The combined use of photoaffinity labeling and surface plasmon resonance-based technology identifies multiple salicylic acid-binding proteins




Salicylic acid (SA) is a small phenolic molecule that not only is the active ingredient in the multi-functional drug aspirin, but also serves as a plant hormone that affects diverse processes during growth, development, responses to abiotic stresses and disease resistance. Although a number of SA-binding proteins (SABPs) have been identified, the underlying mechanisms of action of SA remain largely unknown. Efforts to identify additional SA targets, and thereby elucidate the complex SA signaling network in plants, have been hindered by the lack of effective approaches. Here, we report two sensitive approaches that utilize SA analogs in conjunction with either a photoaffinity labeling technique or surface plasmon resonance-based technology to identify and evaluate candidate SABPs from Arabidopsis. Using these approaches, multiple proteins, including the E2 subunit of α-ketoglutarate dehydrogenase and the glutathione S-transferases GSTF2, GSTF8, GSTF10 and GSTF11, were identified as SABPs. Their association with SA was further substantiated by the ability of SA to inhibit their enzymatic activity. The photoaffinity labeling and surface plasmon resonance-based approaches appear to be more sensitive than the traditional approach for identifying plant SABPs using size-exclusion chromatography with radiolabeled SA, as these proteins exhibited little to no SA-binding activity in such an assay. The development of these approaches therefore complements conventional techniques and helps dissect the SA signaling network in plants, and may also help elucidate the mechanisms through which SA acts as a multi-functional drug in mammalian systems.


Salicylic acid (SA) is a small phenolic molecule that is widely used as a multi-functional drug either directly or as the active component of aspirin (Hart and Harrison, 1996; Bielefeld et al., 2010; Duthie and Wood, 2011; Ekinci et al., 2011; Algra and Rothwell, 2012; Rothwell et al., 2012). SA is naturally produced in plants, and its synthesis is widely distributed throughout the plant kingdom (Duthie and Wood, 2011). In plants, SA serves as a hormone that regulates diverse processes, one of which is the plant defense response against pathogen infection (Vlot et al., 2009). In nature, plants are constantly subjected to pathogen attack. To survive, they have evolved a multi-layer immune system comprising PAMP-triggered immunity and effector-triggered immunity (Chisholm et al., 2006; Jones and Dangl, 2006). In addition to PAMP-triggered immunity and effector-triggered immunity, which occur at the site(s) of infection, plants may also develop a whole-plant immune response called systemic acquired resistance, which confers long-lasting protection against a broad spectrum of pathogens in distal tissues following an initial infection (Durrant and Dong, 2004; Dempsey and Klessig, 2012). Many studies have shown that SA plays a central role in activating PAMP-triggered immunity, effector-triggered immunity and systemic acquired resistance (Lu, 2009; Vlot et al., 2009; An and Mou, 2011). In addition, SA has been shown to regulate many other plant processes, such as resistance to abiotic stresses, flowering, thermogenesis, seed germination, senescence, fruit ripening, photosynthesis, respiration, stomatal closure and cell growth (Vlot et al., 2009; Rivas-San Vicente and Plasencia, 2011).

Despite its importance in so many processes, the underlying mechanisms of SA action remain largely unknown. Most studies have focused on elucidating the SA-mediated defense signal transduction pathways activated by pathogen attack. Genetic screens identified nonexpressor of PR genes1 (NPR1) as a key positive signaling component of one of the SA-mediated signaling pathways (Dong, 2004). NPR1 is a transcription co-activator that interacts with other transcription factors to turn on the expression of defense genes as SA levels increase following infection (Dong, 2004). The function of NPR1 relies on at least two SA-dependent post-translational events (Mou et al., 2003; Spoel et al., 2009). The first occurs in the cytoplasm, where SA-induced cellular redox changes result in reduction of disulfide bridges between NPR1 molecules. This leads to dissociation of oligomeric NPR1 into its monomeric form, which is then translocated into the nucleus (Mou et al., 2003). The second occurs in the nucleus, where NPR1 undergoes SA-dependent phosphorylation. This phosphorylation facilitates recruitment of NPR1 to a Cullin 3-based ubiquitin ligase (CUL3), leading to proteasome-mediated turnover, which is required for activation of defense responses (Spoel et al., 2009). In addition to the NPR1-dependent signaling pathway, a large body of evidence indicates that an SA-mediated NPR1-independent pathway or pathways also exists (Lu, 2009; An and Mou, 2011). Unlike the NPR1-dependent pathway, this pathway is less well characterized.

Identification of the direct targets of SA has proven to be an effective way to reveal SA signaling mechanisms. Several proteins, including catalase (Chen et al., 1993), ascorbate peroxidase (Durner and Klessig, 1995), chloroplast carbonic anhydrase (Slaymaker et al., 2002) and methyl salicylate esterase (tobacco SABP2 and Arabidopsis AtMES9) (Kumar and Klessig, 2003; Forouhar et al., 2005; Vlot et al., 2008), have been identified as SA-binding proteins (SABPs). Catalase and ascorbate peroxidase are scavengers of H2O2; SA inhibits their enzymatic activities, leading to a build-up of H2O2, which acts as a second messenger to activate defense-related genes (Chen et al., 1993; Durner and Klessig, 1995; Orozco-Cardenas et al., 2001). H2O2 also affects the redox status of the cell, which may alter NPR1 cytoplasmic/nuclear distribution. The identification and characterization of methyl salicylate esterase as an SABP has provided better insights into the mechanisms of systemic acquired resistance signaling. Tobacco SABP2 and five of the 18-member family of methyl esterases (MES) in Arabidopsis (AtMES1, 2, 4, 7 and 9) preferentially catalyze conversion of biologically inactive methyl salicylate (MeSA) to active SA (Forouhar et al., 2005; Vlot et al., 2008). In the infected leaf, SA feedback inhibits the activity of these enzymes. This results in accumulation of MeSA, a mobile signal that moves to the distal tissues where it is converted by SABP2 to SA to activate systemic acquired resistance (Park et al., 2007). Very recently, two research groups reported identification of the long-sought SA receptor(s). Fu et al. (2012) showed that the NPR1 paralogs NPR3 and NPR4 directly bind SA. They proposed that NPR3 and NPR4 function as adaptors of CUL3 to mediate NPR1 turnover. Binding to SA modulates their interaction with NPR1 in a SA concentration-dependent manner, thereby fine-tuning NPR1 homeostasis to specify different types of defense responses. They failed to detect SA binding by NPR1, thereby excluding it as the SA receptor. By contrast, Wu et al. (2012) reported that NPR1 itself is the SA receptor. Binding of SA induces a conformational change in NPR1 that relieves the inhibitory effect of the N-terminal auto-inhibitory BTB/POZ domain on the C-terminal transactivation domain. Despite the controversy, addition of NPR1, NPR3 and NPR4 to the list of SABPs provides important insight into the mechanism(s) of SA-mediated, NPR1-dependent plant defense signal transduction.

Although identification and characterization of each of these SA targets has increased our understanding of how SA mediates its effects in plant immunity, many questions remain. For example, the mechanisms by which SA mediates NPR1-independent defense signaling and many other cellular processes remain elusive. Moreover, the very recent discoveries of not only the critical NPR1, NPR3 and NPR4 as SA targets in plants, but also AMP-activated protein kinase, a central regulator of cell growth and metabolism, as a SA target in humans (Hawley et al., 2012), suggest that much remains to be discovered concerning the mechanisms of action of SA in both kingdoms. Therefore, to further our understanding of SA-mediated signaling mechanisms in plants, a search for additional SA targets is required.

Identification of additional SABPs has been challenging. The only reported approach to isolate unknown SABPs from the plant proteome utilized classical protein biochemistry, which involves multiple steps of protein fractionation and purification during which SA-binding activity was followed by size-exclusion chromatography using radiolabeled SA (Chen et al., 1993; Durner and Klessig, 1995; Slaymaker et al., 2002; Kumar and Klessig, 2003). This approach is not only time-consuming but also is generally not sensitive enough to detect proteins that bind SA transiently and/or with low affinity. In the present study, we describe two more sensitive approaches, a photoaffinity labeling-based approach and a surface plasmon resonance (SPR)-based approach, for identifying and validating SABPs in a systematic screen of Arabidopsis total proteins.


Development of a photoaffinity labeling approach to identify candidate SABPs and test their SA-binding activity

Photoaffinity labeling involves covalently cross-linking a photoreactive ligand with an interacting partner that is subsequently identified through the known ligand (Sadakane and Hatanaka, 2006). As formation of the covalent bond prevents dissociation during the detection process, this method may be used to isolate a complex formed by weak and/or transient interactions (Sadakane and Hatanaka, 2006). To identify novel SABPs using this approach, we took advantage of the commercially available photoreactive SA analog 4-azido salicylic acid (4-AzSA), which consists of an SA moiety bearing a photoreactive azide group at position 4 on the phenol ring (Figure 1a). To determine whether 4-AzSA is a biologically active SA mimic, its ability to induce the expression of pathogenesis-related gene 1 (PR-1), which encodes a prototypic SA-inducible defense protein, was assessed. 4-AzSA induced accumulation of PR-1, but to a weaker extent than SA (Figure 1b). To assess the ability of known SABPs, such as AtMES9, to bind 4-AzSA, a competition assay was performed. AtMES9 binding to tritium-labeled SA ([3H]SA) was reduced by 38.4% in the presence of a 100-fold excess of 4-AzSA, compared to a 77% reduction with unlabeled SA (Figure 1c). In contrast, 4-hydroxybenzoic acid (4-HBA), which is an inactive SA analog (Durner and Klessig, 1995), failed to inhibit [3H]SA binding. In summary, although 4-AzSA is less effective than SA at either inducing PR-1 expression or competing for AtMES9 binding, it is a functional SA mimic and may therefore be used as a photoreactive probe to screen for candidate SABPs (cSABPs).

Figure 1.

4-AzSA is a functional photoreactive SA analog. (a) Chemical structure of 4-AzSA. (b) Immunoblot analysis of PR-1 induction (α-PR-1) in tobacco leaves infiltrated with 1 mm SA, 4-HBA or 4-AzSA. Infiltration with 0.16% ethanol (Mock 1) served as a mock treatment for SA and 4-HBA, and infiltration with 0.16% DMSO (Mock 2) served as a mock treatment for 4-AzSA. Non-Inf, non-infiltration control. Proteins stained with Coomassie brilliant blue (CBB) served as a loading control. (c) Binding assay of AtMES9 (0.1 μg μl−1) with [3H]SA (300 nm) in the absence (−) or presence (+) of a 100-fold excess of unlabeled SA or its analogs as competitor, performed by size-exclusion chromatography. Error bars represent SE values calculated from three replications of a single experiment that was representative of three experiments performed.

As detection of proteins cross-linked to 4-AzSA after UV irradiation depends on the ability of anti-SA antibody (α-SA Ab) to recognize 4-AzSA, this interaction was tested indirectly using the competition assay. Binding between the α-SA Ab and [3H]SA was disrupted to a similar extent with 4-AzSA or SA (Figure S1). Thus, the α-SA Ab appears to recognize 4-AzSA and SA with similar efficiency.

Whether the photoaffinity labeling approach could be used to specifically isolate proteins that bind SA was then tested using SABP2 as a positive control and the actin-binding protein AtADF4 (Tian et al., 2009) as a negative control. These recombinant, Escherichia coli-produced proteins were pre-incubated with or without 4-AzSA, irradiated with UV light, and then subjected to immunoblot analysis with α-SA Ab. The α-SA Ab detected an appropriate molecular weight band in the SABP2 sample pre-incubated with 0.05 mm 4-AzSA, but not in the control sample lacking either 4-AzSA or UV irradiation (Figure 2a). By contrast, no specific signal was detected for AtADF4 samples pre-incubated with increasing concentrations of 4-AzSA (0.05–0.5 mm) (Figure 2b). To assess whether photolabeling with 4-AzSA accurately reflects the SA-binding activity of SABP2, the photolabeling reactions were performed in the presence of increasing concentrations of SA. SA effectively inhibited cross-linking of 4-AzSA with SABP2 (Figure 2a), indicating that SABP2 binds 4-AzSA via SA. Together, these results indicate that the photoaffinity labeling approach specifically identifies proteins that bind SA, and that the 4-AzSA–protein complex may be detected using α-SA Ab.

Figure 2.

Evaluation of the photoaffinity labeling approach using 4-AzSA as a probe to identify SABPs. (a) SABP2 was incubated with (+) or without (−) 4-AzSA in the absence or presence of the indicated concentrations of SA for 1 h, and then treated with (+) or without (−) UV light (50 mJ). (b) AtADF4 was incubated with the indicated concentrations of 4-AzSA, and then treated with (+) or without (−) UV light (50 mJ). Proteins labeled with 4-AzSA were detected by immunoblot analysis with the α-SA Ab. Proteins stained with Coomassie brilliant blue (CBB) served as a loading control.

To screen for cSABPs, protein extracts isolated from the uninoculated distal leaves of Arabidopsis after infection with avirulent Pseudomonas syringae pv. maculicola (Psm) ES4326 expressing AvrRpt2 were incubated with 4-AzSA before irradiation with UV. 4-AzSA-cross-linked proteins were pulled down using α-SA Ab-bound Protein G resin, eluted from the resin with SA, and then subjected to SDS–PAGE fractionation. As both the α-SA Ab and Protein G resin may have non-specific cross-reactivity, parallel reactions in which either 4-AzSA or UV irradiation were eliminated from the assay were performed as negative controls. To identify cSABPs, two approaches were then utilized. For the first approach, silver-stained protein bands that were more abundant in the test sample than in the negative controls were cut out of the SDS–PAGE gel and analyzed by mass spectrometry (MS) (Figure 3a). One of the proteins identified using this approach was the E2 subunit of the α-ketoglutarate dehydrogenase enzyme complex (KGDHE2), encoded by At5g55070 (Figure 3a and Table 1). For the second approach, a label-free semi-quantitative proteomics approach was utilized to analyze all proteins from both test and control samples. The samples were subjected to SDS–PAGE, and the complete gel lanes were processed for MS analysis. Putative 4-AzSA-cross-linked proteins were identified based on their abundance in the test sample versus the negative control, as determined based on spectral counts (Friso et al., 2011). Proteins with a minimum of two spectral counts identified in the test sample that were at least twofold more abundant than in the negative control or absent in the negative control were chosen as putative cSABPs. Several dozen proteins were identified as putative cSABPs, including GSTF2 (At4g02520) and GSTF8 (At2g47730), two members of the Phi class of the glutathione S-transferase (GST) family (Table 1). KGDHE2 and its paralog, encoded by a gene at a second locus (At4g26910), were also identified using this approach (Table 1).

Figure 3.

SA-binding activity of KGDHE2 detected by photoaffinity labeling with 4-AzSA. (a) Affinity purification of proteins photolabeled with 4-AzSA. Arabidopsis protein extracts were incubated with (+) or without (−) 4-AzSA for 1 h, and then treated with (+) or without (−) UV light. 4-AzSA photolabeled proteins were pulled down using α-SA Ab-bound Protein G resin, followed by separation by SDS–PAGE and silver nitrate staining. The protein band representing KGDHE2 is indicated by an arrow. (b) Purified recombinant KGDHE2 protein was incubated with (+) or without (−) 4-AzSA (0.5 mM) for 1 h, then treated with (+) or without (−) UV light (30 mJ). (c) KGDHE2 was incubated with (+) or without (−) 4-AzSA in the absence or presence of the indicated concentrations of SA for 1 h prior to UV light irradiation (30 mJ). 4-AzSA-cross-linked proteins were detected by immunoblot analysis with α-SA Ab. Proteins stained with Coomassie brilliant blue (CBB) served as a loading control.

Table 1. Identification of GSTF2, GSTF8, KGDHE2 and its paralog as cSABPs by MS
  1. Experiment 1 was performed by cutting differential protein bands; experiments 2 and 3 were performed by a label-free proteomics approach.

  2. a

    Highest observed score.

Locus IDProtein nameExperimentPeptide sequenceCharge (z)Ion scoreappm (maximum mass error)MS/MS spectral countsPrecursor intensity
At5g55070 KGDHE21 (see Figure 3a)ISTSADAVSHVAPSEK2670.38138 529
DADKMNFADIEK2371.36130 132
3ISTSADAVSHVAPSEK2782.823167 490



KGDHE2 and/or KGDHE2 paralog2GLVVPVIR231−0.121876 434
LGLMoxSGFIK267−0.12672 214
At4g02520 GSTF22NPFGQVPAFEDGDLKLFESR3430.971146 384
VNEWVAEITK259−0.592245 572
At2g47730 GSTF82GMoxFGMoxTTDPAAVQELEGK21210.262100 967

Using purified recombinant proteins tagged with six histidines (His), we tested the SA-binding activity of KGDHE2 and several members of the GST Phi class by photoaffinity cross-linking with 4-AzSA, followed by immunoblot analysis with α-SA Ab. Similar to SABP2, UV irradiation cross-linked KGDHE2 with 4-AzSA (Figure 3b); this cross-linking was inhibited in the presence of excess SA (Figure 3c), confirming the SA-binding activity of KGDHE2. As members of the GST Phi class, we tested not only GSTF2 and GSTF8, but also GSTF10 (At2g30870) and GSTF11 (At3903190). GSTF2, GSTF8 and GSTF10 cross-linked with 4-AzSA in an SA-inhibitable fashion (Figure 4), confirming their interaction with SA. GSTF11 also cross-linked with 4-AzSA, but this was not inhibited by excess SA (Figure S2).

Figure 4.

SA-binding activity of several GSTFs assessed by photoaffinity labeling with 4-AzSA.
Purified recombinant proteins GSTF2 (a), GSTF8 (b) and GSTF10 (c) were incubated with (+) or without (−) 4-AzSA in the absence or presence of the indicated concentrations of SA for 1 h, and then exposed to UV light (50 mJ). Detection of 4-AzSA-cross-linked proteins was performed by immunoblot analysis with the α-SA Ab. Proteins stained with Coomassie brilliant blue (CBB) served as a loading control.

The SA-binding activity of KGDHE2 and the GSTFs was not detected by a size-exclusion chromatography assay

Size-exclusion chromatography assays with [3H]SA have been used previously to identify several SABPs (Chen et al., 1993; Durner and Klessig, 1995; Slaymaker et al., 2002; Kumar and Klessig, 2003; Vlot et al., 2008). We therefore tested whether this assay detected the SA-binding activity of KGDHE2 and the GSTFs. In all cases, little to no SA-binding activity was observed (Figure S3). As size-exclusion chromatography is not sensitive enough to detect transient or weak biomolecular interactions, this result suggests that the interaction between these proteins and SA is weak and/or transient, and therefore cannot be detected unless a more sensitive technique is used.

Development of an SPR-based approach to characterize SABPs

To develop an additional sensitive approach to assess the SA-binding activity of KGDHE2, the GSTFs and other cSABPs, we explored SPR-based approaches. SPR may be used to monitor a wide variety of biomolecular interactions in real time (Cooper, 2002), and therefore may be a more effective approach to detect proteins that interact with SA with low affinity and/or in a transient manner. To detect SA-binding activity by SPR, we initially immobilized SABP2 and AtMES9 onto the sensor chip and then passed an SA-containing solution over the chip. This approach was unsuccessful, presumably due to the small mass of SA. Direct detection of binding of small molecules to immobilized proteins using SPR was previously shown to suffer from low signal and poor sensitivity (Mitchell, 2010).

Therefore, to utilize SPR, we devised a reverse approach in which SA was immobilized on the sensor chip as a ligand and the proteins were passed over the chip. Because the proteins have a large molecular weight, the resonance of the immobilized ligand is significantly altered upon protein binding, and this dramatically increased the sensitivity of the assay. To immobilize SA on the chip, we had synthesized an SA analog, 3-aminoethyl salicylic acid (3-AESA) (Figure 5a), which was affixed to the sensor chip via an amide bond formed between the amine group of 3-AESA and the carboxyl groups on the chip (Cooper, 2002). The ethylamine group was added to the 3 position on the phenol ring based on the PR-1 expression-inducing ability and AtMES9-binding activity of several SA analogs substituted at various positions on the phenol ring with methyl or methoxyl groups. Both 3-methyl SA and 3-methoxyl SA were functional SA analogs (Figure S4). The amine group was extended a little farther from the phenol ring by adding an ethyl rather than a methyl group.

Figure 5.

Evaluation of the SPR-based approach to detect SA-binding activity.
(a) Chemical structure of 3-AESA. (b) Sensorgrams of AtMES9 (0.1 μg μl−1) and maltose-binding protein (MBP, 0.1 μg μl−1) interacting with 3-AESA immobilized on the sensor chip. (c–f) Sensorgrams of AtMES9 (0.1 μg μl−1, c, d) and lactoperoxidase (0.02 μg μl−1, e, f) pre-incubated with the indicated concentrations of SA, 4-HBA or the same volume of 1× PBS buffer for 1 h before injection onto the sensor chip surface. The signal from the mock-coupled surface was subtracted.

After 3-AESA was immobilized to the sensor surface of the CM5 chip, proteins known to bind SA, such as AtMES9, SABP2, bovine lactoperoxidase (PDB ID 2QQT) and α-SA Ab, were passed over the sensor chip surface. A binding signal for AtMES9, bovine lactoperoxidase and α-SA Ab was readily detected in the flow cell to which 3-AESA had been immobilized (Figures 5 and S5a). Surprisingly, we did not detect a signal for SABP2, which binds SA very strongly based on the [3H]SA size-exclusion chromatography assay (Kumar and Klessig, 2003). To determine whether the binding signal detected for AtMES9, bovine lactoperoxidase and α-SA Ab was a specific effect, rather than a non-specific effect generated by the flow of protein solutions over the sensor surface, we passed 0.1 μg μl−1 of maltose-binding protein and a series of 28 randomly chosen additional proteins over the sensor surface; they gave either no response or a very low signal (Figure 5b). Whether binding of AtMES9 and lactoperoxidase to the sensor surface truly reflects their SA-binding activity was further assessed by pre-incubating these proteins with various concentrations of SA or 4-HBA before passing them over the chip. SA inhibited binding of these proteins to the sensor chip in a concentration-dependent manner (Figure 5c,e). In contrast, 4-HBA inhibited their binding to the sensor chip much less effectively (Figure 5d,f). In summary, these data suggest that binding of AtMES9 and lactoperoxidase to the chip is specific and dependent on the SA moiety of 3-AESA, and is therefore a true reflection of their SA-binding activity.

SPR analysis indicates that KGDHE2, GSTF2, GSTF8, and GSTF11 are SABPs

We then utilized our optimized SPR settings to further assess whether KGDHE2 and the GSTFs are true SABPs. A range of concentrations of KGDHE2 (5–100 ng μl−1) was tested, and a strong, concentration-dependent response was observed (Figure S5b). KGDHE2-binding specificity was assayed via competition with SA. Similar to the results obtained with AtMES9 and lactoperoxidase, SA inhibited the binding signal of KGDHE2 in a concentration-dependent manner, while inhibition by 4-HBA was less effective (Figure 6a,b). Similar results were obtained for GSTF2, GSTF8 and GSTF11 (Figure 7a–f). These results provide further evidence that KGDHE2, GSTF2, GSTF8 and GSTF11 are true SABPs. In contrast, no binding signal was generated using GSTF10 or GSTF9 (At2g30860), another member of the Phi class of Arabidopsis GSTs (Figure S5c,d).

Figure 6.

Analysis of the SA-binding activity of KGDHE2 by SPR. Sensograms of KGDHE2 (0.01 μg μl−1) pre-incubated with the indicated concentrations of SA, 4-HBA or the same volume of 1× PBS buffer for 1 h before injection onto the sensor chip surface. The signal from the mock-coupled surface was subtracted.

Figure 7.

Assessment of SA-binding activities of the GSTFs via SPR.
Sensograms of GSF2 (0.1 μg μl−1, a, d), GSTF8 (0.1 μg μl−1, b, e) and GSTF11 (0.1 μg μl−1, c, f) pre-incubated with the indicated concentrations of SA, 4-HBA or the same amount of 1× PBS buffer for 1 h before injection onto the sensor chip surface. The signal from the mock-coupled surface was subtracted.

SA inhibits the enzymatic activities of GSTF8, GSTF10 and GSTF11

To further assess the SA-binding activity of the GSTFs identified by photoaffinity labeling and confirmed by SPR analysis, we analyzed the effect of SA on their enzymatic activity. As the enzymatic activity of GSTF2 was very low using 1-chloro-2,4-dinitrobenzene as the substrate, the effect of SA on GSTF2 could not be assessed. The enzymatic activity of GSTF8, GSTF10 and GSTF11 were readily detected, and these enzymes were inhibited in a concentration-dependent manner by SA but not 4-HBA (Figure 8a–c). In contrast, neither SA nor 4-HBA inhibited the enzymatic activity of GSTF9 (Figure 8d). Thus, inhibition of the enzyme activities of GSTF8, GSTF10 and GSTF11 by SA appears to be specific.

Figure 8.

SA inhibits the enzymatic activity of several GSTFs.
Enzyme activity of GSTF8 (a), GSTF10 (b), GSTF11 (c) and GSTF9 (d) in the presence of the indicated concentrations of SA (squares) or 4-HBA (triangles). Error bars represent SE values calculated from three replications of a single experiment. Experiments were performed at least twice with similar results.


In this study, we describe two approaches to identify cSABPs and/or evaluate their SA-binding activity. Using these approaches, we report the identification of five additional SABPs: KGDHE2, GSTF2, GSTF8, GSTF10 and GSTF11. The interaction between SA and GSTF8, GSTF10 and GSTF11 was confirmed by demonstrating that SA inhibits their enzymatic activity. The effect of SA on GSTF2 enzymatic activity was not assessed due to its low activity. However, as GSTF2 is an SABP, we anticipate that its enzymatic activity, like those of GSTF8, GSTF10 and GSTF11, may be inhibited by SA. In support of KGDHE2, a subunit of α-ketoglutarate dehydrogenase (α-KGDH) enzyme complex, being an SABP, a previous report showed that SA inhibited the enzymatic activity of α-KGDH from rat heart (Nulton-Persson et al., 2004).

Although photoaffinity labeling is a sensitive technology for capturing weak and/or transient interactors, a common problem is non-specific labeling, which may lead to a high percentage of false positives (Kotzyba-Hibert et al., 1995). Although additional analysis is required to distinguish true SABPs from the false positives, this approach generated a workable number of candidates with a reasonable probability of being true SABPs. Compared with classical purification approaches, which were used to identify catalase, ascorbate peroxidase, carbonic anhydrase and methyl salicylate esterase, this approach is much more efficient. Moreover, it does not rely on use of the less sensitive size-exclusion chromatography assay with [3H]SA to follow SA-binding activity during protein fractionation. Another method to identify cSABPs involves a hypothesis-driven approach. For example, NPR1 was identified based on its involvement in SA-mediated defense signaling (Wu et al., 2012), while NPR3 and NPR4 were identified based on their connection with NPR1 (Fu et al., 2012). While the SABPs identified by this hypothesis-driven approach have provided important information about the signaling mechanisms utilized in this well-studied, SA-associated pathway, de novo identification of cSABPs is more likely to reveal novel pathways and new mechanisms of SA action. As the photoaffinity labeling approach using 4-AzSA is more sensitive than existing techniques for identifying SABPs and it can be used to identify proteins whose association with SA is currently unknown, this approach provides a powerful and effective way to screen for cSABPs and to confirm the SA-binding activity of proteins identified through other techniques.

SPR has become the gold standard for measuring biomolecular interactions as it is sensitive and provides both quantitative and qualitative information on binding events (Maynard et al., 2009; Khan et al., 2012). Moreover, it is amenable to adaptation for high-throughput screening. The SPR-based approach that we have developed to assay SA-binding activity is very effective and efficient. With this approach, we not only detected the SA-binding activity of known SABPs, such as AtMES9, bovine lactoperoxidase and the α-SA Ab, but also validated this activity for several cSABPs. Pre-incubating the cSABPs with SA inhibited binding to the chip, suggesting that protein binding is specific and therefore reflective of SA-binding activity. Moreover, because 3-AESA appears to be quite stable once immobilized on the sensorchip, the chip can be sequentially used to test the SA-binding activity of many different proteins after the bound proteins are removed. One potential problem associated with this approach may be false negatives. In this study, although GSTF10 was shown to be a SABP based on both photoaffinity labeling and the ability of SA to inhibit its enzymatic activity, SA-binding activity was not detected using SPR. This was also the case for SABP2, which has high affinity (Kd = 90 nm) for SA (Forouhar et al., 2005). For both proteins, the false-negative results were probably due to steric hindrance. Structural analyses have shown that SA is bound deep in the active-site pocket of SABP2 (Forouhar et al., 2005). The ethyl arm attaching SA to the chip may be too short to allow SA to reach the SA-binding pocket/site in SABP2. In fact, steric hindrance is a common problem in attempts to detect small molecule–protein interactions when the small molecule is bound to the sensor surface (Mitchell, 2010). Therefore, multiple approaches should be used to avoid missing significant SABPs.

The identification of Arabidopsis KGDHE2 as an SABP using the photoaffinity and SPR approaches has opened several avenues for studying the mechanisms of action of SA. As SA was previously shown to be an inhibitor of mitochondrial respiration in plants and mammals (Xie and Chen, 1999; Norman et al., 2004; Nulton-Persson et al., 2004), it may target the mitochondrial electron transport chain and/or components in the TCA cycle which provide the reducing substrates (i.e. NADH and succinate) for electron transport. α-KGDH is a key enzyme complex that catalyzes a rate-limiting step in the TCA cycle and produces a limiting factor for respiration (Tretter and Adam-Vizi, 2000; Araujo et al., 2008). The discovery that SA inhibits the enzymatic activity of rat α-KGDH (Nulton-Persson et al., 2004) suggested that SA inhibits mitochondrial respiration by targeting α-KGDH in mammalian systems. α-KGDH is a large multi-enzyme assembly comprising three enzymes: E1 (α-ketoglutarate dehydrogenase), E2 (dihydrolipoamide succinyl transferase) and E3 (dihydrolipoamide dehydrogenase). Our identification of Arabidopsis KGDHE2 as an SABP suggests that the mechanism through which SA inhibits mammalian α-KGDH is via binding to the E2 subunit. The interaction between SA and KGDHE2 also may contribute to SA-mediated accumulation of reactive oxygen species (ROS), which are key signaling molecules in plant immunity and also help to modulate the cellular redox state (Asai et al., 2010). Supporting this possibility, inhibition of α-KGDH led to the accumulation of ROS in mammalian cells (Tretter and Adam-Vizi, 2004).

Plant GSTs constitute a diverse group of multi-functional proteins whose primary function is to catalyze the conjugation of glutathione (GSH) to a range of electrophilic substrates (Dixon et al., 2002; Moons, 2005). In Arabidopsis, the GST superfamily contains 54 members that have been grouped into six classes, one of which is Phi (Dixon et al., 2009). The biological relevance of SA binding to members of the Phi class of GSTs remains to be explored. One possibility is that this interaction plays a role in hypersensitive response during activation of effector-triggered immunity, as cell death may be facilitated by accumulation of high levels of toxic compounds, such as ROS and lipid peroxidation products (Adam et al., 1989). While the specific biochemical function of individual GSTs varies, it is generally believed that these enzymes are involved in detoxification of cytotoxic compounds of xenobiotic or endogenous origin (Dixon et al., 1998). In addition to catalyzing the conjugation of GSH to electrophilic compounds, some GSTs have GSH-dependent peroxidase activities, through which they detoxify ROS and organic hydroperoxides (Dixon et al., 1998; Moons, 2005). For example, GSTF2 and GSTF8 have peroxidase activity and catalyze conversion of cytotoxic linoleic acid-13-hydroperoxide and cumene hydroperoxide to nontoxic alcohols (Wagner et al., 2002). This finding suggests that these GSTs are scavengers of ROS and other by-products of lipid peroxidation during the hypersensitive response. Therefore, inhibition of their activity by SA may potentiate cell death. A second, non-mutually exclusive possibility is that SA binding by certain GSTFs activates disease resistance by modulating GSH homeostasis. In support of this hypothesis, several studies have shown that SA treatment and pathogen infection increase the GSH content, as well as the ratio between GSH and its oxidized state (GSSG) (Srivastava and Dwivedi, 1998; Mou et al., 2003). As the redox state of cells is largely controlled by the GSH concentration and the ratio between its reduced and oxidized forms (Ghanta and Chattopadhyay, 2011), the enhanced GSH:GSSG ratio upon SA treatment or pathogen infection changes the cellular redox toward a more reducing state, which is required to activate the NPR1-dependent signaling transduction pathway. Interestingly, SA also inhibits catalase and ascorbate peroxidase in plants (Chen et al., 1993; Durner and Klessig, 1995), providing further evidence of its intimate relationship with ROS signaling and redox regulation. In addition to disease resistance, SA binding by specific GSTFs may mediate other SA-regulated processes in plants. For example, GSTF2 has been implicated in auxin binding and transport (Moons, 2005). Thus, binding of SA by GSTF2 may play a role in SA–auxin interplay and its mediated processes.

It is interesting to note that, although the SA-binding activity of each of the five additionally identified SABPs was demonstrated using at least two independent approaches, it was not reliably detected using the size-exclusion chromatography assay with [3H]SA. These results suggest that these proteins bind SA with low affinity and/or transiently. Nonetheless, these interactions are likely to be biologically relevant since SA can reach high concentrations in planta (several hundred micromoles; Huang et al., 2006); moreover, while studies in mammalian systems indicate that SA effectively regulates its targets at concentrations in the millimolar range (Duthie and Wood, 2011; Hawley et al., 2012). It is likely that SA interacts with its many targeted effectors transiently, presumably to facilitate finely tuned regulation of numerous dynamic cellular processes. Thus, the development of these highly sensitive approaches for identifying SABPs should greatly facilitate dissection of the complex SA signaling networks in plants, and help reveal additional modes of action of this multi-functional drug in humans.

Experimental Procedures


SA, 4-HBA and lactoperoxidase were purchased from Sigma-Aldrich (, and 4-AzSA was purchased from Toronto Research Chemicals Inc. ( 3-AESA was synthesized by AstaTech Inc. (

Plant growth and pathogen inoculation

Arabidopsis thaliana and tobacco (Nicotiana tabacum cv. Xanthi-nc) were grown as described previously (Vlot et al., 2008). Four-week-old Arabidopsis Col-0 plants were inoculated by infiltrating three expanded leaves with Psm ES4326 AvrRpt2 at a concentration of 108 cfu ml−1 in 10 mm MgCl2. The uninoculated distal leaves were collected at 48 h post-inoculation.

Plasmid construction, protein expression and purification

The mitochondrial transit peptide of KGDHE2 was predicted by MITOPROT ( to consist of the N-terminal 93 amino acids. The plasmids pET28a-KGDHE2(94–464), pET28a-GSTF9 and pET28a-GSTF11 were constructed by cloning PCR-amplified protein-encoding sequences into the BamHI and SalI sites of pET28a (EMD Millipore Chemicals, to express N-terminally His-tagged proteins in E. coli. pET28a-GSTF2 and pET28a-GSTF10 were constructed by cloning open reading frames without stop codons into the NcoI and XhoI sites of pET28a to express C-terminally His-tagged proteins. GSTF8 was cloned into pET21c at the NcoI and NotI sites to express C-terminally His-tagged protein. The primers used were 5′-gcgggatccGTGGAAGCTGTTGTGCCACACATG-3' (forward) and 5′-gcggtcgacTCATATGTCGAGAAGAAGCCTCTG-3′ (reverse) for KGDHE2, 5′-gcgggatccATGGTGCTAAAGGTGTACGGAC-3′ (forward) and 5′-gcggtcgacTTAAGCTGGGAATGAATACTTG-3′ (reverse) for GSTF9, 5′-gcgggatccATGGTGGTCAAAGTATATGGGCA-3′ (forward) and 5′- GCGgtcgacTTAATAGGCAGCCAATTCCATG-3′ (reverse) for GSTF11, 5′-ccATGGCAGGTATCAAAGTTTTCGG-3′ (forward) and 5′-ctcgagCTGAACCTTCTCGGAAGCTG-3′ (reverse) for GSTF2, 5′-ccATGGTGTTGACAATCTATGCTCC-3′ (forward) and 5′-ctcgagAACAGGTAGTGAGTACTTAGCGG-3′ (reverse) for GSTF10, and 5′- ccATGGGAGCAATTCAAGCTCGTC-3′ (forward) and 5′-gcggccgcCTGCTTCTGGAGGTCAATAACC-3′ (reverse) for GSTF8. Gene-specific sequences are in upper case, and the restriction sites are underlined. The plasmids for expressing His-tagged SABP2 and AtMES9, and the methods to express and purify His-tagged proteins in E. coli, have been described previously (Kumar and Klessig, 2003; Vlot et al., 2009). FLAG-tagged AtADF4 was expressed and purified as described by Tian et al. (2009). Maltose-binding protein was expressed and purified from pMAL-c2X (New England Biolabs, according to the manufacturer's instructions.

PR-1 induction

Fully expanded leaves of 8-week-old tobacco plants were infiltrated with 5 mm Bis/Tris buffer, pH 6.5, containing 1 mm SA or its analogs, or, for the mock controls, the same amount of solvent as was used for dissolving the chemicals. Samples were collected 48 h post-infiltration. Expression of PR-1 was detected by Western blotting using an anti-PR-1 monoclonal antibody (Carr et al., 1987).

[3H]SA-binding assays

[3H]SA-binding assays were performed using a size-exclusion chromatography assay as described by Slaymaker et al. (2002). 4-AzSA binding by MES9 was determined by performing competition-binding assays (Kumar and Klessig, 2003).

SA-binding activity analysis by photoaffinity labeling

Purified proteins (10 μg) were incubated with 4-AzSA (0.05–0.5 mm) in 100 μl 1× PBS buffer containing various concentrations of SA for 1 h on ice, followed by UV irradiation with 254 nm UV light at an energy level of 30–50 mJ, using a GS GENE Linker UV chamber (Bio-Rad, Aliquots (7.5 μl) of reaction mixture were subjected to SDS–PAGE, and 4-AzSA-cross-linked proteins were detected by immunoblot analyses using α-SA Ab (Novus Biologicals,

Isolation of 4-AzSA-cross-linked proteins from Arabidopsis

Uninoculated distal leaves of Arabidopsis Col-0 plants were harvested 48 h post-inoculation after primary infection with Psm AvrRpt2. Leaf tissue (20 g) was ground to fine powder under liquid nitrogen, and homogenized in 50 ml of buffer A [25 mm Tris/HCl, pH 7.5, 1 mm EDTA, 150 mm NaCl, 1 mm dithiothreitol, 10% v/v glycerol, 0.1% v/v Triton X-100, 2% w/v polyvinylpolypyrrolidone, and two Complete mini EDTA-free protease inhibitor cocktail tablets (Roche,]. The homogenate was centrifuged at 20 000 g at 4°C for 20 min, filtered through Millex poly(vinylidene difluoride) Durapore membrane (Millipore, and equally divided into three tubes. The protein extracts were incubated with or without 0.5 mm 4-AzSA at 4°C for 1 h, followed by irradiation at 250 mJ. Excess 4-AzSA was removed by buffer exchange with buffer B (25 mm Tris/HCl, pH 7.5, 1 mm EDTA, 150 mm NaCl, 1 mm dithiothreitol) supplemented with protease inhibitor cocktail (one tablet in 50 ml) using an Amicon ultra filter unit (Millipore) until its concentration was reduced to approximately 1 μm. To expose the SA moiety of 4-AzSA from the binding pocket of the photolabeled proteins and make it fully accessible to the α-SA Ab, the proteins were denatured by adding an equal volume of 8 m urea dissolved in buffer B, followed by incubation on ice for 15 min. Buffer exchange with buffer B was further performed until the concentrations of 4-AzSA and urea reached approximately 30 nm and 300 mm, respectively. The protein preparations were then incubated at 4°C overnight with 100 μl of α-SA Ab-bound Protein G resin, which was prepared by washing Protein G resin (GenScript, with 1 ml buffer B for three times, binding with α-SA Ab for 3–4 h, followed by three-time washing with 1 ml buffer B to remove unbound α-SA Ab. The resin was washed once with 10 ml buffer C [buffer B containing 0.1% v/v IGPAL CA-630 (Sigma-Aldrich)], followed by additional five-time washing with 1 ml buffer C. 4-AzSA-cross-linked proteins were eluted using 100 μl of 5 mm SA in buffer B. The eluates were subjected to 12% SDS–PAGE, and stained with Coomassie brilliant blue or MS-compatible silver stain (Peltier et al., 2000) for identification by MS.

Identification of 4-AzSA-cross-linked proteins by MS

Two approaches described in 'Results' were used to identify 4-AzSA cross-linked proteins. For both approaches, the cut gel bands were reduced, followed by alkylation and in-gel digestion with trypsin; peptides were extracted as described previously (Zybailov et al., 2008). The resulting peptides were dried and suspended in 17 μl of 5% formic acid. All peptide samples were analyzed by reverse phase nano-liquid chromatography electrospray ionization MS/MS as described by Friso et al. (2011). Acquired MS/MS data were searched using Mascot 2.2 (Matrix Science, with a significance threshold of 0.01. Peak lists were searched against the A. thaliana database version 8 or version 10 (, including sequences for known contaminants (e.g. keratin and trypsin) (33 013 and 35 574 total sequences, respectively), and concatenated with a decoy database in which all sequences were in reverse orientation. For each peak list, data were searched as follows: tryptic search with the precursor ion tolerance window set at 6 ppm, with methionine oxidation set as a variable modification and carbamidomethylation (C) as a fixed modification. The ion score threshold was set to 33, which yielded a final peptide false-positive rate below 1%.

SA-binding activity analysis by SPR

SA-binding activity analysis by SPR was performed using a Biacore 2000 instrument. Immobilization of 3-AESA on a CM5 sensor chip (GE Healthcare, was performed using an amine coupling kit (GE Healthcare). After activation of the carboxyl groups using a mixture of N-(3-dimethyl aminopropyl)-N′-ethylcarbodiimide and N-hydroxysuccinimide, 10 mm of 3-AESA dissolved in 0.1 m borate buffer, pH 10, was passed over the surface of the flow cell for a period of 10 min at a flow rate of 5 μl min−1, and the injection was repeated five times before deactivation of excess reactive groups on the surface. Activation, deactivation and preparation of the mock-coupled flow cell were performed according to the manufacturer's instructions and as described previously (Hu et al., 2009). HBS-EP buffer (0.01 m HEPES, pH 7.4, 0.15 m NaCl, 3 mm EDTA, 0.005% v/v Surfactant P20; GE Healthcare, was used as the running buffer. To test SA-binding activity, proteins diluted in HBS-EP buffer were pre-incubated with various concentrations of SA or 4-HBA for 1 h on ice, and then passed over the sensor surface of the 3-AESA-immobilized and mock-coupled flow cells. The binding signals were generated by subtracting the signal for the mock-coupled flow cell from that for the 3-AESA-immobilized flow cell. To re-use the chip, bound proteins were stripped off by injecting either HCl solution (pH 1.5) or NaOH solution (pH 12).

Enzymatic activity analysis of the GSTFs

Enzymatic activity analysis of the GSTFs was performed using 1-chloro-2,4-dinitrobenzene as the substrate with a glutathione S-transferase assay kit (Sigma-Aldrich) according to the manufacturer's instructions. Enzymatic activity was determined by monitoring the absorbance change at 340 nm for 5 min using a SpectraMax Plus 384 spectrometer (Molecular Devices, To test the effect of SA or 4-HBA on enzymatic activity, the GSTFs were pre-incubated with various concentrations of SA or 4-HBA for 30 min at room temperature before the substrate was added to start the reaction.


We thank Dr. D'Maris Dempsey for critically reading the manuscript, Drs. Moonsoo Jin and Xuobo Hu from Department of Biomedical Engineering, Cornell University, and Sorina Popescu for their guidance and/or helpful discussion with SPR analyses, and Drs. Brad Day (Department of Plant Pathology, Michigan State University), Jitae Kim (Department of Plant Biology, Cornell University), Magali Moreau and Hong-Gu Kang for providing negative controls for SPR analyses. C.C.v.D. was supported by Deutsche Forschungsgemeinschaft grant number DA1239/1-1. This work was supported by US National Science Foundation grant number IOS-0820405 to D.F.K. and K.v.W.