CL, Scopel AL. 1997. Phytochrome signalling in plant canopies: Testing its population-level implications with photoreceptor mutants of Arabidopsis. Functional Ecology 11: 441450.
  • Callahan HS, Dhanoolal N, Ungerer MC. 2005. Plasticity genes and plasticity costs: a new approach using an Arabidopsis recombinant inbred population. New Phytologist 166: 129140.
  • Casal JJ, Smith H. 1989. The function, action, and adaptive significance of phytochrome in light-grown plants. Plant, Cell & Environment 12: 855862.
  • DeWitt TJ, Sih A, Wilson DS. 1998. Costs and limits of phenotypic plasticity. Trends in Ecology and Evolution 13: 7781.
  • Donohue K, Messiqua D, Pyle EH, Heschel MS, Schmitt J. 2000. Evidence of adaptive divergence in plasticity: Density- and site-dependent selection on shade-avoidance responses in Impatiens capensis. Evolution 54: 19561968.
  • Dorn LA, Pyle EH, Schmitt J. 2000. Plasticity to light cues and resources in Arabidopsis thaliana: Testing for adaptive value and costs. Evolution 54: 19821994.
  • Dudley SA, Schmitt J. 1995. Genetic differentiation in morphological responses to simulated foliage shade between populations of Impatiens capensis from open and woodland sites. Functional Ecology 9: 655666.
  • Dudley SA, Schmitt J. 1996. Testing the adaptive plasticity hypothesis: Density-dependent selection on manipulated stem length in Impatiens capensis. American Naturalist 147: 445465.
  • Holm L, Doll J, Holm E, Pancho J, Herberger J.1997. Brassica campestris L. In: World weeds: natural histories and distribution. New York, NY, USA: Wiley, 117124.
  • Huber H, Kane NC, Heschel MS, Von Wettberg EJ, Banta J, Leuck AM, Schmitt J. 2004. Frequency and micro-environmental pattern of selection on plastic shade-avoidance traits in a natural population of Impatiens capensis. American Naturalist 163: 548563.
  • Van Kleunen M, Fischer M. 2005. Constraints on the evolution of adaptive phenotypic plasticity in plants. New Phytologist 166: 4960.
  • Van Kleunen M, Fischer M, Schmid B. 2000. Costs of plasticity in foraging characteristics of a clonal plant Ranunculus reptans. Evolution 54: 19471955.
  • Lande R, Arnold SJ. 1983. The measurement of selection on correlated characters. Evolution 37: 12101226.
  • Levins R. 1963. Theory of fitness in a heterogeneous environment. I. Developmental flexibility and niche selection. American Naturalist 97: 7590.
  • Lively CM. 1986. Competition, comparative life histories, and maintenance of shell dimorphism in a barnacle. Ecology 67: 858864.
  • Mitchell-Olds T, Shaw R. 1987. Regression analysis of natural selection: statistical inference and biological interpretation. Evolution 41: 11491161.
  • Neff MM, Fankhauser C, Chory J. 2000. Light: An indicator of time and place. Genes and Development 14: 257271.
  • Poelman JM, Sleper DA. 1995. Breeding field crops. Ames, IA, USA: Iowa University Press.
  • Prakash S, Hinata K. 1980. Taxonomy, cytogenetics and origin of crop brassicas. Opera Botanica 55: 357.
  • Relyea RA. 2002. Costs of phenotypic plasticity. American Naturalist 159: 272282.
  • SAS. 1999. SAS/Stat user's guide. Cary, NC, USA: SAS Institute.
  • Scheiner SM, Berrigan D. 1998. The genetics of phenotypic plasticity. VIII. The cost of plasticity in Daphnia pulex. Evolution 52: 368378.
  • Schmitt J, McCormac AC, Smith H. 1995. A test of the adaptive plasticity hypothesis using transgenic and mutant plants disabled in the phytochrome-mediated elongation responses to neighbors. American Naturalist 146: 937953.
  • Schmitt J, Wulff RD. 1993. Light spectral quality, phytochrome, and plant competition. Trends in Ecology and Evolution 8: 4751.
  • Smith H. 1982. Light quality, photoperception, and plant strategy. Annual Review of Plant Physiology 33: 481518.
  • Steinger T, Roy BA, Stanton ML. 2003. Evolution in stressful environments. II. Adaptive value and costs of plasticity in response to low light in Sinapis arvensis. Journal of Evolutionary Biology 16: 313323.
  • Stinchcombe JR, Dorn LA, Schmitt J. 2004. Flowering time plasticity in Arabidopsis thaliana: a reanalysis of Westerman and Lawrence (1970). Journal of Evolutionary Biology 17: 197207.
  • Sultan SE, Spencer HG. 2002. Metapopulation structure favors plasticity over local adaptation. American Naturalist 160: 271283.
  • Van Tienderen PH. 1991. Evolution of generalists and specialists in spatially heterogeneous environments. Evolution 45: 13171331.
  • Weijschede J, Marinkova J, De Kroon H, Huber H. 2006. Shade avoidance in Trifolium repens: Costs and benefits of plasticity in petiole length and leaf size. New Phytologist 172: 655666.
  • Weinig C. 2000a. Differing selection in alternative competitive environments: Shade-avoidance responses and germination timing. Evolution 54: 124136.
  • Weinig C. 2000b. Plasticity versus canalization: Population differences in the timing of shade-avoidance responses. Evolution 54: 441451.
  • Weinig C, Gravuer KA, Kane NC, Schmitt J. 2004. Testing adaptive plasticity to UV: costs and benefits of stem elongation and light-induced phenolics. Evolution 58: 26452656.
  • Weinig CW, Johnston J, Demink LM, German ZM. 2006. Local and global costs of adaptive plasticity to density in Arabidopsis thaliana. American Naturalist 167: 826836.
  • West-Eberhard MJ. 2005. Developmental plasticity and the origin of species differences. Proceedings of the National Academy of Sciences, USA 102: 65436549.