Structural Evolution of Differential Amino Acid Effector Regulation in Plant Chorismate Mutases*

Background: Chorismate mutase is essential for aromatic amino acid biosynthesis. Results: Structural and biochemical studies of three chorismate mutases from Arabidopsis reveal distinct sets of effector molecules. Conclusion: Key residues in the effector site modulate the regulatory effects of ligands. Significance: Evolution of effector control may lead to specialized regulation of this enzyme in plants. Chorismate mutase converts chorismate into prephenate for aromatic amino acid biosynthesis. To understand the molecular basis of allosteric regulation in the plant chorismate mutases, we analyzed the three Arabidopsis thaliana chorismate mutase isoforms (AtCM1–3) and determined the x-ray crystal structures of AtCM1 in complex with phenylalanine and tyrosine. Functional analyses show a wider range of effector control in the Arabidopsis chorismate mutases than previously reported. AtCM1 is activated by tryptophan with phenylalanine and tyrosine acting as negative effectors; however, tryptophan, cysteine, and histidine activate AtCM3. AtCM2 is a nonallosteric form. The crystal structure of AtCM1 in complex with tyrosine and phenylalanine identifies differences in the effector sites of the allosterically regulated yeast enzyme and the other two Arabidopsis isoforms. Site-directed mutagenesis of residues in the effector site reveals key features leading to differential effector regulation in these enzymes. In AtCM1, mutations of Gly-213 abolish allosteric regulation, as observed in AtCM2. A second effector site position, Gly-149 in AtCM1 and Asp-132 in AtCM3, controls amino acid effector specificity in AtCM1 and AtCM3. Comparisons of chorismate mutases from multiple plants suggest that subtle differences in the effector site are conserved in different lineages and may lead to specialized regulation of this branch point enzyme.

clic Claisen rearrangement of chorismate to prephenate as the committed step in this pathway ( Fig. 1) (6).
Although chorismate mutase activity is found in bacteria, fungi, and plants, the proteins that catalyze this reaction vary in both sequence and overall structure. The chorismate mutases from eukaryotes and most bacteria, also known as the AroQ, are typically dimeric ␣-helical proteins with each monomer consisting of ϳ250 amino acids (7). In some bacteria, such as Bacillus subtilis, the smaller dimeric AroH chorismate mutase consists of ϳ110 amino acid monomers (6,8). The size difference between the two types of enzyme appears to be due to the presence of a regulatory region in the larger AroQ enzymes (9).
Within this regulatory region, an effector site binds aromatic amino acids to modulate enzymatic activity. For example, the effector site of Saccharomyces cerevisiae chorismate mutase (ScCM) 3 can bind either tryptophan or tyrosine (10). Tryptophan binding activates ScCM, and tyrosine leads to attenuation of prephenate synthesis. Thus, downstream metabolites provide reciprocal regulation of flux into either pathway leading from chorismate. This allows ScCM to divert chorismate flow from tryptophan synthesis to phenylalanine/tyrosine synthesis in high tryptophan conditions or to reduce phenylalanine/tyrosine synthesis through inhibition by downstream metabolites. Structural and biochemical studies of the yeast enzyme show that it contains an effector site in each monomer at the dimer interface and that binding of tyrosine and tryptophan alter the conformation of the enzyme between less active (T-state) and more active (R-state) forms, respectively (9 -13).
In Arabidopsis thaliana (thale cress), three different chorismate mutase isoforms have been reported as follows: AtCM1, AtCM2, and AtCM3 (14 -16). AtCM1 and AtCM3 both contain putative N-terminal plastid localization peptides, and AtCM2 is cytosolic (Fig. 2). In various plant species, chorismate mutase activity is found in both the plastid and cytosol (17)(18)(19)(20)(21)(22). In petunia, a chorismate mutase homolog may be involved in the synthesis of phenylalanine-derived volatile phenylpropanoids and benzenoids (23). Biochemical studies using heterologous protein extracts of AtCM1 and AtCM3 suggest that both * This work was supported, in whole or in part, by National Institutes of Health MARC uSTAR undergraduate fellowship (to A. X.). This work was also supported by National Science Foundation Grant MCB-1157771 (to J. M. J.). The atomic coordinates and structure factors ( enzymes are regulated by aromatic amino acids (14 -16); however, an analysis of the purified proteins was not performed. Although AtCM2 contains the putative regulatory effector binding domain, phenylalanine, tyrosine, and tryptophan do not affect its activity (15).
To understand the molecular basis of allosteric regulation in the plant chorismate mutases, here we examine the steady-state kinetic properties of the three Arabidopsis chorismate mutase isoforms and determine the x-ray crystal structures of AtCM1 in complex with phenylalanine and tyrosine. These analyses reveal a wider range of effector control in chorismate mutases of Arabidopsis than previously described for other plants and microbes. Specifically, AtCM1 is activated by tryptophan with phenylalanine and tyrosine acting as negative effectors; however, AtCM3 is activated by tryptophan, cysteine, and histidine. Site-directed mutagenesis of residues in the effector site that differ between AtCM1, AtCM2, and AtCM3 reveals key features leading to differential effector regulation in these enzymes. Moreover, sequence analysis of chorismate mutases from multiple plant species suggests that subtle differences in the effector site are conserved in different lineages.

EXPERIMENTAL PROCEDURES
Materials-All reagents were purchased from Sigma. Clones of the AtCM1 (U83587), AtCM2 (U18739), and AtCM3 (U60550) were obtained from the Arabidopsis Biological Resource Center.
Protein Expression and Purification-Expression constructs were transformed into Escherichia coli Rosetta II (DE3) cells (EMD Millipore). Cells were cultured in terrific broth until A 600 nm ϳ0.6 -0.8 was obtained. Induction of protein expression used a final concentration of 1 mM isopropyl ␤-D-1-thiogalactopyranoside overnight at 18°C. Cells were pelleted by centrifugation and resuspended in 50 mM Tris, pH 8.0, 500 mM NaCl, 20 mM imidazole, 10% glycerol, and 1% Tween. Following sonication, cell debris was removed by centrifugation, and the resulting lysate was passed over a Ni 2ϩ -nitriloacetic acid (Qiagen) column equilibrated in the lysis buffer. The column was then washed with 50 mM Tris, pH 8.0, 500 mM NaCl, 20 mM imidazole, and 10% glycerol. Bound His-tagged protein was eluted with 50 mM Tris, pH 8.0, 500 mM NaCl, 250 mM imidazole, and 10% glycerol. For protein crystallization, incubation with thrombin (1:2000 total protein) during overnight dialysis at 4°C against wash buffer removed the His tag. Dialyzed protein was reloaded on a mixed benzamidine-Sepharose/Ni 2ϩnitrilotriacetic acid column. The flow-through was loaded onto a Superdex-200 26/60 HiLoad FPLC size-exclusion column equilibrated with 25 mM Hepes, pH 7.5, and 100 mM NaCl. Protein concentration was determined by the Bradford method (Protein Assay, Bio-Rad) with bovine serum albumin as standard. Site-directed mutants of AtCM1 and AtCM3 were expressed and purified using the same methods as wild-type protein.
Protein Crystallography-Purified AtCM1 was concentrated to 9 mg ml Ϫ1 and crystallized using the hanging drop vapordiffusion method with a 2-l drop (1:1 concentrated protein and crystallization buffer). Diffraction quality crystals were obtained at 4°C with a crystallization buffer of 30% PEG-400, 0.1 M Hepes, pH 7.5, 0.2 M MgCl 2 , and 1 mM of either phenylalanine or tyrosine. Crystals were flash-frozen in liquid nitrogen with mother liquor supplemented with 25% glycerol as a cryoprotectant. Diffraction data (100 K) was collected at the Argonne National Laboratory Advanced Photon Source 19-ID beamline. The data were indexed, scaled, and integrated with HKL3000 (24). Molecular replacement implemented in PHASER (25) using the yeast chorismate mutase (Protein Data Bank code 4CSM) as a search model was used to determine the structures of each AtCM1 complex. Iterative rounds of manual model building and refinement, which included translationlibration-screen models, used COOT (26) and PHENIX (27). Data collection and refinement statistics are summarized in Table 1. The final model of the AtCM1⅐phenylalanine complex included residues Arg-79 -Val-290 and Val-307-Asp-340, the Kinetic Analysis of Wild-type and Mutant Proteins-Steadystate kinetic assays that monitored the conversion of chorismate to prephenate were performed as described previously (28). Briefly, 100 ng of recombinant protein was added to a 500-l reaction mixture of 50 mM Tris, pH 8.0, and varied concentrations of chorismate (0 -3 mM). The disappearance of chorismate leads to an absorbance decrease at A 274 nm (⑀ ϭ 2630 M Ϫ1 cm Ϫ1 ). Initial velocity data were fit to either Michaelis-Menten or the Hill-modified Michaelis-Menten equation using SigmaPlot. For measuring changes caused by addition of effector molecules, the same reaction was performed with 0.5 mM chorismate and varied concentrations of phenylalanine (0 -10 mM), tyrosine (0 -3.2 mM), or tryptophan (0 -3.2 mM). The resulting data were fit to a dose-response curve, y ϭ max/ (1 ϩ (Ef/EC 50 ) n ), where max is the maximum observed rate; Ef is effector concentration, and n is the Hill slope, using SigmaPlot.

Functional Comparison of Arabidopsis Chorismate Mutases-
Previous work on the three chorismate mutase isoforms from Arabidopsis (AtCM1-3) determined the K m value of chorismate and the general effects of the aromatic amino acids on activity for each enzyme; however, these studies were performed using yeast cell extracts (15,16). To quantify the steady-state kinetic parameters of the AtCM isoforms, each enzyme was expressed in E. coli as N-terminal His-tagged protein and purified using nickel-affinity and size-exclusion chromatographies. All three isoforms were iso-lated as homodimeric forms (ϳ65 kDa; monomer, ϳ32.6 kDa) for biochemical characterization.
Each AtCM isoform converted chorismate to prephenate but with clear differences in kinetic behavior ( Fig. 3; Table 2). Both AtCM1 and AtCM2 followed Michaelis-Menten kinetics (Fig.  3, A and B). In contrast, AtCM3 displayed positive cooperativity with a Hill coefficient of 2.1 ( Fig. 3C; Table 2), indicating that substrate binding at one active site of the homodimer enhanced interaction at the second active site. Weak cooperativity (n ϭ 1.2-1.5) has been reported for chorismate mutases isolated from yeast, Nicotiana silvestris (flowering tobacco), and Solanum tuberosum (potato) (13,19,20). The catalytic efficiency (k cat /K m ) of AtCM2 was 11-and 22-fold higher than that of AtCM1 and AtCM3, respectively. This results from a combination of a more rapid turnover rate and a lower K m value for chorismate displayed by AtCM2 compared with the other two isoforms. In earlier studies (15,16), the inability to measure AtCM protein levels in cell extracts precluded estimation of the k cat value for each isoform. The turnover rates of purified AtCM1 (16.1 s Ϫ1 ), AtCM2 (38.7 s Ϫ1 ), and AtCM3 (13.0 s Ϫ1 ) were up to 20-fold slower than the k cat of the yeast enzyme (387 s Ϫ1 ) (13).
Differential Regulation and Identification of New Effectors of Arabidopsis Chorismate Mutases-The previously reported differential feedback effects of aromatic amino acids on the Arabidopsis chorismate mutases were based on single concentrations of each effector in yeast cell extracts (15,16). Using purified proteins, the effector regulation of each AtCM isoform was re-examined. None of the aromatic amino acids at concentrations up to 10 mM altered AtCM2 activity. Both AtCM1 and AtCM3 were sensitive to effector control but with different sets of amino acids.
To determine the effect of aromatic amino acids on AtCM1, the EC 50 values for tryptophan, phenylalanine, and tyrosine were determined ( Fig. 4A; Table 3). Tryptophan enhanced AtCM1 activity from 19.4 to 55.0 mol min Ϫ1 mg Ϫ1 with an EC 50 ϭ 2.6 M (n ϭ 0.8). Both tyrosine and phenylalanine reduced AtCM1 activity by roughly 20-fold. Although there were similar effects on turnover rates, tyrosine (EC 50 ϭ 10.5 M; n ϭ 0.7) binds 5-fold better than phenylalanine (EC 50 ϭ 49.8 M; n ϭ 1.2). Both of these aromatic amino acids were weaker effectors than tryptophan. Screening of the other 17 amino acids as possible effectors of AtCM1 showed no alterations in prephenate production. Similar results were also observed with AtCM2.
Binding of tryptophan to AtCM3 had a much larger effect than that observed for AtCM1 and led to a 6-fold increase in activity from 4.0 Ϯ 0.1 to 25.0 Ϯ 1.5 mol min Ϫ1 mg Ϫ1 with an EC 50 of 5.50 Ϯ 0.03 M (n ϭ 1.2) (Fig. 4B). In contrast to their negative effects on AtCM1, neither tyrosine nor phenylalanine altered AtCM3 activity. Surprisingly, screening of other amino acids as effectors of AtCM3 revealed that cysteine (EC 50 ϭ 123 Ϯ 1 M; n ϭ 1.6) and histidine (EC 50 ϭ 31.6 Ϯ 0.2 M; n ϭ 1.0) each activated enzymatic activity by ϳ3-fold (Fig. 4B).
Three-dimensional Structure of Arabidopsis Chorismate Mutase 1-To understand effector regulation in the Arabidopsis chorismate mutases, we determined the x-ray crystal structures of AtCM1 in complex with phenylalanine and tyrosine at 2.4 and 2.3 Å resolution, respectively (Table 1). AtCM1 crystallized with one monomer in the asymmetric unit, which forms a crystallographic symmetry-related dimer (Fig. 5A). Eight ␣-helices comprise the core of each AtCM1 monomer. The symmetric dimer interface is made of four helices (␣1, ␣2, ␣4, and ␣7) from monomer A interacting with the same helices from monomer B. The overall structure of AtCM1 is comparable with that of the yeast enzyme (6) with a root mean square deviation of 1.63 Å for 262 C␣ atoms.
Crystallization of AtCM1 with either phenylalanine or tyrosine yielded excellent electron density for each ligand (Fig. 5, C  and D). The position of these ligands in AtCM1 clearly identifies the effector binding site at the dimer interface (Fig. 5A). In each structure, the N-terminal loop (residues 79 -91), ␣2, ␣4, and the ␣2/␣3 loop (residues 148 -187) encompass either aromatic amino acid bound in the effector site. The effector site is largely occluded from solvent, potentially by movement of the two loops during binding of ligands. Each chain contributes residues to each side of the effector site (Fig. 6, A and B). Chain A provides Arg-79, His-145, Val-148, Gly-149, and Arg-150 with Asn-211, Gly-213, Ser-214, and Val-217 coming from chain B.
The structure of the AtCM1⅐phenylalanine complex reveals a set of hydrogen bond interactions that lock the ligand in the effector site (Fig. 6A). The carboxylate of phenylalanine interacts with the side-chain guanidinium group of Arg-79 (2.7 Å), the backbone nitrogens of Gly-213 (3.1 Å), and Ser-214 (2.7 Å), and the carbonyl oxygen of Val-148 (3.3 Å). The side-chain oxygens of Asn-211 (2.9 Å) and Ser-214 (2.8 Å) hydrogen bond to the amine group of the bound amino acid. These contacts position the phenylalanine R-group into a space delineated by Gly-213, Val-217, Val-148, Gly-149, and Arg-150. In the structure of AtCM1 complexed with tyrosine (Fig. 6B), a similar set of interactions are formed but with the addition of a hydrogen bond between the ligand hydroxyl group and N⑀ of Arg-83 (3.3 Å).
Although the structures of AtCM1 complexed with phenylalanine and tyrosine (Fig. 6, A and B) are similar to those of ScCM with tyrosine and tryptophan bound (Fig. 6, C and D) (10), differences exist in the effector sites of these two proteins. In AtCM1, Val-148 and Val-217 replace Ile-74 and Thr-145 from the yeast enzyme. Two striking differences between the plant and yeast enzymes were also observed. First, Gly-149 in AtCM1 replaces Arg-75, which interacts with the carboxylate group of amino acid effectors in ScCM. Second, the N-terminal loop of AtCM1 contributes an alternative basic residue (i.e. Arg-79) to interact with bound effector molecules.
Comparison of residues in the AtCM1 effector site with the corresponding positions in the unregulated AtCM2 and the differentially regulated AtCM3 suggests possible amino acid changes that lead to differences in regulation (Figs. 2 and 6A). Across the three AtCM isoforms, the residues corresponding to Val-148, Arg-150, Asn-211, and Ser-214 are invariant. The res- idue corresponding to Val-217 in AtCM1 is variable, as it is an alanine in AtCM2 and a leucine in AtCM3. Between the regulated AtCM1 and unregulated AtCM2, three changes occur in the effector site, Arg-79, His-145, and Gly-213 of AtCM1 are replaced by aspartate, glutamine, and proline, respectively, in AtCM2. The effector site residues of AtCM3 are nearly identical to those of AtCM1 with the exception of an aspartate substitution for Gly-149.

Functional Analysis of Effector Site Differences on Regulation of Arabidopsis Chorismate
Mutases-To examine the effector site differences, we generated a series of site-directed mutants for kinetic analysis. The first set of AtCM1 mutants probed changes to Arg-79, His-145, Gly-213, and Val-217. The R79K, H145Q, and V217T mutants had varied effects on the EC 50 values for the aromatic amino acid effectors but did not change either positive or negative effects on enzymatic activity ( Table  3). The subtle mutation of Arg-79 to a lysine increased the EC 50 values for phenylalanine and tyrosine by 7-and 10-fold, respectively, but did not alter tryptophan binding. Likewise, the AtCM1 V217T mutant yielded less than 5-fold changes in EC 50 for phenylalanine and tyrosine but retained an EC 50 for tryptophan comparable to wild type. The H145Q mutant led to a    20-fold decrease in tyrosine binding and modest 3-fold changes in EC 50 for phenylalanine and tryptophan. The most dramatic effect was observed in the AtCM1 G213P mutant, which eliminated the effect of aromatic amino acids on enzymatic activity. For comparison, the AtCM1 G213A mutant was also generated and analyzed. This mutation also disrupted effector control. Biochemical analysis of AtCM1 and AtCM3 revealed clear variations in the effects of different amino acids on either protein (Fig. 4). The structure of AtCM1 and sequence comparison between AtCM1 and AtCM3 suggest that the residue corresponding to Gly-149 in AtCM1, which is Asp-132 in AtCM3, may be linked to differential regulation. Mutation of Gly-149 to either aspartate (G149D) or alanine (G149A) in AtCM1 eliminates the effector action of both phenylalanine and tyrosine ( Fig. 7; Table 4), as observed with AtCM3. Although each muta-tion slightly increases the EC 50 of tryptophan, the 13-and 8-fold enhancement in activity observed with the G149D and G149A mutants, respectively, exceeds the 3-fold activation of wild-type AtCM1. Neither mutation introduced regulation by cysteine and histidine. For comparison, the AtCM3 D132G mutant was analyzed and yielded an enzyme that kinetically resembled AtCM1 (Fig. 7; Table 4). This mutant retained activation by tryptophan, although to a lesser extent than observed with AtCM3. In addition, both phenylalanine and tyrosine were now negative regulators of the AtCM3 D132G mutant, which indicates that in the plant chorismate mutases the identity of this residue is critical for effector specificity and responses.

DISCUSSION
Chorismate lies at an important branch point in the synthesis of aromatic amino acids and multiple specialized metabolites that contain aromatic groups (1)(2)(3)(4)(5). The enzymes that function at this branch point, including chorismate mutase, are tightly regulated. In contrast to the bacterial and yeast chorismate mutases, which have been extensively studied as a models for allosteric control (6 -13, 29, 30), the plant chorismate mutases are not well understood.
Earlier reports describe three isoforms in Arabidopsis, two of which are plastid-localized and regulated by aromatic amino acids (14 -16). These previous studies of the Arabidopsis chorismate mutases relied on the analysis of proteins in yeast cell extracts, which contain residual phenylalanine, tyrosine, and tryptophan and complicated accurate assessment of the biochemical properties of the plant proteins. Kinetic analyses of purified AtCM1-3 revealed distinct biochemical and regulatory properties of each enzyme (Table 2; Figs. 3 and 4). For example, AtCM3 showed strong positive cooperativity, whereas the other isoforms followed Michaelis-Menten kinet-   For molecules that inactivate the enzyme, the fold-changes are plotted as negative values (i.e. no effector/ effector ratio). For each protein, bars correspond to ratios for tryptophan (white), phenylalanine (black), tyrosine (orange), cysteine (green), and histidine (blue). Concentrations for assays were 0.5 mM chorismate and 10 mM effector. Ratios were calculated from data with n ϭ 3 and standard errors less than 10% of the mean. ics (Fig. 3). Likewise, both of the plastid-localized forms (i.e. AtCM1 and AtCM3) were much less efficient than the cytosolic AtCM2 for prephenate formation. This may reflect the localization of chorismate and aromatic amino acid biosynthesis to the chloroplast (31), where, presumably, elevated chorismate concentrations may not require the high catalytic efficiency displayed by the cytosolic enzyme. In addition to differences in steady-state kinetic parameters, each AtCM isoform displayed unique responses to effector molecules. The cytosolic AtCM2 was unregulated, and AtCM1 was activated by tryptophan (positive effector) and negatively regulated by phenylalanine and tyrosine (Fig. 4A). Surprisingly, tryptophan, cysteine, and histidine were positive effectors of AtCM3 (Fig. 4B) with tryptophan activation of AtCM3 stronger than observed for AtCM2.
The three-dimensional structures of AtCM1 complexed with tyrosine and phenylalanine were similar to the yeast enzyme, which indicates that these proteins share common mechanistic features (Fig. 5). For example, the active sites of AtCM1 and ScCM are nearly identical (Fig. 5B). Thus, AtCM1 likely uses a catalytic mechanism involving transition state stabilization by Arg-229 and Lys-240, as described for the yeast enzyme (10). In addition, the overall structures of the AtCM1⅐tyrosine and AtCM1⅐phenylalanine complexes closely resemble the T-state of the yeast enzyme. Extensive studies of ScCM show that tyrosine maintains the less active T-state, whereas tryptophan binding leads to the R-state and activation of enzymatic activity (10 -13, 29, 30). The basis of the change between the less active T-state and the activated R-state is that the larger indole side chain of tryptophan bound in the effector site shifts the positions of the ␣-helices corresponding to ␣2 and ␣4 in AtCM1 to alter placement of catalytic residues in the active site to provide for enhanced prephenate formation. The conserved fold of the plant and yeast enzymes maintains the effector binding site at the interface of each enzyme and implies a shared allosteric control mechanism.
Although the plant and yeast chorismate mutases share similar catalytic and regulatory mechanisms, key differences between the effector sites of the plant and yeast enzymes lead to specialized effector responses (Fig. 6). First, phenylalanine has no effect on ScCM (12,13). In the ScCM⅐tyrosine complex, Thr-145 forms a hydrogen bond with the tyrosine hydroxyl and seems to be important for tyrosine binding (10). In AtCM1, this residue is Val-217. The AtCM1 V217T mutant did not enhance tyrosine binding, but instead it modestly increased the EC 50 values of both phenylalanine and tyrosine binding with little effect on tryptophan binding (Table 2). This result suggests that the differential effect of phenylalanine on AtCM1 and ScCM may reside elsewhere.
The second major difference in the allosteric sites of the two enzymes is the presence of an N-terminal loop in AtCM1 that allows Arg-79 to interact with the carboxylate of either phenylalanine or tyrosine. A comparable loop does not exist in ScCM, which uses an arginine positioned approximately where Gly-149 of AtCM1 is located to provide a similar binding contact (10). This remodeling of the effector site may facilitate recognition of phenylalanine by AtCM1, as the AtCM1 R79K mutant shows decreased affinity for both phenylalanine and tyrosine ( Table 3). Repositioning of the charge-charge interaction may partly explain the evolution of phenylalanine binding and its negative effect in AtCM1 compared with ScCM.
The structure of AtCM1 also identified sequence differences with the nonallosteric AtCM2 and allosteric AtCM3, which responds to different amino acids as effectors. Biochemical characterization of site-directed mutants of AtCM1 and AtCM3 indicates that two effector site positions corresponding to Gly-149 and Gly-213 in AtCM1 lead to the different regulatory properties of each chorismate mutase isoform in Arabidopsis.
Substitution of a proline for Gly-213 abolished effector control ( Table 3) and suggests that this residue is important for distinguishing between regulated and unregulated chorismate mutases. Similar results were obtained with the AtCM1 G213A mutant. These mutations likely abolish amino acid binding to the effector site, which results in a loss of allosteric control.
The second position fine-tunes effector specificity and activation responses of AtCM1 and AtCM3. In AtCM1, mutation of Gly-149 to either aspartate, the corresponding residue of AtCM3, or alanine led to mutant enzymes with kinetic properties similar to AtCM3 (Fig. 7). In addition, the reverse mutation in AtCM3 (D132G) yielded a mutant enzyme with properties comparable to AtCM1 (Fig. 7).
To extend the structural and functional insights on the regulation of the Arabidopsis chorismate mutases, we examined the sequences of chorismate mutases from a range of plants (Fig. 8). The chlorophyte chorismate mutases group phylogenetically with ScCM and, like ScCM, have an arginine at position 149 (Fig. 8, red). This is the arginine that is moved to the N-terminal tail in the AtCM1 structure. Members of this clade group retain a glycine at position 213, which suggests that the regulatory properties of this group are probably similar to ScCM, i.e. activated by tryptophan and inhibited by tyrosine with phenylalanine having no effect. The chorismate mutases from plants split into two distinct clades.
The AtCM2-containing clade includes sequences from multiple euphyllophyte plant species (Fig. 8, violet). All of these homologs contain a glycine at position 149 and either a proline (AtCM2) or an alanine at position 213. In addition, all of these sequences lack a plastid localization sequence. Members of this group would likely not be allosterically controlled by the aromatic amino acids and would be cytoplasmic, like AtCM2. Interestingly the bryophyte Physcomitrella patens, the lycophyte Selaginella moellendorffii, and Amborella trichopodo, a flowering plant that diverged from the other flowering species  (32). The second clade of plant chorismate mutases can be divided into three sub-clades (Fig. 8, light green, blue, and dark green). AtCM1 homologs in the first sub-clade contain glycines at positions 149 and 213, and most members contain a putative N-terminal plastid localization sequence. This sub-clade only contains sequences from eudicot species with some species containing multiple isoforms. For example, soybean encodes five members of this sub-clade. The sequence homology suggests that members of this sub-clade will be regulated by tryp-tophan as an activator and with phenylalanine and tyrosine acting as negative regulators.
The second sub-clade includes AtCM3. Although phylogenetically this clade does not appear distinct, chorismate mutases grouped here consistently have an aspartate instead of a glycine at position 149, while retaining a glycine at 213 and the plastid localization sequence. This suggests that the members of this clade likely share the distinct effector control of AtCM3. Interestingly, only members of the family Brassicaceae contain AtCM3-like isoforms, which suggests a specialized role for these proteins in this group of plants. The regulation of AtCM3  (38), and the phylogentic tree was created using MEGA (39). Colors were added based on phylogeny and the presence of Gly-149 and Gly-213. Red indicates the chlorophyte and yeast sequences of the first clade. Pink highlights the AtCM2-like clade sequences. Dark green corresponds to the sub-clade containing the chorismate mutases from mosses, ferns, monocots, and Amborella. Light green and blue highlight the AtCM1-like and AtCM3-like sub-clades, respectively. by tryptophan, cysteine, and histidine, all of which are synthesized in the chloroplast (1,33,34), may provide additional control and/or integration with sulfur and nitrogen metabolism. For example, the synthesis of indole glucosinolates requires both indole and sulfur-containing amino acids and activation of AtCM3, and related isoforms may support specialized metabolism (35). The structural and functional studies presented here suggest that AtCM1-like isoforms are essential for basal phenylalanine/tyrosine biosynthesis and that AtCM3-like isoforms may play a role in specialized metabolite production and stress responses in the Brassicaceae.
Chorismate mutases of the third sub-clade, which contains P. patens and S. moellendorffii, share effector sites that retain the two critical glycines like AtCM1 and have the putative plastid localization signal; however, these homologs are phylogenetically different from the other AtCM1-like sub-clades. This group is interesting as it contains species that diverged quite distantly, including the basal angiosperm Amborella, along with monocots, such as rice and maize. Although the effector site sequences suggest members of this clade would be regulated like AtCM1, the lower ϳ50% amino acid sequence identity makes it interesting to see whether they share biochemical properties or whether they behave differently.
A final question remains. What is the role of the unregulated cytosolic chorismate mutase isoforms in plants? Since aromatic amino acid biosynthesis is localized in the plastid, AtCM2 could be involved in an alternative pathway (15). Besides tyrosine and phenylalanine biosynthesis, there is no other known role for prephenate. Moreover, the enzymes that use prephenate are localized to the plastid (31). Recent work has identified another route to phenylalanine in the cytosol that requires conversion of prephenate to phenylpyruvate followed by transamination to phenylalanine with tyrosine as a donor (36). Interestingly, while the first step of this alternative route appears to be plastidic, the final step is cytosolic. It is possible that AtCM2-like proteins could be essential for this alternative pathway, thus being linked to cytosolic phenylalanine synthesis. Moreover, the high catalytic efficiency of AtCM2 may be required in the cytosol where chorismate levels are likely lower than concentrations in the plastid. Ultimately, our studies of the three Arabidopsis chorismate mutase isoforms suggest that subtle changes may result in evolution of specialized regulation of these enzymes in plants.