Identification of the Folate Binding Sites on theEscherichia coli T-protein of the Glycine Cleavage System*

T-protein is a component of the glycine cleavage system and catalyzes the tetrahydrofolate-dependent reaction. To determine the folate-binding site on the enzyme,14C-labeled methylenetetrahydropteroyltetraglutamate (5,10-CH2-H4PteGlu4) was enzymatically synthesized from methylenetetrahydrofolate (5,10-CH2-H4folate) and [U-14C]glutamic acid and subjected to cross-linking with the recombinant Escherichia coli T-protein using 1-ethyl-3-(3-dimethylaminopropyl)carbodiimide, a zero-length cross-linker between amino and carboxyl groups. The cross-linked product was digested with lysylendopeptidase, and the resulting peptides were separated by reversed-phase high performance liquid chromatography. Amino acid sequencing of the labeled peptides revealed that three lysine residues at positions 78, 81, and 352 were involved in the cross-linking with polyglutamate moiety of 5,10-CH2-H4PteGlu4. The comparable experiment with 5,10-CH2-H4folate revealed that Lys-81 and Lys-352 were also involved in cross-linking with the monoglutamate form. Mutants with single or multiple replacement(s) of these lysine residues to glutamic acid were constructed by site-directed mutagenesis and subjected to kinetic analysis. The single mutation of Lys-352 caused similar increase (2-fold) inK m values for both folate substrates, but that of Lys-81 affected greatly the K m value for 5,10-CH2-H4PteGlu4 rather than for 5,10-CH2-H4folate. It is postulated that Lys-352 may serve as the primary binding site to α-carboxyl group of the first glutamate residue nearest the p-aminobenzoic acid ring of 5,10-CH2-H4folate and 5,10-CH2-H4PteGlu4, whereas Lys-81 may play a key role to hold the second glutamate residue through binding to α-carboxyl group of the second glutamate residue.

bound to the lipoate cofactor of H-protein (reviewed in Ref. 1). T-protein should have the interaction or binding site(s) with H-protein, the folate cofactor, and possibly P-protein and the active site that directly participates in the cleavage of the aminomethyl intermediate. While the kinetic properties of the forward (2) and the reverse (3) reactions catalyzed by T-protein have been extensively studied, little is known in terms of specific amino acid residues of T-protein involved in binding or catalysis.
The primary structures of T-proteins from eight different species have been determined so far (4 -10). The sequence alignment by computer program CLUSTAL W (11) indicates that there are relatively high identities between T-proteins from plants and animals (46 -88%), but there are less identities between animal enzymes and the yeast and Escherichia coli enzymes (ϳ30%). Thus, the regions well conserved among the eight species are very limited (see Fig. 1). A partial sequence homology of T-proteins to ␣ subunit of corynebacterial sarcosine oxidase and rat dimethylglycine dehydrogenase has been reported (12). The homologous region was considered to be the tetrahydrofolate binding domain, since all these enzymes catalyze the formation of 5,10-CH 2 -H 4 folate from different onecarbon donors (12). Physiological folate substrates generally have a string of glutamate residues attached to the benzoyl moiety of the cofactor, and the polyglutamylation enhances the cofactor binding to folate-dependent enzymes (13)(14)(15)(16). The interaction between the polyglutamate moiety and protein has been extensively investigated in thymidylate synthase (TS) that uses 5,10-CH 2 -H 4 folate as a substrate. Crystal structure analysis revealed that TS binds the polyglutamate moiety with a positively charged hydrophilic surface, where only the first glutamate residue (Glu-1) nearest the p-aminobenzoic acid ring is rigidly fixed. The lysine residue at the position 50 of E. coli TS (the numbering of amino acid residues corresponds to that of Lactobacillus casei TS), which is nearly invariant among all TS enzymes, participates in the binding to ␣-carboxyl group of Glu-1 (13).
As part of an overall effort to identify the domains of Tprotein responsible for cofactor recognition, complex association, and catalysis, we investigated specific lysine residues involved in the binding with the glutamyl tail of folate derivatives. In the present study, we utilized a carbodiimide-mediated covalent attachment of 14 C-labeled 5,10-CH 2 -H 4 PteGlu 4 to lysine residues located at or near the folate-binding site of E. coli T-protein. Three lysine residues were identified, and their role was confirmed by replacing these residues with glutamic acid by site-directed mutagenesis. 14 C]glutamic acid (9.29 GBq/mmol) and NaH 14 CO 3 (2.0 GBq/mmol) were obtained from Amersham Pharmacia Biotech. Restriction endonucleases and other DNA modifying enzymes were purchased from New England Biolabs, Roche Molecular Biochemicals, Toyobo (Tokyo, Japan), or Takara Shuzo (Shiga, Japan). Oligonucleotides were from Life Technologies, Inc. Folic acid, lysylendopeptidase, and N-hydroxysuccinimide were purchased from Wako Pure Chemicals (Osaka, Japan) and EDC from Sigma. Pteroyltetra-␥-glutamate was obtained from Schircks Laboratories (Jona, Switzerland). V8 protease was obtained from Roche Molecular Biochemicals. Folic acid and pteroyltetra-␥-glutamate were reduced (17) and used for synthesis of 5,10-CH 2 -H 4 folate and 5,10-CH 2 -H 4 PteGlu 4 , respectively, as described (18).

Materials-[U-
Expression and Purification of Lactobacillus casei Folylpolyglutamate Synthetase-Plasmid pGT3-8.1, a pEMBL vector carrying the L. casei FPGS chromosomal gene (19) kindly supplied by Dr. A. L. Bognar of University of Toronto, was transformed into E. coli MV1190. The bacteria were grown at 37°C in 1000 ml of LB medium (20) containing 100 g/ml ampicillin. At the late log phase, the cells were harvested by centrifugation (8,000 ϫ g, 10 min), resuspended in 50 ml of 50 mM potassium phosphate buffer, pH 7.0, containing 50 mM KCl, and sonicated at 0°C four times for 5 min using a Branson Sonifier 250 with a 1/2-inch disrupter horn in the pulse mode (output 8, 20% duty cycle). The sonicate was centrifuged at 105,000 ϫ g for 60 min, and the supernatant was made 70% ammonium sulfate saturation. The precipitate was resuspended in 20 ml of buffer A (50 mM potassium phosphate, pH 6.8) and extensively dialyzed against buffer A. The sample was applied to a phosphocellulose column (P11, 2.5 ϫ 8 cm, Whatman) equilibrated with buffer A. The column was washed with 10 column volumes of buffer A and eluted with a linear gradient of 0 -500 mM KCl (600 ml) in buffer A. The fractions containing the FPGS activity were pooled and concentrated by 70% ammonium sulfate precipitation. The pellet was resuspended in 3 ml of buffer B (50 mM potassium phosphate, pH 7.0, containing 200 mM KCl), applied to a Sephacryl S-100 column (2.6 ϫ 90 cm, Amersham Pharmacia Biotech) equilibrated with buffer B, and eluted with buffer B. Fractions containing FPGS were concentrated with a Centriprep-10 concentrator (Grace Japan, Tokyo, Japan) and stored at Ϫ20°C. FPGS activity was measured by the incorporation of [U-14 C]glutamic acid into 5,10-CH 2 -H 4 folate essentially as described by Bognar  in a total volume of 4 ml. The mixture was incubated at 37°C for 3 h, and then protein ingredients were removed by filtration with a Centricon-10 concentrator (Grace Japan). The filtrate was divided into three equal portions, and a portion was applied to a reversed-phase C 18 column (TSKgel ODS-120T, 4.6 ϫ 250 mm, Tosoh, Tokyo, Japan) attached to a 655 HPLC system (Hitachi, Tokyo, Japan). The column was developed with stepwise gradients of acetonitrile using solution A (50 mM ammonium acetate, pH 7.1) and solution B (10% acetonitrile in solution A) at a flow rate of 1 ml/min. The absorbance of the effluent was monitored at 295 nm. The collected peak fractions were immediately stored at Ϫ80°C in small aliquots until use. The samples were lyophilized to dryness with a Speed Vac concentrator (Savant) and resolved in water. After the determination of concentration, the material was immediately used for cross-linking with ET. Concentrations of the folate derivative were determined spectrophotometrically using ⑀ 295 ϭ 32.6 ϫ 10 6 cm 2 / mol (22). The material was also analyzed by scintillation counting, spectrophotometry, and matrix-assisted laser desorption ionization/ time-of-flight mass spectrometry.
Expression and Purification of Recombinant E. coli T-protein-DNA manipulations were accomplished by standard techniques (20). The pEGCV(Ϫ)5 plasmid containing the gcv operon (7) was digested with EcoRV and self-ligated. The product, pTZ/ETEH, contains gcvT and gcvH genes and a part of gcvP gene. An NdeI site was introduced adjacent to the codon for N-terminal Ala of gcvT gene in pTZ/ETEH by oligonucleotide-directed mutagenesis with the sense oligonucleotide 5Ј NdeI site originally present in the coding region without altering the amino acid were simultaneously carried out with sense oligonucleotides, 5Ј-TTGGAGATTGATGGATCCGCAACGTACCAG-3Ј (the BamHI site is underlined and modified bases are shown in boldface letters) and 5Ј-GATGTGTCACACATGACCATCGT-3Ј (the modified base is shown in a boldface letter), respectively. The resulted plasmid (pTZ/ETEHnb) was digested with NdeI and BamHI, and the NdeI-BamHI fragment was cloned into pET3a (24). The nucleotide sequence of the resultant expression vector pET/ET was confirmed by nucleotide sequencing employing a 373 DNA sequencing system (Perkin-Elmer).
E. coli BL21(DE3)pLysS cells (24) transformed with pET/ET were grown in 200 ml of LB medium containing 20 g/ml ampicillin and 30 g/ml chloramphenicol. Expression was induced by the addition of 25 M isopropyl-␤-D-thiogalactopyranoside at the start of the culture, and cells were incubated at 30°C for 24 h.
Purification of recombinant ET was carried out as described for the native ET (7) with some modifications. The cell-free extracts were prepared with buffer C (20 mM Tris-HCl, pH 8.0, 2 mM EDTA, 1 mM DTT, 10 M p-amidinophenylmethanesulfonyl fluoride) and subjected to column chromatographies on DEAE-Sepharose CL-6B (2 ϫ 6 cm, Amersham Pharmacia Biotech), hydroxyapatite (2 ϫ 6 cm), and Sephacryl S-100 (2.6 ϫ 90 cm). The elution from the DEAE-Sepharose CL-6B column was performed with a linear gradient of NaCl (0.1-0.3 M in 200 ml of buffer C). The hydroxyapatite column was equilibrated with 5 mM potassium phosphate, pH 7.0, containing 0.5 mM DTT, and the elution was performed by a linear gradient of potassium phosphate (5-100 mM, pH 7.0, in 200 ml) containing 0.5 mM DTT. The Sephacryl S-100 column was developed with 50 mM potassium phosphate, pH 7.0, containing 0.2 M NaCl. The final preparations were pooled, dialyzed against 20 mM potassium phosphate, pH 7.0, concentrated with a Centriprep-10 concentrator and stored at Ϫ20°C.
Purification of Recombinant E. coli P-protein-Recombinant E. coli P-protein was purified from E. coli MV1190 cells transfected with pEGCV(ϩ)5 (7) grown at 30°C for 24 h in 1000 ml of LB medium containing 50 g/ml ampicillin and 25 M isopropyl-␤-D-thiogalactopyranoside. The purification was performed according to the method described previously (7), except that the step with butyl-agarose column chromatography was omitted. The final preparation was more than 90% pure, as judged by SDS-PAGE. P-protein was assayed as described previously (7).
Expression and Purification of E. coli H-protein-For expression of recombinant EH, gcvH gene with an NdeI site adjacent to the codon for the N-terminal Ser residue and a BamHI site 27 bases downstream from the stop codon were amplified by polymerase chain reaction using pEGCV(Ϫ)5 (7) as a template. The following pair of primers was used: a 5Ј-end primer, 5Ј-ACATATGAGCAACGTACCAGCAGAAC-3Ј (the NdeI site is underlined and modified bases are shown in boldface letters); a 3Ј-end primer, 5Ј-CGGATCCTCTCCCGCAGAAGAGG-3Ј (the BamHI site is underlined and the modified base is shown in a boldface letter). The product was purified with a Geneclean II kit (BIO 101) after agarose gel electrophoresis and digested with NdeI and BamHI. The resultant DNA fragment was cloned into pET3a. The nucleotide sequence of the expression vector, pEH, was verified by DNA sequencing.
E. coli BL21(DE3)pLysS cells transfected with pEH were grown in 800 ml of LB medium containing 20 g/ml ampicillin, 30 g/ml chloramphenicol, and 150 M lipoic acid at 30°C for 24 h. The expression of EH was induced by 25 M isopropyl-␤-D-thiogalactopyranoside added at the start of the incubation. Cell-free extracts were prepared as described previously (7) with buffer D (20 mM Tris-HCl, pH 8.0, 2 mM EDTA, 10 M p-amidinophenylmethanesulfonyl fluoride) and heated in a boiling water bath for 2 min. The supernatant of the heated sample was subjected to column chromatographies on DEAE-Sepharose CL-6B (2.5 ϫ 10 cm), hydroxyapatite (2 ϫ8 cm), and Sephacryl S-100 columns (2.6 ϫ 90 cm) essentially as described for the purification of wild-type ET with minor modifications. The elution from the DEAE-Sepharose column was performed with a linear gradient of NaCl (0.28 -0.45 M in 500 ml of buffer D). The buffer employed in the hydroxyapatite column chromatography contained no DTT, and the Centriprep-10 concentrator was replaced by a Centriprep-3 concentrator. The mobility of EH in crude extracts and in the sample from each purification step corresponded to that of lipoylated holoH-protein on SDS-PAGE, indicating that the condition described above led to the full lipoylation of overexpressed EH. The final preparation was more than 95% pure as judged by SDS-PAGE. The activity of H-protein was assayed as described (7).
Location of Polyglutamate Binding Sites on E. coli T-protein-Recombinant wild-type ET (20 nmol) was incubated for 15 min on ice with either 1.2 mM 5,10-CH 2 -H 4 folate or 0.3 mM 14 C-labeled (3.2 ϫ 10 4 dpm/nmol) or non-labeled 5,10-CH 2 -H 4 PteGlu 4 in 50 mM HEPES buffer, pH 7.0, and then treated with 5 mM each of EDC and N-hydroxysuccinimide at 25°C for 1 h. Wild-type ET alone was treated with the cross-linkers as a control. The reaction was quenched by adding 1.0 M Tris-HCl, pH 8.0, to the final concentration of 50 mM, and the reaction mixture was concentrated with a Microcon-10 concentrator (Grace Japan). The material was loaded onto a Superdex 200 HR 10/30 column (Amersham Pharmacia Biotech) equilibrated with 50 mM Tris-HCl, pH 8.0, 0.2 M NaCl, and eluted with the same solution. The resultant cross-linked products were carboxymethylated and digested with lysylendopeptidase as described previously (5). The digested mixture was applied to a reversed-phase C 4 column (RP-304, 4.6 ϫ 250 mm, Bio-Rad), and the column was eluted with four stage gradients of acetonitrile in 0.1% trifluoroacetic acid at a flow rate of 1 ml/min using a Hitachi 655 HPLC system. Absorbance was monitored at 220 nm. Radioactivity, when used, was monitored along with the absorbance at 220 nm. Peptides were analyzed by N-terminal sequencing using Perkin-Elmer 477A and Hewlett Packard G100A protein sequencers. When necessary, cross-linked peptides were digested with Staphylococcus aureus V8 protease (1/18 -1/10, mol/mol) at 25°C for 18 h in 100 l of 50 mM ammonium bicarbonate, pH 7.8, 5% acetonitrile and separated by HPLC as described above but with a gradient of 0 -48% acetonitrile in 0.1% trifluoroacetic acid.
Construction, Expression, and Purification of Mutant E. coli T-proteins-Construction of some mutant ETs (K352E, K352Q, K352R, K75E, K81E, and K75E/K78E/K81E) was performed by in vitro mutagenesis according to the method of Kunkel et al. (23). For this purpose, pTZ/ETEHnb was trimmed by digestion with BamHI and selfligated. The resulted plasmid was employed in the mutagenesis reaction with oligonucleotides listed in Table I (25). For the first round of polymerase chain reaction, pET/ET was used as a template with mutagenic reverse primer in Table I and T7 promoter primer, 5Ј-TAATACGACTCACTAT- a Modified bases are shown in boldface letters. b These oligonucleotides were used for in vitro mutagenesis according to the method of Kunkel et al. (23). c These oligonucleotides were employed as mutagenic reverse primer for the first round of polymerase chain reaction in the megaprimer method (25).
AGGG-3Ј. The double-strand polymerase chain reaction products were gel-purified and used as megaprimers in the second round of polymerase chain reaction along with T7 terminator primer, 5Ј-GCTAGTTAT-TGCTCAGCGGTG-3Ј, using either pET/ET (for K78E and K78E/K81E) or the expression plasmid for K352E (for K78E/K352E and K78E/K81E/ K352E) as a template. Amplified products were purified, digested with NdeI and BamHI, and ligated into pET3a. After verification of the mutations by DNA sequencing, the expression plasmids were transfected into E. coli BL21(DE3)pLysS cells. Expression and purification of mutant ETs were carried out as described for the recombinant wild-type ET.
Kinetic Analysis-Apparent K m values of mutant ETs for 5,10-CH 2 -H 4 folate and 5,10-CH 2 -H 4 PteGlu 4 were determined at varied concentrations of either folate substrates with a constant concentration of other two substrates using the "coupled assay" (3) with some modifications. The reaction mixture contained in 0. 25  Mass Spectrometry-The mass spectroscopic analysis of folylpolyglutamates separated by a reversed-phase C 18 -column was performed by matrix-assisted laser desorption ionization/time-of-flight mass spectrometry using a Voyager RP mass spectrometer (PerSeptive Biosystems). The samples containing ϳ3.6 pmol of folate were employed and 1% ␣-cyano-4-hydroxycinnamic acid in 33% acetonitrile containing 0.1% trifluoroacetic acid was employed as a matrix. Des-Arg 1 -bradykinin (904.04 ϩ 1 Da) and potassium ion (39.00 Da) were used as external standards.
Other Methods-Protein concentration was routinely determined by the method of Bradford (26) with bovine serum albumin as a standard. Concentrations of purified wild-type and mutant ETs and EH were also estimated by amino acid analysis using a Hitachi 830 amino acid analyzer. Comparison of the data obtained by Bradford method and the amino acid analysis revealed that the amounts of proteins estimated with Bradford method were somewhat greater than those obtained by amino acid analysis. Concentrations of ET and EH can be calculated from the data obtained by Bradford method with multiplying 0.7 for ET and 0.9 for EH. SDS-PAGE was carried out by the method of Laemmli (27). 14 C-Labeled 5,10-CH 2 -H 4 PteGlu 4 was enzymatically synthesized from 5,10-CH 2 -H 4 folate and L-[U-14 C]glutamic acid using purified recombinant L. casei FPGS, and the products were separated by HPLC on a reversed-phase C 18 column. Formaldehyde was added to the reaction mixture to prevent the release of methylene-carbon from the folate substrate during incubation. The products and substrate were well separated by stepwise gradients elution of acetonitrile in 50 mM ammonium acetate, pH 7.1 (Fig. 2). Four peak fraction showing characteristic absorption spectra of 5,10-CH 2 -H 4 folate with maximum at around 295 nm were obtained, and their properties were summarized in Table  II. Molecular masses of P1 and P2 were in good agreement with that of 5,10-CH 2 -H 4 PteGlu 4 (844.298 ϩ 1 Da), whereas those of P3 and P4 were consistent with that of 5,10-CH 2 -H 4 folate (457.171 ϩ 1 Da). These results agree well with the amounts of the incorporated radioactive glutamate residues. 5,10-CH 2 -H 4 PteGlu 4 prepared from chemically synthesized pteroyltetra-␥-glutamate gave the same retention time as P2 (data not shown). P1 and P2 gave the same cross-linked lysylpeptides when used to cross-linking with ET. In addition, both P1 and P2 served as a substrate for T-protein. In the following crosslinking experiments, P2 fraction was employed as 14 C-labeled 5,10-CH 2 -H 4 PteGlu 4 .

Preparation of 14 C-labeled 5,10-CH 2 -H 4 PteGlu 4 -
Localization of the Folate Binding Sites by Cross-linking and Lysylpeptide Mapping-Wild-type ET was preincubated with 14 C-labeled 5,10-CH 2 -H 4 PteGlu 4 and then cross-linked using EDC and N-hydroxysuccinimide. The ratio of covalently attached 5,10-CH 2 -H 4 PteGlu 4 /mol of ET was near unity. The cross-linked product was digested with lysylendopeptidase and the resultant peptides were separated by HPLC (Fig. 3B). As a control, ET alone was treated with the cross-linkers and subjected to lysylpeptide mapping (Fig. 3A) and Edman sequencing. The peaks P1, P2, and P4 were radioactive and not observed in the control experiment. The another radioactive peak, P3, was inseparable from peak PЈ, and the absorbance at 220 nm was greater than that of PЈ (Fig. 3, A and B). The initial unnamed radioactive peak (Fig. 3B) did not represent peptide(s) cross-linked with 5,10-CH 2 -H 4 PteGlu 4 since the radioactivity could not be retained on the sequencing column of the Hewlett Packard G100A protein sequencer. The four peaks (P1-P4) were isolated in the parallel experiment with the nonlabeled folate and subjected to Edman sequencing. As shown in Table III, P1 and P2 gave the sequences corresponding to residues 76 -81 with Lys-78 missing and residues 350 -360 with Lys-352 missing, respectively. P4 gave the sequence starting from residue 79 with Lys-81 missing. In the control experiment, these peptides were cleaved at the C-terminal side of the lysine residues not detected in the above Edman degradation. Namely, peptide P1 was cleaved to peptides 76 -78 (K2, Fig. 1) and 79 -81 (K3, Fig. 1), peptide P2 to peptides 350 -352 (K12, Fig. 1) and 353-360 (K13, Fig. 1), and peptide P4 to peptides 79 -81 and 82-119 (K4, Fig. 1). The cleavage of P2 was incomplete due to the resistance of the lysyl-proline bond at 352-353 to lysylendopeptidase (Fig. 3A). K2 and K3 were eluted from the C 18 column in the void volume fraction and K4 was eluted  next to P4 (Fig. 3, A and B). These results indicate that the above three lysine residues could not be detected in the Edman degradation due to the modification and, therefore, are likely the candidates for the cross-linking residues with the polyglutamate moiety of 5,10-CH 2 -H 4 PteGlu 4 .
Peak PЈ (Fig. 3A) contained peptides K1, K5, and K7. N-terminal sequencing of P3 up to 20 residues showed the existence of peptide K6 -7 whose Lys-157 was not modified, in addition to the three peptides mentioned. P3 was digested with S. aureus V8 protease under the specific condition for glutamic acid C-terminal cleavage, and the peptides were separated by HPLC. No radioactive peak was detected and no evidence for the modification of amino acid residues was obtained by amino acid sequencing of all the V8 peptides (data not shown). Peptide K6 -7 with intact Lys-157 was not yielded from the control experiment and from the ET crosslinked with 5,10-CH 2 -H 4 folate. Therefore, the inability of lysylendopeptidase to cleave the C-terminal side of Lys-157 may be attributed to some kind of interaction between Lys-157 and 5,10-CH 2 -H 4 PteGlu 4 .
The parallel cross-linking experiment with 5,10-CH 2 -H 4 folate gave cross-linked peptide peaks P5 and P6 (Fig. 3C), which had the same amino acid sequence as P2 and P4, respectively. The results indicate that Lys-352 and Lys-81 were also involved in the coupling with the carboxyl group(s) of the monoglutamate residue. Peptide peaks corresponding to P1 and P3 were not detected (Fig. 3C).
ET with covalently cross-linked 5,10-CH 2 -H 4 folate has no significant enzymatic activity with or without externally added folate substrate.
Effects of the Mutations of the Folate Binding Sites-To assess the contribution for interaction with folate substrate, above three lysine residues (Lys-78, Lys-81, and Lys-352) and Lys-75 were subjected to site-directed mutagenesis. Lys-352, the residue absolutely conserved among T-proteins from eight various species (Fig. 1), was mutated to glutamic acid, glutamine, or arginine. Other lysine residues were replaced with glutamic acid. In addition to the replacement of a single residue, multiple replacements with glutamic acid were carried out. All mutant proteins were highly expressed in E. coli BL21(DE3)pLysS cells as soluble protein and purified by the same strategy used for the wild-type enzyme. The purified proteins gave a single band on SDS-PAGE (not shown). Apparent K m values of mutant ETs for 5,10-CH 2 -H 4 folate and 5,10-CH 2 -H 4 PteGlu 4 were determined and summarized in Table IV. Wild-type ET showed 6.5-fold higher affinity for 5,10-CH 2 -H 4 PteGlu 4 than that for 5,10-CH 2 -H 4 folate. The single mutation of Lys-352 to Glu and Gln increased K m values for both folate substrates about 2-fold, whereas mutation to Arg caused no effect. Thus, the side chain positive charge appears to be essential for the interaction. Mutation of Lys-78 to Glu increased K m values for both folate substrates 1.4-fold. The K81E mutant showed a 3-fold increase in K m for 5,10-CH 2 -H 4 folate and a 16-fold increase in K m for 5,10-CH 2 -H 4 PteGlu 4. The contribution of Lys-75 was not evidenced by the cross-lining experiment but the mutation of Lys-75 to Glu increased K m for 5,10-CH 2 -H 4 folate 2.5-fold and K m for 5,10-CH 2 -H 4 PteGlu 4 8-fold. The multiple mutations resulted in a 7-13-fold increase in K m values for 5,10-CH 2 -H 4 folate and a 5-252-fold increase in K m values for 5,10-CH 2 -H 4 PteGlu 4 . Mutation of above lysine residues affected slightly the affinity of ET for the protein substrate EH. K m values for EH of wild-type ET and the mutant K75E/K78E/K81E/K353E were 0.45 M and 0.53 M, respectively. Changes in k cat values were not so significant, and even higher values were obtained (Table IV). DISCUSSION In this study, 14 C-labeled 5,10-CH 2 -H 4 PteGlu 4 was used to investigate the folate binding sites on E. coli T-protein. 5,10-CH 2 -H 4 PteGlu 4 whose terminal three glutamate residues were uniformly labeled with 14 C was enzymatically synthesized from 5,10-CH 2 -H 4 folate and [U-14 C] glutamic acid using L. casei FPGS. The L. casei enzyme was chosen because of its preference for 5,10-CH 2 -H 4 folate as the substrate among various folates (21). The radiolabeled products separated by HPLC were solely tetraglutamate derivatives, and no other polyglutamate forms were obtained under the conditions used here (Table II). The products were separated into two peak fractions by HPLC (P1 and P2, Fig. 2), but they behaved similarly as a substrate for ET and gave the same cross-linked lysylpeptides when cross-linked with ET. The differences of these two fractions were not investigated further.
EDC was employed to produce zero-length cross-linking between contacted carboxyl groups of polyglutamate moiety of 5,10-CH 2 -H 4 PteGlu 4 and amino groups of ET in the binary complex. Binding of EDC-activated folate derivatives to folatedependent enzymes has been previously reported with thymidylate synthase (28) and 5,10-methenyltetrahydrofolate synthetase (14). In a preliminary experiment, we activated 5,10-CH 2 -H 4 PteGlu[ 14 C]Glu 3 with EDC and N-hydroxysuccinimide prior to interact with ET under the widely used conditions (29). No significant amounts of labeled peptides were obtained after lysylendopeptidase digestion indicating that the condition employed was not appropriate for the cross-linking. Considerable amounts of cross-linked product were obtained when ET and the folate substrate were mixed first to facilitate the complex formation and then EDC and N-hydroxysuccinimide were introduced to the mixture. The Edman sequencing of covalently radiolabeled lysylpeptides revealed that three lysine residues (Lys-78, Lys-81, and Lys-352) were involved in cross-linking with polyglutamate moiety of 5,10-CH 2 -H 4 PteGlu 4 . Two of them, Lys-81 and Lys-352, were also implicated in the crosslinking with 5,10-CH 2 -H 4 folate. The ␣-carboxylic acid and ␥-carboxylic acid groups of 5,10-CH 2 -H 4 folate were apparently involved in the contact with these two lysine residues.
The contribution of these lysine residues to the binding of the folate substrates with ET was confirmed by site-directed mutagenesis and kinetic analysis. Kinetic analyses were conducted employing saturated amounts of H-, P-, and L-protein (diaphorase). 5,10-CH 2 -H 4 PteGlu 4 was better substrate for wild-type ET than 5,10-CH 2 -H 4 folate (Table IV), in good agreement with other folate-dependent enzymes (13)(14)(15)(16). Mutation of Lys-352 or Lys-78 to glutamate increased K m values for both folate substrates 2-fold or less, whereas mutation of Lys-81 to glutamate affected greatly the K m value for 5,10-CH 2 -H 4 PteGlu 4 (16-fold) than for 5,10-CH 2 -H 4 folate (3-fold) ( Table IV). The results were consistent with the observations about the interaction of folate derivatives with 5,10-methenyltetrahydrofolate synthetase. Maras et al. (14) reported that Lys-18 of rabbit liver 5,10-methenyltetrahydrofolate synthetase interacts with the ␣-carboxyl of the Glu-1 residue of folate derivatives. K m values for 5,10-methenyltetrahydrofolate and 5,10-methenyltetrahydropteroylpentaglutamate are 2.6-and 105-fold, respectively, less than that for 5,10-methenyltetrahydropteroate, indicating that interaction of Glu-1 with the enzyme affected not so greatly on the K m value and positively charged amino acid residues other than Lys-18 may be important for the binding of the polyglutamate form. The double mutant K81E/K352E showed a greater increase in the K m value for 5,10-CH 2 -H 4 PteGlu 4 (99-fold) than for 5,10-CH 2 -H 4 folate (8-fold). These results indicate that the interaction with Lys-81 of 5,10-CH 2 -H 4 PteGlu 4 is more critical than that of 5,10-CH 2 -H 4 folate for the binding to ET. Based on the crystal structure of E. coli TS, Kamb et al. (13) reported that glutamate residues of 5,10-CH 2 -H 4 PteGlu 4 interact with a positively charged hydrophilic surface of the protein where Glu-1 is rigidly fixed through its ␣-carboxyl group with Lys-50, a nearly invariant residue among all TS enzymes. Lys-352 of ET is conserved among T-proteins so far studied, whereas Lys-78 and Lys-81 are not (Fig. 1). Considering these circumstances, it is tempting to postulate that Lys-352 interacts with the ␣-carboxyl group of Glu-1 of both folate substrates and Lys-81 interacts with the ␣-carboxyl group of the second glutamate residue (Glu-2) of 5,10-   substrates may reflect the impaired binding ability to Glu-2 of 5,10-CH 2 -H 4 PteGlu 4 . As mentioned above, negatively charged glutamate residues of folylpolyglutamate lie in a position where they can interact with surface region of E. coli TS having positive potential (13). Maras et al. (14) suggested that a region containing multiple Lys and Arg residues serves as a polyglutamate binding site of rabbit liver 5,10-methenyltetrahydrofolate synthetase. Therefore, we investigated the role of Lys-75 situated in the vicinity of Lys-81 and Lys-78 of ET. Although cross-linking of folate derivatives with Lys-75 could not be demonstrated, K75E mutant showed significant increases in K m values for both folate substrates especially for 5,10-CH 2 -H 4 PteGlu 4 . The result together with the changes in K m values for folate substrates of mutants containing K78E mutation suggests that Lys-75 and Lys-78 are implicated in interaction with polyglutamate tail of 5,10-CH 2 -H 4 PteGlu 4 .
Multisites mutations with substitutions of both Lys-81 and Lys-352 gave higher K m values for 5,10-CH 2 -H 4 PteGlu 4 than for 5,10-CH 2 -H 4 folate. k cat /K m values of these mutants with 5,10-CH 2 -H 4 PteGlu 4 as substrates are 26 -120-fold lower than that of wild-type ET (Table IV). The mutation probably induced electrostatic repulsive effects on the binding of folate substrates. Increase in k cat values of mutant ETs is presumably the results of the repulsive effects that enhance the release of the folate reaction products and accelerate the turnover of ET.
Chlumsky et al. (12) reported that an ϳ380-residue region near the C terminus of ␣ subunit of Corynebacterium sp. sarcosine oxidase exhibits homology with the C-terminal half of rat dimethylglycine dehydrogenase and a region in the middle of T-protein. Because these proteins all catalyze the synthesis of 5,10-CH 2 -H 4 folate from tetrahydrofolate and various onecarbon donors, they suggested that these regions were tetrahydrofolate binding domains. The above mentioned lysyl residues of ET, however, are not situated in the region of T-protein they indicated.
Lys-75, -78, and -81 are not conserved in other T-proteins; therefore, the putative folate-binding site discussed in this study seems to be specific for ET, and there must be distinct folate-binding sites in T-proteins from other species. In this connection, it is worth noting that there is considerable variability in the region of TS that contacts the second and third glutamate residues of polyglutamate tail (13). Details of interaction of T-protein with the folate substrates require further elucidation; crystal structure analysis is essential. However, the present cross-linking study provides the first evidence of the folate binding sites on T-protein.