Product-assisted Catalysis as the Basis of the Reaction Specificity of Threonine Synthase*

Threonine synthase (TS), which is a pyridoxal 5′-phosphate (PLP)-dependent enzyme, catalyzes the elimination of the γ-phosphate group from O-phospho-l-homoserine (OPHS) and the subsequent addition of water at Cβ to form l-threonine. The catalytic course of TS is the most complex among the PLP enzymes, and it is an intriguing problem how the elementary steps are controlled in TS to carry out selective reactions. When l-vinylglycine was added to Thermus thermophilus HB8 TS in the presence of phosphate, l-threonine was formed with kcat and reaction specificity comparable with those when OPHS was used as the substrate. However, in the absence of phosphate or when sulfate was used in place of phosphate, only the side reaction product, α-ketobutyrate, was formed. Global analysis of the spectral changes in the reaction of TS with l-threonine showed that compared with the more acidic sulfate ion, the phosphate ion decreased the energy levels of the transition states of the addition of water at the Cβ of the PLP-α-aminocrotonate aldimine (AC) and the transaldimination to form l-threonine. The x-ray crystallographic analysis of TS complexed with an analog for AC gave a distinct electron density assigned to the phosphate ion derived from the solvent near the Cβ of the analog. These results indicated that the phosphate ion released from OPHS by γ-elimination acts as the base catalyst for the addition of water at Cβ of AC, thereby providing the basis of the reaction specificity. The phosphate ion is also considered to accelerate the protonation/deprotonation at Cγ.

Threonine synthase (TS) 2 catalyzes the last step of the Lthreonine biosynthesis, conversion of O-phospho-L-homoserine (OPHS) into L-threonine and inorganic phosphate (1). TS is a pyridoxal 5Ј-phosphate (PLP)-dependent enzyme and, together with the L-and D-serine dehydratases, threonine de-hydratase, tryptophan synthase, and cysteine synthase, constitutes the ␤-family of PLP enzymes (2,3). Structurally, these enzymes form fold type II PLP enzymes (4). Because TS is found only in bacteria (5), yeasts (6), and plants (4,7), it can be a target for developing antibacterial drugs (5). For this purpose, elucidation of the mechanism of action of TS is crucial.
The reaction mechanism of TS is considered to be the most complicated one catalyzed by the PLP enzymes, proceeding through all the types of intermediates formed during the catalysis of the PLP enzymes (summarized in Scheme 1 (1,3,8,9)). In this mechanism, OPHS reacts with the PLP-Lys aldimine (internal aldimine 1) to form the external aldimine (2) and liberates the side chain of the Lys residue (transaldimination). The intermediate 2 is then converted to the ketimine (3) via a 1,3-prototropic shift. The electron-withdrawing imino group promotes deprotonation at C␤, and after the formation of the enamine (4), the ␥-phosphate group is eliminated, yielding ␤,␥-unsaturated ketimine (5) and a phosphate ion. In 5, deprotonation at C4Ј of the cofactor and protonation at C␥ of the substrate moiety occur to form the PLP-␣aminocrotonate aldimine, 6). The stereospecific addition of a water molecule from the Re face (C␤ being taken as the prochiral center) of 6 to the C␣-C␤ double bond yields the PLP-L-threonine aldimine (8) via the quinonoid intermediate (7). Finally, attack of the ⑀-amino group of Lys on 8 releases the product L-threonine and regenerates the internal aldimine (1).
The remarkable feature of this catalytic reaction is that TS can carry out many regiospecific (C4Ј, C␣, C␤, and C␥) and stereospecific proton transfers at the coenzyme-substrate complexes. Therefore, it is an intriguing problem how such specific proton transfers are enabled to optimize the reaction specificity of TS.
To provide a structural basis for resolving this issue, we previously determined the x-ray crystallographic structures of a thermostable TS from Thermus thermophilus HB8 in the unliganded form and a complex form with the substrate analog, 2-amino-5-phosphonopentanoic acid (AP5) (8). The structure of the complex with AP5 showed an sp 2 hybridization configuration at C␣ and C␤ and an sp 3 hybridization configuration at C4Ј, indicating that it is an enamine analogous to 4. In this structure, the ⑀-amino N of Lys-61 is 3.5, 3.3, and 3.6 Å away from C4Ј, C␣, and C␤, respectively. Considering the fact that the PLP binding lysine residue is the general base catalyst for the 1,3-prototropic shift in aminotransferases (10), we can expect that the ⑀-amino group of Lys-61 is the general base catalyst for the protonation/deprotonation at these atoms, which is involved in steps 2-3, 3-4, and 7-8. In contrast to these steps, however, the catalytic steps involving protonation/deprotonation at C␥ are difficult to understand. The ⑀-amino N of Lys-61 is too far (4.7 Å) from C␥ to directly catalyze the protonation/deprotonation at C␥, and no appropriate protein side chains that can donate/accept a proton are found in the proximity of C␥. Based on the crystal structure of the T. thermophilus HB8 TS complexed with AP5, we pointed out the possibility that the phosphate released during 4 3 5 may reside at the active site and act as a base catalyst for the protonation/deprotonation at C␥ (8). However, the involvement of the phosphate ion in the proton transfer steps and the reaction specificity of TS have not been experimentally studied.
In this study we performed kinetic and spectroscopic analyses of the reactions of the T. thermophilus HB8 TS with OPHS, L-vinylglycine, L-threonine in the presence or absence of phosphate or sulfate and showed that the phosphate ion functions as the catalyst for the addition of water at the C␤ of the PLP-␣-aminocrotonate aldimine, thus enabling TS to undergo the ␤-synthase reaction rather than the ␥-lyase reaction. Additionally, the phosphate was also suggested as the catalyst for the protonation/deprotonation at C␥.

EXPERIMENTAL PROCEDURES
Materials-The recombinant TS was isolated and purified as previously described (8). All other chemicals were of the highest grade commercially available. (E)-4-(3-hydroxy-2methyl-5-(phosphonooxymethyl)pyridin-4-yl)-2-oxobut-3enoic acid (10) was synthesized according to the method of Schnackerz et al. (11). Removal of the cofactor from TS was performed by precipitation with ammonium sulfate at low pH according to the method of Toney and Kirsch (10) except that a 100-fold molar excess of hydroxylamine was added to the enzyme to remove PLP from the Lys residue before the addition of ammonium sulfate. The apoenzyme preparation was passed through a PD-10 (GE Healthcare) column twice to remove ammonium sulfate. The concentration of sulfate in the final preparation, measured after heat denaturation (3 min in boiling water) of the protein, was below the detection limit (1 M) of the photometric method using bromthymol blue (12).
Enzymatic Formation of L-Threonine-TS was incubated with various concentrations of OPHS in a final volume of 100 l of 50 mM PIPES, 100 mM KCl, 0.1 mM EDTA, pH 8.0. After incubation at 298 K for various periods, 100 l of 100 mM perchloric acid was added to terminate the reaction. A 100-l aliquot of 100 mM phenyl isothiocyanate and 100 l of 1 M triethylamine were then added. After incubation for 1 h at 298 K, 400 l of hexane was added and mixed, and the lower layer was loaded onto a Wakosil PTC column (Wako, Japan) using a Beckman System Gold analyzer. The phenylthiocarbamyl L-threonine was detected by monitoring the absorbance at 254 nm.
Enzymatic Formation of ␣-Ketobutyrate-The method used was a modification of that described by Möckel et al. (13). Briefly, TS was incubated with each concentration of substrate in a final volume of 1 ml of 50 mM PIPES, 100 mM KCl, 0.1 mM EDTA, pH 8.0. After incubation at 298 K for 1 h, 1 ml of reagent (1 g of semicarbazide HCl plus 0.9 g of sodium acetate in 100 ml of H 2 O) was added to terminate the reaction. After incubation for 15 min at 298 K, 3 ml of H 2 O was added, and the absorbance due to the semicarbazone was read at 254 nm. Steady-state kinetic parameters were determined by systematically changing the concentration of L-threonine, phosphate, or sulfate. The kinetic data were fitted to the Michaelis-Menten equation by nonlinear regression using KaleidaGraph (Version 3.0, Synergy Software).
Spectrophotometric Measurement-The absorption spectra were measured at 298 K using a Hitachi (Japan) U-3300 spectrophotometer. Stopped-flow spectrophotometry was performed using an Applied Photophysics SX.18MV spectrophotometer. Typically, equal volumes (75 l each) of solution A (28 M enzyme and 100 mM phosphate or sulfate) and solution B (various concentrations of L-threonine) were mixed. The mixing dead time was 2.3 ms under an N 2 gas pressure of 500 kPa. The time-resolved spectra were collected at 298 K using SX.18MV equipped with a photodiode array accessory and XScan (Version 1.0, Applied Photophysics). The spectral data were analyzed by ProKineticist II (Version 1.9, Applied Photophysics) to obtain the spectra of the reaction intermediates and the kinetic parameters. The buffer solution for the spectrophotometric measurements contained 50 mM PIPES, 100 mM KCl, 0.1 mM EDTA, pH 8.0. TS was equilibrated with this buffer by gel filtration using a PD-10 column before the measurements.
Crystallization, Data Collection, and Refinement-The apoenzyme of TS was crystallized by the microdialysis method. The enzyme solution (8 mg/ml) was placed in a 50-l dialysis button and dialyzed at 293 K against 1.45 M ammonium sulfate in 0.1 M MES, pH 6.5. Single crystals with the approximate dimensions of 0.1 ϫ 0.1 ϫ 0.1 mm grew in about 2 weeks, and then the dialysis buttons were transferred to a new reservoir solution supplemented with 40% (v/v) glycerol as a cryoprotectant and kept at 293 K for 5 h. The complex of TS with 10 was crystallized by the hanging drop vapor diffusion method. The drop was composed of 5 l of protein solution (8 mg/ml apoenzyme in 5 mM Tris-HCl, pH 6.2), 5 l of reservoir solution (10% (v/v) PEG3350 in 0.1 M Na 2 HPO 4 and KH 2 PO 4 , pH 6.2), and 1 l of 10 mM 10. Single crystals with the approximate dimensions of 0.2 ϫ 0.1 ϫ 0.1 mm were obtained by equilibration of the mixture solution against the reservoir solution at 293 K for about 1 week. The crystal was soaked for 10 s in a new reservoir solution supplemented with 30% (v/v) glycerol as a cryoprotectant. The crystals were mounted on thin nylon loops (ø, 0.2-0.3 mm) and frozen by flash-cooling to 100 K in a cold N 2 gas stream. The x-ray diffraction data were collected at 100 K by synchrotron X-radiation ( ϭ 0.9 Å) using a DIP6040 detector (Bruker AXS, Madison, WI) in the station BL44XU at SPring-8 (Hyogo, Japan). The diffraction data collected for these crystals were processed and scaled using HKL2000 (14). Both crystals belong to the space group P3 2 21. The details and statistics of the data collection are summarized in Table 4. The programs used for the refinements, calculation of the electron-density maps, and assignment of the solvent molecules were CCP4 program suite Version 6.12 (15) and CNS Version 1.2 (16). Manual re-building was performed using the Xfit module of the Xtal-View software package (17). The structure of TS in the holoenzyme form (PDB code 1UIN, Ref. 8) and the structure of TS complexed with AP5 (PDB code 1V7C, Ref. 8) were used as the initial model for the apoenzyme of TS and the TS complexed with 10, respectively. After being subjected to the rigid-body refinement, the initial structures of the apoenzyme and the complex with 10 were further refined through simulated annealing from 3000 K and several cycles of B-factor and positional refinements. The model of 10 was built using MOE Version 2007.09 (Chemical Computing Group), and then the topology and parameter files for refinement in CNS were generated with XPLO2D in the X-UTIL package (18). Water molecules, phosphate ion derived from the reservoir solution, and residues in the active-site region were examined using 2F o Ϫ F c , F o Ϫ F c , and omit maps. These models were refined through several additional cycles of the B-factor and positional refinements after introducing solvent molecules. The details and statistics of the crystallographic refinement are also summarized in Table 4. The coordinates for the apoenzyme and TS complexed with 10 have been deposited in the PDB with accession codes 3AEY and 3AEX, respectively. nine formation were about 120-and 190-fold higher, respectively, than those for the ␣-ketobutyrate formation, yielding a high reaction specificity of 99% in both the k cat and k cat /K m terms (Table 1). That the ratio of the k cat values for the L-threonine and the ␣-ketobutyrate formation is similar to the corresponding ratio of the k cat /K m values is considered to reflect the fact that L-threonine and ␣-ketobutyrate are derived from the same intermediate bound to the enzyme, i.e. the PLP-␣aminocrotonate aldimine (6).

Catalytic Reaction of TS with OPHS and L-Vinylglycine
A chemical consideration predicts that besides the physiological substrate OPHS, TS may also catalyze the formation of L-threonine from L-vinylglycine via 9 3 6 3 7 3 8. In this context, the reaction of TS using L-vinylglycine was followed. However, no formation of L-threonine was observed, whereas the side reaction product ␣-ketobutyrate was formed with a k cat similar to that when OPHS was used as the substrate (Table 1). This indicates that the formation of 6 alone is not enough for producing L-threonine.
During the reaction of TS with OPHS, a phosphate ion is released (4 3 5) before the formation of 6 (Scheme 1). To investigate whether the released phosphate ion is involved in the catalysis, the reaction starting from L-vinylglycine was studied in the presence of phosphate and its analog sulfate, which has a similar shape and size as phosphate but has a much lower pK a value (1.92 at 298 K). In the presence of 50 mM phosphate, L-threonine was preferentially formed (Table 1). In both the k cat and k cat /K m terms, the reaction specificity was 98%. This is the same pattern as that observed when OPHS was used as the substrate and again reflects the mechanism that L-threonine and ␣-ketobutyrate are derived from the same intermediate (6) bound to the enzyme. The k cat values for the L-threonine and ␣-ketobutyrate formation were similar to those when OPHS was used as the substrate (Table 1). On the other hand, the k cat /K m values for the L-threonine and ␣-ketobutyrate formation were several hundred-fold lower than the corresponding values using OPHS as the substrate (Table 1). This indicates that L-vinylglycine is far more weakly bound to TS than OPHS. According to the previous crystallographic study of TS complexed with AP5 (8), the phosphate group of OPHS is considered to be fixed by Thr-88, Asn-154, Ser-155, Arg-160, and Asn-188 at the active site of TS. The low k cat /K m values for L-vinylglycine (in the presence of phosphate as well as in the presence of sulfate or in the absence of both phosphate and sulfate; see Table 1) as compared with the corresponding values for OPHS indicate the importance of the phosphate group of OPHS for binding to TS. In contrast to the reaction of L-vinylglycine with TS in the presence of phosphate, the reaction in the presence of sulfate showed a pattern essentially similar to the reaction in the absence of both phosphate and sulfate, i.e. only ␣-ketobutyrate was formed without the formation of L-threonine ( Table 1).
The above results indicate that 6 derived from L-vinylglycine is preferentially converted to L-threonine in the presence of phosphate as efficiently as the corresponding reaction for the normal catalysis starting from OPHS, and sulfate cannot substitute for phosphate. To investigate in detail the effect of phosphate ion on the catalytic steps after 6, we analyzed the "reverse" reaction 1 3 8 3 7 3 6 3 1 starting from L-threonine in the presence of phosphate or sulfate.
Reaction of TS with L-Threonine in the Presence of Phosphate or Sulfate Ion-The addition of L-threonine up to 50 mM to TS caused no spectral changes (supplemental Fig. S1, A and B). The CD spectrum also showed no detectable changes upon the addition of L-threonine to TS (data not shown). That change in CD is a sensitive indicator of the formation of the external aldimine (19) strongly suggests L-threonine alone cannot form the external aldimine with PLP (8) in TS. On the other hand, the addition of phosphate or sulfate to TS slightly increased the 416-nm peak and concomitantly decreased the absorption at around 340 nm (supplemental Fig. S1, C and D). This indicates that the phosphate or sulfate ion binds to the active site of TS, causing the PLP-Lys-61 aldimine to undergo a tautomeric shift from enolimine to ketoenamine. From the absorption change at 416 nm, the K d for phosphate and sulfate ions were estimated to be 2.2 and 4.9 mM, respectively.
In the presence of both L-threonine and either of phosphate or sulfate, distinct spectral changes in the TS (supplemental Fig. S1) and ␣-ketobutyrate formation ( Table 2) were observed. However, neither OPHS nor O-sulfohomoserine was detected in the reaction mixture (data not shown). This is interpreted to reflect the fact that the step of phosphate elimination (4 3 5) is essentially irreversible, and neither phosphate nor sulfate can undergo addition to the C␥ of 5 due to their low Lewis basicities. The kinetic parameters for the ␣-ketobutyrate formation are summarized in Table 2. The K d values for phosphate and sulfate kinetically obtained were essentially consistent with the values obtained by the spectro- Product-assisted Catalysis of Threonine Synthase scopic measurements described above. The similarity between the K d values for phosphate and sulfate suggests the similarity of the recognition of the two anions by TS. On the other hand, the k cat and k cat /K m values in the presence of sulfate were about 7-and 2.4-fold higher, respectively, than those in the presence of phosphate.
Stopped-flow Spectroscopy of the Reaction of TS with L-Threonine-To investigate the reaction steps from L-threonine to ␣-aminobutyrate in detail, transient kinetic analyses were carried out. A solution containing TS and 50 mM phos-phate was reacted with L-threonine (Fig. 1A). The absorption at around 330 nm decreased, and that around 410 nm increased within the dead time (2.3 ms), indicating a rapid association of TS and L-threonine. Subsequently, two sharp peaks emerged at 474 and 446 nm, and a broad absorption band at around 330 nm increased. The sharp peaks at 474 and 446 nm are clearly ascribed to the quinonoid intermediate. The absorbance at 331, 416, and 474 nm was plotted versus the elapsed time after mixing ( Fig. 2A). At any wavelengths, the absorbance changes followed a double exponential process, with a k app of ϳ340 and ϳ40 s Ϫ1 at 250 mM L-threonine. These results indicated that the spectral changes in the reaction of TS with L-threonine in the presence of phosphate can be analyzed by a model that includes two steps after the rapid association of the enzyme (A) and the substrate (B): As the rate of ␣-ketobutyrate formation (k cat ϭ 0.034 s Ϫ1 ;    Table 2) is much slower than the k app values of the spectral changes, the regeneration of A from D was not incorporated in this model (Equation 1). The reaction of TS with L-threonine in the presence of 50 mM sulfate was followed in the same way (Fig. 1B). In contrast to the reaction in the presence of phosphate, no significant accumulation of the quinonoid intermediate was observed. The absorption at 416 nm first increased followed by a slow red shift in max and an appearance of a shoulder at around 480 nm. The time course of the absorption changes followed a double exponential process (Fig. 2B) with k app values of ϳ30 and ϳ0.85 s Ϫ1 . The k app values for the slower spectral changes were of the same order as that of the ␣-ketobutyrate formation (k cat ϭ 0.23 s Ϫ1 ; Table 2). In this regard, the accumulation of the spectra of the intermediates was apparently lower than that in the presence of phosphate (Fig. 1, A and B).
These results indicate that, contrary to the case in the presence of phosphate, the regeneration of the free enzyme cannot be neglected. Therefore, the spectral changes in the reaction of TS with L-threonine in the presence of sulfate should be analyzed according to the model Global Analysis of the Reaction of TS with L-Threonine-The time-resolved spectra of TS taken after the reaction with various concentrations (10,20,50,80,120,160,200, and 250 mM) of L-threonine were subjected to a global fitting analysis using the models of Equation 1 (reactions in the presence of phosphate) and 2 (reactions in the presence of sulfate). For the initial values of the kinetic parameters, the apparent rate constants obtained from the double exponential fitting (Fig. 2) were divided by 2, assuming the equilibrium constant of 1 for  (Table 2) were used. After the fitting, the spectra of TS and the intermediates, i.e. A, AB, C, and D, and the kinetic parameters were obtained, each summarized in Fig. 3 and Table 3, respectively.
For both reactions in the presence of phosphate and sulfate, the spectra of the intermediate D showed a large absorption band at 450 nm and a smaller one at around 330 nm. These spectra of D resemble those of the PLP-␣-aminoacrylate aldimine reported for several enzymes in the absorption maxima, strength, and shape (20 -22). Considering the same conjugate system between the PLP-␣-aminoacrylate aldimine and the PLP-␣-aminocrotonate aldimine, we can conclude that D is the PLP-␣-aminocrotonate aldimine (6), whose ketoenamine and enolimine tautomers absorb at 450 and 330 nm, respectively.
For the reaction in the presence of phosphate, the intermediate C showed the typical spectrum of the quinonoid intermediate (Fig. 3A) and was attributed to 7. Accordingly, the spectrum of the intermediate AB, which was similar to that of A, was considered to be that of the Michaelis complex with L-threonine (not shown in Scheme 1 but exists between 8 and 1), the external aldimine (8), or both. On the other hand, for the reaction in the presence of sulfate ion, the spectra of the intermediate C showed a large absorption band at 420 nm and a faint absorption band at around 480 nm (Fig. 3B). The shape and the position of the spectrum showed that the intermediate C is largely the PLP aldimine with no double bond in the

TABLE 3
Transient kinetic parameters for the reaction starting from L-threonine to form ␣-ketobutyrate TS was reacted with various concentrations of L-threonine in the presence of either phosphate or sulfate, and the spectral changes were monitored by a stopped-flow spectrophotometer. The time-resolved spectra were subjected to global analysis based on the models of Equation 1 (phosphate) and Equation 2 (sulfate), and the kinetic parameters were obtained. See "Results" for details.

Reaction system
substrate moiety conjugated to the imine. As compared with the absorption band of AB, which was again similar to A, the absorption band of C was red-shifted (ϳ8 nm) and had a larger intensity at 420 nm and a smaller one at 330 nm. This indicates that AB and C are both PLP aldimines but are clearly distinguished. The most probable interpretation is that AB and C are the Michaelis complex with L-threonine and the external aldimine (8), respectively. Spectral changes similar to this are observed in the transaldimination process of a number of PLP enzymes (19). Assuming that the Michaelis complex and the external aldimine in the reaction in the presence of phosphate have similar absorption spectra as those in the reaction in the presence of sulfate, the absorption spectrum of AB in the presence of phosphate indicates that it is mainly composed of the Michaelis complex. The presence of a small absorption at around 480 nm in the spectrum of C in the presence of sulfate suggests the presence of a trace amount of the quinonoid intermediate (Fig. 3B). Using the ⑀ value at 476 nm of C in the presence of phosphate (Fig. 3A), the fraction of the quinonoid intermediate in C in the presence of sulfate was estimated to be 6.7%. We could not resolve C by global fitting based on the model of a threestep process after the Michaelis complex. This indicates that in the presence of sulfate the concentrations of the external aldimine and the quinonoid intermediate change in parallel, showing that the process of the interconversion between the external aldimine (8) and the quinonoid intermediate (7) is considerably fast when compared with the other processes.
Comparison of the Steady-state and Transient Kinetic Parameters for L-Threonine-The steady-state and the transient kinetic parameters are related to each other by the following equations: where the microscopic kinetic parameters are those shown in Equations 1 and 2. For the reaction in the presence of phosphate, the k ϩ3 value was the only undetermined value. This was calculated to be 0.049 s Ϫ1 using Equation 3 and the observed k cat value for the reaction with L-threonine in the presence of phosphate. Now that the microscopic kinetic parameters have been obtained, we can estimate the steady-state kinetic parameters k cat and k cat /K m using Equations 3 and 4. The k cat /K m value was 1.1 M Ϫ1 s Ϫ1 in the presence of phosphate, and k cat ϭ 0.23 s Ϫ1 and k cat /K m ϭ 4.0 M Ϫ1 s Ϫ1 in the presence of sulfate. The calculated values excellently matched the observed values (Table 2), showing the validity of the kinetic analysis described above.

Free-energy Profile of the Steps of the Latter Part of the TS
Catalytic Reaction-Using the microscopic kinetic parameters obtained above, the free energy profile of the reaction steps between L-threonine and ␣-ketobutyrate could be obtained. The energy profiles in the presence of phosphate and sulfate are shown in Fig. 4. The profiles were drawn according to the direction of the normal catalytic steps shown in Scheme 1. Uncertainties about the energy levels of several intermediates and transition states arising from the inability to dissect the reaction steps are also shown. The profiles indicate that, in the presence of phosphate as compared with sulfate, the transition state between the PLP-␣-aminocrotonate aldimine (6) and the quinonoid intermediate (7) and that between the external aldimine (8) with L-threonine and the Michaelis complex with L-threonine are stabilized (Fig. 4). For the transition state between the PLP-␣-aminocrotonate aldimine (6) and the quinonoid intermediate (7), the stabilization energy was calculated to be 17.7 kJ⅐mol Ϫ1 , corresponding to the rate enhancement of about 10 3 .
Crystal Structure of TS Complexed with a Structural Analog for the PLP-␣-Aminocrotonate Aldimine-10 (Fig. 5B), which is a condensation product of PLP and pyruvic acid, has been used as an analog for the PLP-␣-aminoacrylate aldimine in serine dehydratase (11). Because of the structural similarity of the PLP-␣-aminoacrylate aldimine and the PLP-␣-aminocrotonate aldimine as described above, we expected that 10 can also be used as an analog for the PLP-␣-aminocrotonate aldimine (6) in TS. The apoenzyme of TS was reconstituted with 10 and crystallized in the presence of phosphate. Crystallization of the relevant structures, such as TS, TS complexed with 10, and the apoenzyme of TS either in the phosphate-or sulfate-bound form has been attempted. However, only the crystals of the apoenzyme of TS complexed with sulfate (apoTS-S) were obtained so far. The structures of TS-10-P and apoTS-S were determined by x-ray crystallography at 2.10 and 1.92 Å resolutions, respectively ( Table 4). The overall structure of TS-10-P is shown in Fig. 5A and compared with the structures of TS and TS complexed with AP5 (TS-AP5) (Fig. 5A). The structure of apoTS-S was superimposable on that of TS, showing that the enzyme does not undergo any global conformational change upon the binding of PLP (supplemental Fig.  S2A). However, the binding of the substrate analog AP5 to the holoenzyme induces a large conformational change from the open to the closed form in which the small domain of TS moves as a rigid body to close the active site (8). This closed structure was maintained in TS-10-P (Fig. 5A), indicating that TS is in the closed form in the enamine and the PLP-␣-aminocrotonate aldimine intermediates. These data support the catalytic model that TS is in the closed form, whereas the substrate and the intermediates are bound to the enzyme.
Phosphate Ion Bound to the Active Site-The F o Ϫ F c omit map clearly showed an electron density of 10 in the active site (Fig. 5B). The modeled structure of 10 was almost completely planar, reflecting that all the C atoms of 10 except for C2Ј and C5Ј are sp 2 -hybridized. In addition to this electron density, a distinct electron density having a tetrahedral shape was observed. A phosphate ion could be excellently fit to this density, indicating that a phosphate ion derived from the solvent is bound to the active site of TS-10-P (Fig. 5B). It is important to confirm that this is not a sulfate ion that has been used to prepare the apoenzyme. The solution used for crystallization contained less than 5 pmol of sulfate ion (1 M ϫ 5 l, see "Experimental Procedures"). The number of the cells in a single crystal is estimated to be 1.5 ϫ 10 12 as calculated from the size of the crystal and the unit cell dimensions. Because there  were at least 10 crystals of a size similar to this, the "amount" of the active sites in the crystals are calculated to be more than 50 pmol. Therefore, even if all the sulfate ions remaining in the solution are bound to the crystals, it can occupy less than 10% of the active sites. From these calculations we consider that the observed tetrahedral electron density comes from the bound phosphate ion. The active site structure of TS-10-P was superimposed on that of TS-AP5 (Fig. 6). The PLP moiety of TS-AP5 and that of TS-10 reside in the same plane. However, in contrast to the planar conformation of 10, C4Ј and N␣ of TS-AP5 are sp 3hybridized. As a result, the C␣ of 10 of TS-10-P deviates from the corresponding C␣ of TS-AP5 by 0.4 Å, with a clockwise 21°rotation (viewed from the Re face of AP5, i.e. the Si face of 10, C␣ being taken as the prochiral center) of the three bonds around C␣ about the axis passing through C␣ and perpendicular to the plane of 10. Consequently, the position of the carboxylate group is substantially different from that of AP5. Nevertheless, the carboxylate group of 10 and that of AP5 make hydrogen bonds with the same neighboring groups (Fig.  6), i.e. the main chain NH groups of Thr-85, Asn-87, and Thr-88, the side chain OH group of Ser-84, and a water molecule.
The phosphate ion in TS-10-P is located at exactly the same position as the phosphono group of AP5 in TS-AP5 and is stabilized by interaction with the same groups, i.e. the side chains of Lys-61, Thr-88, Asn-154, Ser-155, Arg-160, and Asn-188 (Fig. 6). The only difference between the phosphate ion and the phosphono group is the possible presence of a hydrogen bond between the carboxylate group of 10 and an O atom of phosphate, which is replaced by a C atom in AP5 (Fig. 6).
The crystal structure of apoTS-S showed that the cavity formed by the removal of PLP is filled with three water molecules and two sulfate ions derived from the solvent (supplemental Fig. S2B). One of the sulfate ions is located at the same position, albeit with an ϳ30°rotation, as the phosphate group of 10 in TS-10-P and that of PLP in the holoenzyme. Another sulfate ion interacts with the ⑀-amino group of Lys-61, a water molecule, and the main chain NH group of Asn-87, which is the residue that interacts with the carboxylate group of 10 in TS-10-P (supplemental Fig. S2B). A third sulfate ion forms hydrogen bonds with the side chains of Thr-88, Arg-160, and Asn-188 and a water molecule. These residues are involved in the binding of the phosphate ion in TS-10-P, and as the result, the sulfate ion is located at a position close to that of the phosphate ion in TS-10-P (supplemental Fig. S2B). The deviation in the position of the sulfate ion is considered to be due to the open conformation of the enzyme protein. Although the crystal structure of TS complexed with 10 and the sulfate ion has not been obtained, these data strongly suggest that the sulfate ion binds to the same position as the phosphate ion during the course of catalysis by TS.

DISCUSSION
Reaction Specificity of PLP Enzymes and TS-PLP catalyzes versatile reactions of amino acids, such as racemization, transamination, elimination, replacement, decarboxylation, aldol cleavage, etc. This indicates that a PLP enzyme must be organized to carry out a specific type of reaction. An important mechanism that accounts for the reaction specificity of the PLP enzymes has been provided by the Dunathan hypothesis (23), which suggests that the conformation of the external aldimine at the active site determines the reaction type; if the bond perpendicular to the plane of the imine-pyridine ring is C-H, C-COO Ϫ , or C-CHOH-, then ␣-deprotonation, decarboxylation, or aldol cleavage occurs, respectively. Although this crucially determines the reaction during the early step of the catalysis, this alone cannot explain all of the reaction specificity of the PLP enzymes. For example, there are many types of reactions that start with ␣-deprotonation, such as racemization, transamination, ␤-elimination, ␤-replacement, ␥-elimination, ␥-replacement, etc. The individual PLP enzymes catalyzing only one of those reactions must control the proton transfer processes to realize their reaction specificity.
TS is a unique PLP enzyme undergoing ␥-elimination and ␤-addition reactions. Therefore, TS shares the reaction pathway 1-6 with ␥-lyases, and this distinguishes them from ␥-synthases, which do not catalyze the step between 5 and 6, the proton transfer between C4Ј and C␥. However, TS is different from the ␥-lyases in that it preferentially catalyzes the addition of water at the C␤ of 6 as compared with the transaldimination of 6 to yield ␣-ketobutyrate (Table 1). Therefore, it is important to know how TS controls these steps to carry out the specific biosynthetic reaction.
Phosphate Ion Promotes the Reaction Steps from 6 to L-Threonine-In this study it was shown that L-vinylglycine, which has been known to be an alternate substrate for the ␥-synthase (24, 25) and 1-aminocycleopropane-1-carboxylate synthase (26) reactions, can be incorporated into and processed at the active site of TS (Table 1). In the presence of phosphate ion, TS forms L-threonine from L-vinylglycine with a k cat value and a reaction specificity comparable with those when OPHS was used as the substrate (Table 1). However, in the absence of phosphate ion or in the presence of sulfate ion, only ␣-ketobutyrate was formed (Table 1). These results indicated that the phosphate ion accelerates the reaction from 6 to the L-threonine formation, and this function is related to a property of the phosphate ion, which is absent in the sulfate ion.
Phosphate as a Base Catalyst for the Addition of Water at the C␤ of 6-To clarify the function of the phosphate ion in the L-threonine formation reaction in detail, the reverse reaction starting from L-threonine was kinetically studied in the presence of phosphate or sulfate. The result showed that the phosphate ion lowers the energy barrier for the addition of the water molecule at the C␤ of 6 and the transaldimination to form L-threonine, as compared with the sulfate ion (Fig. 4). Therefore, we structurally studied how the phosphate ion is involved in the reactions starting from 6 by solving the crystal structure of TS complexed with an analog for 6 (Fig. 5). The phosphate ion was found to bind to the site originally occupied by the phosphate group of OPHS and was supposed to interact with the attacking water molecule in the step 6 3 7. The energy barrier for this step in the absence of phosphate or sulfate was not obtained because of the extremely low affinity of L-threonine in the absence of phosphate or sulfate. However, the finding that the phosphate ion lowers the energy barrier more effectively than the sulfate ion, which has a much lower basicity (pK a ϭ 1.92 versus 7.21; values for the second dissociation at 298 K), strongly indicates that the phosphate acts as a base catalyst during the attack of water on the C␤ of 6. From the difference in the pK a values and the rate constants, the Brønsted ␤ was estimated to be 0.59, indicating that the proton is in an intermediate position between the O atoms of the attacking water molecule and the phosphate or sulfate at the transition state.
In contrast to the effect on the step of the addition of water at C␤, the effect of phosphate ion on the transaldimination 8 3 1 is more difficult to understand. We will return to this problem later.
Possible Involvement of the Phosphate Ion in the Protonation/Deprotonation at C␥-That the phosphate ion locates near the C␥ of 6 and is the base catalyst for the addition of the water molecule to C␤ raises the possibility that this phosphate ion is also involved in the protonation/deprotonation at C␥, as we have suggested in a previous crystallographic study on TS complexed with AP5 (8). This conjecture is attractive as the ⑀-amino group of Lys-61 is too distant from C␥ and other group(s) are considered to be required to protonate/deprotonate the atom. Using the kinetic parameters obtained in this study, we estimated the effect of phosphate on this step as follows.
A simple mechanism for the reaction starting from L-vinylglycine is Scheme 2, where E, S, and AC denote the enzyme, L-vinylglycine, and the PLP-␣-aminocrotonate aldimine (6), respectively. Using the above simple mechanism, the k cat values for L-threonine and ␣-ketobutyrate formation are expressed as, The ratio of these k cat values is equal to k b /k c and was calculated to be 47 in the presence of phosphate (Table 1). Therefore, using the value of k c ϭ 0.049 s Ϫ1 (k c is equal to k ϩ3, which in the presence of phosphate is estimated to be 0.049 s Ϫ1 ; see the calculation shown after Equations 3-4), k a was estimated to be 0.85 s Ϫ1 from Equation 5.
In the absence of phosphate, on the other hand, Equation 6 is simplified to This has a value of 0.015 s Ϫ1 (Table 1). As the PLP-␣-aminocrotonate aldimine was not detected in the reaction of TS with L-vinylglycine (supplemental Fig. S3B), k a is apparently smaller than k c and should be almost equal to the k cat value, 0.015 s Ϫ1 . Therefore, we can consider that k a is increased 57fold in the presence of phosphate.
In the presence of sulfate, k c ϭ 0.43 s Ϫ1 (equal to k ϩ3 ; Table  3). Therefore, k a is almost equal to the k cat value of 0.0034 s Ϫ1 (Table 1). This is even lower than the k a in the absence of phosphate or sulfate (0.015 s Ϫ1 ; see above), indicating that the sulfate ion, unlike the phosphate ion, does not have the ability to enhance k a .
Altogether, these calculations indicate that phosphate promotes the step 9 3 6, probably by functioning as an acid-base catalyst. The rate enhancement (57-fold) is moderate as compared with the value (10 3 -fold) for the addition of water at C␤. The remaining reaction in the absence of phosphate may be ascribed to the catalysis by water molecule(s) near C␥.
The step 9 3 6 is a nonphysiological reaction that exists on the way from the unnatural substrate L-vinylglycine. However, considering the similarity in the reaction (protonation at C␥), we can expect that the phosphate ion functions in the same way in the step 5 3 6 during the normal catalysis.
Effect of Phosphate Ion on the Transaldimination Steps-One of the problems concerning the reaction specificity of TS is the two transaldimination processes starting from 6 and 8. Ideally, the former should be inhibited, and the latter should be promoted during the catalytic reaction of TS. It is, then, important to know the effect of the phosphate ion on these steps. However, as the rate constant for neither of these steps in the absence of phosphate or sulfate could be obtained, the effect of the phosphate ion can be discussed only in comparison with the effect of the sulfate ion.
The rate constant for the transaldimination from 6 is 9-fold higher in the presence of the sulfate ion (Table 3 and the estimated value of k ϩ3 ϭ 0.049 s Ϫ1 in the presence of phosphate), whereas the rate constant for the transaldimination from 8 is higher in the presence of phosphate. The difference between the structure of 6 and 8 is the presence (8) or absence (6) of the ␤-OH group and the hybridization at C␣ (sp 2 for 6 and sp 3 for 8). As the phosphate ion can interact with both the ␤-OH group and the ⑀-amino group of Lys-61 and the latter is the group that directly involved in the transaldimination process, the presence/absence of the ␤-OH group may affect the transaldimination process. As a simple hypothesis, assuming that the ␤-OH group and the ⑀-amino group of Lys-61 compete with each other for binding to the phosphate ion, we can expect that the ⑀-amino group forms a hydrogen bond with the phosphate ion in 6, as observed in the crystal structure of TS-10-P (Fig. 5) but is liberated from the phosphate ion in 8. This may result in attenuating the transaldimination of 6 without affecting the transaldimination of 8. The reverse effect of the sulfate ion may be ascribed to the absence of protons in the sulfate ion.
Product-assisted Catalysis-As shown in the structure of TS-10-P, the phosphate ion is stabilized by many residues at the active site. This is considered to be the cause of the fairly high affinity of the phosphate ion to TS (K d ϭ 2.2 mM). However, comparison of the structure of the unliganded enzyme and TS-10-P indicates that, among the residues interacting with the phosphate ion in TS-10-P, Thr-88, Asn-154, and Ser-155 move toward the phosphate ion upon the conformational change from the open to the closed structure induced by the substrate binding. Therefore, it is expected that during the catalysis the phosphate ion has a much higher affinity to the enzyme than that to the unliganded enzyme. The phosphate ion released at the step of 4 3 5 is then considered to remain at the active site and acts as the catalytic base that directs the reaction from 6 toward the formation of L-threonine.
The product-assisted catalysis in the enzyme has been reported for human 8-oxoguanine DNA glycosylase/lyase (27). The enzyme cleaves 8-oxoguanine from DNA and uses it as a catalyst for the following steps. Although the reactions catalyzed by TS and 8-oxoguanine DNA glycosylase/lyase are different, both reactions are complex and involve many elementary steps. Therefore, during the evolution, these enzymes may have adopted a strategy to expand their catalytic potential by incorporating the product as the catalytic species.
One of the reasons for the presence of the phosphate group in the natural substrate OPHS is to provide a good leaving group at C␥. However, as shown in this study, the released phosphate in turn functions as the catalyst in the subsequent steps, and this mechanism explains how the reaction is controlled in TS at the branching point (6) of the catalysis dividing the ␤-synthase and the ␥-lyase. Although this is a crucial element that warrants the reaction specificity of TS, the catalytic reaction of TS contains other branching points that are subject to side reactions unless there are mechanisms to control the pathway. These points should be studied in the future.