Function of the Active-site Lysine in Escherichia coli Serine Hydroxymethyltransferase*

Serine hydroxymethyltransferase has a conserved lysine residue (Lys-229) that forms the internal aldimine with pyridoxal 5"phosphate. In other pyridoxal 5'-phos-phate enzymes investigated so far, this conserved lysine residue also plays a catalytic role as a base that removes the a-proton from the amino acid Substrate. Three mu- tant forms of Escherichia coli serine hydroxymethyltransferase (K229Q, K!229R, and K229H) were con-structed, expressed, and purified. The absorbance spectra, rapid reaction kinetics, and thermal denaturation of the mutant analogs were studied. Only the K229Q mutant serine hydroxymethyltransferase re- sembled the wild-type enzyme. The results indicate that Lys-229 plays a critical role in expelling the product by converting the external aldimine to an internal aldi- mine. In the absence of Lys-229, ammonia can also catalyze the same function at a much slower rate. However, Lys-229 apparently is not the

1 Will submit part of this work in partial fulfillment of the requirements for the Ph.D. degree at the Universith La Sapienza, Roma. tamate (H4PteGlu) serving as the one-carbon carrier. The enzyme contains covalently bound pyridoxal-P, which forms an internal aldimine with the €-amino group of Lys-229 in the Escherichia coli enzyme (Schirch et al., 1985). SHMTs from rabbit liver cytosol and mitochondria, Neurospora crassa, E.
coli, and Bradyrhizobium japonicum have a conserved 9-amino acid sequence containing this lysyl residue (McClung et al., 1992;Byrne et al., 1992). Previous studies have shown that 2 other residues in this conserved sequence, His-228 and Thr-226, play critical roles in determining reaction and substrate specificity (Stover et al., 1992;Angelaccio et al., 1992). The goal of this study was to determine if Lys-229 also plays a critical catalytic role in addition to its role in forming an internal aldimine with pyridoxal-P.
All pyridoxal-P enzymes appear to have the coenzyme bound to the €-amino group of a lysyl residue as an internal aldimine. This Lys has been changed to another amino acid by site-directed mutagenesis in several pyridoxal-P enzymes (Bhatia et al., 1993;Lu et al., 1993;Ziak et al., 1993;Ilag and Jahn, 1992;Yoshimura et al., 1992;Grimm et al., 1992;Kirsch, 1991, 1992;Nishimura et al., 1991;Planas and Kirsch, 1991;Smith et al., 1989). In most cases, the mutant enzymes contained low observable catalytic activity. One recurring theme in these studies was that substitution of lysine with arginine results in an enzyme with either low levels of activity (<3%) or the ability to interconvert enzyme intermediates. The most extensive study with a lysine-to-arginine mutant was with aspartate aminotransferase, where it was shown that arginine stabilized the key quinonoid intermediate (Toney and Kirsch, 1991). A second recurring theme in these studies is that when alanine is substituted for the active-site lysine, small amines aid in restoring some catalytic activity. For each of the pyridoxal-P enzymes, the low level of activity was interpreted as the activesite Lys amino group serving as the base to remove the a-proton of the substrate in forming a quinonoid intermediate. This view is supported by the three-dimensional structure of aspartate aminotransferase (Arnone et al., 1985;Jansonius and Vincent, 1987). Recently, studies with D-amino-acid transaminase suggested that a second lysine residue could partially substitute for the active-site lysine in the catalytic mechanism (Yoshimura et al., 1992).
Using site-directed mutagenesis, Lys-229 in E. coli SHMT was changed to a Gln, a His, or an Arg residue (K229Q, K229H, and K229R). The spectral and catalytic properties of these three mutant enzymes were studied. Evidence suggests that Lys-229 in SHMT is not the base that removes a proton from the a-carbon of glycine in its conversion to serine.

EXPERIMENTAL PROCEDURES
Materials-All amino acids, coenzymes, and buffers were purchased from Sigma. ~-[U-'~C]Serine (170 mCi/mmol) was purchased from Du  was Purchased from ICN Biochemicals. E. coli SHMT; K229Q, K229H, and m 2 9 R forms of SHMT and C1-tetrahydrofolate synthase were purified as described previously (Schirch et al., 1985;Villar et al., 19851, with the exception that 0.1 m~ pyridoxal-P was included at each stage of the purification of the mutant enzymes, except on the DEAE column. MUtant oligonucleotides for the construction of E. coli K229Q, K229H, and K229R expression vectors were synthesized on a n Applied Biosystem synthesizer by the phosphoramidite method and purified by thin-layer chromatography. ApoSHMT was prepared by using L-cysteine in high salt (Schirch et al., 19853. Site-directed Mutagenesis-Each mutant form of the glyA gene coding for SHMT was generated using the mutagenesis kit from Amersham Corp. The mutant oligonucleotides used to generate K229Q, K229H, and K229R were 21-mers as follows: 5'-CGC CAG GGT TTG GT-G AGT GGT-3', 5'-CGC CAG GGT GTG GTG AGT GGT-3', and 5'-CGC CAG GGT TCT GTG AGT GGT-3' (the underlined bases are the positions of the mismatches). Clones obtained after the mutagenesis procedure were screened by sequencing the gene at the mutated region. M13 clones carrying the mutated gene were purified, subcloned into pBR322, and used to transform E. coli AT2457 cells, which have a deletion of the glyA gene. For each mutant protein, the CNBr peptide containing the active-site sequence was sequenced by the method of Barra et al. (1991) to verify the correct replacement of k g , Gln, or His for the active-site Lys.
The rates of transamination of D-and L-alanine to pyruvate and pyridoxamine phosphate were determined from the first-order plots of the decrease in the absorbance of the external aldimine absorbing at 422 nm (Stover et al., 1992).
Preparation of ApoK229Q, K229Q,Gly, and K229Q.Ser Complexes -K229Q SHMT was added to a 200 mM solution of D-alanine in 50 mM KP;, 10 mM 2-mercaptoethanol, 200 m~ (NH,),SO,, pH 7.6, and incubated at 37 "C for 6-7 h. The enzyme solution was then rapidly desalted in a 5-ml syringe filled with 3 ml of Sephadex G-25 equilibrated with 50 mM KP,, 10 mM 2-mercaptoethanol, pH 7.6. Four-hundred pl were added to each column, followed by a 300-pl wash. The eluates were collected in a test tube after spinning the columns at 4000 rpm for 20 s in a swinging bucket centrifuge. Approximately 90% of the protein was recovered. The K229Q.Ser and K229Q.Gly complexes were prepared by incubating the K229Q apoenzyme with a 1.3-fold excess of pyridoxal-P and either 20 mM L-serine or glycine. After 1 h a t 23 "C, the holoenzyme-amino acid complexes were concentrated with a Centricon filtration system and stored at 4 "C.
Glycine and Serine Off-rates-The rate of dissociation of glycine from the K229Q.12-14C]Gly complex was determined by rapidly desalting on small Sephadex G-25 columns, as described above, 0.5 ml of 0.3 m~ K229Q.12-'4ClGly, 0.1 m~ pyridoxal-P, and 10 m~ [2-14C]Gly. The col-umns were equilibrated with 50 mM KP;, 20 mM nonradiolabeled Gly, 10 m~ 2-mercaptoethanol, pH 7.6. At timed intervals, 300-111 aliquots were removed and rapidly desalted on small Sephadex G-25 columns equilibrated with the same buffer. The concentration of enzyme in the eluate was determined from the pyridoxal-P content. Twenty-five p1 of 4 N NaOH were added to 1.0 ml of enzyme solution, and a spectrum was recorded. The 422 nm absorbing band, characteristic of the pyridoxal-P bound as an external aldimine, shifted to 388 nm, which is free pyridoxal-P. Using a molar absorptivity coefficient of 6550 M -~ Cm" at 388, the concentration of pyridoxal-P was determined ( H a d and Jenkins, 1976) and used as the concentration of enzyme active sites. The amount of radioactive glycine was determined by counting in a Packard Instrument liquid scintillation spectrometer. The off-rate for serine was determined by the same method, except the enzyme was a SHMT.[U-14C]Ser complex. The specific activity of the radiolabeled amino acids was 500 dpdnmol.
Determination of Amino Acids Bound to K229Q SHMT-The method of Kochhar and Christen (1992) was used to determine amino acids bound to the mutant forms of SHMT. Samples (50 1.11) of K229Q SHMT were added to a Sephadex G-25 column in a 1-ml plastic syringe. An additional 50 pl of buffer were added to the column. The syringe was placed in a test tube and centrifuged a t 2000 rpm for 2 min. The enzyme (-90% recovery) was isolated in the test tube in 370 p1.One-hundred pl of the sample were diluted to 800 p1 in 20 mM KP;, pH 7.3. The concentration of enzyme active sites was determined from the pyridoxal-P content, as described above. To the remaining 270 pl of K229Q SHMT were added 25 pl of 9 M perchloric acid to precipitate the protein and to free any bound amino acids. The precipitate was removed by centrifugation and washed with two 100-pl aliquots of water. To the combined supernatants were added 70 pl of 4.5 N KOH to precipitate excess perchlorate. Again, the precipitate was removed by centrifugation and washed twice with 50-pl aliquots of water. The combined supernatants were dried and redissolved in 500 pl of water. Small aliquots of this sample were then analyzed for amino acids. All aqueous solutions described in this procedure used high pressure liquid chromatographygrade water.
ation of K229Q SHMT were obtained with an MC-2 scanning calorim-Differential Scanning Calorimetry-Thermograms for the denatureter from Microcal Inc. (Amherst, M A ) . Each mutant SHMT and apoSHMT were dialyzed for 24 h against 20 m~ KP,, 20 mM L-serine, pH 7.3, and 10 m~ 2-mercaptoethanol prior to analysis. The scanning rate was 30 "Ch. Enthalpy of denaturation ( A H H , ) values were determined a s described previously (Schirch et al., 1991).
Circular Dichroism-Spectra were recorded from 260 to 200 nm using a Jasco Model 500C with a 0.1-cm cell containing 0.2 mg/ml protein in 20 m~ KP,, pH 7.3, a t 25 "C.

Spectral Properties of K229Q, K229R, and K229H Mutant
SHMTs-The spectrum of wild-type SHMT exhibits a single absorption maximum at 422 nm in addition to the protein band at 278 nm. The 422 nm band is due to the protonated internal aldimine of pyridoxal-P bound to Lys-229. The spectra of the three mutant enzymes, as isolated from a final hydroxylapatite column, are shown in Fig. 1. For K229Q SHMT, the spectrum is virtually identical to the spectrum of the wild-type enzyme, with absorption maxima at 278 and 422 nm and a 2781422 ratio Each sample was pretreated with NaCNBH, as described under "Experimental Procedures." of 7.6 (Schirch et al., 1985). Since K229Q SHMT has no activesite lysine residue to form an internal aldimine, the absorption peak a t 422 nm suggests that the enzyme is isolated as an external aldimine with a bound amino acid at the active site. The K229H enzyme exhibits absorption maxima at 278 and 352 nm. However, it also displayed increased absorption at 300 nm compared to the wild-type enzyme, probably due to bound pyridoxal-P. Dialysis of K229H SHMT with 0.2% SDS overnight did not result in significant loss of the absorption band at 352 nm. This suggests that the pyridoxal-P is bound as a covalent complex. The spectrum of K229R SHMT exhibits absorption maxima a t 330 and 390 nm. The 390 nm band is characteristic of the free aldehyde form of pyridoxal-P and suggests that in this mutant enzyme, the coenzyme is not bound as an external aldimine as observed with K229Q SHMT.
Saturation of wild-type SHMT with Gly results in three spectral absorption maxima a t 343, 425, and 495 nm, which have been shown to be complexes on the reaction path (Schirch et al., 1985). Addition of either 50 mM glycine or serine to K229Q SHMT did not change the spectrum of the purified enzyme. Addition of either serine or glycine to K229H SHMT also did not result in any change in its spectral properties. However, addition of either glycine or serine to K229R SHMT resulted in a shift of the absorption maxima from 390 nm to a spectrum with maxima at 406 and 332 nm. The 406 nm band was -25% lower than the 390 nm band, and the 332 nm band was -15% higher than the 330 nm band of the enzyme in the absence of amino acids as shown in Fig. 1. This suggests that this mutant enzyme can form complexes with these amino acids. Rapid removal of the excess amino acids by gel filtration in spin columns resulted in a return to the spectrum, shown in Fig. 1, within 30 s. Prolonged incubation with serine, however, rendered the enzyme inert to further changes, suggesting that some covalent modification had occurred.
Previous studies with SHMT have shown that addition of H,PteGlu, (or its 5-methyl and 5-formyl derivatives) to the SHMT.Gly complex shifts the equilibrium to the quinonoid complex absorbing near 500 nm, resulting in a 2-order magnitude increase in absorbance (Schirch et al., 1985). The addition of glycine and H4PteGlu to the three mutant enzymes did not produce this large increase in absorbance near 500 nm. Only K229Q SHMT showed any evidence of quinonoid formation. Even with this mutant enzyme, the absorbance after saturation with glycine and H4PteGlu was at least 2 orders of magnitude lower than with wild-type SHMT.
Another characteristic spectral property of wild-type SHMT is the shift of the 422 nm absorbing band to 325 nm after addition of NaCNBH3. This is the result of the rapid reduction of the internal aldimine to a secondary amine (Schirch et al., 1985). Complete reduction occurs in a few minutes. Addition of excess NaCNBH3 to the three mutant enzymes resulted in no significant change in spectral properties during a 1-h incubation. Prolonged incubation of K229Q SHMT suggested that a slow reduction was occurring by evidence of the partial disappearance of the 422 nm band and the concomitant appearance of a band a t 325 nm. However, after several hours, this reaction had resulted in only 10-15% change in these spectral bands.
The thermal stability of wild-type SHMT has been extensively studied using differential scanning calorimetry (Stover el al., 1992). These studies have shown that forming the serine external aldimine increases both the T, and A& of the enzyme. This is correlated with the enzyme going from an open conformation to a closed conformation on forming the external aldimine (Stover et al., 1992). We have also used differential scanning Calorimetry to determine the values for T, and A H d of K229Q SHMT.Ser to determine the effect of the mutation on the thermal stability of the enzyme. The results showed that the T, was 75.9 2 0.2 "C and the A H d was 320 2 40 kcalimol. Both of these parameters are only slightly lower than the observed values of 78 "C and 370 kcaVmol observed with the serine complex of wild-type SHMT. This suggests that the K229Q.Ser complex is in the closed form and that the activesite lysyl residue does not play a significant role in stabilizing the enzyme.
Both K229H and K229R SHMTs were also investigated by differential scanning calorimetry in the presence of serine. The thermal denaturation of both of these mutant proteins showed very broad thermal transitions with multiple thermal peaks. With respect to T,, the mutant proteins were intermediate between the wild-type T , of 58.3 2 0.1 "C for apoSHMT and 68 * 1 "C for holoSHMT without serine. The values of AZ& for the K229H and K229R mutants were within experimental error of the 200 kcaVmol for apoSHMT. These results suggest that solutions of these two mutant proteins consist of multiple forms with respect to tertiary structure. To analyze if there were differences in secondary structure, circular dichroism spectra were obtained for the mutant proteins and compared to the wild-type enzyme. These results showed that in the 200-240 nm regions, there were no differences in band shape or intensity of optical activity between the wild-type and the K229Q and K229H mutant proteins. K229R SHMT had the same band shape, but did show an -8% decrease in optical activity.
Catalytic Properties of K229Q, K229R, and K229H SHMTs -The ability of the three mutant enzymes to catalyze the conversion of serine and H4PteGlu to glycine and 5,10-CH2-H,PteGlu was tested.
For each mutant enzyme, a small amount of catalytic activity was found in different preparations, suggesting that the V, , , values of the mutant enzymes were anywhere from 0.05 to 2% of the activity of the wild-type enzyme. However, this activity rapidly disappeared when the mutant enzymes were treated with NaCNBH,. The rate of activity loss was the same as that observed with wild-type SHMT. The low level of catalytic activity was used to determine the K, values for L-serine and allo-threonine. These values were found to be the same as those for wild-type SHMT. These two observations suggest that the small amount of activity of the three mutant enzymes was the result of contamination with a nonmutant form of SHMT. This was further supported by the observation that different preparations resulted in different amounts of catalytic activity. Extended studies on the origin of this small amount of activity resulted in the conclusion that there may be reversion of the mutation at the active site to form the wild-type gene. This was most often found with the K229Q and K229H mutants. We could never obtain a preparation that was completely devoid of a small amount of catalytic activity without treating the enzyme with NaCNBH3. All subsequent studies were therefore performed with purified enzymes that had been pretreated with NaCNBH3, which reduced catalytic activity to <0.01%, but did not cause observable changes in the spectral properties of the enzyme. The only exception to this is that K229H SHMT did show some absorption at 422 nm, which disappeared with the NaCNBH3 treatment. The results of these kinetic studies suggest that none of the three mutant enzymes has steady-state catalytic activity.
All forms of SHMT that have been studied slowly catalyze the transamination of D-alanine and, to a lesser extent, L-alanine to pyruvate and pyridoxamine-P (Shostak and Schirch, 1988). The reaction results in the disappearance of the absorption band of bound pyridoxal-P a t 422 nm with the concomitant appearance of the absorption band of pyridoxamine-P at 325 nm. The pyridoxamine-P has a low affinity for the enzyme and can be removed by either dialysis or gel filtration. These slow transamination reactions are important because they occur by the enzyme proceeding through a normal catalytic pathway of removing the a-proton of the amino acid, but then replacing the proton on the 4"carbon of the coenzyme. For wild-type SHMT, the transamination reactions are first order and can be used to verify that all of the bound pyridoxal-P is reactive. Addition of 200 m M D-alanine, pH 7.6, to K229R and K229H SHMTs showed no evidence of transamination. These are the conditions used to determine the rate of transamination with wildtype SHMT, which has a K, for o-alanine of 30 mM.
Addition of 200 mM o-alanine to K229Q SHMT, pH 7.6, at 37 "C resulted in the biphasic disappearance of the absorption band at 422 nm and the appearance of a new peak at 325 nm. The complete spectral shift suggested that transamination had occurred. Gel filtration resulted in enzyme with no absorbing bands above 300 nm, indicating the formation of apoSHMT. Addition of pyridoxal-P to K229Q apoSHMT resulted in a slow spectral change to a form of enzyme exhibiting absorption maxima at 390 and 335 nm. Addition of either glycine or serine to this solution resulted in K229Q SHMT with a single band absorbing at either 418 or 424 nm, respectively, and exhibiting properties of the purified K229Q enzyme. Analysis of the biphasic rate of the disappearance of absorption at 422 nm during reaction with o-alanine suggested that the reaction could be described as two first-order reactions with rate constants of -1.6 x and 8.5 x min-I. The faster rate is -40% of the rate of transamination observed with wild-type SHMT (21.6 x For transamination to occur with o-alanine and K229Q SHMT, any amino acid bound as an external aldimine would have to be displaced to form the K229Q.o-Ala complex. The displacement of a bound substrate amino acid in pyridoxal-P enzymes lacking an active-site lysine has been shown to be very slow (Toney and Kirsch, 1992). To determine if the slower rate of transamination of o-alanine by K229Q SHMT was determined by the rate of a bound amino acid being displaced rather than the rate of transamination of o-alanine, the purified enzyme was analyzed for bound amino acids. Purified K229Q SHMT was denatured with perchloric acid, and the supernatant was analyzed for released amino acids. Only serine and glycine were found, suggesting that the purified enzyme contained -0.7 eq of bound serine and 0.3 eq of glycine (Table I).
The ability of o-alanine t o convert the holoenzyme to apoenzyme was used to make K229Q SHMT with either only L-serine or glycine bound at the active site (see "Experimental Procedures"). These two forms of the enzyme were reinvestigated for their ability to catalyze the transamination of o-alanine. In each case, the spectral changes could be described by a single first-order reaction with rate constants of 2.9 x m i x ' for the K229Q.Ser complex and 8.9 x min" for the K229Q.Gly complex. This difference in rate between the enzyme-serine and enzyme-glycine complexes accounts for the biphasic rate observed with the purified enzyme since it is a mixture of the serine and glycine complexes.
The ability of K229Q SHMT to transaminate L-alanine was also tested. The wild-type enzyme catalyzes the transamination of this amino acid a t a rate that is about one-fourth the rate of o-alanine (Shostak and Schirch, 1988). However, with  (Shostak and Schirch, 1988).

Enzyme form Serine Glycine eq eq
Purified K229Q SHMT 0.7 K229Q SHMT.Ser K229Q.Gly, spectral changes occurring in the presence of 200 m M L-alanine suggested that a significantly slower transamination was occurring. After 7 h, the absorbance at 422 nm had decreased by only 16%, with a slightly larger increase in absorbance at 325 nm. Normally, the decrease in absorbance at 422 nm is nearly equivalent t o the increase in absorbance at 325 nm since these two forms of the coenzyme have similar molar absorptivity coefficients. Longer incubations could not be followed because of slow precipitation of the enzyme. However, assuming that these spectral changes are the result of the transamination of L-alanine, the rate constant would be -0.4 x min-l, which is 15-fold slower than the rate of transamination of L-alanine found for wild-type SHMT.
Off-rates for Glycine and Serine-Transamination of D-alanine can occur only if the bound serine and glycine can be displaced. The difference in the rates of transamination between the K229Q.Ser and K229Q.Gly complexes suggests that the off-rates for serine and glycine may be different and that for serine, its off-rate may be partially rate-determining in the transamination of o-alanine. To determine the rate at which serine and glycine can be displaced from the active site, apoenzyme was made, and pyridoxal-P was added in the presence of either [ l4C1serine or [ 14Clglycine. After removal of the excess labeled serine and glycine, the enzyme was incubated with an excess of either unlabeled serine or glycine, and the amount of radioactivity remaining on the enzyme was determined with time (see "Experimental Procedures"). The rate of loss of serine was 6.9 2 0.2 x min", and that of glycine was 27 2 4 x min-l (Fig. 2). The slow rate of loss of serine is close to the rate of transamination of D-alanine by the K229Q.Ser complex, suggesting that the rate-determining step in transamination is determined in part by the rate of dissociation of serine.
The conversion of the external aldimine of either the K229Q.Ser or K229Q.Gly complex to the K229Q.o-Ala external aldimine must go either through a gem-diamine complex with both amino acids bound to pyridoxal-P or through an intermediate species involving pyridoxal-P bound as a free aldehyde. Others have found with aminotransferases that buffer amines can catalyze transimination (Toney and Kirsch, 1992; Ilag and P4C1Gly or [I4C1Ser in 50 mM KP,, 10 mM 2-mercaptoethanol, 0.1 rn pyridoxal-P, and a 10 mM concentration of the respective radiolabeled amino acid, pH 7.6, was rapidly desalted into 50 mM KP,, 10 m~ 2-mercaptoethanol, and a 20 mM concentration of the respective nonradiolabeled amino acid. During incubation a t 37 "C, aliquots were removed at timed intervals, and the ratio of radiolabeled amino acid to enzyme was determined a s described under "Experimental Procedures." B , the offrate for serine was determined as described above with either 100 m M Na2S04 ( 0 ) or (NH,),SO, (0) included in the incubation buffer. Jahn, 1992;Nishimura et al., 1991;Yoshimura et aZ., 1992). We tested the effect of ammonia on the rate of release of serine from K229Q.Ser in the presence of either 100 m M Na2S04 or (NH&S04. As shown in Fig. 2, the rate of release of [14C]serine from the enzyme complex in the presence of the ammonium salt is three times the rate observed with the sodium salt. This suggests that small amines can catalyze the release of amino acids from the external aldimine. Single 7hrnover of K229Q.Gly and K229QSer"The ability of K229Q SHMT to catalyze the transamination of D-alanine suggests that this mutant enzyme can form the key quinonoid complex in which the @-proton of o-alanine is removed by a base on the enzyme. Since we have postulated that this is the same base that removes the pro-2s proton of glycine in the interconversion of glycine and serine, it suggests that the enzyme should also be able to interconvert these 2 amino acids. The lack of observable catalytic activity of K229Q SHMT may not be due to its inability to catalyze the interconversion of serine and glycine, but may be due to its inability to release the product (Fig. 3). To determine if the enzyme can catalyze a single turnover of the interconversion of serine and glycine, we looked at the ability of substrate levels of K229QeSer and K229Q.Gly to form their respective products. When 15-20 nmol of the serine complex were first incubated with H4PteGlu and then analyzed for bound amino acids, it was found that nearly all of the serine had disappeared and that a slightly greater than equivalent amount of glycine had appeared ( Table I). The greater than equivalent amount of glycine could be the result of some contamination of glycine in the various reagents since controls of this study always showed varying amounts of glycine.
When the K229Q.Gly complex was incubated with 5,10-CH2-H4PteGlu and analyzed for amino acids, only -0.2 eq of glycine had been converted to serine. This may be due to the equilib- rium constant for this reaction, which favors the formation of glycine over serine by a factor of 10 (Schirch et al., 1977). To determine if it was an equilibrium problem, we coupled the reaction to the conversion of the product H4PteGlu3 to 10-CHO-H4PteGlu3 by adding formate, MgATP, and the enzyme 10formyltetrahydrofolate synthetase, which is one activity of the trifunctional enzyme C1-tetrahydrofolate synthase (Strong et al., 1987). By removing the product H4PteGlu3, we were able to convert nearly all of the bound glycine to serine (Table I). Previous studies have shown that this assay works only if a polyglutamate form of the folate is used (Strong and Schirch, 1989). We also found that to be true in this study. Using H,PteGlu instead of H4PteGlu3 resulted in only -20% of the glycine being converted to serine. The rate at which K229Q,Ser was converted to K229Q.Gly and 5,10-CH2-H4PteGlu was determined by adding an excess of the enzyme methylenetetrahydrofolate dehydrogenase-cyclohydrolase and NADP+, which converts the CHz-H4PteGlu to 10-CHO-H4PteGlu and NADPH (Strong et al., 1987). The rate of the reaction was monitored at 340 nm. The results showed that 1.0 eq of NADPWeq of K229Q.Ser was formed in this reaction (Fig. 3). A trace of the absorbance change a t 340 nm at limiting dehydrogenase-cyclohydrolase concentration is shown in Fig. 3 A . The amplitude of the absorbance change was proportional to the amount of K229Q.Ser used in the reaction. To determine the rate of conversion of K229Q.Ser to K229Q.Gly, increasing amounts of the dehydrogenase-cyclohydrolase enzyme were added until no further increase in rate was observed. The results are shown in Table I1 and suggest that the rate of conversion of K229Q.Ser to K229Q.Gly is 0.21 s-', which is -2% of the value of Keat for wild-type SHMT in the conversion of serine to glycine (Stover et al., 1992).
The rate at which K229Q.Gly and 5,lO-CH2-H4PteGlu3 are converted to K229Q.Ser was investigated by coupling the conversion of the product H4PteGlu3 back to 5,10-CH2-H4PteGlu3 by the addition of formate, MgATP, NADPH, and the trifunctional enzyme C1-tetrahydrofolate synthase. This is the same system used to determine the conversion of bound glycine to bound serine in the amino acid analysis experiments described above. The first reaction in the coupled assay is the conversion of H,PteGlu, to 10-CHO-H4PteGlu3 by the synthetase activity of the trifunctional enzyme. The kcat for this reaction is an order of magnitude slower than the rate of conversion of glycine to serine by wild-type SHMT (Strong and Schirch, 1989). Fig.  3B shows a stopped-flow spectrophotometer trace at 340 nm after flowing K229Q SHMT.GIy against 5,10-CH2-H4PteGlu3 and the other components of the coupled enzyme system. In this experiment, the C1-tetrahydrofolate synthase was ratedetermining. The rate of decrease in absorbance at 340 nm was first-order. The magnitude of the absorbance change at 340 nm was proportional to the concentration of K229Q SHMT.GIy. A linear rate of decrease in absorbance followed the first-order burst and was also observed in the absence of K229Q.Gly. The results reported in Fig. 3   change at 340 nm being the result of 0.9 eq of K229Q SHMT.Gly being converted to K229Q SHMT.Ser. To determine the rate of this reaction, increasing amounts of C1-tetrahydrofolate synthase were added to show that the rate was the result of the conversion of glycine to serine and not the conversion of the product H4PteGlu3 to 10-CHO-H4F'teGlu3 by the synthetase reaction. As shown in Table 11, the rate of decrease in absorbance at 340 nm was dependent on the amount of C1tetrahydrofolate synthase that was added. We were unable to add enough of this enzyme to make the SHMT-catalyzed reaction rate-determining. The greatest value of the first-order rate constant determined suggests that K229Q.Gly is converted to K229Q.Ser with a rate that may be as large as the value of kcat for wild-type SHMT (Stover et al., 1992).

DISCUSSION
In aspartate aminotransferase, considerable evidence indicates that the active-site Lys plays a dual role in the mechanism of the reaction (Toney and Kirsch, 1991). First, it is the base that accepts and donates the proton in the 1,3-prototrophic shift in the interconversion of the aldimine and ketimine intermediates. Second, it also is required to expel the product amino acid from the external aldimine intermediate by forming an internal aldimine (Toney and Kirsch, 1992). Several other pyridoxal-P enzymes have now been studied by site-directed mutagenesis with respect to the function of the active-site Lys (Bhatia et al., 1993;Lu et al., 1993;Ilag and Jahn, 1992;Yoshimura et al., 1992;Grimm et al., 1992;Kirsch, 1991, 1992;Nishimura et al., 1991;Planas and Kirsch, 1991;Smith et al., 1989). Some of these mutant enzymes can slowly interconvert the aldimine and ketimine intermediates. However, in most cases, it appears that the active-site Lys residue is also the base that accepts the a-proton from the amino acid substrate. In each of these enzymes, expulsion of the product amino acid external aldimine is very slow, as found with aspartate aminotransferase.
Scheme I shows the reactions that appear to be involved with K229Q SHMT. The purified enzyme is a mixture of external aldimines of the two amino acid substrates glycine and serine (structures I and 111). A key intermediate in the interconversion of these two complexes is the quinonoid complex shown as structure 11. The interconversion of structures I1 and I11 involves a proton transfer between a base on the enzyme and the a-carbon of glycine. The failure to observe a strong absorbance of the ternary complex K229Q.Gly.H4PteGlu at 500 nm, as observed with wild-type SHMT, suggests that Lys-229 is the base involved in this proton abstraction. However, there are several experiments that indicate that this conclusion is wrong. First, K229Q SHMT catalyzes the transamination of D-alanine at 40% of the rate of wild-type SHMT (when starting with the glycine complex). The transamination of D-alanine involves forming the same quinonoid complex (structure 11) with subsequent placement of the proton at the 4"carbon of pyridoxal-P (structure IV). With the wild-type enzyme, the transamination of D-alanine is accompanied by a significant absorption band at 502 nm, which shows that structure I1 accumulates during the reaction. With K229Q SHMT, there is no observable absorbance at 502 nm, suggesting that the equilibrium between structures I1 and I11 favors structure 111 in this mutant.
A second experiment that suggests that Lys-229 is not the base involved in the interconversion of structures 111 and I is that when 5,10-CH2-H4PteGlu is added to the K229Q.Gly cornplex, it is converted to the K229Q.Ser complex (Table I). We could not determine a true rate for this reaction because ofthe low kcat value for the coupling enzyme. However, it appears that the rate in this direction at least approaches the kcat value for wild-type SHMT. The rate-determining step in the intercon- version of glycine and serine by wild-type SHMT is not known, so a direct comparison of the rates of the interconversion of structures I11 and I is not possible. However, the rate of conversion of Gly to Ser is sufficiently fast for K229Q SHMT that Lys-229 cannot be the base involved in this reaction. K229Q SHMT also catalyzes the conversion of the bound serine to glycine when H4PteGlu is added ( Table I). The rate of this conversion is 2% of the kcat value for the wild-type enzyme (Table 11). If Lys-229 was the base accepting the proton in the conversion of glycine to serine, then it must be the acid that donates the proton in the conversion of the quinonoid complex to the glycine external aldimine. Even when the rate is only 2% of the wild-type rate, it would be difficult to argue that Lys-229 is the acid involved as the proton donor.
The second role played by the active-site Lys residue in aspartate aminotransferase appears also to be a role in SHMT. The observation that K229Q SHMT is purified with a mixture of serine and glycine bound at the active site suggests that these amino acids cannot be expelled readily by the mutant enzyme. Indeed, the off-rates of 6.9 x and 27 x min-l are similar to the slow release of the amino acid substrates found with aspartate aminotransferase Kirsch, 1991, 1992). Small amines were found to accelerate the off-rate for amino acid products with aspartate aminotransferase. Likewise, we have found that 200 l~l~ ammonia at pH 7.6 gives a n -3-fold increase in the rate of release of serine. This may be due to the ability of ammonia to form an imine by the mechanism shown in Scheme I (structures VI and VI). An unanswered question is if the ammonia attacks the external aldimine from the same or opposite face of the 4"carbon as the active-site lysine.
We have previously shown that in the transamination of D-alanine and L-alanine by wild-type SHMT, different bases on the enzyme were involved in the removal of the (2S)-proton of Serine Hydroxymethyltrunsferuse D-alanine and the (2R)-proton of L-alanine (Shostak and Schirch, 1988). In aspartate aminotransferase, it is the R-protons of aspartate and glutamate that are removed by the activesite lysine. Since it appears that the same face of the pyridoxal-P ring is facing the solvent in SHMT as in aspartate aminotransferase, one might expect that the expelled Lys-229 would be on the side of the (2R)-proton of L-alanine (Dunathan and Voet, 1974). It would then be in a position to be a base to remove the (2R)-proton. In support of this conclusion is the observation that L-alanine is transaminated extremely slowly by K229Q SHMT. However, because there appeared to be a very slow rate of transamination, we cannot conclude without reservation that the base that removes the (%)-proton of L-alanine is Lys-229.
Substitution of Lys-229 with Arg results in the isolation of an enzyme with at least part of the pyridoxal-P bound as the free aldehyde (Fig. 1). The addition of serine or glycine does convert some of the bound pyridoxal-P to the external aldimine. However, the affinity for serine must be low because as soon as the serine is removed, the external aldimine is converted in a few seconds back to the aldehyde form. The substitution of Arg for Lys-258 in aspartate aminotransferase results in an enzyme form that can form the quinonoid intermediate (Toney and Kirsch, 1991). The observation that K229R SHMT cannot even form a stable external aldimine suggests that the active site does not tolerate a positive charge next to the imine-positive charge. This may further substantiate that Lys-229 is not the active-site base and that it never goes through a positive charged intermediate, as required for the aspartate aminotransferase mechanism.