Reevaluation of Abscisic Acid-Binding Assays Shows That G-Protein-Coupled Receptor2 Does Not Bind Abscisic Acid

The phytohormone abscisic acid (ABA) plays an integral role in plant growth and development, but only recently have putative receptors been identified (Razem et al., 2006; Shen et al., 2006; Liu et al., 2007; Pandey et al., 2009). Our results, however, suggest that the putative extracellular ABA receptor, G-Protein-Coupled Receptor2 (GCR2), does not bind ABA. We propose that an improved method for measuring ABA binding is required and that the search for ABA receptors should continue.

ABA regulates dormancy, seed germination, and maturation as well as promoting adaptation to abiotic stress and regulating stomatal aperture (for review, see Leung and Giraudat, 1998; Finkelstein et al., 2002; Himmelbach et al., 2003). Genetic approaches have led to the isolation of genes involved in the ABA biosynthetic and signal transduction pathways (Finkelstein et al., 2002; Nambara and Marion-Poll, 2005). Yet, only recently have supposed ABA receptors been identified in Arabidopsis (Arabidopsis thaliana; Verslues and Zhu, 2007). These are the nuclear flowering-time protein FCA (Razem et al., 2006), the plastid-associated magnesium-protoporphyrin IX chelatase H subunit (CHLH; Shen et al., 2006), a protein originally identified as a membrane-bound GCR2 (Liu et al., 2007), and, recently, two novel G-protein coupled receptors (GPCRs), GPCR-Type G-Protein1 (GTG1) and GTG2 (Pandey et al., 2009). The identification of multiple ABA receptors was expected, as both extracellular and intracellular ABA perception occurs (for review, see Finkelstein et al., 2002) and different ABA analogs induce distinct gene expression patterns (Huang et al., 2007). Moreover, ABA-binding activities have been found in discrete cellular fractions, including the plasma membrane, cytosol, and endomembranes, suggesting that multiple receptors exist (Wan and Hasenstein, 1996; Pedron et al., 1998; Zhang et al., 1999, 2001; Kitahataa et al., 2005).

Recently, we demonstrated that FCA does not bind ABA (Risk et al., 2008), and the paper reporting FCA as an ABA receptor has been retracted (Razem et al., 2008). As FCA was used as the positive control in the ABA-binding assays that demonstrated that GCR2 bound ABA, we also reevaluated the ABA-binding properties of GCR2. Here, we show that the putative extracellular ABA receptor, GCR2, also does not bind ABA.

To investigate the ABA-binding properties of GCR2, we obtained the GCR2 expression plasmid used by Liu et al. (2007) and prepared recombinant protein from Escherichia coli extracts. Initially, GCR2 was prepared in the presence of 0.1% Triton and purified as described by Liu et al. (2007; Fig. 1A). However, the resulting protein did not bind ABA (Fig. 1B). GCR2 was also purified in the absence of any detergent and using 0.5% SDS in the lysis buffer (Razem et al., 2004; Ligeng Ma, personal communication). In each preparation, only a small proportion of His-tagged GCR2 bound to Ni²⁺ resin, and the eluted protein from all protein preparations failed to bind ABA (Fig. 1B). In addition, high counts were observed in the buffer-only controls (approximately 20% of total counts). As Liu et al. (2007) showed ABA binding to purified GCR2 as relative binding activity (%) with the buffer-only counts subtracted (see figure 3A in Liu et al., 2007), we were unsure if the lack of binding and high background counts were due to the inability of GCR2 to bind ABA or to problems with the ABA-binding assay. To address this question, like Liu et al. (2007), we included FCA residues 444 to 747 (FCAΔN443), which is slightly longer than the FCA protein (residues 500–747) they used, as a control. The purified soluble proteins used in ABA-binding experiments are shown in Figure 1C. ABA-binding assays were carried out in triplicate, with three different protein preparations, using the method based on that of Zhang et al. (1999) as reported previously (Zhang et al., 2002; Liu et al., 2007), except for some slight modifications (Fig. 1D). Using this assay, we failed to detect the binding of ABA to GCR2 or FCA (Fig. 1D). Subsequent communication revealed that high background counts had been observed by others (Ligeng Ma, personal communication), but they had subtracted the buffer-only counts (Zhang et al., 2002; Razem et al., 2006; Liu et al., 2007). When high background counts are observed, subtraction of the buffer-only counts can substantially alter the perceived binding (Fig. 1E). We conclude that GCR2 does not bind ABA and suggest that the reported ability of GCR2 to bind ABA is likely due to poor-quality protein combined with data analysis.

Figure 1.

ABA does not specifically bind GCR2. A, Recombinant proteins used in B are shown. GCR2 was expressed in Rosetta2 E. coli for 3 h at 37°C following induction with 0.2 mm isopropyl β-d-1-thiogalactopyranoside. Cells were lysed according to Liu et al. (2007), with the addition of 250 mm Suc and 15% (v/v) glycerol as recommended by Ligeng Ma (personal communication), and for the indicated samples 0.1% (v/v) Triton or 0.5% (w/v) SDS was added to the lysis buffer. GCR2 was then purified and eluted in the absence of any detergent, as described previously (Liu et al., 2007). Protein concentrations were established using the Bradford assay, and samples were analyzed by SDS-PAGE. BSA, Bovine serum albumin. B, Specific binding of ABA was not observed (mean ± sd; n = 3). The assays were carried out in triplicate using freshly prepared proteins for each assay. A total of 50 ng of purified protein and 50 nm [³H](±)-ABA (Amersham; 46 Ci/mmol) was used in each assay. Protein stability was maintained by the addition of 950 ng of bovine serum albumin. Assays were carried out in 100 μL of Suc-binding buffer (Liu et al., 2007) and incubated for 1 h in the dark at 4°C, following which 50 μL of 0.5% (w/v) dextran-coated charcoal was added. Samples were then incubated on ice for 10 min before the supernatant was recovered and the dpm was measured using an LKB Wallac 1217 Rackbeta Liquid Scintillation Counter. C, Recombinant proteins used in D are shown. Proteins were expressed and purified as described for A except that cells containing FCAΔN443 and GST were lysed in phosphate-buffered saline, pH 7.4, and the resulting supernatant was affinity purified on glutathione-Sepharose. The GST tag was removed from FCAΔN443 with PreScission protease, and GST alone was eluted with 10 mm reduced glutathione and 50 mm Tris, pH 8. GCR2 was purified with 0.5% SDS in the lysis buffer as described above. D, GCR2 does not bind ABA (mean ± sd; n = 3). The assays were carried out as described for B. E, Binding data shown in D following subtraction of the background counts observed in the buffer-only control.

At this point, we reevaluated the methods used to identify the known ABA receptors, as it was surprising to discover that two of the reported receptors did not bind ABA. In the case of GCR2, it was predicted that a GPCR might function as an ABA receptor, because null mutants in components of the heterotrimeric G-protein complex (gpa1 and agb1) have ABA-response phenotypes (Pandey et al., 2006). As only one GPCR (GCR1) had been characterized in plants, Liu et al. (2007) used bioinformatic approaches to identify novel GPCRs that might function as extracellular ABA receptors. GCR2 fitted the predicted role of an ABA receptor, as the gcr2 mutant was shown to exhibit an ABA-insensitive phenotype and to interact with the Gα-subunit protein, GPA1 (using four different methods: surface plasmon resonance spectroscopy, yeast split ubiquitin assays, in planta bimolecular fluorescence complementation, and coimmunoprecipitation assays; Liu et al., 2007). However, three recent papers raise questions about the function of GCR2 (Gao et al., 2007; Johnston et al., 2007; Guo et al., 2008). It now appears that GCR2 is not a seven-transmembrane GPCR but a member of the LanC protein superfamily, as it shares 40% sequence identity with human LANCL1 and LANCL2, which are similar to bacterial lanthione synthetases (Gao et al., 2007). Moreover, GCR2 can be modeled on the LanC family member nisin cyclase, indicating a very similar tertiary structure, with just one membrane-bound helix (Johnston et al., 2007). This suggests that GCR2 is a peripheral protein and not a transmembrane protein. In addition, it has now been shown that gcr2 mutants, as well as mutants in two related genes, gcr2-like1 and gcr2-like2, and the various double and triple mutant combinations all failed to show the reported ABA response in stomatal regulation, seed germination, or early seedling development, unlike that seen in loss-of-function gpa1 mutants (Gao et al., 2007; Guo et al., 2008). These results raise questions about the validity of the methods used to demonstrate an interaction with GPA1; in particular, others have emphasized the need for care when assessing the interaction of membrane proteins using “split-protein sensors” (Müller and Johnsson, 2008). Our results, together with these reports that contradict the findings of Liu et al. (2007), demonstrate that GCR2 is not an extracellular ABA receptor.

FCA was identified as an ABA receptor based on sequence similarity to ABAP1 (Razem et al., 2004). Although ABAP1 was reported to lack RNA-binding domains, our analysis suggests that rather than being a novel protein it is a fragment of barley (Hordeum vulgare) FCA (encoded within a truncated cDNA) that contains part of the second RNA-binding domain (Fig. 2). ABAP1 was isolated from barley using polyclonal anti-idiotypic antibodies (AB2), which bound the heavy chain of the anti-(+)-ABA monoclonal antibody (15-I-C5; Razem et al., 2004). ABA binding to ABAP1 was then demonstrated directly using a [³H](+)-ABA binding assay similar to that used in our studies. We attempted to directly test the ability of ABAP1 to bind ABA. Although we obtained the ABAP1 expression plasmid from R.D. Hill and colleagues, we were unable to purify protein of sufficient purity and stability to allow us to perform binding experiments (data not shown). Therefore, although we were unable to assess the ability of ABAP1 to bind to ABA, we suggest that the protein used in the assays was of uncertain quality and that this may have contributed to the observed binding (Razem et al., 2004). Furthermore, given the sequence similarity between ABAP1 and FCA (Fig. 2) and the fact that FCA does not bind ABA (Risk et al., 2008), it is unlikely that ABAP1 binds ABA.

Figure 2.

ABAP1 is a truncation of barley FCA. An alignment was made between Arabidopsis AtFCA (Z82989), barley ABAP1 (AF127388), and barley HvFCA (TA38508_4513) using ClustalW and shaded using Boxshade. The alignment shows identical amino acids in black and similar amino acids in gray. The sequences underlined in dark gray represent the positions of the RNA recognition motifs, while the light gray line identifies the WW domain. This alignment suggests that the ABAP1 protein is a truncation of the longer HvFCA protein. The entire FCA sequence (838 amino acids) with the alternate start site has been included here (Macknight et al., 2002), as it highlights the similarity between AtFCA and HvFCA. The position of the Met start codon from the original gene annotation is indicated with an arrow. This gives rise to the shorter 747-amino acid Arabidopsis FCA.

CHLH was identified as a homolog of the putative ABA-binding receptor from broad bean (Vicia faba; Zhang et al., 2002). CHLH is a subunit of the magnesium-protoporphyrin IX chelatase (Mg-chelatase). CHLH is involved in retrograde signaling to the nucleus following plastid damage, while Mg-chelatase plays an integral role in chlorophyll biosynthesis (Walker and Willows, 1997; Strand et al., 2003). CHLH overexpression and RNA interference lines showed altered ABA responses, specifically changes in seed germination and postgermination growth and altered stomatal opening and closure. In addition, the expression of ABA-signaling genes is affected by changes in CHLH expression (Shen et al., 2006). To demonstrate direct binding of ABA to CHLH, however, the ABA-binding assays used were similar to those of Razem et al. (2006), raising questions about the authenticity of this receptor. Further experiments are required to validate the role of CHLH.

The most recent ABA receptors to be reported, novel GPCR proteins GTG1 and GTG2, were also identified using bioinformatic approaches (Pandey et al., 2009). Consistent with their proposed role in ABA signaling, Arabidopsis gtg1/gtg2 double mutants show typical hyposensitivity to ABA, including reduced dormancy, greater seedling development in the presence of ABA, and compromised expression of ABA-responsive genes. The stomata from these double insertion mutants also show reduced sensitivity to ABA. Intriguingly, the GTG proteins also bind GDP/GTP, and the bound nucleotide influences ABA signal transduction. Thus, Pandey et al. (2009) concluded that GTG1 and GTG2 bind ABA and execute the first steps of the GPCR-G-protein signaling cascade (Pandey et al., 2009). This is surprising, as no other GPCR proteins have been found to both perceive the signal and carry out the initial steps of the transduction pathway.

Identification of GTG1 and GTG2 as GPCRs relied in part on their similarity (approximately 65% amino acid similarity) with GPR89/GPHR, a human protein that was reported to be an orphan GPCR (Pandey et al., 2009) but has now been shown to function as an anion channel that modulates Golgi functions through the regulation of acidification (Maeda et al., 2008). As the sequence motifs that identify GPCR are poorly defined, evidence that the GTG proteins are functional GPCRs comes from their interaction with the Gα-subunit protein GPA1 and their ability to bind ABA. Interaction of GTG1 and GTG2 with GPA1 was demonstrated using the split-ubiquitin system and by coimmunoprecipitation assays. Both methods can give rise to spurious results, especially when using overexpressed membrane-localized proteins, and further validation is desirable (Mackay et al., 2007; Müller and Johnsson, 2008).

GPCRs are notoriously difficult to purify (Sarramegna et al., 2003; Mancia and Hendrickson, 2007), making in vitro experiments difficult. In the case of Pandey et al. (2009), ABA-binding experiments were carried out with purified protein, but they report that only 1% of the purified GTG proteins were capable of binding ABA. One must cautiously interpret binding affinities when such low stoichiometry has been seen (approximately 0.01 mol of ABA per mol of protein), and the authors suggest that further optimization of protein purification methods are needed to allow a more detailed analysis of ABA binding (Pandey et al., 2009). The ABA-binding pocket of the GTG proteins has not been identified, but a direct role for the GTG proteins in binding ABA would be strongly endorsed if residues required for binding were identified using site-directed mutagenesis. This is especially important as GTG1 and GTG2 were reported to have K_d values of 35.8 and 41 nm, respectively. Other reported ABA-binding proteins also have high affinities for ABA: K_d values of 19, 20.1, and 32 nm for FCA, GCR2, and CHLH, respectively (Razem et al., 2006; Shen et al., 2006; Liu et al., 2007). This is somewhat surprising, as McCourt and Creelman (2008) suggest that ABA receptors with binding affinities in the nanomolar range would be saturated under normal physiological conditions.

Why were GCR2 and FCA previously reported to bind ABA? Based on our results with various protein preparations, we propose that the differences in apparent ABA binding are probably due to the variable quality of the purified proteins used. We generally observed greater interreplicate variability in the [³H](±)-ABA-binding assays when less stable proteins of modest purity were used. In the absence of a positive control, we suggest that the currently used assay is not sufficiently sensitive to detect binding of ABA, in part because high background counts are routinely observed. This can be seen in Figure 1D, where there is no significant difference between ABA binding for the different proteins. Pandey et al. (2009) modified the ABA-binding method when characterizing GTG1 and GTG2. [³H](±)-ABA was incubated with either GTG1 or GTG2, and then this mix was filtered through a GF/C membrane and subsequently washed with buffer. This resulted in significantly lower background counts.

Radiolabeled hormones have been used to show ligand binding to both the TRANSPORT INHIBITOR RESPONSE1 (TIR1) auxin receptor (Dharmasiri et al., 2005; Kepinski and Leyser, 2005) and the GIBBERELLIN-INSENSITIVE DWARF1 (GID1) GA receptor (Ueguchi-Tanaka et al., 2005). In these studies, more rigorous binding assays were utilized. Auxin binding was measured using glutathione S-transferase (GST)-TIR1 in a pull-down experiment. The resin was then washed to remove unbound auxin (Dharmasiri et al., 2005; Kepinski and Leyser, 2005). To assess binding of GA, Ueguchi-Tanaka et al. (2005) incubated GID1 with radiolabeled GA₄. This mix was then fractionated on a NAP-5 column and radioactivity was measured. To definitively prove that the remaining ABA receptors, CHLH and the GTGs, bind ABA, we suggest that a second method of determining ABA binding, such as isothermal titration calorimetry, is needed. However, isothermal titration calorimetry requires significant quantities of purified protein and would only be suitable for analyzing candidate proteins and not for conducting screens.

In conclusion, we find no evidence supporting the claim that GCR2 is an ABA receptor. The putative Arabidopsis ABA receptors were all identified based on sequence analysis and bioinformatics approaches rather than using direct biochemical or genetic approaches. Thus, the ABA-binding assays have been used to confirm the hypothesis that these proteins bind ABA. In the absence of a positive control, we suggest that more rigorous analysis of ABA binding is required when receptors have been identified based on a candidate gene approach.