An atypical heterotrimeric G-protein γ-subunit is involved in guard cell K+-channel regulation and morphological development in Arabidopsis thaliana


  • David Chakravorty,

    1. Plant Genetic Engineering Laboratory, School of Agriculture and Food Sciences, University of Queensland, Brisbane, Qld 4072, Australia
    Search for more papers by this author
    • Present address: Biology Department, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA 16802, USA.

  • Yuri Trusov,

    1. Plant Genetic Engineering Laboratory, School of Agriculture and Food Sciences, University of Queensland, Brisbane, Qld 4072, Australia
    Search for more papers by this author
  • Wei Zhang,

    1. Biology Department, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
    • Present address: The Key Laboratory of Plant Cell Engineering and Germplasm Innovation, Ministry of Education; School of Life Science, Shandong University, Jinan, Shandong, 250100, China.

  • Biswa R. Acharya,

    1. Biology Department, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
  • Michael B. Sheahan,

    1. School of Environmental and Life Sciences, The University of Newcastle, Callaghan, NSW 2308, Australia
    Search for more papers by this author
  • David W. McCurdy,

    1. School of Environmental and Life Sciences, The University of Newcastle, Callaghan, NSW 2308, Australia
    Search for more papers by this author
  • Sarah M. Assmann,

    1. Biology Department, 208 Mueller Laboratory, Pennsylvania State University, University Park, PA 16802, USA
    Search for more papers by this author
  • José Ramón Botella

    Corresponding author
    1. Plant Genetic Engineering Laboratory, School of Agriculture and Food Sciences, University of Queensland, Brisbane, Qld 4072, Australia
      (fax +61 7 3365 1699; e-mail
    Search for more papers by this author

(fax +61 7 3365 1699; e-mail


Currently, there are strong inconsistencies in our knowledge of plant heterotrimeric G-proteins that suggest the existence of additional members of the family. We have identified a new Arabidopsis G-protein γ-subunit (AGG3) that modulates morphological development and ABA-regulation of stomatal aperture. AGG3 strongly interacts with the Arabidopsis G-protein β-subunit in vivo and in vitro. Most importantly, AGG3-deficient mutants account for all but one of the ‘orphan’ phenotypes previously unexplained by the two known γ-subunits in Arabidopsis. AGG3 has unique characteristics never before observed in plant or animal systems, such as its size (more than twice that of canonical γ-subunits) and the presence of a C-terminal Cys-rich domain. AGG3 thus represent a novel class of G-protein γ-subunits, widely spread throughout the plant kingdom but not present in animals. Homologues of AGG3 in rice have been identified as important quantitative trait loci for grain size and yield, but due to the atypical nature of the proteins their identity as G-protein subunits was thus far unknown. Our work demonstrates a similar trend in seeds of Arabidopsis agg3 mutants, and implicates G-proteins in such a crucial agronomic trait. The discovery of this highly atypical subunit reinforces the emerging notion that plant and animal G-proteins have distinct as well as shared evolutionary pathways.


Heterotrimeric GTP-binding proteins (G-proteins) are signal transduction components that mediate the action of integral membrane receptors in most eukaryotic organisms (New and Wong, 1998). G-proteins consist of alpha (α), beta (β) and gamma (γ) subunits, and commonly localize to the cytoplasmic face of the plasma membrane. G-proteins transduce signals from membrane-spanning G-protein-coupled receptors (GPCRs), mediating a variety of intracellular responses (McCudden et al., 2005; for review). In plants few G-protein-mediated molecular pathways have been elucidated to this point; however, several phenotypes have been described in G-protein mutants. G-proteins have been implicated in developmental processes controlling leaf shape, hypocotyl length (Ullah et al., 2001), flower and silique shape (Lease et al., 2001), stomatal density (Zhang et al., 2008a), cell proliferation (Izawa et al., 2010) pollen tube growth (Wu et al., 2007), lateral root formation (Ullah et al., 2003) and root apical meristem growth (Chen et al., 2006), as well as germination responses to glucose, and hormones such as gibberellins, brassinosteriods and abscisic acid (ABA; Ullah et al., 2002; Oki et al., 2009). G-proteins have also been implicated in ABA regulation of post-germination development (Pandey et al., 2006) and blue-light-induced production of phenylalanine (Warpeha et al., 2006, 2007). Roles in processes of crucial agronomic importance have also been reported for G-proteins, including ABA-mediated regulation of stomatal aperture to control water loss (Wang et al., 2001), and resistance to some necrotrophic pathogens (Trusov et al., 2006) by a mechanism independent of the salicylic acid-, jasmonic acid-, ethylene- and ABA-mediated pathways (Trusov et al., 2009).

Mammalian cells employ numerous combinations of subunits to distinguish between different signals with 23 α-, 5 β- and 12 γ-subunits reported in humans (McCudden et al., 2005). In contrast, one canonical α-subunit (GPA1), one β-subunit (AGB1) and two γ-subunits (AGG1 and AGG2) have been identified in the model plant Arabidopsis thaliana (Ma et al., 1990; Weiss et al., 1994; Mason and Botella, 2000, 2001). Such a disparity of subunit combinations leads to questions about to how plant G-proteins can be involved in so many distinct processes. Furthermore, Gβγ acts as a functional monomer according to the prevailing paradigm of G-protein signalling and any process mediated by AGB1 should require a γ-subunit; therefore simultaneous knockout of the two γ-subunits in Arabidopsis should phenocopy β-null (agb1) mutants. However, we have previously reported this is not the case. An agg1agg2 double mutant did not display a number of well-documented gpa1 or agb1 phenotypes (Trusov et al., 2008). The most likely explanation for this disparity is the existence of additional undiscovered γ-subunits. However, even when accounting for an extreme primary sequence variability between γ-subunits, homology-based searches had previously failed to reveal any γ-like proteins of a similar size to AGG1 and AGG2.

We present here the characterization of a novel γ-like gene, AGG3, in Arabidopsis that represents a new class of G-protein γ-subunits specific to plants but not present in animals. We show that AGG3 binds AGB1 in vivo and in vitro. Additionally, agg3 T-DNA mutants reproduce all but one of the ‘orphan’agb1 mutant phenotypes not previously observed in agg1, agg2 or agg1agg2 mutants. Our study shows that AGG3 is involved in the regulation of stomatal aperture by ABA as well as developmental processes.

Results and discussion

Identification and bioinformatic analysis of a new putative G-protein γ-subunit in Arabidopsis

Two γ-subunits (AGG1 and AGG2) have previously been characterized in Arabidopsis but fail to explain all the phenotypes shown by β-deficient mutants (Mason and Botella, 2000, 2001; Trusov et al., 2007, 2008). Aside from AGG1, BLAST searches using AGG2 as a query identified an additional unknown protein (over twice as large as AGG1 and AGG2), encoded by the Arabidopsis gene At5g20635 (GenBank locus AAT85756), with a relatively low level of homology (score 40.8 and expect value 3 × 10−4 over a 55-amino-acid region). Alignment of AGG1, AGG2 and AAT85756 showed that homology to the two known Arabidopsis γ-subunits resides in the N-terminal region of AAT85756, while the C-terminal region is extremely rich in cysteine residues, with 44 Cys in the last 128 residues (Figure 1a).

Figure 1.

 AGG3 contains an N-terminal γ-domain with homology to AGG1 and AGG2.
(a) Sequence alignment of AGG1, AGG2 and AAT85756. Intron positions are indicated by an asterisk within each sequence. Boxes indicate sequence features: 1, coiled-coil motif predicted by the web-based COILS software (Lupas et al., 1991); 2, DPLL/I box; 3, C-terminal CaaX motif. AGG3 truncation constructs (trunc1 and trunc2) used for complementation are indicated by black arrows and the dotted line indicates the region of the protein not included in the trunc2 construct.
(b) T-DNA insertion sites in the agg3 mutants. Grey boxes represent exons. Arrows show positions of forward and reverse primers used for real-time PCR. T-DNA insertions are not drawn to scale. T-DNA left (LB) and right (RB) border orientations were determined by PCR and are shown.
(c) Real-time PCR analysis of the mutant lines. Total RNA extracted from flowers was used for cDNA synthesis. Forward and reverse primers depicted in (b) were used to perform relative quantification. Arabidopsis ACTIN 2 was used as a control.

At5g20635 contains an open reading frame of 753 bp encoding a theoretical 251-amino-acid protein with a predicted molecular weight of 27.2 kDa. AAT85756 possesses several important features that qualify it as a putative γ-subunit. Six out of the eight AGG1 residues identified by Temple and Jones (2007) as being important for contact with AGB1 are conserved (L37, E40, S51, D66, P67 and L68), one is conservatively substituted (L69), and only N77 is not conserved in AAT85756 (highlighted in Figure 1a and Figure S1 in Supporting Information for homologues). Secondly, AAT85756 is predicted to contain an N-terminal coiled-coil domain (box 1 in Figure 1a), an important structural characteristic of γ-subunits providing strength to β/γ dimerization (Pellegrino et al., 1997; McCudden et al., 2005). Thirdly, the positions of the first three introns in AGG1, AGG2 and At5g20635 are identical (asterisks in Figure 1a), suggesting that At5g20635 may be the result of an ancient gene duplication event. Fourthly, AAT85756 contains a C-terminal isoprenylation (CaaX) motif (box 3 in Figure 1a), an important element conserved within G-protein γ-subunits (Simonds et al., 1991; Chakravorty and Botella, 2007). Finally, using Phyre, a software package that predicts the most similar available tertiary structures to a query using primary sequence information, the only hits with significant homology to AAT85756 (40–55% estimated precision), were γ-subunit structures from the heterotrimeric G protein (with no other hits over 20%;; Kelley and Sternberg, 2009). We therefore decided to explore the possibility that AAT85756 might function as an unconventional G-protein γ-subunit, and tentatively named it AGG3.

AGG3 shows strong interaction with the Arabidopsis β-subunit

One of the most important features characterizing γ-subunits is their strong interaction with β-subunits that can only be disrupted under denaturing conditions (Pellegrino et al., 1997). Initial proof of the interaction between AGG3 and AGB1 was provided by blue/white colony screening using the yeast two-hybrid (Y2H) system. The strength and nature of the interaction was further studied by performing quantitative Y2H β-galactosidase activity assays using AGB1 and several N- and C-terminal AGG3 deletions (Figure 2a,b). Full-length AGG3 showed a strong affinity for AGB1 (compared to the standard p53/SV40 positive control; Li and Fields, 1993), and the γ-domain (AGG31−135) showed higher affinity for AGB1 than either AGG1 or AGG2. Conversely, the C-terminal Cys-rich domain of AGG3 (AGG3136–251) displayed no affinity for AGB1. Residues 1–78 of the AGG3 γ-domain, containing the coiled-coil domain, were not sufficient to bind AGB1; however, their removal did significantly decrease the signal from the β-galactosidase reporter gene (see AGG379−251 in Figure 2b; = 0.0148 when compared to full-length AGG3). This suggests that the AGB1/AGG3 interaction is not due to non-specific binding of the coiled-coil domains, rather the coiled-coil interaction enhances the strength of the binding. Comparison of AGG31−78 and AGG31−99 indicates that the region containing residues 78–99 is important for binding to the β-subunit. This region most notably contains the highly conserved DPLL/I motif (box 2 in Figure 1a) which has been identified as an important contact patch between β- and γ-subunits (Temple and Jones, 2007). The interaction observed between AGG3 and AGB1 was abolished by the overexpression of either AGG1 or AGG2 in the same yeast cells, indicating that AGG1 and AGG2 compete with AGG3 for the same binding sites on AGB1 (Figure 2e). No interaction was detected between AGG3 and either AGG1, AGG2 or GPA1. Confirmation of the AGB1/AGG3 interaction was also provided using the mating-based split-ubiquitin system (see the NubG-AGG3/AGB1-Cub interaction in Figure 2c).

Figure 2.

 AGG3 interacts with AGB1 in vivo and in vitro.
(a) Sequence features of AGG3 and schematic of AGG3 deletion constructs used in yeast two-hybrid. Residue numbers are presented with deletions.
(b) Quantitative measurements of yeast two-hybrid interactions. AGB1 was co-transformed with full-length γ-subunits (dark grey bars) or deletions (light grey bars) of AGG3. Values are mean ± SE of o-2-nitrophenyl-β-d-galactopyranoside (ONPG) assays of β-galactosidase activity conducted on at least two independent colonies assayed in duplicate or triplicate. p53/SV40 and p53/AGG3 (white bars) were used as positive and negative controls, respectively.
(c) Interaction of AGB1–AGG3 using the split-ubiquitin system, determined by growth on interaction selective media containing 500 μm methionine. Yeast carrying NubG-AGG3/AGB1-Cub and the positive control NubWT-AGG3/AGB1-Cub grow, indicating interaction, while the negative control (NubG/AGB1-Cub) does not.
(d) An in vitro binding assay of AGB1/AGG3 confirms that AGG31–112 binds to a GST-AGB1 fusion protein, but not to GST alone.
(e) Yeast three-hybrid system interaction assays using AGB1 fused to the GAL4-binding domain and AGG3 fused to the GAL4-activation domain (i), and with the additional expression of AGG1 (ii) or AGG2 (iii) as competitors of AGB1-binding.

The interaction between the N-terminal fragment of AGG3 (residues 1–112), containing the entire γ-domain, and AGB1 was also confirmed using in vitro binding assays. [35S]-Methionine-labelled AGG31−112 was found to bind an immobilized glutathione S-transferase (GST)–AGB1 fusion protein, but not GST (Figure 2d).

AGG3 is involved in the control of guard cell ion channels and germination responses to ABA

In order to test if agg3 mutants could account for the ‘orphan’agb1 phenotypes not explained by mutants in the two known γ-subunits (Trusov et al., 2008), three insertion mutants (agg3-1, agg3-2 and agg3-3) were obtained from publicly available T-DNA collections and the insertion sites (Figure 1b) confirmed by sequencing. We initially used Northern blotting and RT-PCR (using multiple primer pairs) to confirm agg3-1 and agg3-2 were transcript null alleles. However, we obtained contradictory results for agg3-3, and subsequently demonstrated by quantitative (q)PCR that agg3-3 produced approximately 2.5% of the wild-type level of AGG3 transcript in flowers, while agg3-1 and agg3-2 are transcript null alleles. (Figure 1c).

We previously observed that gpa1 and agb1 mutants exhibit ABA-hyposensitivity in several aspects of guard cell response, including ABA-inhibition of stomatal opening and ABA-inhibition of the inward K+-channels which mediate the K+-influx that drives water uptake into guard cells and stomatal opening (Wang et al., 2001; Fan et al., 2008). Activation by ABA of the slow anion channels that mediate anion loss during stomatal closure was also abrogated in gpa1 and agb1 mutants under conditions where strong buffering of cytosolic pH was applied, but not under conditions of weak pH buffering. While agg1 and agg2 single and double mutants exhibited wild-type stomatal responses to ABA (Trusov et al., 2008), in all three agg3 mutant lines stomatal opening and inward K+ currents were clearly hyposensitive to ABA (Figure 3a,c,d and Figure S2a). As observed in gpa1 and agb1 mutants, but not in agg1 and agg2 mutants, anion currents in agg3 mutants were hyposensitive to ABA under strong cytosolic pH buffering (Figure 3e,f) but showed wild-type ABA responsiveness under weak cytosolic pH buffering conditions (Figure S2c,d). In contrast to the results observed for inhibition of stomatal opening, ABA-mediated promotion of stomatal closure was wild-type in the agg3 lines (Figure 3b and Figure S2b) as well as in all other subunit mutants (Wang et al., 2001; Fan et al., 2008; Trusov et al., 2008).

Figure 3.

agg3 mutants are hyposensitive to ABA inhibition of both stomatal opening and guard cell K+in currents, and hypersensitive to ABA during germination.
(a) Inhibition of light-induced stomatal opening by 20 μm ABA in wild-type (Col-0) and agg3 mutants. Data are the mean of five replicates and ≥120 stomatal apertures were measured for each experiment.
(b) Induction of stomatal closure in wild-type (Col-0) and agg3 mutants in response to 20 μm ABA. Data are the mean of three replicates and ≥150 stomatal apertures were measured in each replicate.
(c) Typical whole-cell K+ current traces of guard cells with or without 50 μm ABA treatment. Time and voltage scales as shown in the top right panel apply to all measures.
(d) Current/voltage relationship of time-activated whole-cell K+ currents as illustrated in panel (c). The following voltage family was used for whole-cell K+ current recordings: the holding potential was −79 mV and the test voltage steps were from −219 mV to 61 mV with +20 mV increments. A duration of 3.9 sec was used for each test voltage. All the recordings used for the analysis were obtained 10 min after the formation of the whole-cell configuration. Numbers of guard cells analysed were: Col-0 (20), Col-0 + ABA (32), agg3-1 (18), agg3-1 + ABA (16), agg3-2 (14), agg3-2 + ABA (17), agg3-3 (12), agg3-3 + ABA (20).
(e) Typical whole-cell anion currents in guard cells (Col-0 and agg3 mutants) with or without 50 μm ABA treatment using strong pH buffering in the pipette (internal) solution. Time and voltage scales as shown in the top left panel apply to all measures.
(f) I/V curves of steady-state whole-cell anion currents as recorded in (e). = number of cells assayed. Col-0 control (= 8) and ABA treatment (= 8); agg3-1 control (= 6) and ABA treatment (= 7); agg3-2 control (= 7) and ABA treatment (= 8); agg3-3 control (= 8) and ABA treatment (= 8). Symbols for genotype and treatment are as designated in panel (e). The following voltage family was used for anion channel recordings: the holding potential was +30 mV and voltage steps were applied from −145 to +35 mV with +30 mV increments, with a duration of 50 sec for each test voltage. Whole-cell anion currents were recorded 12 min after achievement of the whole-cell configuration and used for analysis.
(g, h) Germination of selected genotypes on 0.5 × MS supplemented with (g) 2 μm and (h) 5 μm ABA (without sugar). After 2 days of stratification, seeds were transferred to a growth cabinet and germination scored daily. *Significant differences from Col-0 under the same treatment in (a), (b), (g) and (h) (*P < 0.05). All values are shown as the mean ± SE.

It is well established that Gβγ dimers directly regulate K+-channels in mammals (Reuveny et al., 1994; Krapivinsky et al., 1995). Since AGG3 is involved in stomatal regulation and agg3 mutants show a clear loss of responsiveness of guard cell K+-channels to ABA, we tested the ability of AGG3 to interact with Shaker-type K+-channel proteins of Arabidopsis, which are voltage-gated inward- or outward-rectifying channels that, among other roles, mediate guard cell K+-fluxes (Lebaudy et al., 2007). No conclusive positive results were obtained using the yeast split-ubiquitin system (Table S1), therefore we cannot demonstrate at this time a direct physical interaction between AGG3 and any of the channels studied. Nevertheless, alternative protein interaction methods such as bimolecular fluorescence complementation might be more useful in proving such interaction and it cannot be ruled out that the full GPA1/AGB1/AGG3 heterotrimer could be required for the interaction to occur.

Furthermore, the hypersensitivity to ABA during germination reported for agb1 and gpa1 mutants (Pandey et al., 2006) and not observed in agg1 or agg2 mutants (Trusov et al., 2008) is clearly visible in all three agg3 lines (Figure 3g,h). Small amounts of sucrose can rescue ABA-mediated inhibition of germination (Garciarrubio et al., 1997; Price et al., 2003), with a stronger rescue effect observed in gpa1, agb1 and all three agg3 line mutants than in the wild-type (Figure S2g) or the agg1 and agg2 mutants (Trusov et al., 2008). All lines used in this experiment germinated uniformly on standard 0.5 × MS (Figure S2h), confirming the ABA-relatedness of the mutant phenotypes.

AGG3-deficient mutants recapitulate developmental phenotypes shown by AGB1, but not AGG1 and/or AGG2 mutants

We have previously reported that AGG1 and AGG2 expression matched AGB1 expression in most tissues except flowers where AGG1 and AGG2 expression was almost nonexistent (Anderson and Botella, 2007; Trusov et al., 2008). Northern blot analysis showed that AGG3 expression was highest in reproductive tissues, with comparatively low expression observed in all other tissues analysed (Figure 4a). Performance of RT-PCR on guard cell mRNA confirmed that AGG3 is expressed in these cells (Figure S2e). Several AGG3pro:GUS reporter lines were produced but displayed aberrantly high expression in most tissues assayed. No microarray expression data are available for AGG3.

Figure 4.

agg3 mutants recapitulate agb1 mutant phenotypes that could not be observed in agg1, agg2 or double agg1agg2 mutants.
(a) Northern blot of AGG3 using total RNA extracted from various tissues of Arabidopsis (Col-0). RNA (10 μg) from each tissue was separated by electrophoresis in a 1% agarose gel and transferred to a nylon membrane. The blot was hybridized with 32P-labelled AGG3 cDNA and washed at high stringency. The blot was then stripped and probed with a 25S wheat ribosomal gene to assess equal loading.
(b) Comparison of morphological traits in Col-0 and G-protein mutants: (i) average rosette leaves from a 30-day-old plant; (ii) mature open flowers; (iii) mature siliques. Silique (c) length and (d) width of 10 mature representative siliques. Width is measured at the base of the head of the silique (i.e. the narrowest point in wild-type siliques which gives a numerical representation of the flat top siliques seen in b).
(e) Flower length of 10 opened flowers.
(f) Ratio of seed length:width of at least 30 seeds.
(g) Length of 15–20 hypocotyls after 4 days’ growth in darkness.
(h) Plant height from the base of the rosette to tip of the main bolt at cessation of flowering.
(i) Stomatal density on the abaxial epidermis side of cotyledons from 7-day-old seedlings.
(j) Seed number per mature silique. White bars are Col-0, light grey bars are gpa1-4 and agb1-2 controls and dark grey bars are agg mutants. *Significant differences from Col-0 (*P < 0.05–0.01; **P < 0.01–0.001; ***P < 0.001). All values are mean ± SE and assays were repeated two to four times with similar results.

In addition to the stomata and ABA response phenotypes analysed above and to further establish whether AGG3 is the ‘missing’γ-subunit we examined all other previously described G-protein phenotypes that were not observed in agg1, agg2 or agg1agg2 mutants (Trusov et al., 2008). If AGG3 is the previously unidentified γ-subunit we would expect it to recapitulate most of the unaccounted for Gβγ-phenotypes observed in agb1 mutants. All three agg3 lines display the characteristic flower and silique morphologies (Figure 4c–e) including the distinctive shorter petals, flat siliques and rounder leaves observed in agb1 mutants but not in agg1, agg2 or agg1agg2 mutants (Lease et al., 2001; Trusov et al., 2008; Figure 4b). Etiolated hypocotyls of agb1, gpa1 and agg3 mutants are shorter than the wild-type (Ullah et al., 2001, 2003; Figure 4g), while agg1, agg2 or agg1agg2 mutants do not display this trait (Trusov et al., 2008). As described for agb1-2 (Ullah et al., 2003), agg3 mutants are shorter than the wild-type at maturity (Figure 4h). As previously described by Zhang et al. (2008a), stomatal density in gpa1 mutants is lower than in the wild-type while agb1 mutants show higher stomatal density. All three agg3 mutants show wild-type levels of stomatal density (Figure 4i). In addition to the results above we have also included a graphical summary of our initial pilot assays comparing these traits with the wild-type on a percentage basis (Figure S2k). These assays included agb1, agg1, agg2 and agg3 mutants, demonstrating that recapitulation of even the most subtle of the ‘orphan’ phenotypes was only observed in agg3 mutants under our conditions.

The results presented above show that agg3 mutants can account for most of the ‘orphan’ phenotypes described in agb1 mutants, one exception being the reduction in rosette diameter observed in agb1-2 (Figure S3c,d). As neither agg1 nor agg2 mutants display this phenotype it is possible that a triple agg1/agg2/agg3 mutant will be required to recapitulate this phenotype. However most other phenotypes are clearly partitioned among the γ-subunits. For example, AGG3 is not involved in some of the phenotypes previously attributed to either AGG1 or AGG2 (Trusov et al., 2007) such as lateral root proliferation and resistance to Fusarium oxysporum (Figure S3a,b; Trusov et al., 2006, 2009).

AGG3 subcellular localization

To experimentally examine the subcellular localization of AGG3 we created N- and C-terminal GFP fusions (named GFP–AGG3 and AGG3–GFP, respectively). Both GFP–AGG3 and AGG3–GFP showed fluorescence clearly limited to the periphery of the cell in roots and leaves of stably transformed Arabidopsis lines (Figure 5a–c), although they also showed small motile vesicle-like spots (Figure 5c), much like the Golgi-localization reported for AGG1, but seldom seen for AGG2 (Adjobo-Hermans et al., 2006). When cells from GFP fusion transgenic lines were partially plasmolysed the fluorescence shrank away from the cell walls and was confined to the periphery of the protoplast (Figure 5d,e), demonstrating that AGG3 is located on the plasma membrane, not the extracellular space. Membrane localization was also observed when tobacco leaves were infiltrated with the GFP fusion constructs (Figure 5f–h). The intensity of the GFP signal almost doubled when tobacco leaves were co-infiltrated with a 35Spro:AGB1 construct (compare Figure 5g with 5h and quantified in 5i), indicating that the presence of an additional β-subunit helps stabilize AGG3, as observed for other Arabidopsis γ-subunits (Adjobo-Hermans et al., 2006; Zeng et al., 2007). However, unlike AGG1, localization of AGG3 in the motile vesicular-like spots appeared to increase when co-infiltrated with AGB1. Such an increase may indicate retention of, and therefore possibly a function for, AGB1/AGG3 dimers on or within internal vesicles.

Figure 5.

 AGG3 localizes to the plasma membrane.
(a, b) Cells from plate-grown Arabidopsis roots expressing (a) AGG3–GFP and (b) GFP–AGG3.
(c) Stably transformed Arabidopsis seedling leaves expressing AGG3–GFP, showing the small motile vesicles observed.
(d) GFP fluorescence of root cells grown on 600 mm sucrose showing the GFP receding away from the propidium iodide stained cell walls (red).
(e) Brightfield image corresponding to panel (d).
(f–h) Tobacco leaf discs transiently expressing (f) GFP–AGG3, (g) AGG3–GFP and (h) AGG3–GFP and 35Spro:AGB1 simultaneously.
(i) Quantification of fluorescence from tobacco cells shown in (g) and (h), showing the increase in fluorescence when AGB1 is co-infiltrated with AGG3–GFP. Values represent the mean of over 100 individual cell measurements ± SE.

The C-terminal Cys-rich domain is important for AGG3 function

The presence of the C-terminal Cys-rich domain makes AGG3 quite unique and different from all other Gγ subunits studied so far. It is therefore important to determine whether the C-terminal domain is essential for the correct function of the AGG3 protein. For that purpose we performed complementation studies in the agg3-3 background using two truncated AGG3 proteins lacking the C-terminal region. Truncations as well as the full-length complementation constructs were amplified from genomic DNA and contain 2 kb of the native promoter as well as all four introns present in the native gene. The first truncation construct (trunc1) contained the entire γ-domain (residues 1–135) including a CaaX motif present at the end of the AGG1/2 alignment (residues 132–135; Figure 1a). The second truncation construct (trunc2) also contained the γ domain (residues 1–129), but differed from trunc1 in that we added the CaaX motif present at the C-terminus of the native gene (residues 246–251). Importantly, both truncated AGG3 constructs lack the Cys-rich region but contain a CaaX isoprenylation motif, so aberrant subcellular localization was not expected to be an issue.

Five independent homozygous lines were analysed for each of the two AGG3 truncation constructs, of which we show two representative lines for simplicity. We measured silique length and width, flower and hypocotyl length, mature plant height, and seed germination in 2 μm ABA (uniform germination on standard 0.5 × MS plates is shown in Figure S2i). Complementation lines containing the full-length AGG3 construct showed statistically significant differences to the agg3-3 line, although several of the phenotypes were not fully restored to wild-type levels (Figure S3e–j). In contrast, lines containing either of the truncated AGG3 constructs were indistinguishable from the agg3-3 line and unable to confer even partial complementation. Therefore without the Cys-rich domain, the truncated AGG3 constructs were not able to provide functional complementation in the agg3-3 mutant background, underscoring the importance of the Cys-rich domain for the function of the AGG3 protein.

AGG3 homologues in rice are important quantitative trait loci for grain yield

Homology searches using BLAST were only able to identify AGG3 homologues in plants. A multiple alignment of several representative AGG3 homologues is included in Figure S1 with the approximate location of the γ-domain indicated. Homology among the different proteins is restricted to the γ-domain and appears to deteriorate in the C-terminal region. However, despite the divergence, all C-terminal domains are characterized by the abundance of Cys residues. Two AGG3 homologues, GS3 and DEP1, are present in rice (Oryza sativa; Figure S1), and mutations in either of them result in enhanced yield (Fan et al., 2006; Huang et al., 2009; Takano-Kai et al., 2009). GS3 was identified as ‘a major QTL [quantitative trait locus] for grain length and weight and a minor QTL for grain width and thickness’ (Fan et al., 2006), and was recently isolated using map-based cloning. All long-grain cultivars analysed contained a nonsense mutation in GS3 introducing a stop codon that resulted in premature termination of the protein; this mutation is absent in all short-grain varieties (Fan et al., 2006). Recent complementation studies have confirmed the link between the GS3 mutation and grain size in rice (Takano-Kai et al., 2009). The second AGG3 rice homologue, DEP1 is also a major yield QTL and the mutant allele has a strong effect in the determination of panicle architecture. Interestingly, dep1 seems to be a gain-of-function mutation due to its dominant character and the fact that knockout of the gene by RNA interference (RNAi) did not reproduce the dep1 mutant phenotype (Huang et al., 2009). Due to their atypical nature, neither of these proteins was previously identified as a G-protein γ-subunit, therefore our work clearly establishes the involvement of G-proteins in this crucial agronomic trait.

The observations in rice prompted us to study the morphological characteristics of seeds from mutants in all G-protein subunits. Seed size in Arabidopsis is extremely sensitive to the health of the maternal plant, and while length and width measurements displayed a trend in the mutants relative to the wild-type, it was neither clear nor reproducible. However, a clear trend emerged when the length:width ratio was calculated. Seeds from agg1 and agg2 mutants (including the double mutant) displayed no statistically significant differences from the wild-type (Figure S2f); however, gpa1-4, agb1-2 and agg3 mutant seeds were shorter and wider, and therefore displayed a lower length:width ratio than the wild-type (Figure 4f). Interestingly our results are in sharp contrast to the phenotype observed for the GS3 mutant in rice but are not totally unexpected. In fact, the rice d1 mutant, lacking a functional G-protein α-subunit, also shows a very different phenotype from the Arabidopsis gpa1 mutants. Rice d1 mutants are gibberellin-insensitive, showing an extreme dwarf phenotype, and have impaired resistance to infection by pathogens (Fujisawa et al., 1999; Ueguchi et al., 2000; Suharsono et al., 2002), while gpa1 mutants do not show dwarfism and are not hypersensitive to any of the pathogens tested so far (Ullah et al., 2003; Llorente et al., 2005; Trusov et al., 2006). The accumulated evidence suggests that G-proteins might have followed somewhat divergent evolutionary paths in dicotyledonous and monocotyledonous plants. We also analysed the number of seeds per silique in all G-protein mutant genotypes and Col-0 (Figure 4j). Our data show that gpa1-4, agb1-2 mutants contain fewer seeds per silique than the wild-type and while AGG1- and AGG2-deficient mutants showed wild-type levels of seeds, AGG3-deficient plants showed identical seed numbers to agb1-2 mutants.

New and unusual elements in the G-protein canonical model

Currently, researchers are broadening the scope of their investigations into G-proteins as we realize that this common signal transduction mechanism is far more complex than the established dogma suggests. Recent ‘unconventional’ discoveries in mammalian systems include human Gβ5 dimerization with γ-like regulator of G-protein signalling (RGS) proteins instead of a canonical γ-subunit (Sondek and Siderovski, 2001), and the existence of G-protein-coupled receptor (GPCR) heterodimers (Minneman, 2007). In plants, the present study and others illustrate that G-protein signalling components can deviate considerably from the canonical structures as demonstrated by the GPCR-like regulator of G-protein signalling, AtRGS1 (for review see Temple and Jones (2007)). Additionally, Arabidopsis possesses two unique proteins (GTG1 and GTG2) with GPCR-like structures able to bind and hydrolyse GTP. Both proteins interact with the Arabidopsis G-protein α-subunit and regulate signal transduction of the plant hormone ABA (Pandey et al., 2009). Furthermore, in some circumstances a plant-specific, extra-large α-subunit (XLG2) associates with AGB1 in planta and is involved in disease resistance (Zhu et al., 2009). AGG3-type γ-subunits are ubiquitous across the plant kingdom; however, BLAST searchers have failed to identify any animal homologues, suggesting that AGG3 is yet another unique component of plant G-protein signalling.

The next crucial step is to identify the significance and role of the novel AGG3 C-terminal region in G-protein signalling. Cysteine-rich regions are often involved in protein–protein interactions and protein complex formation (Voorberg et al., 1991; Okada et al., 1999; Labunskyy et al., 2005). Such a role would fit with the emerging model that G-proteins are found in large receptor, effector and regulator complexes such as the 400 kDa and 700 kDa G-protein complexes identified in rice and Arabidopsis, respectively (Kato et al., 2004; Wang et al., 2008). Furthermore, there are many examples of γ-subunits being crucial for receptor specificity (Kleuss et al., 1993) and it is possible that AGG3’s Cys-rich region could interact with receptor complexes.

In conclusion, we present here a highly atypical G-protein γ-subunit, previously unidentified due to its divergence from the well-established paradigm. We propose that AGG3 works as a bona fide γ-subunit based on multiple lines of evidence: (i) bioinformatic analysis shows that the N-terminus of AGG3 possesses important structural characteristics conserved in other γ-subunits, including conservation of very important amino acid residues; (ii) the positions of the first three introns in AGG1, AGG2 and AGG3 are identical; (iii) AGG3 strongly interacts with the β-subunit in vivo and in vitro; (iv) the interaction between AGG3 and the β-subunit is centred on the γ-domain whilst the Cys-rich domain shows no interaction; (v) AGG3 competes with the two other known γ-subunits (AGG1 and AGG2) for the same binding sites on the β-subunit; (vi) AGG3 is involved in established G-protein pathways in plants; (vii) AGG3-deficient mutants account for all but one of the ‘orphan’ phenotypes unexplained by the two previously known γ-subunits in Arabidopsis; and (viii) the intensity of the GFP signal almost doubled when leaves were co-infiltrated with AGG3–GFP and AGB1, indicating that the presence of additional β-subunit helps stabilize AGG3, as observed for other Arabidopsis γ-subunits. The discovery of AGG3 probably provides a complete set of G-protein subunits in Arabidopsis. AGG3 exhibits features not previously described in either plants or animals, including an extended C-terminal Cys-rich domain. While we have shown here that AGG3 is involved in the classic G-protein pathway of K+-channel regulation, we predict that AGG3 will also be found to account for novel mechanisms of G-protein action given its unusual nature.

Experimental procedures

Arabidopsis growth and measurement

Seed was stratified for 2 days at 4°C and grown under long-day conditions (16 h light at 60–100 μmol m−2 sec−1) between 21 and 23°C in a modified University of California (UC) soil mix (60% UC: 40% vermiculite). Height measurements were made from the soil to the top of the main bolt at the cessation of flowering. Rosette diameter was measured at the widest point after 6 and 17 days of growth. Ten opening (stage 13 according to Smyth et al., 1990) flowers and 10 mature siliques per line were photographed next to a ruler. Dry mature seeds (approximately 30 per line) were photographed under a microscope next to a ruler. Measurements were made using ImageJ software (NIH; All morphological traits presented were measured on at least two independent occasions for all lines. Plate assays were performed as outlined in Trusov et al. (2008).

Mutant lines

Three independent T-DNA insertion mutants in the Col-0 ecotype were obtained for AGG3 (At5g20635). agg3-1 (Salk_018024) from the Salk Institute (Alonso et al., 2003), agg3-2 (CS807967) and agg3-3 (CS844088) from the SAIL collection (McElver et al., 2001) were received from the Arabidopsis Biological Resource Center (ABRC; Homozygous lines were identified by PCR, and the insert position was verified by sequencing.

Transcriptional analysis

The RNA was isolated from various tissues and northern blotting was performed as described by Trusov et al. (2006), using the coding region of AGG3 as a probe. Real-time PCR was performed as described by Pandey and Assmann (2004). Primers used were AGG3F1 which binds to exon 1, and AGG3R2 which binds across the junction of intron 3 (see Table S2b for sequences).

Stomatal responses

Arabidopsis plants were grown and assayed as previously described (Trusov et al., 2008). Each stomatal bioassay was performed with five independent replicates for opening and three independent replicates for closure, respectively, and at least 120 stomatal apertures were measured in each replicate. Stomatal density was measured as described in Zhang et al. (2008a).

Guard cell protoplasts for electrophysiology experiments were isolated using the method published previously (Wang et al., 2001; Coursol et al., 2003; Fan et al., 2008; Trusov et al., 2008; Zhang et al., 2008b). For whole-cell K+-current recordings, the bath solution and pipette solutions were used as indicated in Trusov et al. (2008) and Fan et al. (2008), and K+-currents were analysed as described by Wang et al. (2001). For whole-cell anion current recordings, solutions (including the bath solution and both strong and weak pH buffering pipette solutions) were as indicated previously (Pei et al., 1997; Wang et al., 2001; Fan et al., 2008). Five millimolar Mg-ATP and 5 mm 2-amino-2-(hydroxymethyl)-1,3-propanediol (Tris)-GTP were both added daily before the start of the experiments (Marten et al., 1992; Pei et al., 1997). Steady-state whole-cell anion currents were acquired by subtracting the basal currents at a holding potential of +30 mV from the average currents between 42.5 and 50.0 sec. The software for data recording and analysis was pCLAMP 6.0.3 and figures were drawn with SigmaPlot 8.0.

Cloning of complementation and gene fusion constructs

All complementation, fusion and reporter gene constructs were created using the pCR®8/GW/TOPO® entry vector and Gateway® recombination system (Invitrogen; In many cases the pMDC series of binary vectors were used (Curtis and Grossniklaus, 2003). Primer sequences are shown in Table S2a. From floral Col-0 cDNA AGG3 was amplified without a stop codon using AGG3f-EcoRI and AGG3r-XhoI primers. The entry clone was recombined into pMDC84 to generate a 35Spro:AGG3–GFP fusion construct. A genomic copy of AGG3 with the terminator region was amplified using the AGG3cdsf-SalI primer and the AGG3ter-NotI primer. The entry clone was recombined into pMDC44 for a 35Spro:GFP–AGG3 fusion construct.

Three complementation constructs were cloned, one full length and two truncations without the C-terminal domain. All copies were amplified by PCR using the AGG3prf-NotKpn primer. The full length construct was amplified with the AGG3r-BamHI primer. The first truncation was amplified with the AGG3rm1 primer which amplifies up to R135 and adds a stop codon, and the second with AGG3rm2 which amplifies up to K129 but adds residues P246–F251 (putative isoprenylation motif) and the stop codon. All three entry clones were recombined into a Gateway-ready modified pCambia2380 binary vector with a plant terminator (Gao et al., 2003).

Yeast two-hybrid and in vitro binding assay

Yeast work and in vitro binding was carried out as described by Mason and Botella (2000). pAS1-CHY2-AGB1 and pACT2-AGG1 and pACT2-AGG2 from Mason and Botella (2000, 2001) were used as positive controls. Full-length and deletion constructs of AGG3 were amplified using the primers presented in Table S2b, and cloned inframe into pACT2 using the terminal NcoI and BamHI restriction sites incorporated during PCR.

Split-ubiquitin assay

The mating-based split-ubiquitin system (mbSUS) was used to check the interaction between AGG3 and nine different K+-channel proteins: AKT1, AKT2/3, AKT5, KAT1, KAT2, GORK, SKOR, ATKC1 and SPIK. In order to generate Nub and Cub constructs for each gene, clones with error free sequence in pCR®8/GW/TOPO® were mobilized into the appropriate yeast expression vector by Gateway cloning methods (Lalonde et al., 2010).

For this assay, Nub-fusion proteins of AGG3 (AGG3-NubG or NubG-AGG3) were used to test the interaction of AGG3 with the different K+-channel Cub-fusion proteins. The Nubwt-AGG3 fusion protein was used as a positive control as published by Obrdlik et al. (2004), and an empty vector containing NubG only was used as a negative control, to check the specificity of interaction of AGG3 with the K+-channel Cub-fusion proteins. To test the interaction between AGG3 and the different K+-channels, we used four different specific combinations of fusion proteins (or constructs). For example, to test the interaction between AGG3 and KAT1, interaction analysis was performed in four different combinations: AGG3-NubG + KAT1-Cub (test interaction), NubG-AGG3 + KAT1-Cub (test interaction), empty vector + KAT1-Cub (negative control), and Nubwt-AGG3 + KAT1-Cub (positive control). In addition, reciprocal interaction was also tested between the Cub-fusion protein of AGG3 and Nub-fusion proteins of K+-channels (e.g. KAT1) with appropriate positive and negative controls. Each entire set of interaction assays was repeated three times.

In vitro binding assay

A GST-AGB1 construct was produced by amplifying AGB1 with primers AGB1f-EcoRI and AGB1r-XhoI (Table S2a), and cloning inframe into pGEX-KG (Guan and Dixon, 1991). pGEX-KG-AGB1 and empty pGEX-KG vector were transformed into BL21 Escherichia coli expression cells and used to produce GST–AGB1 and GST proteins respectively. in vitro binding was performed by first binding GST–AGB1 and GST (in PBS + 1% Triton X-100) to glutathione-agarose resin (Sigma Aldrich; AGG3 (residues 1–112) was then labelled with [35S]-Methionine using a T7 TnT® (transcription and translation) Coupled Reticulocyte Lysate System (Promega; according to the manufacturer’s instructions and incubated with the GST–AGB1 and GST resins. Bound proteins were eluted and separated on a 12% PAGE gel which was dried and exposed to a phosphorimager plate.

GFP analysis

The GFP fluorescence from root epidermal cells of 1-week-old Arabidopsis seedlings [grown on 0.5 × MS medium with pH adjusted to either 5.5 or 8.1 with 2-(N-morpholine)-ethanesulphonic acid (MES) or HEPES, respectively] was imaged and quantified as described (Sheahan et al., 2004), but without optical sectioning. Agroinfiltration experiments were performed as described (Sheahan et al., 2004). Agrobacteria (harbouring binary vectors encoding AGB1 and/or AGG3 fusions to GFP) were grown and the leaves of tobacco plants agroinfiltrated as described (Sparkes et al., 2006). Leaf discs were excised 4 days after agroinfiltration and GFP fluorescence imaged using a confocal laser-scanning microscope (LSM 510, Zeiss; equipped with a 40 × C-Apochromat water-immersion objective (NA 1.2; Zeiss) and 488 nm argon laser and BP500-530IR filter. The GFP fluorescence intensity was quantified from midplane cell sections as described (Sheahan et al., 2004).

Statistical analysis

All statistical analyses were performed using a Student’s t-test. In most cases < 0.05 was considered as significantly different and gradations of significance past that were indicated, except for stomatal response assays where < 0.01 were considered as significantly different.


We thank Dr Mike Mason, Dr Emily McCallum and Tim Gookin for technical advice and Antony Martin for assistance with Agroinfiltration. Work in J. R. Botella’s lab was supported by Australian Research Council grants DP0772145 and DP1094152, work in S. M. Assmann’s lab was supported by United States Department of Agriculture grant 2006-35100-17254, and M. B. Sheahan’s work was supported by Australian Research Council Australian Postdoctoral Fellowship (DP0770679). The materials described in this work will be available on request provided that the requesters comply with all local and international requirements for the transport and import of genetically modified plant material.