Author for correspondence: Sarah M. Assmann Tel: +1 (814) 863 9579 Email: email@example.com
•Phenotypic plasticity is the ability of one genotype to display different phenotypes under different environmental conditions. Although variation for phenotypic plasticity has been documented in numerous species, little is known about the genetic mechanisms underlying phenotypic plasticity. Given their widespread roles in hormonal and environmental signaling, we examined whether genes which encode heterotrimeric G proteins are plasticity genes.
•We grew multiple alleles of heterotrimeric G-protein mutants, together with wild-type Arabidopsis thaliana, under different watering regimes to determine the contributions of G-protein genes to phenotypic plasticity for a number of developmental and reproduction-related traits.
•G-protein mutations did not affect significantly the amount of phenotypic variation within an environment for any trait, but did affect significantly the amount of phenotypic plasticity for certain traits.
•AGB1, which encodes the β subunit of the heterotrimeric G protein in Arabidopsis, is a plasticity gene and regulates reproductive trait plasticity in response to water availability, resulting in increased fitness (defined as seed production) under drought stress.
Heterotrimeric G proteins are multisubunit guanosine triphosphate (GTP)-binding proteins that function in the transduction of external signals into cellular responses. Because G proteins regulate a large array of cellular and developmental processes in both plants and animals, it is of interest to evaluate their potential impact on developmental plasticity and fitness. According to the paradigm of G-protein signaling, the G protein is activated following the binding of a ligand to an associated membrane-bound G-protein-coupled receptor (GPCR). This binding results in a conformational change in the alpha subunit (Gα) and the subsequent exchange of GTP for guanosine diphosphate by Gα, resulting in the dissociation of Gα from the beta gamma dimer (Gβγ). Gα and/or Gβγ are then free to interact with downstream signal effectors until the intrinsic GTPase activity of Gα results in the reassembly of the inactive trimer (Assmann, 2002).
In mammals, there are a number of genes which encode heterotrimeric G-protein subunits and hundreds of GPCRs have been predicted, resulting in a large, diverse assortment of potential G-protein signaling pathways (Fredriksson & Schioth, 2005; McCudden et al., 2005). Numerous ligands have also been identified for mammalian GPCRs, including light, sensory molecules including odors and tastes, hormones, neurotransmitters and bacterial toxins (Civelli, 2005), and mutations in mammalian G-protein subunits often result in genetic disorders (Spiegel & Weinstein, 2004; Weinstein et al., 2006).
Despite the paucity of heterotrimeric G-protein subunits in the Arabidopsis genome, functional studies of G-protein mutants have shown diverse roles for heterotrimeric G proteins in germination, development, phytohormone responses [abscisic acid (ABA), auxin, brassinosteroids, gibberellins], stress responses (ozone, reactive oxygen species, pathogens) and stomatal aperture regulation (Ullah et al., 2001, 2002, 2003; Wang et al., 2001, 2007; Pandey & Assmann, 2004; Joo et al., 2005; Llorente et al., 2005; Pandey et al., 2006; Fan et al., 2008; Zhang et al., 2008a;Zhang et al., 2008b). How a limited number of heterotrimeric G-protein subunits can transduce such a large number of hormonal and environmental signals is a fundamental question in plant G-protein signaling (Assmann, 2004). Further characterization of GTGs and other unconventional G proteins, such as the XLGs (Lee & Assmann, 1999; Ding et al., 2008), and the identification of GPCR ligands, potential tissue-specific GPCRs, and G-protein signaling effectors in plants may help to elucidate this question. An additional model which has been proposed is that G proteins may serve as signal modulators instead of direct transducers of signals (Fig. 1). By functioning as ‘cross-talk hubs’, G proteins could fine tune a phenotype or physiological response based on multiple signals/environmental inputs (Assmann, 2004). This model also addresses another paradox in plant G-protein signaling: why, if G proteins are important to plant physiology, are G-protein mutations not lethal in plants? A mutation may not be lethal if additional copies or similar versions of the gene exist in the genome or if the expression of the gene is specific to a certain stress (e.g. cold shock) or to a less vital plant tissue or cell type (e.g. trichomes). However, none of these situations applies to Arabidopsis as Gα and Gβ are both encoded by single, canonical genes which are widely expressed throughout the plant (Ma et al., 1990; Huang et al., 1994; Lease et al., 2001; Anderson & Botella, 2007). Alternatively, if G proteins function in directing hormonal and environmental cross-talk, they may be required only in the production of the ‘optimal’ phenotype, and the direct transducers of the signal would still function in the absence of functional G-protein subunits.
Plants, being sessile organisms, are hypothesized to have evolved increased phenotypic plasticity, the ability of one genotype to display different phenotypes under different environmental conditions, compared with their mobile animal counterparts (Bradshaw, 1972; Schlichting, 1986; Sultan, 1987; Huey et al., 2002). Heightened plasticity in plants would allow plants to compensate for inescapable and inhospitable environments. Although variation for plasticity has been documented in numerous plant and animal species, and many theories have been proposed concerning the ecological and evolutionary significance of this variation, very little is known about the explicit genetic machinery which underlies phenotypic plasticity.
The phenotypic plasticity (or lack thereof) of a trait can be graphically represented by a reaction norm, which is a plot of the mean phenotypic value of the trait in different environmental conditions (Fig. 2). A horizontal reaction norm indicates that the trait lacks plasticity, whereas a line with a nonzero slope or a curved line is indicative of phenotypic plasticity. As plasticity genes control the shape or slope of the reaction norm of a trait, when these genes are mutated, it is expected that the shape/slope of the reaction norm will diverge from that of the wild-type. Mutations may alter the height of the reaction norm without affecting the overall plasticity for a trait (line C), affect the amount of plasticity for a trait (line B) or may change the direction of plasticity for a trait (line D). Differences in reaction norm shapes (plasticities) among genotypes can be detected and tested using analysis of variance (ANOVA)-based statistical methods. Specifically, a significant genotype × environment/treatment interaction term for a trait indicates that there is variation for plasticity among the genotypes, that is, the response curves have different shapes/slopes.
Although functional analysis of specific genic mutants is a widespread method to determine gene function, this technique has been applied to phenotypic plasticity studies only in a few instances, which have focused mainly on photoreceptor genes. For example, in Arabidopsis, this approach has been employed to study the genetic basis of phenotypic plasticity in photomorphogenetic responses (Pigliucci & Schmitt, 1999, 2004). Arabidopsis hy1 and hy2 photoreception mutants, which display constitutively active shade avoidance responses, showed reduced fitness under some environmental conditions relative to the more plastic wild-type (Pigliucci & Schmitt, 1999). This result suggests that HY1 and HY2 (which encode a plastid heme oxygenase and a phytochromobolin synthase, respectively) (Muramoto et al., 1999; Kohchi et al., 2001) are plasticity genes. In addition, naturally occurring polymorphism of the photochrome PHYB locus has been associated with altered light responses in Arabidopsis; however, phenotypic plasticity was not measured explicitly (Filiault et al., 2008). In field experiments, plants harboring mutations in phototropin blue light photoreceptors showed reduced fitness relative to the wild-type under a range of light conditions; however, significant genetic variation for phenotypic plasticity was observed only for the seedling emergence rate (Galen et al., 2004).
Genetic variation for plasticity has also been documented among wild-type accessions of Arabidopsis (Pigliucci & Kolodynska, 2002; Schmuths et al., 2006; Brock & Weinig, 2007), and naturally occurring variation for plasticity within plant species has been assessed in quantitative trait locus (QTL)-based phenotypic plasticity studies. Using recombinant inbred lines in lieu of genetic mutants, these studies have lent additional support for the existence of plasticity genes. Studies on recombinant inbred lines of Arabidopsis (Kliebenstein et al., 2002; Ungerer et al., 2003; Juenger et al., 2005), barley (Lacaze et al., 2009) and poplar (Wu, 1998) found significant QTL × environment interaction for a number of traits. Specifically with regard to drought stress in Arabidopsis, Hausmann et al. (2005) found significant recombinant inbred line × environment interactions for a number of water use traits, further supporting the existence of genes which regulate phenotypic plasticity to drought stress in Arabidopsis.
The signaling cross-talk mechanism discussed above may be an important component of phenotypic plasticity, as phenotypes could be tweaked on the basis of multiple environmental inputs. A mutation in a G-protein subunit might decrease cross-talk and therefore reduce the plant’s ability to adjust a phenotype or response to changing environments based on multiple signals. Specifically, a G-protein mutation might affect the degree of plasticity of a trait (Fig. 2, line B), the amount of variation within an environment for a trait (that is the ‘noisiness’ of the trait) or might simply shift the mean value of the trait away from the wild-type value (Fig. 2, line C) (Assmann, 2004). Therefore, by studying populations of G-protein mutants under multiple environments, we might reveal whether G-protein genes are plasticity genes, as well as evaluate the importance of G proteins to plant fitness. In addition, previous studies of heterotrimeric G proteins have, to a large extent, focused on guard cell physiology (Wang et al., 2001; Pandey & Assmann, 2004; Fan et al., 2008) and cell division (Ullah et al., 2001, 2003; Chen et al., 2006). By studying whole-plant phenotypic plasticity responses of G-protein mutants, we might reveal novel functions of G proteins that could not be identified by previous cell-centered approaches.
1Do G-protein mutations alter the shape/slope of reaction norms in response to water availability, that is are G-protein genes ‘plasticity genes’?
2Does a G-protein mutation affect the level of phenotypic variation within an environment?
3What are the fitness consequences of G-protein mutations in different environments?
Using multiple alleles of gpa1, agb1 and gcr1 mutants, we found significant variation for plasticity for a number of reproduction-related traits in response to water availability. agb1 mutants showed significantly reduced plasticity for inflorescence height, number of fruits and seed number per fruit. Interestingly, agb1 mutants showed enhanced fitness under drought stress compared with the wild-type, but all G-protein mutants showed reduced fitness under ample water conditions. These data support the hypothesis that heterotrimeric G-protein genes are indeed plasticity genes in plants.
Materials and Methods
Plant growth conditions and water treatments
All Arabidopsis thaliana (L.) Heynh seeds used in this experiment were collected from parent plants that were grown together under uniform conditions. gpa1-3, gpa1-4, agb1-1, agb1-2, gcr1-1, gcr1-2 and gpa1-4agb1-2 mutants have all been described previously and were generated using the ecotype Col (Lease et al., 2001; Jones et al., 2003; Ullah et al., 2003; Chen et al., 2004). All mutants are T-DNA insertional mutants, with the exception of agb1-1, which is an ethylmethanesulfonate-generated point mutation (Lease et al., 2001). It has been determined that gpa1-3 and gpa1-4 are null mutants at both the transcript and protein levels (Jones et al., 2003). agb1-2, gcr1-1 and gcr1-2 are all transcript null alleles (Ullah et al., 2003; Chen et al., 2004). agb1-1 produces a larger and less abundant AGB1 transcript compared with the wild-type as a result of a destabilizing point mutation, which results in a failure to splice out the first intron and the introduction of a premature stop codon (Lease et al., 2001). All alleles were backcrossed once in our laboratory and the genotypes of parent plants were confirmed via PCR of genomic DNA. Seed storage was identical for all seed lots. Cold stratified seeds (stratified at 4°C for 48 h in darkness on wet filter paper) were sown directly on the surface of a soil mix composed of Miracle-Gro potting mix (The Scotts Co, Marysville, OH, USA), Turface Greens Grade fritted clay (Profile Products LLC, Buffalo Grove, IL, USA) and perlite in a 16 : 8 : 1 volume ratio. The plants were grown in Kord 90 mm press-fit pots (Kord Products Inc. Brampton, Ontario, Canada) in a walk-in Conviron growth chamber (Conviron Inc. Winnipeg, Manitoba, Canada). The photoperiod was 12 h light (140 μmol m−2 s−1, 21°C) and 12 h dark (19°C) and the relative humidity was 60%.
Three weeks after sowing, the plants were treated with one of three watering regimes: ample water, moderate drought or severe drought. The moderate and severe drought designations are relative to the ample water treatment for our experiment. Plants were individually watered using a bottle-top volumetric dispenser. Plants subjected to ample watering had continually moist soil (c. 95% of the soil water-carrying capacity) and weekly water applications ranging from 55 to 170 ml depending on the plant age. Ample water-treated plants never wilted between waterings, showed no signs of waterlogging and were healthy in appearance. Severe drought-treated plants had soil which dried out completely between watering (c. 20% of the soil water-carrying capacity) and had weekly water applications ranging from 10 to 50 ml. Severe drought-treated plants displayed considerable wilting between waterings. Plants receiving the moderate drought treatment received approximately twice the volume of water applied to the severe drought-treated plants (c. 40% of the soil water-carrying capacity). Moderate drought-treated plants showed some turgor loss between watering, but to a lesser extent compared with the severe drought-treated plants. Water application was adjusted for treatment and plant age, and all plants within a treatment received the same amount of water. Relative water content measurements of 5-wk-old fully expanded leaves from three blocks showed no significant differences between the genotypes for any watering regime, indicating that the levels of drought stress were consistent across genotypes (data not shown). The average leaf relative water contents for each treatment were 76% for ample water, 65% for moderate drought and 61% for severe drought. These values are within the wide range of leaf relative water content values utilized for Arabidopsis drought stress in other published reports (Gigon et al., 2004; Rizhsky et al., 2004).
Experimental design and response variables
Plants were arranged in a split-plot design. Three trays, each representing one water level, were clustered in a block, and 12 blocks were placed on separate shelves in the growth chamber. Genotypes were randomly assigned a position within a tray with two genotype replicates per tray. There were two replicate plants × 8 genotypes × 3 treatments × 12 blocks for a total population of 576 plants. The transition from vegetative to reproductive growth was assessed by recording the flowering time (days from sowing), when the first open flower was visible, for each plant. Because of the large population size and the fact that flowering times were affected by treatment and genotype, each plant was individually harvested 4 wk after the plant began to flower and the following variables were recorded: inflorescence height (cm), number of primary lateral branches and number of fruits plus any pistil which showed elongation or swelling. Excised rosettes were dried at 70°C until a constant mass was achieved, and the dry mass was determined. For three blocks of plants, five fruits were harvested (two from the main inflorescence and three from the lateral branches), and the seed number per fruit was determined using a dissecting scope. Aborted or shriveled seeds were excluded from the seeds per fruit measurements. Seed production was estimated for each plant in the three blocks which had seeds per fruit measurements as (total fruit number × seeds per fruit). Relative fitness (total seed number/mean total seed number for water level) was calculated post hoc for the three blocks for which total seed production was determined.
Experiment-wide variances for genotype means under ample water and drought stress were calculated using Minitab 15 (Minitab Inc. State College, PA, USA). F test equal variance tests were performed in Minitab between mutant and wild-type trait variances from ample water and severe drought stress treatments. Eighty-four F tests were performed and the sequential Bonferroni correction was applied to keep a table-wide α of 0.05 (Rice, 1989).
Multivariate and univariate analyses were performed using Proc GLM in SAS 9.1 (SAS Inc. Cary, NC, USA). As plant mortality resulted in an unbalanced design, the two genotype × treatment replicates within a block were averaged to enable analysis by Proc GLM. Multivariate analysis of variance (MANOVA) was first performed to identify whether there were significant effects of genotype, treatment, and genotype × treatment interaction for a suite of reproduction-related traits, including rosette mass, inflorescence height, lateral branch number and fruit number. A second MANOVA was performed using data only from the three blocks for which the seed number per fruit and total seed production were calculated. The second MANOVA included the following response variables: rosette mass, inflorescence height, lateral branch number, fruit number, seed number per fruit and total seed production. Univariate ANOVAs were performed following the MANOVAs in order to determine which traits showed significant variation for phenotypic plasticity (genotype × treatment interaction). Flowering time was analyzed using ANOVA only. For both the multivariate and univariate analyses, the split-plot experimental design required that the whole-plot factor, treatment, be tested over the whole-plot random error term, treatment × block. Genotype and genotype × treatment interaction were tested over the residual error. Data and residuals were examined to ensure that all ANOVA assumptions were satisfied. Fruit number and total seed production required transformation (square root) in order to satisfy ANOVA assumptions of normality and stable variance. Square root transformation was used, because it is the recommended method for normalizing counted data (Kuehl, 2000) and was the most effective transformation for meeting the ANOVA assumptions. The sequential Bonferroni correction was applied to the univariate P values to minimize inflation of table-wide error from multiple tests (Rice, 1989).
To determine whether G-protein mutants showed significantly different plasticities relative to Col, contrasts were performed using SAS 9.1 to test a priori-selected comparisons on all traits which had a significant genotype × water level interaction. Because two alleles of each mutant were studied, the contrasts were designed to simultaneously test the plasticity of Col against the plasticities of both mutant alleles of each gene. Combining alleles limited the inflation of α and, at the same time, increased the biological validity of the experiment: if the two mutant alleles of the same gene showed divergent responses, statistically significant differences with Col would probably not be detected. However, reaction norms were also examined individually to ensure that alleles of the same gene had similar plasticity responses.
The following contrasts were performed for the univariate analyses: Col against both alleles combined for gpa1, agb1 and gcr1 mutants for ample water vs moderate drought and ample water vs severe drought. The double mutant gpa1-4agb1-2 was tested against Col, and against both alleles combined of gpa1 or agb1 for ample water vs moderate drought and ample water vs severe drought. The sequential Bonferroni correction was applied to adjust for the inflation of type 1 error and to maintain a table-wide α of 0.05 (Rice, 1989). It has been suggested that the sequential Bonferroni correction can be overly stringent when applied to ecological experiments; therefore, we also applied biological reasoning when interpreting each contrast (Moran, 2003). Because relative fitness (Stanton & Thiede, 2005) was a post hoc addition to our analysis, relative fitness was analyzed independently from the other response variables. ANOVA was performed to test whether genotype, treatment and genotype by treatment interaction were significant.
Variation for plasticity among genotypes (genotype × environment interactions)
Both MANOVAs showed that there is significant genotype × treatment interaction (Tables 1 and 2) for the reproduction-related traits (both P values were < 0.0001). Significant genotype and treatment effects were also observed; however, the treatment effect for the seven variable MANOVA performed on only three blocks (Table 2) could not be estimated as a result of insufficient error degrees of freedom. The MANOVAs indicated that significant genetic variation exists for reproductive trait plasticity. Univariate ANOVAs were then performed in order to determine which specific traits showed significant genotype × treatment interaction. The results from the univariate ANOVAs are shown in Table 3. We found significant genotype × environment interactions for all traits, even after the sequential Bonferroni correction was applied to the P values. Least-squares means and standard errors are listed for all alleles and water levels in Table S1 (see Supporting Information). F tests of specific contrasts between any of the G-protein mutants and Col for ample water vs moderate drought and ample water vs severe drought revealed that, for rosette mass, number of lateral branches, total seed production and flowering time, there were no significant differences in genotype × water level interactions (plasticities) after the application of the sequential Bonferroni correction (Table 4). However, significant differences in plasticities were observed between Col and the G-protein mutants for a number of reproduction-related traits (Table 4). Although the contrasts were performed on pooled alleles, reaction norms for mutant alleles of the same gene were examined in cases in which the contrasts were significant; such examination confirmed that both alleles of each gene did indeed show similar responses for all plasticity traits discussed here (see, for example, Figs S4–S7; Supporting Information). Interestingly, agb1 mutants showed significantly reduced plasticities relative to Col for inflorescence height (Fig. 3), fruit number (Fig. 4) and seeds per fruit (Fig. 5), suggesting that AGB1 functions as a plasticity gene, mediating phenotypic plasticity in the reproductive phase of plant growth in response to water availability. The plasticities of the gpa1-4 agb1-2 double mutant for the most part resembled those of the single agb1 mutants. The exception was the inflorescence height, for which agb1 mutants showed significantly reduced plasticity relative to the double mutant (Fig. 3, Table 4). gpa1 mutants showed increased plasticity for inflorescence height relative to Col (Fig. 3). The double mutant showed reduced plasticity relative to gpa1 for inflorescence height (Fig. 3), indicating that the double mutant shows an intermediate phenotype. gcr1 mutants showed increased plasticity for square root fruit number relative to the wild-type (Fig. 4).
Table 1. Wilks’ lambda, F and P values from multivariate analysis of variance (MANOVA) including the following response variables: rosette mass, inflorescence height, number of lateral branches and square root number of fruits; data from all 12 experimental blocks
Genotype × treatment
Table 2. Wilks’ lambda, F and P values from multivariate analysis of variance (MANOVA) including the following response variables: rosette mass, inflorescence height, number of lateral branches, square root number of fruits, seed number per fruit and square root total seed production; data from the three experimental blocks for which seed number per fruit was obtained
1, Multivariate analysis of the significance of the treatment effect could not be performed because of insufficient error d.f. (treatment × block is the appropriate error term given the split-plot experimental design).
Genotype × treatment
Table 3. F and (P values) for all fixed effects for univariate analyses of variance (ANOVAs); treatment was tested over the treatment × block error term, and genotype and genotype × treatment were tested over the residual error; all P values were significant before and after sequential Bonferroni correction
Genotype × treatment
1, Only three blocks were used to assess seeds per fruit and seed production.
28.08 (≤ 0.0001)
744.91 (≤ 0.0001)
8.43 (≤ 0.0001)
40.89 (≤ 0.0001)
541.7 (≤ 0.0001)
12.79 (≤ 0.0001)
Number of lateral branches
108.9 (≤ 0.0001)
272.7 (≤ 0.0001)
6.06 (≤ 0.0001)
Square root number of fruits
9.35 (≤ 0.0001)
320.91 (≤ 0.0001)
13 (≤ 0.0001)
Seeds per fruit1
10.47 (≤ 0.0001)
256.7 (≤ 0.0001)
4.23 (≤ 0.0001)
7.79 (≤ 0.0001)
220.07 (≤ 0.0001)
50.37 (≤ 0.0001)
Table 4. P values of univariate contrasts on genotype × treatment means (ample, ample water; moderate, moderate drought; severe, severe drought)
No. lateral branches
Sq. rt. no. fruit
Seeds per fruit1
Total seed production1
1, Data from three blocks only.
Values in bold were significant before application of the sequential Bonferroni correction; values in italic and with underlines were significant after application of the Bonferroni correction.
Col vs both agb1 ample vs moderate
Col vs both agb1 ample vs severe
Col vs double ample vs moderate
Col vs double ample vs severe
Col vs both gcr1 ample vs moderate
Col vs both gcr1 ample vs severe
Col vs both gpa1 ample vs moderate
Col vs both gpa1 ample vs severe
Double vs both agb1 ample vs moderate
Double vs both agb1 ample vs severe
Double vs both gpa1 ample vs moderate
Double vs both gpa1 ample vs severe
For total seed production, no significant differences were found in the plasticities of Col and any of the G-protein mutants following the sequential Bonferroni correction (Table 2). However, it should be noted that, before statistical correction, contrasts between Col and agb1 in ample water vs moderate drought (P =0.0027) and ample water vs severe drought (P =0.003) were significant. Given that two independent alleles of agb1 were used in the experiment and their reaction norms showed reduced plasticity for total seed production (Fig. 6; see also Fig. S7, Supporting Information) relative to Col, and also that the sequential Bonferroni correction can be overly conservative (Moran, 2003), we feel confident that AGB1 also mediates plasticity for total seed production. All G-protein mutants showed reduced fitness, defined here as total seed production, under ample water conditions, but the agb1 mutants and the double mutant gpa1agb1 showed increased fitness under both moderate and severe drought stress relative to Col (Fig. 6). The relative fitness data also corroborate rank changing among the genotypes in different environments (Fig. 7). A significant genotype effect (P <0.0001) and significant genotype × treatment interaction (P =0.0009) were observed for relative fitness. Although all mutant genotypes showed reduced relative fitness relative to Col under ample water growth conditions, agb1 mutants and the double mutant gpa1-4agb1-2 showed increased relative fitness relative to Col under drought stress.
Phenotypic variance of within genotype × treatment trait means
The within-treatment, experiment-wide variances of the trait means for the mutants and wild-type for ample water and severe drought treatments are shown in Table S2 (see Supporting Information). Equal variance tests were performed to determine whether the mutants showed increased or decreased phenotypic variation relative to the wild-type within a particular environment (Table S3). After correcting for multiple F tests using the sequential Bonferroni correction, only one null hypothesis was rejected. agb1-2 mutants showed significantly reduced variance (P <0.0005) for seed number per fruit under ample water relative to the wild-type. However, this reduction in variance was not observed in the second agb1 allele, agb1-1 (P =0.168), which brings into question the biological relevance of this statistically significant observation. Overall, phenotypic variance for traits within a given environment was not impacted significantly by G-protein mutations.
AGB1 functions as a plasticity gene for a number of reproduction-related traits
Our data show that mutation of the sole Gβ subunit of Arabidopsis, AGB1, results in pleiotropic effects on the extent of plasticity in response to water availability. agb1 mutants showed reduced plasticity for the number of fruits, inflorescence height, seed number per fruit and total seed production relative to Col. Interestingly, agb1 mutants showed increased seed production per fruit under drought stress relative to the wild-type, and reduced seed production per fruit relative to the wild-type under well-watered conditions. The reduced plasticity of agb1 mutants resulted in enhanced fitness under drought stress, but was maladaptive under well-watered conditions. This conclusion was also supported when relative fitness was assessed. All G-protein mutants, that showed reduced plasticity for reproductive traits, including agb1, showed lower relative fitness relative to Col under ample water conditions. Under drought stress, agb1 mutants and the double mutant showed increased relative fitness relative to Col. The relative fitness data support the conclusion that the significant genotype × treatment interaction observed for total seed production can be attributed to rank changing among the genotypes, and is not a consequence of changes in variance or the square root transformation. It has also been noted that differences in phenotypic variation can be a consequence of age- or size-dependent ontogenetic drift (McConnaughay & Coleman, 1999). It should be noted that we observed no significant differences in rosette growth rate (as determined by projected leaf area calculations) among the genotypes within a given water treatment (Figs S1–S3, see Supporting Information), suggesting that the significant variation for plasticity observed cannot be attributed to ontogenetic drift. These results suggest that AGB1 is a plasticity gene, as it contributes to the shape of the phenotypic response under the environments tested in our experiment.
Multiple alleles have not been used frequently in mutant studies of phenotypic plasticity; one exception is the study by Galen et al. (2004). The use of two independent mutant alleles of agb1 in this study strengthens our conclusion that AGB1 functions as a plasticity gene, at least in the environments tested. These results suggest that it would be worthwhile to pursue additional experiments to determine the extent to which G proteins function in regulating phenotypic plasticity under different stresses and in natural environments, where other resource limitations might influence the extent of plasticity observed. It has been recognized that there may also be costs associated with plasticity under some circumstances (Callahan et al., 2005; Ghalambor et al., 2007; Van Buskirk & Steiner, 2009). Although additional research under different environments is required, plasticity costs may be illustrated under our conditions by the fact that, under water limitation, the least plastic genotype, agb1, showed greater fitness (by both absolute and relative measures) than the more plastic genotypes.
AGB1 has been shown previously to function in inflorescence and fruit development. agb1-1 was originally isolated in a screen for erecta-like mutations, where it showed a slightly reduced inflorescence height and significantly shortened, blunt-tipped fruits relative to the wild-type, phenotypes which were later also observed in agb1-2 (Lease et al., 2001; Ullah et al., 2003). In addition, it was shown that AGB1 was expressed ubiquitously throughout the plant, but its expression was elevated in flowers and highest in fruits (Lease et al., 2001). The shortened fruit phenotype corresponds to our findings that agb1 shows reduced seeds per fruit. This short phenotype is specific to agb1 mutants and is not observed in gpa1, agg1 and agg2 mutants (Ullah et al., 2003; Trusov et al., 2008). Functional selectivity of the Gβ subunit has been reported for other G-protein-mediated responses, including necrotrophic pathogen resistance (Llorente et al., 2005; Trusov et al., 2006), sugar inhibition of seed germination and lateral root formation (Chen et al., 2006).
According to the paradigm of G-protein signaling, activation of GPA1 results in a conformational change in Gα and the release of the Gβγ dimer, and signal propagation. As the reduced plasticity phenotype is present in agb1 mutants and in gpa1agb1 double mutants, but not in gpa1 mutants, we can conclude that Gβ is responsible for signal integration or transduction, resulting in wild-type plasticity for the number of fruits, seed number per fruit and total seed production. In classical G-protein signaling, the α subunit typically requires the β subunit, not only for trimer reassembly, but also for GPCR association. Therefore, GPA1 activation is eliminated by both gpa1 and agb1 mutations, but AGB1 activation is abolished only by the agb1 mutation. For inflorescence height, gpa1 and agb1 display opposite phenotypes (enhanced and reduced plasticity, respectively), which also suggests that AGB1 is responsible for mediating wild-type plasticity, as free AGB1 is active in gpa1 mutants (potentially more so than in the wild-type), but is absent from agb1 mutants. gcr1 mutants showed increased plasticity relative to the wild-type for fruit number and, although GCR1 might function in the mediation of plasticity for fruit number, there are probably additional unknown GPCRs that contribute to the perception and/or integration of environmental input signals with regard to the regulation of phenotypic plasticity.
G-protein mutations do not affect significantly the level of phenotypic variation within an environment
Within a given environment, genetically identical organisms can show divergent phenotypes as a result of stochasticity in gene expression. These random events among cellular molecules can modify cell status and consequently result in phenotypic changes at the organismal level (often thought of as experimental ‘noise’). Stochasticity is generally thought to be detrimental to fitness; however, it can also be a source of heterogeneity, which can provide a fitness benefit in fluctuating environments (Kaern et al., 2005; Raser & O’Shea, 2005). A G-protein mutation might increase or decrease the ability of a cell to buffer itself against stochasticity, and therefore it has been hypothesized that G-protein mutations might affect the amount of phenotypic variation within an environment (Assmann, 2004). To test this hypothesis, we compared the phenotypic variances (Table S2) of the G-protein mutants within an environment with the phenotypic variance of the wild-type within the same environment. We found that the within-environment phenotypic variation was not altered significantly by mutations in G-protein subunit genes (Table S3, Supporting Information), suggesting that this hypothesis is not supported: G-protein mutations do not affect the range of possible phenotypes within our environments.
agb1 mutants show enhanced fitness under drought stress
Based on their previously reported altered stomatal sensitivities to ABA (Wang et al., 2001; Pandey & Assmann, 2004; Fan et al., 2008), it is possible to make predictions concerning the fitness benefits or costs that G-protein mutants may incur when grown under drought stress conditions. However, although gpa1 and agb1 are hyposensitive with regard to the ABA inhibition of stomatal opening, but show wild-type ABA promotion of stomatal closure (Wang et al., 2001; Fan et al., 2008), and gcr1 mutants are hypersensitive towards both ABA inhibition of stomatal opening and promotion of closure (Pandey & Assmann, 2004), our detailed phenotypic analysis of whole-plant traits under controlled water stress conditions did not support the predictions that would be made based on these stomatal response phenotypes. gcr1 mutants, despite their stomatal hypersensitivity to ABA and their reported improved recovery following drought stress (Pandey & Assmann, 2004), showed no fitness advantage relative to the wild-type under drought stress or well-watered conditions. Although we predicted that gpa1 and agb1 mutants would show reduced fitness under drought stress based on the partial ABA insensitivity of their stomatal phenotypes, agb1 mutants (but not gpa1 mutants) exhibited increased fitness under drought stress and reduced fitness under well-watered conditions. The pleiotropic nature of G-protein mutations may, in part, explain why the stomatal response phenotypes were not predictive of plant fitness under drought stress. Recently, it has been shown that G proteins regulate stomatal density, GPA1 as a positive modulator and AGB1 as a negative regulator (Zhang et al., 2008a), which may have implications for water loss and carbon assimilation under different environmental conditions. In addition, agb1 mutants and, to a lesser degree, gpa1 mutants are hypersensitive to ABA inhibition of germination and root growth (Pandey et al., 2006), which may contribute to survival under drought stress. Nevertheless, although the lack of plasticity of agb1 in reproductive-related traits resulted in a fitness advantage under drought stress, lack of plasticity was maladaptive for agb1 under well-watered conditions. Therefore, G proteins do contribute to plant fitness, in part via the regulation of phenotypic plasticity. Given that climate change can lead to increased variability in environments and resources, this finding could have important agronomical implications, as agb1 mutants show reduced plasticity and therefore more stable yields across environments.
Phenotypic analysis of G-protein mutants under multiple environmental conditions has identified novel functions of plant heterotrimeric G proteins in the regulation of the phenotypic plasticity of inflorescence height, seed number per fruit, fruit number and total seed production. We also found that the known altered guard cell sensitivities towards ABA did not predict the fitness outcomes of the mutants under drought stress. All G-protein mutants studied showed reduced fitness under well-watered as well as drought environments, with the exception of agb1-containing mutants which showed improved fitness under drought stress conditions. These results thus attest to one of the apparent paradoxes of plant G-protein signaling: although G-protein mutation is not lethal, it does, in fact, result in nonoptimal phenotypes under some environmental conditions.
G proteins mediate responses to ABA, auxin, brassinosteroids, gibberellins and sugars, environmental signals, including ozone and pathogens, and intrinsic, unknown developmental cues mediating leaf shape, stomatal density and fruit shape. A fundamental question in G-protein signaling is how only two heterotrimeric G-protein combinations (GPA1/AGB1/AGG1 and GPA1/AGB1/AGG2) can transduce such a diversity of signals. Our results, implicating heterotrimeric G proteins in the mediation of phenotypic plasticity responses, support the model that G proteins function, at least in part, as cross-talk hubs, integrating signals, rather than directly transducing them (Assmann, 2004), thereby tweaking a phenotype relative to the environment at hand. More studies are warranted in which G-protein regulation of plasticity is examined across additional environmental gradients, as well as across generations, in order to further elucidate this novel contribution of heterotrimeric G proteins to plant development and fitness. It will be interesting to assess the extent to which the results presented here would extrapolate to field studies: studies in highly controlled environments are important steps towards the design of such experiments. Although QTL-based studies have supported the existence of plasticity genes, the present study is one of a few that has directly tested the regulation of phenotypic plasticity by specific genes. Additional plasticity studies with other environmental signaling mutants will be integral to our understanding of whether plasticity genes are rare or common in plant and other genomes, and to gaining an insight into the genetic basis of phenotypic plasticity.
We thank Dr Naomi Altman, Associate Professor of Statistics, Penn State University, for her guidance on MANOVA analyses and data transformation.