Plasticity and evolution in drought avoidance and escape in the annual plant Brassica rapa


Author for correspondence:
Steven J. Franks
Tel: +1 718 817 3609


  • A key question in ecological genetics is to what extent do plants adapt to changes in climatic conditions, such as drought, through plasticity or evolution.
  • To address this question, seeds of 140 maternal families of Brassica rapa were generated from collections made before (1997) and after (2004) a natural drought. These seeds were planted in the glasshouse and grown under low-water and high-water conditions.
  • Post-drought lines flowered earlier than pre-drought lines, showing an evolutionary shift to earlier flowering. There was significant genetic variation and genotype by environment (G × E) interactions in flowering time, indicating genetic variation in plasticity in this trait. Plants that flowered earlier had fewer leaf nodes and lower instantaneous (A/g) and integrated (δ13C) water use efficiency than late-flowering plants.
  • These results suggest that B. rapa plants escape drought through early flowering rather than avoid drought through increased water use efficiency. The mechanism of this response appears to be high transpiration and inefficient water use, leading to rapid development. These findings demonstrate a trade-off between drought avoidance and escape, and indicate that, in this system, where drought acts to shorten the growing season, selection for drought escape through earlier flowering is more important than phenotypic plasticity.


Drought is a major physiological stress that influences the abundance and distribution of species, as well as the growth, performance and fitness of individual plants (Stebbins, 1952; Bohnert et al., 1995; Bray, 1997; Gurevitch et al., 2006; Taiz & Zeiger, 2006). Climatic data show an increase in the frequency of drought over the past century, and models predict further impacts of drought, as well as substantial variation in their global occurrence (IPCC, 2007; Portmanna et al., 2009). To anticipate how plants will respond to changes in climatic conditions, such as drought, it is important to understand dehydration physiology, as well as how drought responses are shaped by natural selection (Heschel et al., 2004; Ludwig et al., 2004; Sherrard & Maherali, 2006; Agrawal et al., 2008; Wu et al., 2010).

Plants can cope with drought either through escape or avoidance (Ludlow, 1989). Escaping drought entails the completion of the life cycle in advance of the effects of drought (Bazzaz, 1979; Heschel & Riginos, 2005; Wu et al., 2010). By contrast, plants avoid drought by maintaining high fitness despite drought conditions. This strategy, sometimes referred to as drought tolerance (Heschel & Riginos, 2005), is generally accomplished through a high water use efficiency (WUE), measured as photosynthetic carbon gain (A) over transpirational water loss (stomatal conductance, g) (Schulze, 1986; Arntz & Delph, 2001). Although plants could potentially escape and avoid drought, theory and previous findings suggest that there are likely to be trade-offs between these strategies (Bazzaz, 1979; McKay et al., 2003; Heschel & Riginos, 2005). The reason for this is that a trait that allows for greater drought avoidance, such as high WUE, may reduce the rate of growth and development and constrain or prevent drought escape.

Drought can also potentially cause either plastic or evolutionary changes in avoidance or escape. With plasticity, the expression of the phenotype is shaped by environmental conditions (Via et al., 1995; Schlichting & Pigliucci, 1998). A plastic response to drought would mean that the plants alter their phenotype by increasing avoidance or escape traits in drought relative to nondrought conditions (Mal & Lovett-Doust, 2005; Caruso, 2006). By contrast, drought could also act as an agent of selection and cause genetically based evolutionary changes in avoidance or escape (Fox, 1990; Ludwig et al., 2004). To what extent adaptive plasticity and local adaptation enable organisms to deal with spatial and temporal environmental variation remains one of the central questions of ecological genetics (Conner & Hartl, 2004; Kawecki & Ebert, 2004; Ghalambor et al., 2007). However, rarely have both plasticity and evolution been examined in a single study.

Seasonal drought is a regular feature of environments with a Mediterranean climate, such as southern California, which has a wet season in the winter and spring and a dry season in the summer and autumn (Minnich, 2007). In these habitats, the length of the growing season is determined by the duration of rains in the spring, which varies greatly between years (Franks et al., 2007). In such environments with high variation in the seasonal pattern of precipitation, plants could potentially respond via an avoidance or escape strategy, and could show either plastic adjustments or evolution following changes in precipitation patterns (Sherrard & Maherali, 2006).

In this study, a set of populations of the annual plant Brassica rapa was used that had been shown previously to have evolved earlier flowering following a natural drought comprising five consecutive years of low precipitation and short growing seasons (Franks et al., 2007). This previous work was the first study to demonstrate a rapid evolutionary change in natural plant populations in response to a change in climate (Franks & Weis, 2008). However, the following questions remain. Did avoidance as well as escape contribute to the drought response? Is there plasticity in drought response strategies? What are the physiological and developmental mechanisms of the drought response? Did the drought cause evolutionary changes in plasticity?

To address these questions, seeds from two populations were used (a wet site and a dry site), collected before and after the drought; 140 maternal families were created, which were grown under low-water (drought) and high-water (nondrought) conditions to examine plasticity in drought escape. For all of these individuals, the time of first flowering was recorded and the number of leaf nodes at the time of first flowering was measured, which has been shown previously to relate to the rate of development and to drought escape in this species (Franks & Weis, 2008). For a subset of individuals, data were collected on instantaneous WUE using an infra-red gas analyzer. Data were also collected on carbon stable isotope ratios (δ13C, hereafter abbreviated as δ), which have been shown to provide a long-term average estimate of WUE because of the slower rate of diffusion and enzymatic discrimination against the heavier 13C isotope in CO2 (Farquhar et al., 1989; McKay et al., 2003). The following hypotheses were tested: that drought caused escape through either a plastic adjustment or an evolutionary shift to earlier flowering, and that there is a trade-off between drought avoidance and escape.

Materials and Methods

Study system

Brassica rapa (syn. campestris) L. (Brassicaceae) is an introduced annual plant with C3 photosynthesis that is widespread throughout North America. In southern California, the plants germinate in winter and throughout the wet season and flower in the spring and early summer. The length of the growing season varies widely among years depending on the amount and temporal pattern of precipitation, with the length of the growing season variable and determined by the duration of spring precipitation (Franks et al., 2007). The study populations included Back Bay (BB), located in Newport Harbor, and Arboretum (Arb), located in the San Joaquin Marsh Reserve, both in Orange County, California. The two sites are c. 2.7 km apart (Franke et al., 2006). BB is a consistently drier site and Arb is a wetter and also more variable site (Franke et al., 2006). A previous study estimated soil moisture content at saturation to be 41.5% for Arb soil and 21.5% for BB soil (Franke et al., 2006). Seeds were collected in 1997 (after five wet years) and 2004 (after five drought years), with 400–1800 seeds collected per generation at each site, and raised for a generation in the glasshouse to reduce maternal and storage effects (Franks et al., 2007) and to generate the maternal family groups for the experiments.

Experimental design

From each of the groups (Arb 1997, BB 1997, Arb 2004 and BB 2004) that had been crossed within groups for one previous generation, 40 plants were randomly selected from the original 100 lines per group to create new experimental maternal lines. Of these 40, any families with germination rates too low to be used in the study were eliminated and a random choice was made among the remaining families, so that there were 35 families within each group used for the study. These maternal individuals were crossed with one randomly selected individual from the same group, with the two individuals mated as pairs. Twenty seeds were collected from each maternal individual. A total of 2800 seeds was planted. Of these, 2238 plants germinated and survived and were used in the study. Seeds were planted in 3.75-l pots in soil made up of a ratio of four parts Sunshine Mix (Sun Gro Horticulture, Vancouver, BC, Canada) to one part sterilized sand, with c. 14 g of slow release 14-14-14 Osmocote (Scotts, Marysville, OH, USA) fertilizer per pot. There were four plants to a pot (one from each group). The pots were given a high-water treatment, in which plants were watered daily to soil capacity for the duration of the experiment, or a low-water dry-down treatment, in which plants were watered fully for 14 d to allow establishment and then watered only once per week thereafter. Based on field observations of germination patterns and precipitation records for the study sites (Franks et al., 2007), it would be possible for plants to germinate and experience 2 wk of regular precipitation followed by light to no precipitation in the field. After 3 wk of growth (1 wk of drought treatment), the water content of the soil relative to saturated soil (RWC), as measured with a time-domain reflectometer (TDR) using 6-cm probes, was significantly lower in the low-water (18.9 ± 0.18) than high-water (20.5 ± 0.19) treatment (F1,100 = 14.1; < 0.001). After another 3 wk, the difference was even greater, with RWC significantly lower in the low-water (4.43 ± 1.18) than high-water (24.1 ± 1.15) treatment (F1,100 = 108; < 0.0001).

Families and treatments were assigned to pots in a randomized complete block design with 12 pots per block. Half of the individuals in each family (10) were given a high-water treatment and the other half were given a low-water treatment. Plants were grown in the glasshouse under light with a 16-h light : 8-h dark photoperiod. For each plant, the date of germination, date of first flowering and number of leaf nodes at the time of first flowering were recorded. The time to first flowering and the number of leaf nodes at first flowering are both related to drought escape (Franks & Weis, 2008).

After previous calibration and test runs, gas exchange measurements were made on a single day, 33 d after the average germination date and 1 d after the average days to first flowering, between 10:00 and 12:00 h, with an Li-Cor 6400 infra-red gas analyzer (Li-Cor, Lincoln, NE, USA). Light response curves (Supporting Information Fig. S1) showed photosynthesis saturation beginning at a photosynthetic photon flux density (PPFD) of 500 μmol m–2 s−1 of photosynthetically active radiation (PAR). A light chamber PAR of 700 μmol m–2 s−1 was used for all measurements. Leaf temperature was set to 26°C. In all cases, the leaf filled the entire chamber, so that measurements did not have to be corrected for leaf area. Plants were obtained for gas exchange measurements by randomly selecting 40 plants, with 10 plants from each group (Arb 1997, Arb 2004, BB 1997, BB 2004) of different maternal families, all in the high-water treatment, and taking the data in a randomized order (data from one plant could not be used, for a total sample size of 39). Only the high-water treatment was used for this analysis to maintain uniformity of conditions and to maximize the ability to compare between populations and generations in inherent WUE under optimal conditions. For each measurement, data were collected three times in a row (30–60 s apart) and the average was taken for all values.

Fifty-nine days after the start of the experiment, when all plants had flowered but none had yet senesced, leaves were collected for stable isotope analysis. Entire leaves were collected, placed in coin envelopes and dried at 60°C for 8 d. For each plant, the first noncotyledon basal rosette leaf produced and the last stem leaf produced were collected. The reason for collecting leaves produced early in the experiment and leaves produced later in the experiment is that early leaves could show inherent differences in water use when all plants are grown under common conditions, whereas leaves produced later in the experiment could show differences in water use in response to different treatments, which take time to show an effect in a dry-down experiment. Samples were sent to the stable isotope facility at the University of California, Davis, where they were ground using a ball mill and analyzed using isotope ratio mass spectrometry. The results are reported as δ13C (‰) relative to the PDB standard (Hubick et al., 1986).

Statistical analyses

All analyses were performed with SAS 9.1 (The SAS Institute, Cary, NC, USA). Time to flowering was analyzed with a failure time analysis with the Lifereg procedure using a gamma distribution with treatment, population source, family (nested within source population), block and treatment by family as terms in the model. The treatment term estimates plasticity and the treatment by family term estimates genotype by environment (G × E). The family and family by treatment interaction terms were evaluated with log likelihood tests (Littell et al., 1996). Data on WUE, A, g and δ were analyzed with general linear models. The relationships between time to flowering and leaf nodes and between time to flowering and δ were analyzed with linear regressions. For the gas exchange data, the effect of reproductive status (in flower or not yet in flower) on WUE was analyzed with a general linear model, and reproductive status was also included as a covariate in analyses of population, collection year and treatment on WUE, A and g. Leaf nodes and WUE were square-root transformed; these transformed and other untransformed variables met model assumptions.


Evolution and plasticity of drought escape

Flowering time (days between germination and first flowering) was compared in pre- and post-drought lines to test for evolutionary changes in drought escape. There was a significant effect of collection generation on time to first flowering (Table 1). Descendants (2004 lines) flowered earlier than ancestors (1997 lines) in both populations and both treatments (Fig. 1). In addition, dry site (BB) plants flowered earlier than wet site (Arb) plants (Fig. 1).

Table 1.   Plasticity and genetic variation in time to first flowering
  1. Results are from a failure time analysis with flowering time (number of days between germination and flowering) as the dependent variable. df, degrees of freedom; Gen, generation (1997 or 2004); Pop, source population (Arboretum or Back Bay); Trt, treatment (low-water or high-water). Significances of family terms are from log-likelihood tests. Factors shown in bold have < 0.05.

Pop1219.5< 0.0001
Gen123.0< 0.0001
Trt × Pop10.60.4362
Trt × Gen10.20.6609
Pop × Gen10.70.3956
Trt × Pop × Gen10.20.6789
Family (Pop)1382.5< 0.0001
Trt × Family (Pop)156.8< 0.0001
Block9112.6< 0.0001
Figure 1.

 Flowering time. Shown are the least-squared means (± 1SE) of days between germination and first flowering for Brassica rapa plants grown under glasshouse conditions. For each bar, the first letter indicates the experimental treatment (H, high-water; L, low-water) and the following letters indicate the source populations (Arb, Arboretum, wet site; BB, Back Bay, dry site). Black bars, plants from the 1997 (pre-drought) lines; gray bars, plants from the 2004 (post-drought) lines.

To examine plasticity in drought escape, the effects of drought treatment, maternal family and the interaction between these factors on flowering time were examined. There was a significant effect of family (Table 1), indicating genetic variation for flowering time. There was no effect of drought treatment on the time to first flowering (Table 1). There was a significant family by treatment interaction term (Table 1), indicating genetic variation in plasticity (G × E) for this trait (Fig. 2).

Figure 2.

 Reaction norms. Shown are the norms of reaction for flowering time for the 140 maternal Brassica rapa plant families grown under low-water and high-water conditions. Each point represents a family mean.

To determine whether the degree of plasticity was different between the populations or whether plasticity changed following drought, the interactions between population and treatment and between generation and treatment were examined. There was no difference in plasticity between the populations or between generations (treatment by population and treatment by generation terms not significant; Table 1).

If plants that flower early do so by shifting from vegetative growth to reproduction at an earlier developmental stage, the number of leaf nodes may be correlated with flowering time (Franks & Weis, 2008). There was a significant positive relationship between days to first flowering and the number of leaf nodes (= 2137; r2 = 0.20; < 0.0001; Fig. 3) when all plants were pooled. When grouped by treatment or by collection generation, there was a significant positive relationship between days to first flowering and the number of leaf nodes for each subgroup (< 0.0001). The average number of true leaves produced before flowering was 7.3 (± 0.04), and the mean number of days between germination and first flowering was 31.5 (± 0.12).

Figure 3.

 Relationship between days to first flowering and number of leaf nodes in Brassica rapa lines. This analysis is pooled across the two populations, two collection years and two watering treatments (= 2137; r2 = 0.20; < 0.0001). Pooling of data was shown to indicate the overall trend, but the patterns were similar when the groups were analyzed separately (see text for details).

WUE and drought avoidance

Drought avoidance was examined by measuring WUE with gas exchange and stable isotope methods. When the gas exchange measurements were taken, 26 of the 39 plants in the subset recorded were already in flower, and the rest had not yet flowered. The instantaneous gas exchange measurements showed that WUE was significantly lower (less efficient) for plants that had flowered relative to plants that had not yet flowered (F1,38 = 16.6, < 0.001; Fig. 4a). This appeared to be mainly a result of the difference in stomatal conductance (g), with conductance significantly greater in plants that had flowered than in plants that had not flowered (F1,38 = 13.0, < 0.001; Fig. 4b). There was no significant difference in the rate of photosynthesis (A) between plants that had flowered and those that had not flowered (F1,38 = 1.7, > 0.05; Fig. 4c).

Figure 4.

 Instantaneous gas exchange measurements for Brassica rapa plants in flower (open bars) and plants not yet in flower (gray bars). Shown are the least-squared means and standard errors for water use efficiency (WUE: A/g) (a), rate of transpiration (g) (b) and rate of photosynthesis (A) (c).

To determine whether the natural drought caused a change in drought avoidance and whether the populations differed in this trait, the effect of generation and population on WUE (A/g) was examined. Although average WUE (in μmol CO2 μmol−1 H2O) was greater (more efficient) in the pre-drought (1997) lines (70.9 ± 5.0) than in the post-drought (2004) lines (60.3 ± 4.9), and greater in the wet site (68.3 ± 4.8) than in the dry site (62.9 ± 5.2) populations (pooling plants in flower and not in flower), these differences were not significant (> 0.05). This was also the case when the plants in flower and not in flower were analyzed separately (> 0.05).

To investigate the effect of drought on integrated WUE (δ), the effect of the experimental drought treatment, collection generation and population on δ taken from samples of the first and last leaves produced before flowering was examined. There were differences among leaves, with WUE significantly lower in samples of the last leaf produced compared with first leaf samples (F1,68 = 41.2, < 0.0001; Fig. 5). In the first leaf samples, there was no effect of treatment, population or generation on δ (all > 0.05; Fig. 5). However, there was a significant effect of watering treatment on δ for samples of the last leaf (F1,43 = 4.8, = 0.0342), with WUE greater in low-water than in high-water conditions (Fig. 5). This indicates that the low-water treatment caused an increase in drought avoidance.

Figure 5.

 Integrated measurements of water use efficiency using stable isotope ratios. Shown are the mean δ13C (‰) values relative to the PDB standard (± 1SE). A lower (more negative) number indicates lower water use efficiency. Closed bars, high-water conditions; open bars, low-water conditions. Brassica rapa first leaf samples are from the first produced basal leaf, and last leaf samples are from the last stem leaf produced before flowering. There were no significant differences between populations or generations, which were pooled for this analysis. There was a significant difference in δ between the first and last leaves collected and between the low- and high-water treatments for the last leaf but not for the first leaf samples (see text for details and statistics).

If there is a trade-off between drought escape and avoidance, days to first flowering and WUE should be correlated. There was a significant positive phenotypic correlation between δ and days to first flowering based on samples from the first (= 33, r2 = 0.30, = 0.0009) and last (= 41, r2 = 0.12, = 0.0301) leaves (Fig. 6). This indicates that plants flowering later had greater WUE, and that plants flowering early did so with lower WUE.

Figure 6.

 Relationship between δ13C (integrated water use efficiency) and days to first flowering. δ13C (‰) is given relative to the PDB standard. A lower (more negative) value of δ13C indicates lower water use efficiency. This analysis is pooled across the two populations of Brassica rapa, two collection years and two watering treatments. Closed circles, samples taken from the first leaf; open circles, samples taken from the last leaf. Sample sizes and regression statistics are reported in the text.


The results of this study show that a natural drought caused an evolutionary shift to earlier flowering in natural populations of B. rapa, and demonstrate that the mechanism of this drought escape response is rapid development facilitated by high stomatal conductance, which maximizes carbon gain at the expense of transpirational water loss. The findings indicate that there is a trade-off between drought avoidance and escape, and that these plants escape drought through early flowering rather than avoiding or tolerating drought or through plasticity in drought response. This study shows that the approach of comparing ancestors with descendants under common conditions (Franks et al., 2008) is highly useful for testing hypotheses about plasticity and evolution in natural populations.

For two B. rapa populations grown in common environments under both wet and dry conditions, the post-drought (2004) plants flowered earlier than the pre-drought (1997) plants, indicating an evolutionary shift to earlier flowering. This is in keeping with the results of a previous study (Franks et al., 2007), which is an important validation considering that the previous study was the first to show an evolutionary response in a plant population to a natural change in climate. The study reported here shows that the difference in flowering time was maintained through a second generation of crosses in the glasshouse. This finding indicates that rapid evolution can occur in natural plant populations following changes in environmental conditions, as has also been demonstrated in several studies on animals (Thomas et al., 2001; Berteaux et al., 2004; Umina et al., 2005; Balanyáet al., 2006; Bradshaw & Holzapfel, 2008).

The results of this study provide the physiological and developmental mechanism of a rapid evolutionary change in a natural population. According to this study, the mechanism of the evolutionary shift to earlier flowering in B. rapa following drought was rapid development and reproductive maturation, facilitated by inefficient water use in early flowering plants. Evidence to support this mechanism includes the fact that plants that flowered earlier had fewer leaf nodes, indicating that plants that flower early do so at an earlier stage (Franks & Weis, 2008), plants that flowered early had lower integrated WUE (δ), and plants already in flower had lower instantaneous WUE and higher stomatal conductance than plants that had not yet flowered. This indicates that early flowering plants maintain high stomatal conductance, which allows them to rapidly gain carbon as they grow and develop at a cost of heavy water use. This has been called the ‘live fast, die young’ approach (Wu et al., 2010) and appears to fit the description of the favored escape strategy in this system following drought. The results are consistent with the hypothesis of a trade-off between drought avoidance and escape (McKay et al., 2003; Heschel & Riginos, 2005; Wu et al., 2010).

The reason that the drought favored escape rather than avoidance is probably a result of the interaction between environmental conditions and plant life history in this system. It is well known that conservative water use can be adaptive, particularly in environments that are consistently water limited (Ludlow, 1989; Bray, 1997; Taiz & Zeiger, 2006). However, in seasonally dry and variable environments, such as the Mediterranean climate of southern California where this study was conducted, the strategy of heavy and inefficient water use during periods of abundance in order to rapidly develop, reproduce and escape drought can also be a successful strategy (McKay et al., 2003). In a previous study, Heschel & Riginos (2005) found selection for lower WUE and early flowering under early season drought in Impatiens capensis. By contrast, Heschel et al. (2002) found positive selection for greater WUE in I. capensis under late season drought. The pattern from these two previous studies is in contrast with the results of this study, where late season drought selected for earlier flowering and lower WUE. The reason for this difference is probably a result of differences between the Mediterranean environment of this study and the temperate environment of the previous studies. In the environment of this study, where there is generally at least some precipitation early in the growing season and droughts are characterized by shorter growing seasons and potentially a complete lack of precipitation late in the season, it makes sense that an annual plant, such as B. rapa, would show selection for drought escape. Using Avena barbata, another plant occurring in the Mediterranean climate of California, Sherrard & Maherali (2006) also found selection for earlier flowering under drought conditions. Interestingly, drought did not select for lower WUE in their study.

Although there was a general pattern of escape following drought, it is important to recognize that there are a range of physiological responses among individuals within the populations in this study. This is illustrated by the significant G × E in flowering time across watering treatments, with some genetic lines flowering earlier under low-water conditions and other lines flowering earlier under high-water conditions. There also seems to be some capacity for physiological adjustment to drought conditions, given that tissue samples collected from the last leaves produced at the time of flowering in the drought treatment tended to have higher δ values than leaves in the well-watered treatment, indicating that these plants can respond to drought conditions by increasing WUE. This finding is in keeping with a previous study showing lower WUE in alpine Ipomopsis plants in flower than those still in a vegetative state (Campbell et al., 2005). However, in desert shrubs, juveniles have been shown to be less water use efficient than adults (Donovan & Ehleringer, 1992), with this increased WUE leading to greater photosynthesis but not, in this case, an increase in growth rate (Donovan & Ehleringer, 1994). These studies indicate that major physiological shifts may occur during development and, in particular, at the transition to a reproductive state.

Plants can respond to drought through evolution or plasticity. Although the main response to drought in this system seemed to be evolution, there was some evidence for a limited degree of plasticity in flowering time. Although average flowering time did not differ between the low- and high-water treatments, there was genetic variation in flowering time plasticity (G × E). However, there was not an evolutionary change in plasticity following drought, as determined by comparison of the plasticity in pre-drought (1997) and post-drought (2004) families. This indicates that drought did not act to increase or decrease flowering time plasticity in these populations. There was also no difference in plasticity between the wet site and dry site populations. This was somewhat surprising as the Arb site shows considerably more variation than the BB site in both soil moisture and flowering time within the site and over time (Franke et al., 2006). Theory predicts that more heterogeneous sites should produce greater plasticity (Baker, 1965; Schlichting & Pigliucci, 1998). However, few studies have compared plasticity in populations from areas that differ in the degree of environmental heterogeneity. In a notable exception, Sultan (2001) found greater plasticity in Polyganum species with a wider distribution and greater ecological breadth than in species more narrowly distributed. The lack of plasticity in flowering time in B. rapa in this study may be a result of the fact that drought occurs late in the growing season, making it difficult for plants to adjust the timing of flowering in response to drought. By contrast, traits, such as stomatal conductance, which is constantly regulated and adjusted by most plants (Taiz & Zeiger, 2006), are perhaps more likely to be adjusted by drought.

Despite widespread interest in plasticity, this is the first study to use the ancestor–descendant comparison approach (Franks et al., 2008) to investigate evolutionary changes in plasticity in a natural plant population, although similar approaches have been taken with studies of bacteria (Bennet et al., 1992). In addition, relatively few studies have measured plasticity in drought responses. Previous studies have shown significant plasticity and G × E to soil moisture variation for morphological and reproductive traits in Lobelia siphilitica (Caruso, 2006) and Lythrum salicaria (Mal & Lovett-Doust, 2005). In experiments with swamp milkweed (Asclepias incarnata), Agrawal et al. (2008) found selection for reduced WUE in the field (under wet conditions), and plasticity for traits, such as specific leaf area, but not WUE, when wet and dry treatments were compared. Heschel et al. (2004) found positive selection for WUE under drought conditions and plasticity, but no G × E to soil moisture variation, in populations of Polygonum persicaria from wet and dry sources. Even from this small number of studies, it is evident that plasticity, genetic variation and selection on drought response traits are quite variable depending on the environmental conditions and species. Whether or not there are any general patterns, such as increased plasticity with increasing spatial or temporal variation in soil moisture, remains to be determined (Sultan, 2001).


I thank Chris Herman, Esther Ko, Phong Le, Angelene Ng, Karo Torosian and many other students for help with data collection. Sarah Kimball provided help with the Li-Cor 6400. John McKay assisted with the stable isotope data collection. Art Weis provided advice on the study design. Niamh O’Hara and Ellen van Wilgenburg provided helpful comments on the manuscript. I thank the University of California Reserve System for permission to work on the sites. This research was supported by grants from Fordham University.