Growth form evolution and shifting habitat specialization in annual plants

Authors


S. P. Bonser, School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney 2052, Australia.
Tel.: +61 02 9385 3863; fax: +61 02 9385 1158;
e-mail: s.bonser@unsw.edu.au

Abstract

Optimal plant growth form should vary across environments. We examined the potential for mutations causing large changes in growth form to produce new optimal phenotypes across light environments. We predicted that the upright growth form would be favoured in a light limiting environment as leaves were in a position to maximize light interception, while a rosette (leaves in a basal position) growth form would be favoured in a high light environment. Growth form genotypes of Brassica rapa (upright wild-type and rosette mutants) and Arabidopsis thaliana (large rosette wild-type and increasingly upright growth form mutants) were grown in a greenhouse in control (ambient) and filtered (low) light treatments. Compared to upright genotypes, rosette genotypes had relatively high fitness in control light but had a relatively large fitness reduction in filtered light. Our results demonstrate the potential importance of rapid growth form evolution in plant adaptation to new or changing environments.

Introduction

Variability in plant growth form is central to our understanding of plant adaptation and diversity. Key growth form innovations such as the evolution of nonwoody (herbaceous) stem tissue are correlated with diversification rates across flowering plant families (Eriksson & Bremer, 1992; Ricklefs & Renner, 1994,2000; Dodd et al., 1999). Variability in growth form can significantly contribute to differences in competitive success across environments (e.g. Geber, 1989; Weiner & Thomas, 1992; Tremmel & Bazzaz, 1995). In addition, plant growth form is correlated to life history functions such as allocation to reproduction and has been used to identify plant strategies across environments (Bonser & Aarssen, 1996,2003).

Small genetic changes can produce large changes in plant growth form. For example, growth form polymorphisms within species are often the result of one or two mutations (Gottlieb, 1984). Mutations in key regulatory genes often precipitate dramatic differences in plant structure (e.g. Purugganan, 1998; Barrier et al., 2001; Fishman et al., 2002; Remington & Purugganan, 2002,2003). In addition, closely related species can diverge radically in growth form (Mes & Hart, 1996; Möller & Cronk, 2001; Fishman et al., 2002; Verboom et al., 2004). The range of growth forms found within many taxa suggests that plant growth form is evolutionarily flexible (sensuRicklefs & Renner, 1994) and can occur rapidly. Since growth form can influence competitive success and life history (Tremmel & Bazzaz, 1995; Valladares & Pearcy, 1998) growth form flexibility should be ecologically significant as the evolution of novel forms could promote the exploration of variable ecological landscapes (Thompson, 1998; Bone & Farres, 2001).

Plant growth form responses to contrasting light environments have become a model system for studying plant adaptation (Dudley & Schmitt, 1996; Pigliucci & Schmitt, 1999; Dorn et al., 2000). Light filtered through a canopy of neighbouring plants reduces both light intensity and spectral quality (through reducing the ratio of red to far red light). These neighbour effects tend to induce a suite of responses in growth form that maximizes height extension and places leaves in a position where they can best intercept available light (e.g. Schmitt & Wulff, 1993; Dudley & Schmitt, 1996). These well-documented plastic responses in growth form demonstrate that integrated traits within the plant body plan can allow for shifts in the placement of modules, and these changes are critical to fitness maximization across variable light environments. However, there may also be limits on adaptive plasticity as well as costs associated with plasticity (van Tienderen, 1990; Givnish, 2002). The combination of frequent growth form mutations and limits to adaptive plasticity could favour ecologically specialized growth form genotypes occupying habitats of variable light quality.

In this study, we examined effects of light environment on growth form in two short lived annual species Brassica rapa L. and Arabidopsis thaliana Heynh (Brassicaceae), henceforth referred to as their generic name only. Growth form mutants in Brassica shift the primary leaf placement from a cauline position (on an upright stem) to a basal position as a rosette. Growth form mutants in Arabidopsis shift the relative leaf placement from a basal rosette to cauline stem leaves. We test the following predictions: (1) Genotypes of both upright and rosette growth forms will display plastic responses consistent with a shade avoidance response to changing spectral light quality. However, this plasticity will be limited in that growth form genotypes will not converge on a single optimal phenotype in either environment. (2) If integration of traits within environments is high and results in correlations between traits advantageous in a given environment (e.g. in shade) and traits that are disadvantageous in the same environment, integration will constrain rapid adaptive evolution to that environment (e.g. Pigliucci, 2003; Merilä & Björklund, 2004). On the other hand, if integration between traits is minimal or produces a pattern of trait covariation already seen in plastic shade avoidance responses, then integration should not inhibit the evolution of new growth form genotypes. (3) Growth form genotypes will differ in relative fitness across environments. The basal leaf position of rosette genotypes should result in relatively low fitness in low light treatments, as the primary leaves on these plants will not be displayed in a position to maximize light interception. In contrast, upright genotypes should be relatively more fit in low light treatments, but the resource cost of placing leaves in an elevated position should result in relatively low fitness in the high light treatment. Fitness differences between growth forms across light environments demonstrate that rapid growth form evolution can shift ecological specialization of new growth form genotypes and should facilitate adaptation in heterogeneous environments.

Materials and methods

Experiment 1

We examined growth form and fitness in full-sib families of Brassica. Two growth forms were used in this experiment: Wild-type (upright growth form), and Rosette mutants (rosette growth form). Wild-type Brassica genotypes produce leaves supported on an upright stem. Flowers are borne in terminal bractless elongating racemes (inflorescences). Rosette mutant genotypes are homozygous for the recessive rosette (ros) allele and mutants are gibberellic acid deficient. In the absence of gibberellic acid, internode extension is suppressed and leaves are displayed at the basal position as a rosette (Rood et al., 1989). Seeds for Brassica growth form genotypes were obtained from the Crucifer Genetics Cooperative (University of Wisconsin, Madison, WI, USA).

Generation of experimental seed

To minimize maternal environmental effects on growth form, a parental generation was grown in the greenhouse to produce seed for the experiment. Seeds were germinated on moist filtre paper in sterilized Petri dishes. Since gibberellic acid is critical in seed germination, seeds of mutants were soaked in a solution of gibberellic acid (100 parts per million) for 20 min and subsequently washed with distilled water prior to germination. Addition of gibberellic acid to the seeds will have a primary activation of the germination processes in the endosperm and should not have a significant or a lasting effect on the developing embryo. Any unintended residual effects of the gibberellic acid treatment on the emerging seedling would tend to make the rosette mutants more similar to the wild-type phenotype, and interpreting differences between growth forms would be more conservative.

Plants were transplanted individually into pots (3.8 cm diameter, 21 cm deep ‘cone-tainers’ placed in racks 18 cm tall, Stuewe and Sons Inc, Corvallis, OR, USA) filled Metro Mix (The Scott's Company, Marysville, Ont., Canada) at the emergence of both cotyledons (2–3 days old). Racks were placed in a greenhouse with supplemental light where photosynthetically active radiation ranged from 250 to 350 μmol m−2 s−1. Plants were watered daily and fertilized bi-weekly with 20-20-20 N-P-K nutrient solution. Individuals were staked for support, where required.

The parent generation was initiated from 30 plants of each growth form genotype (upright and rosette) sorted into 15 pairs. Dehiscent anthers from each plant were used to conduct reciprocal crosses within each pair, producing full-sib seeds on each maternal plant. One plant from 10 of the 15 pairs was selected at random for the experiment, establishing 10 family lines for both upright and rosette growth forms. Seeds (30–50) each of the selected maternal plants were germinated and transplanted individually into cone-tainers (as described above) for the experiment.

Experimental design

Racks of cone-tainers were placed in a greenhouse and were assigned to one of two light treatments. Control plants received full light available in the greenhouse. Plants assigned to the light filter treatment were grown inside a cylinder (27 cm tall) of green plastic (Lee Filters, Andover, UK; number 121 Lee green) open at the top. These cylinders simulated the effect of light being filtered through a canopy of competing plants by reducing the photosynthetic flux density by 35% and reduced the red : far-red light ratio from 1.0 to 0.20.

Experimental units (i.e. blocks) consisted of one individual of each family line from upright and rosette growth forms in both control and filtered treatments. The positions of treatment and family were randomized within the block. The experiment consisted of 10 fully replicated blocks for a total of 400 plants (i.e. two growth forms × 10 family lines per growth form × 2 light treatments × 10 blocks).

We measured the following architectural, morphological, and fitness traits: Total biomass, plant height, an estimate of leaf area, leaf relative growth rate, and total fruit production. Leaf area was measured at the age of plant maturity (i.e. the age of first reproduction), and was an estimate based on the length and width of each leaf on the main stem. Leaf relative growth rate was measured as the average daily increase in leaf area for plants measured as juveniles (2 weeks old) and adults (age at first reproduction). All other traits were measured at final development.

Experiment 2

A similar experiment was conducted on a range of growth form genotypes of Arabidopsis. Natural populations of Arabidopsis develop a large basal rosette of leaves. Upright flowering stems are initiated (bolting) from the apical and axillary meristems of rosette leaves. Bolting ends the production of new rosette leaves in Arabidopsis. Additional leaves and leaf bracts are supported on the upright stems. Flowers are produced in terminal elongating racemes. Mutant genotypes ranged in form from nearly wild-type in form to phenotypes with small rosettes. All seeds were obtained from the Arabidopsis information resource (http://www.arabidopsis.org). Genotype numbers refer to the designation for each genotype within the seed catalogue.

We examined body plan and fitness traits in plants derived from a natural population (wild-type genotype, en-2 (genotype cs1138)), seeds originally collected from a field edge near Frankfurt, Germany. Nine additional genotypes were examined, eight of which are rosette growth form mutants originally derived from wild-type en-2 plants (genotypes cs316, cs331, cs333, cs379, cs400, cs410, cs446 and cs453). The ninth genotype (cs304) is a growth form mutant derived from the an-1 wild-type (originally collected from Antwerp, Belgium). The an-1 wild-type plants are very similar in size and form to the en-2 wild-type plants. Genotypes were selected to span a range of growth forms from plants with large rosettes and a less upright form (e.g. cs1138 (the wild-type phenotype), cs453), to profoundly reduced rosettes (e.g. cs304, cs316).

Generation of experimental seed

To minimize maternal environmental effects on growth form, a parental generation was grown in the greenhouse to produce seed for the experiment. Arabidopsis is a fully inbreeding species, seeds from a maternal plant are genetically uniform and growth form mutations are maintained across generations. For the parent generation, seeds for each genotype were placed in Petri dishes on moist filter paper, and cold treated (4 °C) for 72 h. Petri dishes were then placed in a controlled environment chamber (16 h light, 8 h dark) to promote seed germination. Seedlings were transplanted to 8 cm diameter pots filled with Metro Mix at the emergence of both cotyledons (2–3 days old). Plants were watered daily and fertilized bi-weekly with 20-20-20 N-P-K nutrient solution.

Experimental design

Seeds from one maternal plant of each genotype were germinated and transplanted into pots following the method of the parent generation. Pots were placed in a greenhouse and were assigned to the same two light treatments as the Brassica experiment (see above). Experimental units (i.e. blocks) consisted of a single individual of each genotype in control and filtered treatments. The positions of growth form and treatment combinations were randomized within the block. The experiment consisted of 10 full replicated blocks for a total of 200 plants (i.e. 10 genotypes × 2 light treatments × 10 blocks).

We measured the following architectural, morphological, and fitness related traits: Biomass, height, number of primary stems, rosette size, stem leaf allocation, number of co-fluorescence branches and total fruit number. Estimates of rosette and stem leaf allocation were measured at the age of first flowering. All other traits were measured at final developmental stage. Reproductive effort was measured as the relative allocation to reproduction (fruit number relative to plant biomass).

Data analysis

Light and growth form effects on phenotypes

Variables were log-transformed, where necessary, to meet the normality assumptions of statistical analyses. For each trait, a mixed model two-factor anova was used to assess the significance of growth form and light treatments. The following factors were examined: Growth form (fixed effect), light treatment (fixed effect), Growth form by Treatment interaction (fixed effect), block (random effect) and (for the Brassica experiment only) Family nested within Growth Form (random effect).

Trait integration

Correlations and covariances among traits were used to assess patterns of trait integration. Pearson product moment correlation coefficients were used to test for significant (P < 0.05) correlations between pairs of traits. P-values were adjusted using sequential Bonferoni corrections (Rice, 1989) to control for multiple comparisons.

For Brassica, trait correlations were calculated from mean trait values for each family line within a growth form (upright and rosette) and light treatment. Patterns of trait correlation were compared between growth forms and light treatments. For Arabidopsis, trait correlations were calculated from mean trait values of the 10 genotypes in each light treatment. The effect of light treatment on integration was assessed by comparing the pattern of trait correlation between the two light treatments. In addition, a Pearson product moment correlation coefficient was used to examine the relationship between rosette size relative reproductive effort (the difference in reproductive effort in control vs. light filtered treatments relative to reproductive effort in the control treatment) in order to test for a relationship between growth form and fitness reduction induced by the light filter treatment.

Two separate analyses were used to test for differences in patterns of correlation (integration within traits) across the two light environments. First, Mantel tests (R-package Version 4, Casgrain & Legendre, 2001) were used to test for significant relationships between pairs of correlation matrices. 1000 randomizations were used to derive potential distributions each pair of matrices. Second, variance of each trait and covariances between each pair of traits were used to construct variance-covariance matrices in upright and rosette genotypes (Brassica only) in control and light filtered treatments. Similarity of variance-covariance matrices was assessed using common principle components (CPC) analysis, using the jump-up test for hierarchical analysis (Phillips & Arnold, 1999; Camera et al., 2000). This analysis compares the structure of two variance-covariance matrices through a hierarchical series of tests for whether the matrices share one or more principle components, whether they share all principle components, whether they share all components but differ in proportion (different Eigen values) or whether they are equal (Phillips & Arnold, 1999).

Results

Growth form evolution and plant responses to light environments

For Brassica, family and positional (block) effects were significant for each trait. In addition, growth form and fitness traits were generally significantly different between growth forms and across light environments (Table 1). The growth form by treatment interactions were significant for height, leaf growth rate, and fruit number, but were not significant for the other traits (Table 1). Upright plants had greater biomass but lower leaf growth rate than rosette plants (Fig. 1). Plants in filtered light had lower biomass and leaf growth rate compared to the control light treatment (Fig. 1). Upright plants were taller than rosette plants in both light treatments. However, the light filter treatment induced an increase in plant height in rosette plants, while decreasing height in upright plants (Fig. 1).

Table 1.  Two-way anova results for each of the growth form and fitness traits measured in growth form genotypes of Brassica across light treatments. Mean squares, demoninator degrees of freedom and P-values are presented. Numerator degrees of freedom for each effect are shown in parentheses.
TraitGrowth form (1)Light treatment (1)Block (9)Family (form) (18)Growth form by treatment (1)Error
Biomass
 MS0.030.400.010.0020.00030.32
 df den19306306306306 
 P0.0014<0.0001<0.00010.0040.58 
Height
 MS4092.90.1864.6266.70750.4312.22
 df den18306306306306 
 P<0.00010.90<0.0001<0.0001<0.0001 
Branch number
 MS54.8910.732.442.570.391.37
 df den19306306306306 
 P0.00020.0050.070.020.59 
Leaf size
 MS3010.227532.1485737.960.580.04
 df den182692692690.44 
 P0.052<0.0001<0.0001<0.0001  
Leaf growth rate
 MS48.031.750.0080.030.0010.01
 df den19262262262262 
 P<0.00010.190.670.00050.73 
Fruit number
 MS31.64218.876.056.736.932.28
 df den18305305305305 
 P0.04<0.00010.006<0.00010.009 
Figure 1.

Mean (±SE) values across family lines of upright (filled circles) and rosette (open circles) growth forms of Brassica in control light and light filtered treatments for (a) plant biomass, (b) leaf area and (c) plant height.

In Arabidopsis, block effects were significant for many traits (Table 2). In addition, genotype, light treatment, and genotype by light treatment interaction were significant for all traits. Allocation to rosette leaves differed by nearly an order of magnitude across the ten genotypes in both light treatments (Fig. 2), relative allocation to basal leaves was greater in large rosette genotypes. The genotype differences define a continuum of growth forms from a large rosette to small rosette, more upright forms. Similarly, geometric and architectural traits such as the number of main stems and plant height varied across genotypes and tended to be greater in genotypes with larger rosettes (Fig. 2). Branch number on main stems differed across genotypes but was not consistently related to rosette size.

Table 2.  Two-way anova results for each of the growth form and fitness traits measured in growth form genotypes of Arabidopsis across light treatments. Mean squares, denominator degrees of freedom and P-values are presented. Numerator degrees of freedom for each effect are in parentheses. Denominator degrees of freedom for each model effect were 165.
TraitGrowth form (9)Light treatment (1)Block (9)Growth form by treatment (9)Error
Log Biomass
 MS2.5924.930.080.131.79
 P<0.0001<0.0001<0.0001<0.0001 
Height
 MS940.0506.427.848.313.4
 P<0.0001<0.00010.040.0004 
Log Rosette size
 MS2.7313.360.090.250.06
 P<0.0001<0.00010.11<0.0001 
Log Basal stem number
 MS0.34.970.020.040.02
 P<0.0001<0.00010.480.03 
Leaf position
 MS1.188.540.390.180.09
 P<0.0001<0.0001<0.00010.05 
Branch number
 MS5504.12702.8672.91273.0292.7
 P<0.0001<0.00010.021<0.0001 
Fruit number
 MS1.6 × 1061.5 × 1071.3 × 1053.0 × 1053.4 × 104
 P<0.0001<0.00010.0002<0.0001 
Figure 2.

Reaction norms for genotypes of Arabidopsis in control light and light filtered treatments for (a) rosette size, (b) plant height, (c) number of main stems and (d) plant biomass. Differences in symbol size indicate the relative rosette size differences across growth form genotypes.

On average, rosettes were significantly smaller in filtered compared to control light. Two genotypes (cs376, cs400) showed a large decline in rosette size in filtered light, while the other genotypes were less sensitive and did not change in relative rosette size. For all genotypes, relative allocation to basal leaves decreased in filtered light. Biomass was significantly reduced in filtered light and the reduction in biomass tended to be more pronounced in the smaller rosette genotypes (Fig. 2). The number of main stems and branches on main stems also decreased in filtered light but the decrease was not related to rosette size. Height was reduced in some genotypes while other genotypes showed no change or even marginal increases in height (Fig. 2).

Trait correlations and phenotypic integration

Several growth form and fitness traits were correlated across family lines within growth forms (Brassica) and across growth form genotypes (Arabidopsis). In upright family lines of Brassica, only height and fruit number were positively correlated in control light (Table 3a); in filtered light, height, biomass and fruit number were all positively correlated (Table 3b). For the rosette growth form in control light, height and fruit number were positively correlated and both were negatively correlated with leaf growth rate (Table 3c). In the light filter treatment, height was again positively correlated with fruit number (Table 3d).

Table 3.  Correlation coefficients (r) for all traits across family lines of Brassica for upright (a, b) and rosette (c, d) families in control (a, c) and light filtered (b, d) treatments. Significant correlations after sequential bonferoni are indicated in bold.
 BiomassHeightBranch numberLeaf sizeLeaf growth rate
(a)
Height
r0.56    
P0.10    
Branch number
r0.260.14   
P0.460.69   
Leaf size
r−0.28−0.44−0.23  
P0.940.190.51  
Leaf growth rate
r0.40−0.48−0.680.18 
P0.610.160.030.61 
Fruit number
r0.480.96−0.13−0.38−0.30
P0.160.00010.720.270.40
(b)
Height
r0.92    
P0.0002    
Branch number
r0.550.46   
P0.100.17   
Leaf size
r0.01−0.17−0.12  
P0.980.640.75  
Leaf growth rate
r0.250.160.23−0.23 
P0.490.650.510.52 
Fruit number
r0.810.940.13−0.140.09
P0.0040.00010.730.700.80
(c)
Height
r0.57    
P0.08    
Branch number
r0.530.71   
P0.110.03   
Leaf size
r−0.15−0.66−0.39  
P0.670.040.27  
Leaf growth rate
r−0.32−0.85−0.650.71 
P0.360.0020.040.02 
Fruit number
r0.550.990.61−0.67−0.83
P0.10<0.00010.060.030.003
(d)
Height
r0.37    
P0.30    
 
Branch number
r−0.11−0.35   
P0.750.32   
Leaf size
r0.19−0.490.35  
P0.590.150.34  
Leaf growth rate
r0.370.120.490.31 
P0.300.960.150.38 
Fruit number
r0.290.97−0.46−0.58−0.05
P0.42<0.00010.180.080.87

Across genotypes of Arabidopsis in control light, rosette size was positively correlated to height and coflorescence branching (i.e. secondary branches produced on the main stems) was positively correlated with fruit number (Table 4a). In filtered light, biomass was correlated with height, rosette size, and number of main stems. Rosette size was also positively correlated with main stem number (Table 4b).

Table 4.  Correlation coefficients (r) for all traits across genotypes of Arabidopsis in a) control light and b) light filtered treatments. Significant correlations after sequential bonferoni are indicated in bold.
 BiomassHeightRosette sizeMain stem numberLeaf positionBranch number
(a)
Height
r0.81     
P0.004     
Rosette size
r0.880.86    
P0.00070.002    
Main stem number
r0.770.680.79   
P0.0090.030.006   
Leaf position
r−0.030.260.29−0.06  
P0.940.470.410.87  
Branch number
r0.660.260.420.61−0.51 
P0.040.470.230.060.14 
Fruit number
r0.740.450.480.63−0.30.86
P0.010.190.160.050.40.002
(b)
Height
r0.83     
P0.003     
Rosette size
r0.920.94    
P0.00020.0001    
Main stem number
r0.940.730.86   
P<0.00010.020.001   
Leaf position
r−0.30−0.23−0.31−0.3  
P0.390.520.380.39  
Branch number
r0.760.650.760.78−0.8 
P0.010.040.010.0080.006 
Fruit number
r0.540.320.450.62−0.650.78
P0.100.370.160.050.040.008

Tests on similarities in patterns of trait integration between growth forms (Brassica) and light treatments (Brassica and Arabidopsis) yielded mixed results. In Brassica, three of the six pair-wise comparisons of correlation matrices were significantly related (Mantel test, P < 0.05): (a) Upright Control (UC) vs. Upright Filter (UF), (b) UC vs. Rosette Control (RC) and (c) UF vs. RC. The remaining three matrices were not significantly related (UC vs. Rosette Filter (RF), UF vs. RF, and RC vs. RF, Mantel tests, n.s.). Only the UF and RF matrices differed significantly from proportionality and UC and UF matrices differed from equality (Fleury Hierarchy Jump Up test, P < 0.05) but not in proportionality.

In Arabidopsis, Correlation matrices in control and filtered light treatments were significantly related (Mantel r = 0.92, P < 0.01). Variance-covariance matrices did not differ from equality (Fluery Hierarchy Jump Up test, n.s.).

Genotype fitness across light environments

In Brassica, fruit number (plant fitness) was significantly greater in the upright growth form than in the rosette growth form in the filtered light treatment (Fig. 3a). Fruit number decreased greatly in the rosette form in filtered relative to control light while fitness in the upright growth form was only marginally reduced (Fig. 3a).

Figure 3.

(a) Mean (±SE) values for fruit number across family lines of upright (filled circles) and rosette (open circles) growth forms of Brassica in control light and light filtered treatments. (b) Reaction norms of fruit number across growth form genotypes of Arabidopsis in control light and light filtered treatments. Differences in symbol size indicate the relative rosette size differences across growth form genotypes.

There was no clear relationship between rosette size and fruit number (fitness) in Arabidopsis. Relatively small rosette genotypes had both the greatest and the lowest fruit production (Fig. 3b). The cs331 genotype is a small rosette genotype that produced more fruits on average than all other genotypes. However, these fruits were small compared to the fruits of the other genotypes. Fruit size differences across the other nine genotypes (anova, P < 0.05) were not related to rosette size. In control light treatment, large rosette genotypes tended to have higher fruit production than small rosette genotypes. In filtered light, fruit production was reduced in all genotypes and the reduction was greater (a steeper slope of the reaction norm) in large rosette genotypes (Fig. 3b). In addition, the crossing of reaction norms demonstrates that relative fitness of growth form genotypes may change across light environments (Fig. 3b). For example, genotype cs333, a small-rosette genotype, had the third lowest fruit production in control light, but the second highest in the light filter treatments. Mean reproductive effort (number of fruit relative to plant biomass) tended to be negatively correlated with rosette size across genotypes both in control (r = –0.55, P = 0.1) and filtered (r = −0.77, P < 0.01) light. Relative reproductive effort (reproductive effort in filtered vs. control light) was negativity correlated with rosette size (r = −0.79, P < 0.01).

Discussion

Plant responses to light environments

Both wild-type and growth form mutant genotypes of Brassica and Arabidopsis displayed highly significant plastic responses in growth form traits. Plants were smaller (lower biomass) in filtered compared to control light. Response to light quality differed between growth form mutants in Brassica. Rosette mutants became taller in filtered light, consistent with the shade avoidance response, while upright family lines decreased in height (Fig. 1). The dramatic reduction in biomass for plants of both growth forms suggests the relative allocation to plant height (i.e. height relative to plant size) was high in both upright and rosette genotypes in the low light treatment. In Arabidopsis, low light intensity and spectral quality induced plants to display a more upright phenotype through reduced rosette size and number of basal branches, and little or no reduction in height (Fig. 2). Trait plasticity in this study was consistent with shade avoidance responses in previous studies for Brassica (Schmitt et al., 1995) and Arabidopsis (Pigliucci & Schmitt, 1999).

Phenotypic plasticity through shade avoidance responses has been repeatedly demonstrated to be adaptive across variable light environments. Genotypes displaying shade avoidance responses maintain relatively high fitness across light environments, but genotypes lacking these responses have relatively low fitness in unfavourable light environments (Schmitt et al., 1995; Dudley & Schmitt, 1996; Pigliucci & Schmitt, 1999). If there are limits to plasticity, however, phenotypic responses to changing environments will often not bridge the morphological gap between genotypes from contrasting environments (van Tienderen, 1990; Pigliucci, 2001). In this study, growth form genotypes remained morphologically distinct in both species and light treatments. For example, upright plants of Brassica were taller than rosette plants in control light and light filtered treatments (Fig. 1), and Arabidopsis genotypes with relatively large rosettes in control light also had relatively large rosettes in low light (Fig. 2). Thus, growth form genotypes did not converge on a single optimum phenotype in either light environment. Limits to plasticity during development can prevent the expression of an optimal phenotype across light environments (van Hinsberg & van Tienderen, 1997). Shade avoidance responses should be seen as adaptive when compared to no phenotypic response, but genetic differences in growth form may be required in order for plants to display an optimal phenotype in a given environment.

Adaptation to heterogeneous environments may favour either generalist (plastic) or specialist genotypes, depending on the nature of environmental variability, and genotypes that are more plastic are generally seen as less specialized (e.g. van Tienderen, 1991; Weinig, 2000; Pigliucci, 2001). A product of this view is that increasing plasticity decreases environmental specialization. We suggest that these tradeoffs need not play a pivotal role in plant adaptation, as growth form mutations can produce genotypes pre-adapted to different resource environments. Evolution of novel growth forms is not related to the degree of phenotypic plasticity (i.e. the slopes of the reaction norms) for experimental lines of Brassica or Arabidopsis. Thus, selection could continue to favour the evolution of plastic generalization even as genotypes become specialized to new environments. We are uncertain if rapid growth form evolution demonstrated in experimental lines will also be observed in natural populations. In addition, we are uncertain if plasticity will be maintained in natural populations as they specialize to new environments. Future studies across a range of species are required in order to establish the frequency and significance of growth form mutations and plasticity to plant adaptation in heterogeneous environments.

Growth form evolution and trait integration

Phenotypic integration is the pattern and magnitude of covariation among a set of traits (see Merilä & Björklund, 2004). Complex patterns of phenotypic integration can constrain the evolution of new trait combinations where these new combinations are either not viable or relatively unfit (Merilä & Björklund, 2004). Trait integration should be most likely to limit the expression of new trait combinations over short evolutionary time scales (Merilä & Björklund, 2004), and may be particularly important in limiting the viability and success of growth form mutants. We found significant pair-wise correlations between family lines within growth form genotypes of Brassica (Table 3) and between growth form genotypes of Arabidopsis (Table 4). However, only a few traits were significantly correlated to fitness (fruit number). In Brassica, height was highly correlated to fruit number in both growth forms and light treatments, because a major component of individual height is the length of the terminal inflorescence. Thus, plant height and flower production are highly integrated traits throughout plant development. In Arabidopsis, biomass, rosette size and number of main stems tended to be correlated across growth form genotypes. However, fruit number was only significantly correlated to branch number in the control light treatment (the correlation was not significant after bonferoni correction in the light filter treatment). This correlation arises because infloresences are produced at both the terminal and axillary positions on branches.

We found no consistent changes in phenotypic integration across light environments. However, some interesting patterns suggest that phenotypic integration may be environmentally dependent. Across upright family lines of Brassica, correlation coefficients between traits tended to be greatest (and more often significant) in the low light environment. Similarly, correlation coefficients across Arabidopsis growth form genotypes generally increased in the low light environment. These results are consistent with the prediction that integration should be strongest in the most stressful environments (Pigliucci, 2004). However, increased integration with decreasing light availability was not a general phenomenon as correlations tended to be greater in the high light environment for rosette family lines of Brassica.

Growth form mutations generally resulted in plants with less biomass and lower plant fitness (lower fruit number) in a given environment. In spite of this effect, fitness of growth form mutants in their favoured environments was sometimes equal to or greater than fitness of the wild-type genotypes. For example, an Arabidopsis genotype with a relatively small rosette had the third lowest fruit production in high light but second highest fruit production in low light (the predicted favoured environment of an upright genotype) (Fig. 3b). Differences in relative fitness in growth forms across environments demonstrate that growth form mutations can result in genotypes preadapted to environments of changing light availability.

Reproductive effort, and relative reproductive effort between low light and high light environments were negatively correlated with rosette size across genotypes of Arabidopis. Growth form genotypes producing an upright phenotype also allocate relatively more resources to reproductive output than a genotype with a large rosette. This shift in allocation is greatest in the low light environment, the environment predicted to favour increasingly upright phenotypes. Changing patterns of reproductive effort can counteract some of the negative fitness consequences of the relatively low plant biomass of the upright genotypes. Future experiments are required to test how the changing relative allocation to reproduction can affect competitive outcomes between genotypes (and species) across resource gradients (e.g. Aarssen & Taylor, 1992).

Optimal growth form should vary across environments. Growth form strategies may evolve as a result of long term selection on plants in new resource or competitive environments. However, we believe that rapid growth form evolution will both permit the expression of new phenotypic optima in changing environments and promote the expression of adaptive phenotypes across heterogeneous landscapes.

Acknowledgments

This research was supported by an NSERC Post Doctoral Fellowship to S.P.B. and an Andrew Mellon Foundation Grant to M.A.G.

Ancillary