Author for correspondence: Alice A. Winn Tel: +1 850 644 9833 Fax: +1 850 644 9829 Email: firstname.lastname@example.org
• Phenotypic traits differ between plants in different environments and within individuals as they grow and develop. Comparing plants in different environments at a common age can obscure the developmental basis for differences in phenotype means in different environments. Here, we compared trait means and patterns of trait ontogeny for perennial (Viola septemloba) plants growing in environments that differed in quality either naturally or due to experimental manipulation.
• Consistent with predictions for adaptive stress resistance, plants grown in lower-quality environments allocated proportionately more biomass to roots and rhizomes, and produced smaller, thicker and longer-lived leaves. The developmental trajectory of almost all traits differed between environments, and these differences contributed to observed differences in trait means.
• Plants were able to alter their initial developmental trajectory in response to an increase in resources after 8 wk of growth. This result contrasts with previous findings, and may reflect a difference in the way that annual and perennial species respond to stress.
• Our results demonstrate the complexity of interactions between the environment and the development of the phenotype that underlie putatively adaptive plastic responses to environment quality.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The integration of ecology and development will require linking gene action and function to the development of the phenotype in an ecologically realistic setting (Gilbert, 2001). Plant evolutionary ecologists contribute to this endeavor by describing how plant development differs among natural environments and how natural selection shapes development and its interaction with the environment. Recent work on phenotypic plasticity in plants has focused on specific responses to particular environmental characteristics, such as stem elongation in response to light quality (e.g. Dudley & Schmitt, 1996; Weinig, 2000) and induced defense in response to attack by herbivores (e.g. Zangerl & Berenbaum, 1990; Agrawal et al., 1999; Baldwin, 1999). This focused approach has permitted the identification of potentially adaptive responses and the initial dissection of the pathways by which plants can perceive and respond to specific environmental stimuli.
Plants in natural environments do not face one environmental factor or selective agent at a time, however (see e.g. Gomez, 2003; Haag et al., 2004; Huber et al., 2004). Recent work showing that the selective advantage of stem elongation in response to foliar shade is reduced or eliminated in microsites with low soil moisture (Huber et al., 2004) illustrates that plants must be able to integrate different kinds of information to respond adaptively. A full understanding of the interaction between natural environments and the development of the phenotype must ultimately account for plant responses to the myriad factors that simultaneously influence development and its consequences under natural conditions.
Traits expected to differ between plants grown in environments of different quality may also change as individuals grow and develop (e.g. Evans, 1972; Coleman et al., 1994; Wright & McConnaughay, 2002; Niinemets, 2004). Consequently, when plants growing in different environments are compared at a common age, traits may differ between them either because plants in the poorer environment are growing more slowly and are at an earlier point along a fixed developmental trajectory or because plants express different patterns of change in trait values with growth in different environments (Evans, 1972; Coleman et al., 1994; Coleman & McConnaughay, 1995; Gunn et al., 1999; McConnaughay & Coleman, 1999; Wright & McConnaughay, 2002; Niinemets, 2004). For example, McConnaughay and Coleman (1999) found that Polygonum pensylvanicum grown in low- and high-water treatments differed in the ratio of root biomass to shoot biomass (root : shoot) because root : shoot decreased as plants grew, and plants in the low-water treatment grew more slowly along the same developmental trajectory as those in the high-water treatment. By contrast, a similar difference in root : shoot in Abutilon theophrasetes was due to differences in the relationship between size and root : shoot between low- and high-water environments, as well as differences in growth (McConnaughay & Coleman, 1999).
In the context of understanding the interactions between the environment and the genotype during the development of the phenotype, the distinction between phenotype differences resulting from differences in growth rate along a fixed developmental trajectory and those due to differences in the pattern of trait change with size during development (i.e. differences in developmental trajectory) is significant because the latter implies a more complex interplay between environmental signals and plant development than would a delay in growth alone. Surprisingly few studies have distinguished the roles of differences in size and differences in developmental trajectory as the basis for plastic responses to the environment (reviewed by Wright & McConnaughay, 2002), and responses of annual plants in controlled environments have been emphasized (but see Niinemets, 2004). Differences in stress response and how it develops might be expected for annual and perennial plants, given that annuals can often evolve to avoid stress but perennials must be able to tolerate it (Stanton et al., 2000; Steinger et al., 2003).
Some studies that quantified effects of environment on developmental trajectories have found that the trajectory of development can only be altered by environmental conditions early in life, suggesting a possible constraint of fixed development on adaptive phenotypic plasticity (Gedroc et al., 1996; McConnaughay & Coleman, 1998; cf. Weinig & Delph, 2001). However, a recent study reported greater response of development to environmental variation for reproductive individuals than for juveniles of the herbaceous perennial Leontodon hispidus (Niinemets, 2004). The scarcity of data on the ontogeny of developmental flexibility prohibits general conclusions regarding either the extent or timing of the potential for the environment to alter patterns of development.
We compared traits implicated in the stress response syndrome for a perennial plant growing in environments that differed naturally in the rate of growth they supported in the field, which we defined as environment quality. We also conducted a glasshouse study in which we manipulated resource availability to determine how traits change as plants grow and develop and whether ontogenetic trajectories differ between high- and low-quality environments. In addition, we subjected plants to changes in resource availability to determine whether the developmental trajectory is fixed early or remains responsive to the environment later in development. These are the first steps towards elucidating adaptive developmental responses to natural variation in overall environment quality.
Materials and Methods
We measured response to environment quality in Viola septemloba Leconte, a perennial, rhizomatous violet common in sandy pinelands of the south-eastern coastal plain of the USA (Fernald, 1970). In north Florida, V. septemloba bears leaves all year round, and individuals produce dimorphic leaves and flowers in a seasonal pattern. From August through January, the majority of new leaves produced are entire and cordate. From February to July, individuals produce primarily deeply lobed leaves (Winn, 1999). Open, chasmogamous flowers are produced in late winter and reduced, obligately selfing, cleistogamous flowers are produced in the late summer and fall. Seedlings produce only cordate leaves, and the smallest individuals produce only cordate leaves and only cleistogamous flowers, if they produce flowers at all. All but the smallest individuals possess a clearly delineated underground rhizome, which serves as a storage organ.
At our study population in the Saint Marks National Wildlife Refuge in Wakulla Co., Florida, USA, widely spaced longleaf pine, Pinus palustris, form a sparse evergreen tree canopy, and a diversity of herbaceous species compose an open ground cover. Several aspects of the site suggest that plant growth is limited by soil resources. In addition to the open canopy, the soil is deep, well-drained sand, and individuals often wilt during long periods without rain. Plants in the field show dramatic positive responses in growth and flowering when watered and fertilized (K. S. Moriuchi, unpublished data).
Based on the stress resistance syndrome described for plants (Lambers & Poorter, 1992; Chapin et al., 1993), we expect violets in lower-quality environments to allocate proportionately more biomass to roots and to storage at the expense of leaves. We predict that they will maximize water and nutrient use efficiency and decrease tissue turnover rate by producing smaller, thicker and longer-lived leaves than those of plants in higher quality environments.
We measured traits related to stress response in 458 plants from third-generation inbred lines of V. septemloba that were planted in the field as part of a study of seasonal change in leaf morphology (A. A. Winn & K. S. Moriuchi, unpublished). A total of 6100 plants comprising at least 100 replicates of each of 60 inbred lines were transplanted to the field as 3-month-old seedlings in November 2001. Individuals were planted in sets of 50 at 122 locations that we judged suitable for growth of V. septemloba. At each location, individuals were planted into undisturbed vegetation at a density within the natural range at this site (30 cm apart). We collected all fruits produced by these transplants, counted and weighed the seeds, and dried and weighed fruits and their supporting peduncles.
One yr after transplanting, we harvested 20 randomly selected individuals from each inbred line without regard to planting location. At the time of harvest, we measured the area of the most recent fully expanded leaf for each harvested plant. Each plant was separated into roots, rhizome, leaves and reproductive structures (seeds and fruits plus their peduncles), and each part was dried and weighed to the nearest 0.01 mg. The average total dry biomass of the plants harvested from a location was calculated as an index of environment quality at that location, and we identified the 25 lowest and 25 highest quality locations based on this measure. We used only the plants harvested from these 50 locations to compare traits expressed in low- (n = 209 plants) and high-quality (n = 249 plants) environments.
We calculated specific leaf area (SLA) as the ratio of lamina area of the most recent fully expanded leaf to its dry weight. We quantified per cent allocation to roots, rhizome, leaves and reproduction for each individual as the ratio of the dry weight of each structure to total dry weight. Root : leaf was calculated as the ratio of the total dry weight of roots to that of leaves.
Trait means from low- and high-quality environments were compared in mixed model anova including the effects of environment quality and inbred line. Environment quality was treated as a fixed effect because we selected locations to represent opposite extremes of environment quality. The inbred line effect was included as a random effect to reduce potential bias due to unequal replication of lines among planting locations. Unequal replication of lines dictated that models be fit with restricted maximum likelihood. Trait distributions were tested for normality and homoscedasticity and were transformed where necessary to meet the assumptions for anova.
We grew individuals from seed in environments manipulated to be of low and high quality in the glasshouse and harvested subsets every 2 wk to determine the effects of both ontogeny (changes in plant phenotype with growth and development) and environment quality on traits expected to respond to environment quality. Seeds from 50 fruits from the field study were chosen without regard to inbred line and each was weighed to the nearest 0.001 mg. All seeds were germinated in a growth chamber, and transplanted as they germinated into soil collected from the site of the field study. Seedlings were grown in an unheated glasshouse in individual 4 × 4 × 6 cm3 cells of 72-cell flat inserts.
One half of the seedlings (n = 262) were assigned at random to a low soil resource treatment in which water was provided only when the plants appeared wilted, and no nutrients were added. The remaining seedlings (n = 259) received daily watering and weekly addition of a 20–20–20 (N : P : K) fertilizer. Because the glasshouse is shaded for part of the day, light availability is similar to the partial canopy at the field site (A. Winn, pers. obs.).
To estimate leaf turnover rates, we recorded the date of appearance and of death (the first date on which a leaf was completely brown) of the first true leaf produced by each seedling. Beginning 14 d after the initiation of resource treatments, we harvested 20 randomly selected seedlings per treatment every 2 wk for 6 wk. After the fourth harvest, half of the plants remaining in each treatment were transferred to the alternate environment to determine if individuals could change their initial responses to their resource treatment. All plants were transplanted into larger pots at this time to prevent restriction of root growth. At each harvest, we measured SLA and resource allocation as described for the field study. In the first harvest, we could not distinguish rhizome tissue from root tissue, so all underground tissue was weighed together as root tissue.
Comparisons of traits in low- and high-resource treatments were restricted to the plants from the first four harvests, plus those that remained in their initial treatment for the final two harvests, except where noted. For each trait, ancova for the effects of treatment with total biomass as a covariate were used to evaluate plasticity and ontogenetic variation. Type III sums of squares were used for significance tests because mortality resulted in unequal sample sizes. A significant treatment effect indicates a plastic response to resource treatment and a significant biomass effect indicates that the trait changes as plants grow. A significant interaction indicates that the trajectory of development differs between the treatments. When such interactions were significant, natural-log-transformed plant biomass was regressed on the trait to describe the ontogenetic trajectory within each treatment. All traits were natural-log-transformed to meet the assumptions of ancova. When assumptions were not met by transformation, ancova on ranked data were also performed (Huitema, 1980). Because analyses of ranked and log-transformed data produced the same results, only those from the latter are presented.
Survival analysis (Fox, 1993) was performed to compare leaf longevity between resource treatments. Leaves that did not die prior to harvest were assigned an age equal to the time between their initiation and harvest date (i.e. were censored). We used the Wilcoxon χ2 statistic to test for the effect of resource treatment on leaf longevity with individual biomass as a covariate (Lee & Wang, 2003).
Data from the last two harvests were analyzed to determine if plants could change their initial responses to resource availability. Because we were interested in the response of plants switched either from low- to high-resource availability or from high- to low-resource availability, we performed two separate analyses: we compared trait means for plants that remained in the low-resource treatment for the entire experiment with those switched from the low- to the high-resource treatment after 8 wk of growth, and we compared plants that remained in the high-resource treatment with those switched to the low-resource treatment. Effects of biomass, treatment and their interaction were compared with ancova as described earlier in this section.
Analyses were performed with sas v. 7.0 (SAS, 1999) MIXED, GLM, REG, NPAR1WAY and LIFETEST procedures.
The average biomass of plants harvested from high-quality field locations was nearly three times that of plants harvested from low-quality locations (Table 1). Consistent with the sizes of plants and the seasonal pattern of leaf dimorphism, only 6% of all leaves at harvest in November were lobed. None of the plants produced chasmogamous flowers in the first year after transplanting, but 1.9% of those in the low-resource locations produced at least one cleistogamous flower and 20% did so in the high-quality locations.
Table 1. Trait means (se) for Viola septemloba plants harvested in the field from the 25 locations with the greatest (high-quality environment) and least (low-quality environment) mean total plant dry biomass
Statistics are from mixed-model anova for the effects of environment quality (fixed) and inbred line (random). Unequal replication of lines among blocks precluded a reliable test for the interaction. ***, P < 0.001; **, P < 0.01; *, P < 0.05.
All traits differed significantly between low- and high-quality environments, but only total biomass and per cent allocation to roots varied among inbred lines (Table 1). On average, plants in the low-quality locations allocated 22% less biomass to leaf tissue, 11% more to rhizomes and 25% more to root tissue than those in high-quality locations. Root : leaf was nearly twice as large in the low-quality locations. Based on lamina biomass, leaves produced in the low-quality environment were smaller, and SLA was 15% greater in the low-quality environments (Table 1).
Because field plants were harvested at only one time, we could not describe how traits changed as plants grew or the effects of environment quality on ontogenetic trajectories. Sequential harvest of plants in the glasshouse experiment allowed us to address these issues.
In the glasshouse experiment, 21% of the plants in the low-resource treatment and 6.1% in the high-resource treatment died before the final harvest. No chasmogamous flowers were produced, and only plants in the high-resource treatment (37%) produced cleistogamous flowers. Of the plants from the first four harvests, none produced lobed leaves, and less than 3.5% of leaves in the last two harvests were lobed. Plants in the high-resource treatment grew nearly twice as fast as those in the low-resource treatment (slope of the regression of loge (total biomass) on age = 0.029 for the low-resource and 0.053 for the high-resource treatment). Initial seed mass had no significant effect on growth rate (F = 0.04, P > 0.83) and consequently was not included as a covariate in the remaining analyses.
With the exception of SLA, all traits averaged across the six harvests differed significantly between the low- and high-soil-resource treatments (Table 2). Plants grown in the low-resource treatment allocated 8% less of their biomass to leaves, 14% more to rhizome tissue and 28% more to roots. Root : leaf biomass was 45% greater in the low-resource treatment, and plants produced laminas that were 21% lighter than in the high-resource treatment.
Table 2. Trait means (se) for Viola septemloba plants grown in the glasshouse before (raw) and after (size-corrected) the effects of total biomass were accounted for
B × T
ancova results are F-values for the effects of resource treatment with total biomass as a covariate. T-statistics compare size corrected means.
Several traits varied significantly with total biomass, indicating ontogenetic changes, and the ontogenetic trend differed between treatments for all traits except per cent allocation to roots, for which there was no variation with total biomass (Table 2, Fig. 1). As plants grew larger in the high-resource treatment, they allocated proportionately less biomass to leaves, but allocation to leaves did not change with biomass in the low-resource treatment (Fig. 1a). Root : leaf also increased with biomass in the high-resource treatment only (Fig. 1c). The most dramatic difference between treatments in ontogeny was for per cent allocation to rhizome. There was a significant decrease in allocation to rhizome as biomass increased in the high-resource treatment and a significant increase in the low-resource treatment (Fig. 1d). The average size of leaves as measured by dry biomass increased in both treatments, but the increase was slower in the low-resource treatment (Fig. 1e). Specific leaf area decreased as plants grew in both treatments, but the decrease was twice as great in the low-resource treatment (Fig. 1f).
Differences between size-corrected trait means for the two treatments (Table 2) paralleled the differences for uncorrected means and for the differences between low- and high-quality field environments for all traits except SLA (compare Tables 1 and 2). Plants in the low-resource treatment allocated proportionally less to leaves and more to roots and rhizomes, had substantially greater root : leaf, and produced smaller leaves than those in the high-resource treatment (Table 2). Specific leaf area did not differ significantly between resource treatments in the glasshouse when differences in biomass were not taken into account, but did differ significantly after differences in biomass were accounted for (Table 2).
In both the field and the glasshouse, many more plants in the higher-resource environments flowered than in the low-resource environments, but we cannot distinguish between a plastic delay in flowering and delayed development as explanations for this pattern because no plants in the glasshouse low-resource treatment grew as large as the smallest flowering plants in the high-resource treatment (data not shown).
In the low-resource treatment, 92.6% of first true leaves survived to the time of harvest, vs 73.5% of those in the high-resource treatment (Fig. 2). Leaf longevity was significantly greater in the low-resource treatment (χ2 = 14.87, P < 0.001) and also for smaller plants (χ2 = 15.42, P < 0.001).
Analysis of data from the last two harvests indicated that plants did respond to the change in resource treatment imposed after the fourth harvest. Plant biomass and all traits measured differed significantly between plants that remained in the same resource environment and those switched to the alternate environment (Table 3, Fig. 3). Regardless of the direction of the change in resource environment, plants achieved trait values that indicated convergence toward the mean in the environment to which they had been switched (Fig. 3). For example, SLA was less for the plants that remained in the low-resource treatment than for those that remained in the high-resource treatment, but the SLA of the plants switched from the low- to the high-resource treatment was significantly greater than plants that remained in the low-resource treatment. Likewise, plants switched from the high- to the low-resource treatment developed leaves with significantly lower SLA than those that remained in the high-resource treatment (Fig. 3f).
Table 3. Comparison of trait means and results of ancova for the effect resource treatment with biomass as a covariate for each trait measured after some Viola septemloba plants were switched between resource treatments
LL vs LH
HH vs HL
LL vs LH
HH vs HL
B × T
B × T
ancova statistics are F-values for the effects of total biomass, resource treatment, and the interaction between them. Separate comparisons were made between plants that remained in the low-quality environment (LL) and plants switched to the high-quality environment (LH), and between plants that remained in the high-quality environment (HH) and plants switched to the low-quality environment (HL). ***, P < 0.001; **, P < 0.01; *, P = 0.05.
For plants switched from the low- to the high-resource treatment, there was a significant interaction between the effect of biomass and treatment for allocation to roots, leaves and rhizomes and for root : leaf (Table 3), indicating differences in the developmental trajectory of these traits. This interaction was not significant for any comparisons between plants that remained in the high-resource treatment and those switched to the low-resource treatment (Table 3).
We found a common pattern of plastic response to lower-quality environments, defined as those that supported a lower growth rate, for plants grown in environments that differed naturally in quality, those grown in environments in which we manipulated resource availability, and those that were transferred between resource availability treatments (Tables 1 and 2, Fig. 3). In all cases, plants in the lower-quality environment allocated proportionately less biomass to leaves, and more to roots and rhizomes than those in the better environment, consistent with the expectation of greater storage in more stressful environments (Chapin et al., 1990). The production of smaller, longer-lived leaves in lower-quality environments accords with putatively adaptive increased resource use efficiency in response to stress (Chapin et al., 1987; Lambers & Poorter, 1992; Chapin et al., 1993; Kikuzawa, 1995). Although the concordance between the responses we observed and the stress response syndrome constitutes circumstantial evidence for adaptive phenotypic plasticity, quantitative measures of natural selection would be required to confirm this conclusion (e.g. Dudley & Schmitt, 1996; Dorn et al., 2000).
The consistency of the responses we observed to low-quality environments, along with the similarity of the stress response syndrome across a diversity of species and environments (Chapin et al., 1993), suggests that the generalized response to stress can provide a framework for extending our understanding of phenotypic plasticity beyond the realm of responses to single environmental factors. The complementary literature on mechanisms by which plants sense and respond to stress (Chapin et al., 1993, reviewed by Knight & Knight, 2001; Zhu, 2002; Farnsworth, 2004) also suggests a common underlying mechanism of response, and offers a bridge to link the effects of natural environmental variation on the development of the phenotype to the genetic and physiological mechanisms that underlie them.
Sequential harvests in the glasshouse revealed that most traits changed as individuals grew, and that plants in the low-resource treatment were not simply developmentally delayed miniatures of those in the high-resource treatment. Plants developed along distinct trajectories in the two environments with respect to all traits except allocation to roots (Fig. 1), and some of these differences were dramatic. Specific leaf area declined with biomass twice as fast in the low-resource treatment as in the high treatment (Fig. 1f), and the ontogenetic change in allocation to rhizome was of opposite sign in the two treatments (Fig. 1c).
Whether plants in different environments arrive at different phenotypes due to differences in developmental trajectory or by being at different stages or sizes along a fixed trajectory may not influence the ecological consequences of the observed differences in phenotype. However, if plants develop along different trajectories in different environments, understanding how the phenotype develops will require examination of the trajectory of trait development and not just the endpoint. Although we could only measure developmental trajectories in the glasshouse, and our own results showing that development differs among environments supports the need for caution in extrapolating results beyond the environments in which they were measured, the qualitative similarity we found between responses to low-quality environments in the field and the glasshouse suggests that ontogenetic trajectories were similar in the field and glasshouse. If so, then differences in ontogenetic trajectory contributed to plastic responses to environment quality in the field as well.
Recognizing that plants in different environments develop along different trajectories enhances our understanding of how environmental effects on the phenotype arise. For example, the difference in mean SLA between our low- and high-resource treatments was nearly twice as large after means were corrected for differences in plant size (Table 2), indicating that both slower growth and a faster decrease in SLA with plant size (Fig. 1f) contributed to the difference. Thus some of the plastic response in SLA was effectively hidden when plants were compared at a common age.
When switched from the low- to the high-resource treatment after 8 wk of growth, plants were able to alter the developmental trajectory of several traits (Table 3, Fig. 3). Our observation that development remains responsive to the environment beyond the earliest stages of development conflicts with the results of some previous studies that have compared the potential for plastic response between plants of different age or developmental stage in annual species. For example, Gedroc et al. (1996) found for two annual species that developmental trajectories of several traits were not sensitive to changes in nutrient availability after 3 wk of age (cf. McConnaughay & Coleman, 1998). Weinig and Delph (2001) found that plastic response to irradiance in Arabidopsis thaliana at early developmental stages exacted an opportunity cost that constrained response at later stages. The contrast with our results for a perennial suggests that longer-lived species may have more flexible development than short-lived annuals. The failure of plants switched from the high- to the low-resource treatment to change their developmental trajectory (Table 3) may indicate that these plants were buffered by their larger size at the time of the switch, or it may be that flexibility does eventually decline as plants grow. We cannot distinguish between these alternatives with our data.
Future efforts to understand how natural environmental variation influences the development of the phenotype need to consider both the trajectory of development and its flexibility, and that these may differ between annual and perennial species.
We thank T. E. Miller, A. B. Thistle, S. Sultan and three anonymous reviewers for constructive suggestions for improving the paper. This work was supported by the National Science Foundation (DEB−9903878).