Brassicaceae-specific Gretchen Hagen 3 acyl acid amido synthetases conjugate amino acids to chorismate, a precursor of aromatic amino acids and salicylic acid

To modulate responses to developmental or environmental cues, plants use Gretchen Hagen 3 (GH3) acyl acid amido synthetases to conjugate an amino acid to a plant hormone, a reaction that regulates free hormone concentration and downstream responses. The model plant Arabidopsis thaliana has 19 GH3 proteins, of which 8 have confirmed biochemical functions. One Brassicaceae-specific clade of GH3 proteins was predicted to use benzoate as a substrate and includes AtGH3.7 and AtGH3.12/PBS3. Previously identified as a 4-hydroxybenzoic acid-glutamate synthetase, AtGH3.12/PBS3 influences pathogen defense responses through salicylic acid. Recent work has shown that AtGH3.12/PBS3 uses isochorismate as a substrate, forming an isochorismate-glutamate conjugate that converts into salicylic acid. Here, we show that AtGH3.7 and AtGH3.12/PBS3 can also conjugate chorismate to cysteine and glutamate, which act as precursors to aromatic amino acids and salicylic acid, respectively. The X-ray crystal structure of AtGH3.12/PBS3 in complex with AMP and chorismate at 1.94 Å resolution, along with site-directed mutagenesis, revealed how the active site potentially accommodates this substrate. Examination of Arabidopsis knockout lines indicated that the gh3.7 mutants do not alter growth and showed no increased susceptibility to the pathogen Pseudomonas syringae, unlike gh3.12 mutants, which were more susceptible than WT plants, as was the gh3.7/gh3.12 double mutant. The findings of our study suggest that GH3 proteins can use metabolic precursors of aromatic amino acids as substrates.

Plants have adapted and evolved a number of chemicals to respond to the world around them (1). By modifying common growth and defense hormones, plants are able to activate or inactivate hormone molecules rapidly in response to environmental cues (2,3). One set of phytohormone modifying enzymes are the Gretchen Hagen 3 (GH3) 2 family of acyl acid amido synthetases. These enzymes conjugate amino acids to acyl acids, notably the plant hormones indole-3-acetic acid (IAA; the major auxin hormone) and jasmonic acid (2,4). In some cases, the amino acid-conjugated hormone is a storage or degradation form of the molecule, but in other cases, the conjugate is the active form of the molecule that binds the hormone receptor (5)(6)(7).
Although GH3 proteins are found in all plants, the roles of many of these proteins remains to be investigated. In Arabidopsis thaliana (thale cress), there are 19 GH3 proteins, and to date, only 8 of these have been biochemically characterized: AtGH3.1, AtGH3.2, AtGH3.3, AtGH3.5/WES1, AtGH3.11/ JAR1, AtGH3.12/PBS3, AtGH3.15, and AtGH3.17/VAS2 (8 -13). Using the X-ray crystal structures of AtGH3.11 and AtGH3.12, the acyl acid substrate-binding pocket was compared with the active site residues across all GH3 proteins (10). With this information, the function of a given GH3 protein can be predicted based on the known function of other clade members.
Another clade of GH3 proteins with a characterized homologue is Brassicaceae-specific and has two members in Arabidopsis: AtGH3.7 and AtGH3.12/PBS3, which share 75% amino acid sequence identity (10). AtGH3.12/PBS3 was originally identified in a mutant screen for increased susceptibility to both avirulent and virulent Pseudomonas syringae strains and was named pbs3-1 (avrPphB susceptible 3) (14). Later work identified a biochemical activity with the nonphysiological substrate 4-hydroxybenzoic acid (4-HBA) (15). In Arabidopsis, knockout of pbs3 resulted in reduced levels of salicylic acid (SA), a plant pathogen defense signaling molecule, and showed decreased accumulation of pathogen responsive gene transcripts and SAglucosides after infection (15)(16)(17). Although SA biosynthesis is predicted to occur in the chloroplast (18 -20), SA-glucosides can be actively transported to the vacuole, where these molecules presumably function as a storage form of SA (21)(22)(23). Interestingly, treatment of pbs3-1 plants with exogenous SA leads to normal induction of pathogen responsive transcripts and SA-glucosides after infection and the SA-treated pbs3-1 plants are as resistant to P. syringae infection as WT plants (17). From these studies, it was proposed that AtGH3.12/PBS3 functions upstream of SA, potentially in its biosynthesis. Recent work identified a plastidial membrane transporter and AtGH3.12/PBS3 as required proteins in SA production (24). In that study (24), AtGH3.12/PBS3 forms an isochorismate-glutamate conjugate, which degrades into SA. This work also identified chorismate-glutamate as a secondary product of the enzyme (24). To date, AtGH3.7, the other member of this clade, has escaped in vivo and biochemical characterization. SA can be synthesized by two routes in plants, both of which use the aromatic amino acid branch point metabolite, chorismate, as a precursor (25). In one pathway, the chorismate-derived aromatic amino acid phenylalanine is converted into SA through benzoate intermediates or coumaric acid (26). The second pathway converts chorismate into SA using isochorismate synthase and AtGH3.12/PBS3 (18,27). There are two isochorismate synthase genes in Arabidopsis, but the ics1/ics2 double mutant line still produces SA, suggesting that this pathway is not the only source of SA in this model plant (20). AtGH3.12/ PBS3 forms isochorismate-glutamate, which then degrades into SA (24); however, homologs of AtGH3.12/PBS3 and AtGH3.7 are only found in the Brassicaceae (10), which raises a question of how widespread this new biosynthetic pathway actually is in the plant kingdom.
Here, we re-examined the possible biochemical function of the Brassicaceae-specific clade proteins AtGH3.12/PBS3 and AtGH3.7 and, in the absence of commercially available isochorismate, identified each protein as chorismate-conjugating enzymes. Using a combination of steady-state kinetics, X-ray crystallography, and site-directed mutagenesis, the molecular basis of chorismate binding in AtGH3.12/PBS3 was revealed, which also provides insight on the isochorismate-conjugating activity of this protein. The in vivo role of AtGH3.7 was analyzed through P. syringae pv. tomato DC3000 infection assays, and a double mutant gh3.7/gh3.12 Arabidopsis line was generated to determine the combined role of this unique class of GH3 proteins in the Brassicaceae family. Because of the relaxed substrate specificity that is typical of GH3 proteins, it is not surprising that AtGH3.12/PBS3 can recognize both chorismate and isochorismate as substrates (11)(12)(13)24).
Screening of AtGH3.12/PBS3 confirmed the previously identified activity for the conjugation of 4-HBA, as well as benzoic acid, prephenate, and p-coumaric acid, with glutamate ( Fig. 1a) (10,15,16); however, activity for the enzyme as a chorismateconjugating enzyme was identified with a specific activity 8-fold higher than that determined for 4-HBA (Fig. 1a). No activity was observed with the other acyl acid/amino acid combinations. Formation of the chorismate-glutamate conjugate AtGH3.7 and AtGH3.12 conjugate amino acids to chorismate formed by AtGH3.12/PBS3 was confirmed by LC-MS/MS. Incubation of the enzyme with chorismate, ATP, and glutamate led to formation of a peak with a m/z of 353.9, which matches the expected deprotonated chorismate-glutamate conjugate (Fig. 2). As a negative control, assays in the absence of protein or any one substrate did not lead to observation of the product peak. AtGH3.12/PBS3 displayed the highest specific activity using glutamate as the amino acid substrate (2,400 nmol min Ϫ1 mg protein Ϫ1 ; Fig. 1b), which is consistent with earlier reports of this proteins second half-reaction specificity (10,15,16). This activity was 4.5-fold higher than the next highest specific activity, which was with leucine (520 nmol min Ϫ1 mg Ϫ1 ).
Steady-state kinetic analysis of AtGH3.12/PBS3 revealed similar K m values for 4-HBA (270 M) and chorismate (170 M), but the turnover rate for chorismate was nearly 10-fold higher than that observed for 4-HBA (Table 1). This leads to chorismate having a roughly 15-fold higher catalytic efficiency (k cat / K m ) than 4-HBA (4,220 and 278 M Ϫ1 s Ϫ1 , respectively).
The activity of AtGH3.12/PBS3 with chorismate as substrate led us to examine AtGH3.7, the other enzyme in the same clade that shares a conserved acyl acid-binding site with AtGH3.12/ PBS3 (10). Using the same panel of acyl and amino acid substrates, the chorismate-conjugating activity of AtGH3.7 was identified with the highest activity observed using cysteine as the amino acid substrate (Fig. 3a). The specific activity of AtGH3.7 with chorismate was ϳ50-fold higher than that of either p-coumaric and cinnamic acid substrates. No activity with other acyl acids in the panel was observed. Screening of the chorismate conjugation with each amino acid identified the highest activity with cysteine as a substrate, along with serine and glutamine as other possible substrates (Fig. 3b). Steadystate kinetic analysis of AtGH3.7 indicates that the chorismate and cysteine substrate pair is the preferred combination com-pared with either serine or glutamine as the amino acid substrate (Table 2). Compared with AtGH3.12/PBS3, AtGH3.7 shows a comparable K m value and a 2-fold lower turnover rate, which results in protein that is 3-fold less efficient than AtGH3.12/PBS3 for conjugating chorismate (Table 2). Of the 20 amino acids, cysteine, serine, and glutamine had the highest activities, but further kinetic analysis showed that the preferred amino acid substrate is cysteine ( Table 2). With cysteine, saturation kinetics was observed, but kinetic analysis with both serine and glutamine displayed linear behavior up to 10 mM concentrations. This indicates that the K m values for these amino acids are substantially higher and not in a physiologically relevant range.
The identification of chorismate-conjugating activity of these two GH3 proteins led us to re-examine why this activity was missed in previous studies, which used a phosphate-release assay to monitor the first half-reaction (i.e. adenylation) activity of AtGH3.12/PBS3 (15). To understand the failure of the adenylation assay in detecting the chorismate activity, we repeated the assay in the absence and presence of glutamate (Fig. 4). Using this assay in the absence of an amino acid, AtGH3.12/PBS shows a higher specific activity with 4-HBA than with chorismate (120 Ϯ 4 and 24 Ϯ 1.2 nmole min Ϫ1  AtGH3.7 and AtGH3.12 conjugate amino acids to chorismate mg Ϫ1 , respectively), as previously observed (15). Addition of glutamate to the assay mix leads to a modest increase in the adenylation rate with 4-HBA (140 Ϯ 1.5 nmol min Ϫ1 mg Ϫ1 ). In contrast, the observed activity with chorismate increases 35-fold (620 Ϯ 14 nmol min Ϫ1 mg Ϫ1 ) (Fig. 4).

SA inhibition of chorismate conjugate formation in AtGH3.12/ PBS3 and AtGH3.7
Previously published AtGH3.12/PBS3 kinetic data showed that SA was a competitive inhibitor in the reaction with 4-HBA (15). With the new data showing that 4-HBA is not the preferred substrate of AtGH3.12/PBS3, we wanted to confirm that the reaction was still inhibited by SA when chorismate was the acyl acid substrate. SA inhibited AtGH3.12/PBS3 with an IC 50 of 7 Ϯ 1 M, which is similar to what was found previously with 4-HBA as a substrate (IC 50 ϭ 15 M; Ref. 15). Because AtGH3.7 also uses chorismate as a substrate, we hypothesized that the chorismate-cysteine conjugate formation by AtGH3.7 would also be inhibited by SA. In contrast to AtGH3.12/PBS3, SA less effectively inhibited AtGH3.7 with an IC 50 value of 2.4 Ϯ 0.2 mM, which is 340-fold higher than the IC 50 value of SA for AtGH3.12/PBS3 with chorismate as an acyl acid substrate.

Structural analysis of AtGH3.12/PBS3 in complex with chorismate
To date, three-dimensional X-ray crystal structures of AtGH3.12/PBS3 have been solved in complex with AMP, AMP, and SA, and a nonhydrolyzable ATP analog and SA (10,28). Interestingly, the active site pocket space is bigger than SA, which suggested a possible larger substrate for the enzyme (10,28). To gain an understanding of how chorismate potentially fits in the acyl acid-binding pocket of AtGH3.12/PBS3, the 1.94 Å resolution X-ray crystal structure of the protein in complex with chorismate and AMP was solved by molecular replacement ( Table 3).
The overall structure of the AtGH3.12/PBS3⅐chorismate⅐ AMP complex was similar to the previously solved structures with the C-terminal flexible domain adopting the "closed" active site conformation (Fig. 5a). Within the active site, clear

AtGH3.7 and AtGH3.12 conjugate amino acids to chorismate
electron density for AMP was observed, as was weaker, but interpretable, electron density for chorismate (Fig. 5b). The position of chorismate in the structure was in the same area of the active site as SA in the earlier structures (10,28). Compared with other phytohormones or hormone analogs, including jasmonyl-isoleucine, a nonhydrolyzable IAA-AMP analog, SA, and IAA, that have been co-crystallized with various GH3 proteins (10 -13, 28, 30), chorismate contains multiple polar and charged moieties.
In the AtGH3.12/PBS3 active site, chorismate is oriented with the ring carboxylate pointed toward the phosphate group of the bound AMP molecule (Fig. 5c). This carboxylate forms charge-charge interactions with the side chain of Arg-123. An additional hydrogen bonding interaction occurs via a watermediated contact with Tyr-120. Ile-217 is positioned below the chorismate ring and fits into the ring pucker. The distal chorismate carboxylate forms charge-charge interactions with Arg-213 and a hydrogen bond with Tyr-181. Although it does not directly interact with chorismate, Tyr-178 forms charge-charge interactions with the side chain of Arg-123 to position it to interact with the ring carboxylate of chorismate. This set of active site residues is also invariant in AtGH3.7, which is consistent with the use of chorismate as an acyl acid substrate of both enzymes.
This structure provides a view of how chorismate can bind in the acyl acid site of AtGH3.12/PBS3; however, the location of the ligand likely does not represent where it binds during catalysis, as the ring carboxylate of chorismate is ϳ10 Å away from the AMP. As noted for other GH3 protein structures (11)(12)(13), charge repulsion between the carboxylate of an acyl acid substrate and the nucleotide phosphate can alter positioning of ligands in the site. During the reaction sequence, the presence of Mg 2ϩ is an essential counterion that balances active site charges and allows the first half-reaction to occur (10, 30, 31).

Site-directed mutagenesis of AtGH3.12/PBS3 chorismateconjugation activity
To gain a better understanding of the contributions of residues in the acyl acid-binding site of AtGH3.12/PBS3, several previously generated site-directed mutants (Y112F, Y120F, R123K, T161S, I217A, and F218Y) and three new point mutants (Y178F, Y181F, and R213K) were analyzed for their effect on steady-state kinetic parameters with chorismate ( Table 4). The Y112F, T161S, and Y181F substitutions had modest Ͻ2.5-fold changes on catalytic efficiency (k cat /K m ) compared with WT. As observed in the X-ray structure (Fig. 5c), these three residues are not in direct contact with the chorismate ligand. The Y120F mutant, which removes a hydrogen bond donor, displayed a 13.5-fold increase in K m with a corresponding 22-fold reduction in catalytic efficiency. Mutation of Arg-123 to a lysine had a ϳ5-fold effect on k cat /K m . Mutations of the two residues at the   With chorismate bound in the structure, several residues that were not previously analyzed emerged as potential mutagenesis targets, Tyr-178, Tyr-181, and Arg-213. At the distal side of the bound chorismate, each tyrosine was mutated to phenylalanine. The Y181F mutant exhibited a less than 2-fold effect on catalytic efficiency (Table 4); however, the Y178F substitution, which removes a positioning interaction with Arg-213, had no detectable activity (Table 4). Likewise, the R213K variant was inactive. These results suggest that the combination of Tyr-178 and Arg-213 are critical for the chorismate-conjugation activity of AtGH3.12/PBS3.

Arabidopsis gh3.7 mutants show WT pathogen response when infected with PstDC3000
Because AtGH3.7 and AtGH3.12/PBS3 can both use chorismate as a substrate, we hypothesized that the two proteins could have redundant and/or overlapping functions in planta. Previous physiological investigations of gh3.12/pbs3 knockout mutants in Arabidopsis showed that the gene was involved in pathogen defense, and the knockout showed enhanced disease susceptibility to virulent and avirulent strains of P. syringae pv. tomato DC3000 (14 -17). Moreover, gh3.12/pbs3 mutants are compromised in pathogen-induced SA-glucoside accumulation, and this accumulation could be restored by exogenous application of SA, which also enhanced resistant to virulent P. syringae DC3000 (17).
Here we infected gh3.7-1 mutants by leaf syringe infiltration with WT P. syringae DC3000 and found no statistically significant difference between WT and mutant plants (Fig. 6). The Arabidopsis isochorismate synthase mutant, sid2-2, was included as a positive control because it shows enhanced disease symptoms and is more susceptible to pathogen infection (18). In contrast to gh3.7-1 mutants, gh3.12/pbs3 mutant plants were significantly more susceptible to P. syringae DC3000 on day 4 (p Ͻ 0.0001), as was sid2-2 (p ϭ 0.012). This confirmed that AtGH3.7 has a different role than AtGH3.12/PBS3 in defending aerial tissue against pathogens. This result suggests that the product of AtGH3.12/PBS3 may be important for pathogen resistance, whereas the product of GH3.7 appears to not be involved. There was no visible growth phenotype in the gh3.7-1 plants compared with WT Arabidopsis plants.

Arabidopsis gh3.7 gh3.12 double mutants are more susceptible to infection than WT
Although gh3.7-1 had neither a disease resistance nor a pathogen susceptibility phenotype, a double mutant A. thaliana line was generated by crossing gh3.7-1 and gh3.12/pbs3 mutant lines. Overall, there were no noticeable growth phenotypes between the gh3.7-1 gh3.12/pbs3 double mutant and WT plants. Also, to confirm that gh3.7-1 knockout lines showed WT pathogen responses, a second allele was included in these experiments, gh3.7-2. Plants were infected by syringe infiltration using WT P. syringae DC3000, and sid2-2 was included as a positive control. In these experiments, gh3.7-2 showed some resistance to P. syringae DC3000 compared with WT, but this difference was not statistically significant (p ϭ 0.33) (Fig. 7). Therefore, AtGH3.7 and AtGH3.12/PBS3 do not have evidently redundant roles in pathogen responses in infected leaf tissue. On day 4 after infection, the gh3.7-1 gh3.12/pbs3 double mutant was more susceptible to infection compared with WT (p Ͻ 0.0001) and was more susceptible compared with gh3.12/ pbs3 (p Ͻ 0.0001) and sid2-2 (p Ͻ 0.0001) knockout lines (Fig.  7). This suggests that whereas AtGH3.7 alone does not function in defense in leaf tissues, the combination of both AtGH3.7 and AtGH3.12/PBS3 aids plants in pathogen response. Figure 6. Arabidopsis gh3.7-1 knockout lines do not have a pathogen infection phenotype. Leaves from infected plants were sampled 0, 2, and 4 days post-inoculation for quantifying P. syrinagae pv. tomato DC3000 growth in the tissue (cfu/mg leaf; log-scale on the y axis). a, this graph depicts three biological replicates with 4 -6 leaves per replicate per genotype. Statistical analyses were conducted using a two-way ANOVA with statistical significance between control (WT; Col-0) and samples are indicated by asterisks (* ϭ p Ͻ 0.05; **** ϭ p Ͻ 0.0001). b, zoomed-in view of day 4 data from panel a. All means were compared with all other means using a two-way ANOVA (**** ϭ p Ͻ 0.0001).

Discussion
The GH3 family of acyl acid amido synthetases found in all plants is critical for modulating the concentration of acyl acid hormones in plants by conjugating them to amino acids (2-3). AtGH3.12/PBS3 was previously shown to conjugate glutamate to 4-HBA, which is not a plant hormone (15). When AtGH3.12/ PBS3 was first discovered, it was found to have an SA-deficiency phenotype that could be rescued by exogenous SA, which suggested that the enzyme works upstream of SA signaling (16,17). Recent work identified AtGH3.12/PBS3 as essential for isochorismate-derived biosynthesis of SA (24).
With this knowledge, we screened aromatic and aromatic precursor metabolites as substrates for AtGH3.12/PBS3 and for the previously uncharacterized AtGH3.7, a homologue of AtGH3.12/PBS3. This approach identified chorismate-conjugating activity for each protein (Figs. 1-3, Tables 1 and 2). Rescreening of AtGH3.12/PBS3 revealed that it had higher activity with chorismate than with 4-HBA and can conjugate chorismate to glutamate (Figs. 1 and 2, Table 1). Likewise, biochemical analysis of AtGH3.7, which is from the same clade of GH3 proteins in Arabidopsis (10), identified a similar chorismate-conjugating activity with cysteine as the amino acid substrate (Fig. 3). Interestingly, despite the 75% amino acid sequence identity and use of the same acyl acid substrate, SA inhibited AtGH3.12/PBS3 with a 340-fold lower IC 50 value than that of AtGH3.7. Although recent analysis of the isochorismate-conjugation activity of AtGH3.12/PBS3 suggests this as the major function of the enzyme (24), the use of commercially available chorismate (in contrast to the unstable isochorismate molecule) for structure-function studies remains a valuable tool. Moreover, the potential role of AtGH3.7 as an isochorismate-conjugating enzyme remains to be examined.
The GH3 acyl acid amido synthetases follow a two-step pingpong mechanism where the acyl acid substrate binds and is adenylated and pyrophosphate is released, followed by a transfer step where the amino acid is added to the acyl acid to form the conjugate and AMP is released (10,30). This mechanism helps explain why 4-HBA was identified as a substrate of AtGH3.12/PBS3. Early assays only measured pyrophosphate release in the first half-reaction (15). The assay used to identify chorismate as a substrate measured AMP release after the full reaction sequence. As noted here (Fig. 4), the difference between pyrophosphate release from the 4-HBA and chorismate reactions in the absence and presence of an amino acid substrate (i.e. glutamate) with AtGH3.12/PBS3 suggests that the substrates are held in the active site until the amino acid transferase step. This is a similar mechanism to ensure proper product formation as observed in other adenylating enzymes (32,33).
Having chorismate bound in a structure provided further evidence that chorismate is preferred over 4-HBA as a substrate of AtGH3.12/PBS3 (Fig. 5). Compared with the X-ray crystal structures of other GH3 acyl acid amido synthetase that use either auxins or jasmonic acid, key differences among the structures reveal how the AtGH3.12/PBS3 active site recognizes chorismate (10 -13, 28, 29). The active site of AtGH3.11/JAR1 has only one polar residue, a histidine, that interacts with the jasmonic acid ketone through a water molecule (10). The auxinconjugating GH3 proteins have polar residues situated to interact with their substrates (11-13, 28, 29). In comparison, the AtGH3.12/PBS3 active site contains two arginines and two tyrosines, with one of each positioned near each carboxylate of chorismate (Fig. 5). As indicated by kinetic analysis of site-directed mutants of AtGH3.12/PBS3 (Table 4), the charge-charge Figure 7. Arabidopsis gh3.7-1 gh3.12/pbs3 double mutant lines are more susceptible to pathogen infection than WT or sid2-2. Leaves from infected plants were sampled 0, 2, and 4 days post-inoculation for quantifying P. syrinagae pv. tomato DC3000 growth in the tissue (cfu/mg leaf; log-scale on the y axis). a, this graph depicts four biological replicates with 4 -6 leaves per replicate per genotype. Statistical analyses were conducted using a two-way ANOVA with statistical significance between control (WT; Col-0) and samples indicated by asterisks (**** ϭ p Ͻ 0.0001). b, zoomed-in view of data from day 4 from panel a. All means were compared with all other means using a two-way ANOVA (**** ϭ p Ͻ 0.0001).

AtGH3.7 and AtGH3.12 conjugate amino acids to chorismate
interactions between the arginines and the carboxylates appear critical for positioning chorismate or isochorismate in the active site for the ensuing adenylation and amino acid transfer reactions.
Considering that amino acid sequence identity of AtGH3.12/ PBS3 and AtGH3.7, we wondered if they would have physiologically redundant roles in planta. Because AtGH3.12/PBS3 has an SA-deficiency response and both proteins can use chorismate as an acyl acid substrate, we hypothesized that the enzyme products may have similar roles in pathogen defense. To determine the role of AtGH3.7 in pathogen defense, knockout lines were infected with P. syringae. Next, we examined if the roles of AtGH3.7 and AtGH3.12/PBS3 were additive and generated an Arabidopsis double mutant. The gh3.7-1 gh3.12 double mutants were even more susceptible to pathogen infection than gh3.12 mutants alone than the positive control, the isochorismate synthase sid2-2 knockout line, which is deficient in SA biosynthesis (Fig. 7). This suggests that having both AtGH3.7 and AtGH3.12/PBS3 is important in Brassicaceae species for defense against pathogenic microorganisms.
Knowing the promiscuous biochemical activity of AtGH3.12/ PBS3 presents several possibilities for the role of the chorismate-related metabolites in planta. In the absence of this protein, plants are more susceptible to pathogen attack and were found to have less SA-glucoside accumulation after infection (15)(16)(17). These data suggests that the enzyme is upstream of SA biosynthesis, as was confirmed for the isochorismate-glutamate conjugates that spontaneously degrade into SA (24). One possibility is that chorismate-glutamate is siphoning chorismate away from other biosynthetic pathways, including that of the auxins IAA and phenylacetic acid as a storage form of the molecule; however, a hydrolase that would release free chorismate remains to be identified. Both AtGH3.12/PBS3 and AtGH3.7 are cytoplasmic proteins (as each has no clear chloroplast localization sequence), whereas SA biosynthesis occurs in the plastid of plants (18). How chorismate is transferred from the chloroplast is an open question as a chorismate transporter has yet to be identified. Evidence now suggests that EDR5 (Enhanced Disease Susceptibility 5) transports isochorismate from the plastid to the cytosol (24), which raises the possibility of this protein as a transporter of chorismate, as well.
Interestingly, AtGH3.12/PBS3 and AtGH3.7 are not the only cytosolic enzymes that can use chorismate as a substrate; chorismate mutase 2 in Arabidopsis is found in the cytosol and converts chorismate into prephenate, the precursor to tyrosine and phenylalanine via the arogenate pathway (34 -38). The chorismate conjugates could be diverting chorismate from the cytosolic chorismate mutase that would convert it into prephenate. There may be separate, nonredundant physiological roles for the two conjugates in planta. Future studies are needed to decipher the function of these metabolites in pathogenesis in Brassicaceae.
In conclusion, we have identified chorismate, a hormone biosynthetic intermediate, as a substrate for the Brassicaceaespecific clade of GH3 proteins that contains AtGH3.7 and AtGH3.12/PBS3. Although AtGH3.12/PBS3 was previously shown to be involved in pathogen defense responses (15)(16)(17)24), the physiological role and potential isochorismate-conju-gating activity of AtGH3.7 remains to be understood as it does not exhibit a statistically significant effect during pathogen attack, suggesting a different role from AtGH3.12/PBS3. Taken together, these data suggest that regulation and modification of SA biosynthetic intermediates is a critical component of pathogen responses in the Brassicaceae.

Enzyme assays
Activity assays were performed using an enzyme-coupled system containing 10 units of myokinase, pyruvate kinase, and lactate dehydrogenase, 1 mM phosphoenolpyruvate, and 1 mM NADH to couple AMP generation to conversion of NADH to NAD ϩ , which leads to a decrease in absorbance at 340 nm (30). Reaction mixtures contain 2 mM ATP, 2 mM MgCl 2 , 0 -2 mM acyl acid, 0 -10 mM amino acid, and purified AtGH3.12/PBS3 (1 g) or AtGH3.7 (10 g). For assays with chorismate, the substrate was isolated and characterized as previously described (38). For determination of kinetic parameters, reactions were performed with either 4-HBA or chorismate varied from 0 to 1 mM. All data were fit to the Michaelis-Menten equation using GraphPad Prism. Inhibition of AtGH3.12/PBS3 and AtGH3.7 was analyzed using 10 g of protein with 1 mM chorismate and 10 mM amino acid (glutamate or cysteine for AtGH3.12/PBS3 or AtGH3.7, respectively) in the presence of varied (0 -10 mM) SA. IC 50 values were determined by fitting data to y ϭ max/(1 ϩ (I f /IC 50 )), where max is the maximum observed rate difference and I f is the inhibitor concentration, in GraphPad Prism.

Pyrophosphate release assays
Adenylation assays were performed using the pyrophosphate reagent assay kit (Sigma) to couple pyrophosphate release to NADH depletion with 2 mM ATP, 2 mM MgCl 2 , and purified AtGH3.12/PBS3 (30 g) with either 4-HBA or chorismate (1 mM) in the absence or presence of glutamate (10 mM) (15). Assays were performed at 25°C and were initiated by the addition of enzyme.

Protein crystallography
Crystals of AtGH3.12/PBS2 in complex with AMP and chorismate were grown by the vapor diffusion method in hanging drops of a 1:1 mixture of protein (15 mg ml Ϫ1 ) and crystallization buffer (15% PEG-3350 (w/v), 0.25 M ammonium acetate, 100 mM sodium acetate (pH 4.4), 5 mM tris(2-carboxyethyl-)phosphine, 5 mM MgCl 2 , 10 mM chorismate, and 4 mM AMP). Crystals were flash frozen with mother liquor supplemented with 30% PEG-3350 as a cryoprotectant. Diffraction data were collected at ESRF ID23-2 and were indexed and integrated with XDS and scaled with XSCALE (40,41). Molecular replacement was performed with PHASER (42) using the previously solved AtGH3.12 structure (PDB 4EQL)(10) for the search model. Structures were refined with BUSTER (43) and manual model building was done in COOT (44). Data collection and refinement statistics are shown in Table 3. Coordinates and structure factors for AtGH3.12/PBS3 in complex with AMP and chorismate were deposited in the PBD (PDB 6OMS).

Plant and pathogen materials and growth conditions
All A. thaliana transgenic lines and mutants used in this study were in the Col-0 background. The sid2-2 mutant (18) was obtained from Barbara Kunkel. Plants were grown on soil in a growth chamber with a short-day photoperiod (8 h light/16 h dark) at 21°C with a relative humidity of 75% and a light intensity of ϳ130 Einsteins s Ϫ1 m Ϫ1 . P. syringae strain DC3000 (45) were grown on Nutrient Yeast Glycerol medium at 28°C (46). The gh3.7 gh3.12 double mutant was generated using gh3.7 SALK_106726 and gh3.12 SALK_018225C lines as parents in the crosses (47). PCR verified that each SALK line and the double-mutant were knockout lines. To confirm the phenotype of gh3.7, the SAIL_755_E09 line was also used in plant infection experiments.

P. syringae inoculation and quantification of bacterial growth
Leaves of 5-week-old A. thaliana plants were infected by syringe infiltrations using a solution containing 10 5 cells ml Ϫ1 in 10 mM MgCl 2 and a 1-ml needleless syringe. To quantify bacterial growth in the plant, whole leaves were sampled at days 0, 2, and 4 post-inoculation, weighed to determine leaf mass (mg), ground in 10 mM MgCl 2 , and then plated in serial dilutions on nutrient yeast glycerol medium supplemented with rifampicin. Between four and six leaves were sampled per treatment (48).