The plant-specific phi class of glutathione transferases (GSTFs) are often highly stress-inducible and expressed in a tissue-specific manner, suggestive of them having important protective roles. To date, these functions remain largely unknown, although activities associated with the binding and transport of reactive metabolites have been proposed. Using a sensitive and selective binding screen, we have probed the Arabidopsis thaliana GSTFs for natural product ligands from bacteria and plants. Uniquely, when overexpressed in bacteria, family members GSTF2 and GSTF3 bound a series of heterocyclic compounds, including lumichrome, harmane, norharmane and indole-3-aldehyde. When screened against total metabolite extracts from A. thaliana, GSTF2 also selectively bound the indole-derived phytoalexin camalexin, as well as the flavonol quercetin-3-O-rhamnoside. In each case, isothermal titration calorimetry revealed high-affinity binding (typically Kd<1 μM), which was enhanced in the presence of glutathione and by the other heterocyclic ligands. With GSTF2, these secondary ligand associations resulted in an allosteric enhancement in glutathione-conjugating activity. Together with the known stress responsiveness of GSTF2 and its association with membrane vesicles, these results are suggestive of roles in regulating the binding and transport of defence-related compounds in planta.
- phi class glutathione transferase 2 (GSTF2)
- plant defence
In plants, GSTs (glutathione transferases; EC 22.214.171.124) are a diverse group of proteins, with the respective superfamilies described in Arabidopsis thaliana , rice [2,3], poplar , maize and soya bean . Plant GSTs exist in seven distinct subfamilies, termed the phi (F), tau (U), theta (T), zeta (Z), lambda (L), DHAR (dehydroascorbate reductase) and TCHQD (tetrachlorohydroquinone dehalogenase) classes [1,6]. Whereas the DHARs and GSTZs have roles in ascorbic acid and tyrosine metabolism respectively , the endogenous functions of the other classes, notably the large and plant-specific groups of phi and tau GSTs, are poorly understood . This is in contrast with the known roles of GSTFs and GSTUs in catalysing the conjugation of xenobiotics with the tripeptide glutathione (GSH), with the resulting detoxification determining the selectivity of several major herbicides .
Using Arabidopsis as a model system, the functional genomics of the GSTFs and GSTUs are currently the subject of considerable interest. Proteomic and transcriptomic studies in Arabidopsis plants and cultures have shown that these proteins accumulate in response to xenobiotics [9,10], plant hormones  and infection [12,13], as well as illumination and environmental stress . In the few instances where progress has been made in defining function, the Arabidopsis AtGSTs have been found to be involved in signalling and transport, rather than in conjugating natural products. For example, AtGSTF12 is involved in anthocyanin and proanthocyanidin accumulation , whereas AtGSTU20 modulates responses to light reception . On the basis of these observations, the latest evidence suggests that plant GSTs have evolved regulatory non-catalytic functions, which are most probably an extension of their ability to selectively bind biologically active ligands. For example, tau class GSTs isolated from Arabidopsis and maize (Zea mays) have recently been shown to selectively bind fatty acid and porphyrin natural products respectively . When AtGSTUs were incubated with plant extracts, family members bound oxidized fatty acid derivatives in a protein–ligand-specific manner . Binding partners included unstable oxylipins with known roles as stress signalling agents [18,19]. Similarly, the expression of maize ZmGSTUs in the chloroplasts of tobacco plants resulted in the accumulation of porphyrins due to these proteins binding unstable porphyrinogen precursors [17,20]. In both cases, similar types of metabolites were found to hyper-accumulate in bacteria when these same enzymes were overexpressed in Escherichia coli. Thus, when ZmGSTU enzymes were overexpressed in E. coli, the bacteria accumulated porphyrin intermediates in a de-regulated manner, owing to their binding of the haem precursor harderoporphyrinogen . When AtGSTUs were overexpressed in E. coli, a range of unusual oxidized fatty acids were determined, which showed similarities in terms of chain length and oxygenation to the metabolites identified in binding studies with plant extracts . These studies have demonstrated that a combination of determining metabolite perturbation following GST expression in bacteria and the characterization of bound ligands on incubation with plant extracts are powerful and complementary approaches in identifying classes of chemicals which are selectively recognized by these proteins. Importantly, the identification of binding partners by ‘ligand fishing’ can determine new functions for GSTs. For example, using such approaches we have recently identified a subset of flavonols as specific binding partners of Arabidopsis and wheat GSTLs that are associated with a novel glutathionyl-flavonol reductase activity .
Using the combination of metabolic perturbation following bacterial expression and ligand fishing using plant extracts, the present paper reports on the identification of a new set of binding partners of the Arabidopsis phi class proteins. In particular, AtGSTF2 has been identified as selectively recognizing a range of biologically active natural products, which is of interest based on the well-described regulation of this protein by pathogen attack, as well as by treatment with plant hormones, heavy metals and xenobiotics [10,12,22].
GST cloning, expression, purification and ligand analysis
Constructs for the bacterial expression of N-terminally Strep-tagged AtGSTs were synthesized and expressed in E. coli as described previously , with any disruption to metabolism in the cultures determined by HPLC-MS as described previously . Site-directed mutagenesis was performed using overlap-extension PCR and appropriate mutagenic oligonucleotide primers. Strep-tagged proteins were purified using standard procedures . For expression of untagged GSTF2, the coding sequence was excised from pET-STRP3 with NdeI and SalI, and ligated into pET-24a (Novagen) NdeI and XhoI sites, generating pET24a-F2. Untagged GSTF2 was expressed as described for tagged proteins and purified from soluble extracts using Orange A–agarose (Millipore), with elution using 5 mM GSH in HBS [Hepes-buffered saline; 20 mM Hepes/NaOH, 150 mM NaCl and 1 mM EDTA (pH 7.5)]. To remove GSH and any bound ligands, purified AtGSTF2 was buffer-exchanged into HBS acidified to pH 5.5 with acetic acid and then repeatedly concentrated by ultrafiltration prior to dilution with acidified HBS. Finally, AtGSTF2 was desalted into the required buffer by gel filtration through a HiTrap desalting column (GE Healthcare). As an alternative method, AtGSTF2 was extensively dialysed against HBS prior to use.
Enzyme and ligand-binding assays
GST activity was determined with CDNB (1-chloro-2,4-dinitrobenzene) as a substrate . GST(CDNB) assays in the presence of ligands were performed using 1 mM GSH in HBS for consistency with the binding studies. ITC (isothermal titration calorimetry) experiments were performed at 25 °C in HBS and analysed using a VP-ITC instrument with Origin 7.0 software (GE Healthcare). In addition to titrating ligand into protein, control titrations of ligand into buffer and buffer into protein were performed. Ligand concentrations were determined from absorbance values using molar absorption coefficients which were empirically determined for the ligands in HBS as ϵ348(norharmane)=3.98 mM−1, ϵ348(harmane)=4.44 mM−1, ϵ352(lumichrome)=9.13 mM−1 and ϵ317(camalexin)=1.23 mM−1. For other ligands, molar absorption coefficients described in the literature were used, whereas the concentration of GSH was determined gravimetrically. The content of Strep-AtGSTF2 protein was based on a calculated ϵ280 of 20.34 mM−1.
Gel filtration was performed using a Superdex 200 10/30 GL column (GE Healthcare), with HBS plus 2.5 mM DTT (dithiothreitol) and 1 mM GSH as running buffer at a flow rate of 0.5 ml/min. The effect of ligands on the elution of AtGSTF2 was tested by adding 10 μM lumichrome to the running buffer. Protein elution was monitored by determining the absorbance at 280 nm, after calibrating the column with aldolase, BSA, ovalbumin, chymotrypsinogen and ribonuclease A.
Ligand binding to AtGSTFs in E. coli
As an extension to our previously reported study with the AtGSTUs , all AtGSTFs that had been shown to be expressed as soluble Strep-tagged recombinant proteins, namely AtGSTs F2–F10 and F14 , were expressed in E. coli. After growth under inducing conditions to stationary phase, the bacteria were harvested and solvent-extracted prior to analysis by HPLC, coupled with UV and MS detection. Unlike the GSTUs, the expression of GSTs F4, F5, F6, F7, F8, F9, F10 and F14 did not give any major perturbations in metabolic profiles as compared with the control bacteria. In contrast, the metabolic profiles from E. coli expressing AtGSTF2 [AGI (Arabidopsis Genome Initiative) code At4g02520] and AtGSTF3 (AGI code At2g02930) revealed an identical group of five novel UV-absorbing peaks (1–5; Figure 1A), which were not observed in the extracts from the control bacteria. In each case, these novel metabolites were tentatively identified by tandem and accurate mass electrospray MS as N-containing heterocyclic natural products (Figure 2); namely norharmane (1), harmane (2), lumichrome (3), indole-3-aldehyde (4) and 1-acetyl-β-carboline (5). Identities were subsequently confirmed by comparing their spectroscopic properties with those of authentic standards. To distinguish between metabolite accumulation due to a selective binding to GST and perturbations caused by a non-specific ectopic protein expression, Strep-tagged AtGSTF2 and AtGSTF3 were affinity-purified from E. coli and then extracted with methanol and analysed for co-purifying ligands. Compounds 1–4 were recovered bound to both AtGSTF2 and AtGSTF3 (Figures 1B and 1C), showing that their accumulation was due to their selective associations with the respective proteins. In view of the identical ligand-binding profiles obtained with the two AtGSTFs, further detailed characterization was focused on AtGSTF2, which on the basis of its known regulation and stress inducibility was the best characterized of the two proteins .
Identification of Arabidopsis ligands of AtGSTF2
To complement and extend the bacterial screening, plants were screened for compounds that bound strongly to AtGSTF2. Two global metabolite profiling approaches were adopted, namely an in vivo method where the Strep-tagged enzymes were plant-expressed and bound ligands identified after affinity purification of the proteins, and an in vitro method where the bacterially expressed purified enzymes were incubated with plant extracts . In both cases, a Strep-tagged AtGSTU19 was used as a positive control, representing a distinct class of GST (tau) which had been previously shown to bind oxidized fatty acids present in tobacco . Using the in vivo ligand fishing method, Strep-tagged AtGSTF2 was stably expressed in Arabidopsis, but poor expression levels hampered protein and ligand recovery. As a result, this approach was not pursued further. Instead, the in vitro method was employed, with extracts from Arabidopsis plants passed over affinity-immobilized AtGSTF2, which had been purified from bacteria grown in M9 medium (see below). For comparison, AtGSTF6 was also included in the screen as it represented a phi enzyme which had not caused any compounds to accumulate when expressed in E. coli. When the AtGSTU19 control was tested under these conditions, a similar range of compounds to those previously identified in tobacco were determined . In contrast, this approach readily identified plant-derived ligands of AtGSTF2 that were not ligands of AtGSTU19, and these were surprisingly similar to those determined with AtGSTF6 (Figure 3). In particular, the abundances of two compounds, 6 and 7, were enhanced at least 100-fold on binding to AtGSTF2 and AtGSTF6 compared with their retention by the AtGSTU19 control (Figure 3). Compound 6 had a distinctive UV–Vis absorbance spectrum (Figure 2) and was tentatively identified as the well-known Arabidopsis flavonoid quercetin-3-O-rhamnoside (quercitrin). Compound 7 had a much weaker UV signature that nevertheless gave abundant ions by MS, and was tentatively identified as the indole-derived antifungal phytoalexin camalexin (Figure 2). The identities of both compounds were subsequently confirmed by comparison with authentic standards.
AtGSTF2–ligand binding studies
ITC was then used to examine the affinity and stoichiometry of binding of the ligands identified in E. coli and Arabidopsis to AtGSTF2. In preparing the proteins for ITC studies, it became apparent that when the Strep-tagged AtGSTF2 was purified from E. coli under standard conditions, the protein was unsuitable for titration assays, due to the high levels of pre-bound ligands. To minimize such contamination, AtGSTF2 was expressed in bacteria grown in M9 minimal broth instead of the normal ‘rich’ medium used, to reduce the availability of preformed indole compounds. In an alternative approach, when AtGSTF2 was purified from bacteria grown in rich medium, the protein was either extensively dialysed prior to use, or acidified to pH 5.5 with acetic acid to release associated ligands, prior to rapid desalting and renaturation. Binding studies with lumichrome, norharmane and harmane showed that all three ligands bound with high affinity to AtGSTF2, with the association being strongly promoted in the presence of 1 mM glutathione (Table 1). However, increasing the concentration of glutathione (GSH) to 5 mM did not further enhance ligand binding. In contrast, indole-3-aldehyde bound with more moderate affinity, whereas other N-containing heterocyclic molecules including riboflavin, kinetin, trans-zeatin, indole-3-acetic acid, harmine and tetrahydroharmane carboxylic acid showed no obvious association by ITC. To rule out any effect of the Strep-tag on ligand binding, untagged native AtGSTF2 was purified from recombinant bacteria using Orange A affinity chromatography . When assayed by ITC with lumichrome, binding to the native protein was essentially identical with that observed with the Strep-tagged form, confirming that the affinity tag did not affect the results obtained. It was then of interest to determine the stoichiometry of ligand binding to AtGSTF2, which is composed of subunits that normally associate to form homodimers . In the case of norharmane, harmane and indole-3-aldehyde, each ligand bound in an approximately 1:1 ratio with the AtGSTF2 polypeptide subunits, consistent with one molecule of ligand binding per GST monomer. In contrast, lumichrome bound to AtGSTF2 with a lower stoichiometry (0.36:1). To examine the possibility that this binding ratio arose from AtGSTF2 forming trimeric or other unexpected oligomers, the protein was analysed by gel-filtration chromatography in the presence and absence of lumichrome. In both cases, AtGSTF2 eluted as a single peak immediately following an ovalbumin standard (43 kDa), consistent with the protein remaining as a dimer on ligand binding. It was concluded that a single lumichrome molecule was binding per protein dimer, although the determined stoichiometry was sufficiently low that inaccuracies in determining active protein and ligand concentrations could not satisfactorily account for the deviation from the expected 0.5:1 ratio for associations with a dimer.
Further binding studies with AtGSTF2 focused on the optimal ligands harmane and lumichrome, and the effect of their binding on GST activity towards the model substrate CDNB. In both cases stoichiometric binding by each ligand enhanced the rate of CDNB-conjugating activity by 60%. Kinetic analysis showed that the addition of harmane (52.9 μM) did not cause a major decrease in substrate binding affinity, with the apparent Km towards GSH being unchanged (15.5 μM), whereas the apparent Km towards CDNB increased slightly from 1.6 mM to 2.4 mM. Instead, the observed activation was due to an increase in turnover (kcat) in each case. There was no obvious explanation for this non-competitive activation, but possibilities include enhanced catalysis, a reduction in the binding affinity for the dinitrobenzylglutathione product of CDNB conjugation, or displacement of an unobserved non-competitive inhibitor from the enzyme active site. The observed increase in apparent Km(CDNB) pointed to decreased product binding being the most likely explanation. To confirm that the association of these ligands with the protein did not involve covalent modification, electrospray MS of intact ligand-treated AtGSTF2 was performed. These experiments demonstrated that the protein had not undergone any irreversible modifications.
The binding of other ligands to AtGSTF2 was then examined. First, the effect of lumichrome and GSH on the binding of harmane to AtGSTF2 was investigated using ITC (Table 1). The presence of lumichrome and GSH both increased harmane-binding affinity, but by different mechanisms. Lumichrome reduced the stoichiometry and associated heat release of harmane binding to AtGSTF2 by one-third, whereas GSH showed no such effect. Complementary binding studies also demonstrated that both lumichrome and harmane increased binding affinity for GSH from Kd=22 μM to Kd=~ 5 μM (Figure 4 and Table 1). From these studies it was clear that the binding of ligands and GSH to AtGSTF2 was co-operative. The binding of the phytoalexin camalexin to AtGSTF2 showed a different response, with an apparent unusual stoichiometry of 1.45:1. Intriguingly, camalexin prevented any thermally visible binding of harmane to AtGSTF2. Similarly the presence of harmane prevented any ITC-detectable association with camalexin, confirming that the two ligands shared a (presumably single) binding site. Quercitrin showed stoichiometric binding to AtGSTF2, albeit with a somewhat weaker affinity (Kd=6 μM) than determined with the heterocyclic ligands. Binding by quercitrin was abrogated by pre-binding of harmane, again suggesting that both ligands were binding at the same site.
To better understand the results obtained from binding studies, the active sites of AtGSTF2 and AtGSTF3 were examined, making use of the available crystal structure for AtGSTF2 (PDB code 1BX9). Of particular interest were residues around the active site that were conserved in these two GSTs but altered in related enzymes that did not show the selective binding to harmane or lumichrome and could therefore be involved in the ligand associations. Phe123 was a promising candidate since this residue is substituted with an isoleucine residue in the enzymes AtGSTF4 and AtGSTF5 which do not bind these heterocyclic compounds. On generating the mutant AtGSTF2–123I, the purified enzyme showed considerably higher CDNB-conjugating activity than the parent AtGSTF2. In addition, unlike AtGSTF2, the conjugating activity of the mutant was inhibited by high levels of harmane (Figure 5). As compared with AtGSTF2, AtGSTF2–123I bound harmane and lumichrome with only slightly reduced affinity and with similar heat release (Table 1), showing that this mutation had not disrupted ligand binding. However, overall a lower binding stoichiometry was observed, suggesting that only approximately half of the protein was active, which was indicative of the inefficient conformational folding of the mutated protein.
The functional characterization of GST superfamily members in Arabidopsis and other plants remains challenging, at least in part due to the potential functional redundancy resulting from the expansion of GST family gene numbers. Genetic approaches have had some limited success in unravelling GST function. For example, disrupting the expression of AtGSTF12 caused an obvious pigment phenotype in Arabidopsis seeds that could be linked to transport of anthocyanins and proanthocyanidins , although even here the precise function of the GST remains unknown. However, altering the expression of other GSTs to perturb function has generally been less successful. For example, no major disruption in the metabolism or physiology of Arabidopsis was determined following the RNAi (RNA interference) knockdown of multiple GSTF family members . Instead, progress on defining plant GST function has been through determining unexpected associations with other proteins of interest, such as the discovery of AtGSTU20 interacting with phytochrome signalling components . In an attempt to apply an alternative systematic strategy to define GST function, we have developed a ligand fishing approach that uses global metabolite profiling of extracts purified by virtue of their affinity to a protein of interest, to identify tightly binding ligands. We have already shown that this approach works to identify not only ligands of lambda class GSTs, but also to subsequently identify a natural function for these enzymes . In the present study, we have applied this approach to Arabidopsis phi class GSTs, in particular AtGSTF2, and have identified some interesting potential functions for this protein in planta.
An initial metabolite profiling screen of bacterially expressed AtGSTFs identified AtGSTF2 (and its close homologue AtGSTF3) as causing an unusual accumulation of heterocyclic compounds in the host bacteria, due to their tight binding to the ectopically expressed proteins. None of the identified ligands were bacterial in origin and instead originated from the rich culture medium and showed similarities to plant natural products. The four indole derivatives identified were probably formed through tryptophan reacting with other medium components during the autoclaving of the medium [25,30]. Lumichrome was most probably present as a photodegradation product of the riboflavin present in yeast extract . In each case, harmane, norharmane and lumichrome all bound to AtGSTF2 far more tightly than had been described for previously identified interactions of this protein with auxins and flavonoids . Although lumichrome was not identified as an AtGSTF2 ligand in Arabidopsis, it is a well-known plant metabolite with an established activity as a rhizobial signalling agent . Therefore AtGSTF2 could regulate the biological activity and associated photoactivation of lumichrome in planta by being involved in its transport and storage. For example, AtGSTF2 is known to be up-regulated in response to pathogens, whereas riboflavin is implicated in the induction of systemic disease resistance in plants , most probably following its transformation to lumichrome. The identified indole derivatives harmane, norharmane and 1-acetyl-β-carboline are unusual natural products that are not known to occur in Arabidopsis. However, Arabidopsis does contain a variety of indole-derived defence-related metabolites that these ligands could be mimicking, and the availability of pure standards made these useful for in vitro binding studies.
The binding studies with lumichrome and AtGSTF2 were mechanistically revealing, as they demonstrated that the association with one ligand caused the protein to both increase its enzymic conjugating activity and enhance its binding to secondary ligands, notably harmane. Both harmane and lumichrome increased AtGSTF2 enzyme activity, without affecting the affinity of the enzyme for its substrates. Further studies with the wild-type and mutant AtGSTF2–123I enzymes showed that low concentrations (<5 μM) of harmane enhanced activity in both cases (Figure 5). However, although higher levels further increased activity in the wild-type enzyme, a slow reduction in conjugation was determined with the AtGSTF2–123I mutant (Figure 5). On the basis of the kinetic data, we postulate that binding of harmane/lumichrome to AtGSTF2 promotes the release of the DNB-GSH reaction product, thereby enhancing enzyme turnover. It is unclear whether the non-substrate ligands bind adjacent to the active site and directly influence catalysis, or act further away as allosteric effectors. The competitive nature of ligand binding suggests that these compounds associate at the same location. In addition, ligand and GSH binding are co-operative, with the association enhancing catalysis without obviously altering substrate affinity. The two-phase effects seen on catalysis with increasing harmane concentration suggest the presence of a high-affinity harmane-binding site that activates the enzyme and a much lower affinity binding site (or sites) that activates the wild-type enzyme but inhibits the P123I mutant. The nature of these binding sites awaits further investigation.
The association of AtGSTF2 with multiple ligands had previously been observed in structural biology studies, which showed each monomer to bind two molecules of the inhibitor S-hexylglutathione . Biochemical results consistent with AtGSTF2 having separate binding sites for the ligands IAA (indole-3-acetic acid) and NPA (N-1-naphthylphthalamic acid) have also been reported . The present study provides direct evidence of how ligand-binding interactions with AtGSTF2 could modulate signalling events, with one group of compounds (e.g. lumichrome) influencing the binding to other biologically active natural products (e.g. indole derivatives). Alternatively, such binding could directly modulate the conjugating activity of AtGSTF2 toward molecules involved in signalling, thereby altering their activity. Consistent with AtGSTF2 having a signalling role, previous studies have shown the respective gene to be responsive to diverse stimuli, including plant hormones (auxins and ethylene) and pathogen attack [12,32]. A regulatory role for AtGSTF2 in development and stress tolerance has also been suggested by the observation that Arabidopsis plants expressing sense and antisense constructs of the orthologous gene from Brassica juncea showed subtle changes in flowering time, along with differences in shoot regeneration and stress resistance .
The binding of AtGSTF2 to indoles, such as camalexin in Arabidopsis extracts, and the β-carboline alkaloids harmane, norharmane and 1-acetyl-β-carboline, and indole-3-aldehyde in E. coli cultures was particularly interesting. AtGSTF2 was originally identified as a membrane-associated protein which could be labelled with azido-activated IAA in photoaffinity experiments . Our own ITC studies showed that neither IAA, kinetin nor trans-zeatin were tightly bound by AtGSTF2, confirming that this protein is unlikely to have a physiological role in binding these plant hormones . However, AtGSTF2 could bind other plant indole-derived metabolites sufficiently well to exert a physiological effect. For example the AtGSTF2 ligand indole-3-aldehyde is a known biologically active plant metabolite , with its application causing an inhibition of lateral bud growth . More strikingly, tight (Kd=1.2 μM) binding of AtGSTF2 to camalexin is very suggestive of a physiological role, perhaps in the intracellular transport of this reactive and defence-inducible secondary metabolite. A strong link between camalexin and GSH already exists, with good evidence that the sulfur in camalexin is introduced through glutathionylation of a cytochrome P450-activated precursor [39,40]. This conjugation may occur spontaneously, but it is likely that in vivo the reaction is more carefully regulated using a GST, with AtGSTF6 implicated in this role . Intriguingly, AtGSTF2 shows a very similar pattern of ligand binding to AtGSTF6 and shares common features in its expression . As such, it is therefore possible that AtGSTF2 may also catalyse the glutathione conjugation of a related camalexin precursor. A similar conjugation also occurs during (indole) glucosinolate synthesis , providing an alternative potential role for AtGSTF2. GSTs may therefore be involved both in the synthesis of these indole-derived defence compounds, and in the subsequent intracellular transport of these compounds, as illustrated using camalexin as an example (Figure 6).
AtGSTF2 was also shown, along with AtGSTF6, to selectively bind the flavonol quercitrin in Arabidopsis extracts, even though this is a very minor metabolite in the plant extracts. AtGSTF2 has also been shown to interact with flavonoids in earlier studies , with the binding of kaempferol inhibiting associations with the ligand NPA. This binding was shown to be competitive, with 100 μM kaempferol inhibiting binding to either NPA or IAA by two-thirds . In the present study, the binding of AtGSTF2 to quercitrin was weaker than that observed with the N-containing ligands, suggesting that binding to camalexin and related molecules would be more likely than associations with flavonols. However, previous studies in Arabidopsis have shown that AtGSTF2 had an altered subcellular localization in flavonoid-deficient mutants as compared with wild-type plants, suggesting that flavonol-binding was physiologically relevant . This dual binding activity may suggest that AtGSTF2 functions to transport flavonoids under normal conditions, but in response to pathogen attack then preferentially transports the transiently abundant indole-derived phytoalexins, with its expression level increased to cope with the increased demand. If AtGSTF2 did have a binding/transport function for natural products this would help to explain the relative abundance of the respective protein, which can be easily affinity-purified from relatively small amounts of Arabidopsis tissue .
GSTs are typically thought of as detoxifying conjugating enzymes. However, neither AtGSTF2 nor AtGSTF6 have been shown to form conjugates in planta, with the recombinant enzymes showing little activity towards model substrates such as CDNB . Similarly, neither AtGSTF conjugated any of the ligands identified, although their binding to the protein was shown to be promoted in the presence of GSH. This provides further evidence that such GSTs retain a functional dependence on GSH, even when performing non-catalytic binding functions. Although the studies described in the present paper cannot unequivocally identify a function for AtGSTF2 and other GSTFs in planta, they do give a mechanistic context for how this protein could perform a regulatory role in responding to different classes of ligands through alterations in structure and function. When considered along with the strong and rapid induction of AtGSTF2 in response to pathogens  and its association with plasma membrane vesicles , this does suggest that this protein performs a regulatory transport function involving the binding and export of small bioactive natural products during plant stress.
Robert Edwards and David Dixon conceived the study and wrote the paper, with David Dixon performing the experiments. Jonathan Sellars assisted with chemical synthesis, data interpretation and revision of the paper.
This work was supported by the Biotechnology and Biological Sciences Research Council [grant number BBC51227X1]. R. E. is supported by a research development fellowship awarded by the Biotechnology and Biological Sciences Research Council.
Abbreviations: CDNB, 1-chloro-2,4-dinitrobenzene; DHAR, dehydroascorbate reductase; GST, glutathione transferase; HBS, Hepes-buffered saline; IAA, indole-3-acetic acid; ITC, isothermal titration calorimetry; NPA, N-1-naphthylphthalamic acid
- © The Authors Journal compilation © 2011 Biochemical Society