Thiamin Biosynthesis in Escherichia coli

ThiFSGH and ThiI are required for the biosynthesis of the thiazole moiety of thiamin in Escherichia coli. The overproduction, purification, and characterization of ThiFS and the identification of two of the early steps in the biosynthesis of the thiazole moiety of thiamin are described here. ThiS isolated from E. coli thiI + is post-translationally modified by converting the carboxylic acid group of the carboxyl-terminal glycine into a thiocarboxylate. ThethiI gene plays an essential role in the formation of the thiocarboxylate because ThiS isolated from athiI − strain does not contain this modification. ThiF catalyzes the adenylation by ATP of the carboxyl-terminal glycine of ThiS. This reaction is likely to be involved in the activation of ThiS for sulfur transfer from cysteine or from a cysteine-derived sulfur donor.

The thiC gene complements pyrimidine (4-amino-5-hydroxymethylpyrimidine) requiring mutants, the thiFGH genes complement thiazole-requiring mutants, and thiE codes for thiamin phosphate synthase (16). An additional gene (thiI) required for thiazole biosynthesis in Salmonella typhimurium has recently been identified (17). This gene maps at 9.5 min on the E. coli chromosome and complements a nuvC mutation. 2,3 NuvC is required for the biosynthesis of thiouridine in tRNA (18) as well as the thiazole moiety of thiamin (19) (Fig. 2). This suggests that ThiI may play a role in the sulfur transfer chemistry involved in the thiazole biosynthesis.
When the gene product, originally assigned as ThiG, was purified from an overexpression strain, mass spectrometry and Edman sequencing both demonstrated that the protein consisted of two subunits. One of the subunits, which we will call ThiS, has a mass of 7310.74 Da. The second subunit, which will retain the ThiG name, has a mass of 26896.5 Da. 4 Although the anomalous mass of ThiG was previously noted (15), ThiS was missed in the original reading frame assignment because of errors in both published sequences (15,21) and also because of the difficulty of detecting ThiS, which stains poorly with Coomassie Blue and migrates at the dye front during normal SDS-PAGE 5 analysis. The carboxyl terminus of ThiS has the -Gly-Gly sequence found at the carboxyl terminus of human erythrocyte ubiquitin (GenBank accession number 1070588). This sequence similarity was striking because ThiF shows significant sequence similarity to the ubiquitin-activating enzyme (GenBank accession number 731039) including the ATPbinding site. This suggested that ThiF might catalyze the adenylation of ThiS and that ThiS-COAMP might react with cysteine (or a cysteine-derived sulfur donor) to give ThiS-COSH (Fig. 3). In addition, the dual role of ThiI in thiazole and in 4-thiouridine biosynthesis suggested that ThiI might play a role in this sulfur transfer reaction. In this paper, we describe experiments to test these hypotheses.

EXPERIMENTAL PROCEDURES
LB broth was purchased in dehydrated form from Life Technologies, Inc. Tryptose blood agar base was purchased from Difco (Detroit, MI). Ampicillin and isopropyl-␤-D-thiogalactopyranoside were from Jersey Lab and Glove Supply (Livingston, NJ). Tris, DTT, EDTA, (NH 4 ) 2 SO 4 , and ATP were from Sigma. Sodium chloride was from Fisher (Pittsburgh, PA). Acetic acid and MeOH were from Aldrich. Acrylamide/Bis (37.5:1) was purchased from Bio-Rad. Dialysis membrane was from Spectrum (Houston, TX). All buffers were prepared from distilled, deionized water and were filtered through Millipore type HA 0.45-m filters before use. Wizard PCR preps were purchased from Promega (Madison, WI). BL21(DE3) and the DE3 Lysogenization kit were purchased from Novagen (Madison, WI).
Protein purifications were performed on a Waters 650 chromatography instrument (Milford, MA). Gel filtration column (Superdex 75) was from Amersham Pharmacia Biotech. Reverse-phase mass spectrometric preparations were on Michrom BioResources (Auburn, CA) reversephase peptide traps. Protein concentration was assayed using Coomas-sie Plus protein assay reagent from Pierce. Concentrations were calibrated against a standard curve generated using bovine serum albumin. SDS-PAGE gels were prepared and run using standard methods (22). Pyrophosphate assays were conducted using the Enz-Chek pyrophosphate assay kit from Molecular Probes Inc. (Eugene, OR).
Mass values reported correspond to the most abundant isotopic peak (the difference in 1.0034 Da units from the monoisotopic peak is given in italics following the value). All spectra were calibrated externally using bovine ubiquitin as standard (8564.64 Ϫ5).
Construction of thiI Ϫ Strain, VJS2890 -A nuvC E. coli B strain was supplied by Hans Bremer (Dept. of Molecular Cell Biology, University of Texas, Dallas, TX). This strain is equivalent to the original nuvC strain, RYH101, described by Ryals et al. (19). Plasmid pVJS728, which was able to complement this nuvC mutant, was isolated from an E. coli K-12 HindIII library in pHG165 (23). The thiI456::Km (E. coli K-12) mutation was constructed by inserting a kanamycin cassette into the PstI site of pVJS728. This was back crossed to the E. coli chromosome by recombination in the sbcBC strain VJS2889 (24,25). Genetic crosses were performed via bacteriophage P1 kc-mediated transduction (26).
Phenotype Analysis of VJS2890 -The VJS2890 strain was tested for growth with 5-methyl-4-(␤-hydroxyethyl)thiazole and screened for UV sensitivity (nuvC ϩ phenotype). VJS2890 was grown in thiamin-limiting defined medium supplemented with 0.2% glucose (26) and then plated as a lawn on defined medium. 1 l of 100 M 5-methyl-4-(␤-hydroxyethyl)thiazole was spotted on the plate, and growth around or between the 5-methyl-4-(␤-hydroxyethyl)thiazole spots was determined after overnight incubation at 37°C. UV sensitivity was determined as described previously by Ryals et al. (19).
Construction of thiI Ϫ Overexpression Strain, VJS2890(DE3)-Sitespecific integration of the DE3 prophage into the chromosome of the ThiI mutant, VJS2890, was accomplished using the DE3 lysogenization kit (Novagen). The basal and induced expression levels of T7 RNA polymerase in a representative set of 12 lysogens were tested following the protocol provided with the kit. The lysogen (VJS2890(DE3)) that showed the lowest level of basal expression and a high level of isopropyl-␤-D-thiogalactopyranoside-induced expression of T7 RNA polymerase was chosen for use as an overexpression strain.
Construction of the ThiFS Overexpression Vector-The 3 base pairs immediately upstream of the ThiF start codon on plasmid pVJS716 (15) were mutated to insert an NdeI site (CATATG) using PCR. The 5Ј primer was 5Ј-CT GGA AAT TGC AGG AGT TGC ATA TG ATG ACC GTG ACT TTA-3Ј and the 3Ј primer was 5Ј-CCA GAT AGC CAC TGG CGG-3Ј. The desired 1.8-kilobase pair fragment was purified using Wizard PCR preps (Promega) and the NdeI/SalI fragment cloned into the corresponding sites on pET-22b (Novagen) to yield plasmid pG201. The MscI/SalI fragment (549 base pairs) was exchanged with the corresponding DNA from pVJS720 (15). A representative plasmid was named pCAC111. The remaining PCR-derived DNA (520 base pairs) was sequenced, and no mutations were observed.
Overexpression and Purification of ThiFS from E. coli BL21(DE3) (thiI ϩ )-Plasmid pCAC111 was transformed into E. coli strain BL21(DE3), and transformants were selected on tryptose blood agar plates containing ampicillin (200 g/ml). A single colony of E. coli containing pCAC111 was grown at 37°C for 12 h in 3 ml of LB broth supplemented with ampicillin (200 g/ml); 1 ml of this starter culture was diluted into 1 liter of the same medium and grown to an A 595 of approximately 0.6. Expression was induced by adding isopropyl-␤-Dthiogalactopyranoside to a final concentration of 0.3 mM, and growth was continued at 37°C for 12 h. Cells were harvested by centrifugation (8000 ϫ g, 30 min) and stored at Ϫ70°C. A frozen cell pellet from 1 liter of cell culture was thawed and resuspended in 25 ml of loading buffer (50 mM Tris, 2 mM DTT, 2 mM EDTA, pH 7.5), treated with lysozyme (5 mg, 40 min, 4°C), sonicated at 4°C for 4 min, and centrifuged (27,000 ϫ g, 15 min). The resulting cell free lysate was taken to 20% (NH 4 ) 2 SO 4 saturation at 4°C by slow addition of solid (NH 4 ) 2 SO 4 and gentle stirring over 45 min. This mixture was centrifuged (27,000 ϫ g, 15 min), and the supernatant was taken to 50% (NH 4 ) 2 SO 4 saturation at 4°C by slow addition of solid (NH 4 ) 2 SO 4 with gentle stirring over 45 min. This mixture was centrifuged as before, and the resulting pellet was stored at Ϫ70°C. The 20 -50% (NH 4 ) 2 SO 4 pellet was thawed, resuspended in 10 ml of loading buffer, and dialyzed (12,000 -14,000 molecular weight cut-off) for 12 h at 4°C against 4 liters of loading buffer to exchange out (NH 4 ) 2 SO 4 from the solution. This was then passed through a 0.45-m filter and loaded onto a Protein-Pak DEAE HR 8-m 1000 Å column (20 ϫ 100 mm) (Waters). The ThiF/ThiS protein mixture was purified by gradient elution (flow rate was 4 ml/min). The column with bound protein was washed with loading buffer for 10 min and then taken from 0% elution buffer (50 mM Tris, 2 mM DTT, 2 mM EDTA, 1 M NaCl, pH 7.5) to 30% elution buffer over 70 min. Under these conditions, ThiF and ThiS eluted together at 15% elution buffer, suggesting that ThiS is bound to ThiF. The ThiFS-containing fractions were identified by SDS-PAGE analysis, pooled, concentrated, exchanged into gel filtration buffer (50 mM Tris, 2 mM DTT, 2 mM EDTA, 20 mM NaCl) using a YM10 Amicon ultrafiltration membrane, and loaded onto a Superdex 75 gel filtration column (Amersham Pharmacia Biotech). At a flow rate of 1.2 ml/min, ThiF and ThiS again eluted together after 1.5 h. The ThiFScontaining fractions were identified by SDS-PAGE analysis and pooled, concentrated, and exchanged into 50 mM Tris, 2 mM DTT, pH 7.5, using a YM10 Amicon ultrafiltration membrane (Millipore). The protein solution was diluted by adding glycerol to 10% and frozen at a final concentration of 16 mg/ml until use. SDS-PAGE analysis of the purification is shown in Fig. 4A.
Pyrophosphate Assay-Pyrophosphate was assayed using the Enz-Chek pyrophosphate assay kit (27). The assay volume was 1 ml. The reaction buffer was 20 mM Tris-HCl, 1 mM MgCl 2 , pH 7.5. 75 l of ThiFS isolated from BL21(DE3) (thiI ϩ ) (approximate concentration 16 mg/ml; the molar ratio of ThiF to ThiS was not determined) was added to initiate the reaction. ATP was added to a concentration of 2.5 mM. Assays were run for 1 min.
Mass Spectrometric Detection of a Covalent ThiS-AMP Adduct-Reaction buffer (50 l, 10 mM ATP, 10 mM MgCl 2 , 50 mM Tris, 2 mM DTT, pH 7.5) was added to ThiFS (25 l of a 16 mg/ml solution) isolated from E. coli VJS2890(DE3) (thiI Ϫ ) incubated at 25°C for 10 min and then acidified to pH 3 by addition of glacial acetic acid (2 l). The assay mixture was diluted with water (100 l) and loaded onto a reversephase peptide trap (Michrom BioResources, Inc.). The trap was washed with 1 ml of 98% H 2 O/1% MeOH/1% acetic acid, and the protein was eluted with 150 l of 80% methanol/20% acetic acid (the first 50 l was discarded). 2 l of this eluent was loaded into a borosilicate glass capillary (1.5 mm inner diameter) pulled to a 2-m tip at one end (28). A platinum wire inserted into the distal end of the glass capillary made contact with the solution and a voltage of 0.7-1.4 kV on the wire determined the flow rate (Ϸ25-75 nl/min). Droplets formed by ESI were sampled by the heated metal capillary (Ϸ110°C) of a 6T Fourier transform mass spectrometer described previously (29,30). Briefly, ions are guided through five stages of differential pumping by a skimmer and three quadrupole ion guides into the magnet bore held at 10 Ϫ9 torr. Ions of interest were isolated in the trapped ion cell by the stored wave form inverse Fourier transform (SWIFT) technique (31) and collisionally activated by sustained off-resonance irradiation (SORI) (32) in the presence of Ϸ10 Ϫ6 torr N 2 .
Mass Spectrometric Detection of ThiS Thiocarboxylate-ThiFS (25 l, 16 mg/ml), isolated from E. coli BL21(DE3) (thiI ϩ ) was thawed and acidified to pH 3 by the addition of glacial acetic acid (2 l). The mixture was diluted with water (100 l) and loaded onto a reverse-phase peptide trap (Michrom BioResources, Inc.). The trap was washed with 1 ml of 98% H 2 O/1% MeOH/1% acetic acid, and the protein was eluted with 150 l of 80% methanol/20% acetic acid (the first 50 l was discarded). 2 l of this eluent was loaded into a borosilicate glass capillary and electrosprayed as described above. ThiFS isolated from E. coli VJS2890(DE3) (thiI Ϫ ) was analyzed in an identical manner.
Iodoacetic Acid Derivatization of ThiS Thiocarboxylate-ThiFS (25 l, 16 mg/ml) isolated from E. coli BL21(DE3) (thiI ϩ ) was thawed, and iodoacetic acid was added (25 l of a 10 mM solution in 25 mM Tris, pH 7.5). This mixture was incubated for 5 min at 25°C and then acidified to pH 3 by addition of glacial acetic acid (2 l). The mixture was diluted with water (100 l) and loaded onto a reverse-phase peptide trap (Michrom BioResources, Inc.). The trap was washed with 1 ml of 98% H 2 O/1% MeOH/1% acetic acid, and the protein was eluted with 150 l of 80% methanol/20% acetic acid (the first 50 l was discarded). 2 l of this eluent was loaded into a borosilicate glass capillary and electrosprayed as described above. Fig. 4. Both of the ThiFS overexpression strains provided high levels of soluble protein. The average yield of purified (Ͼ90%) ThiFS was 5 mg/liter of cell culture. ThiF and ThiS copurify through several steps, demonstrating that ThiF and ThiS form a stable complex. ThiS stains poorly with Coomassie Blue and is not visible on the gel. The mass spectra discussed below clearly demonstrate its presence in all ThiFS samples examined.

Purification of ThiFS-The purification of ThiFS is summarized in
Pyrophosphate Production by ThiFS Isolated from E. coli BL21(DE3) (thiI ϩ )-ThiFS catalyzed the formation of pyrophosphate from ATP (Fig. 5). This reaction required Mg 2ϩ as a cofactor. Pyrophosphate production reached a plateau, presum- ably because of the complete consumption of ThiS, after the formation of approximately 23 nmol pyrophosphate. If the ratio of ThiF:ThiS was 1:1, 35 nmol of pyrophosphate would be expected. This suggests that the ThiF:ThiS stoichiometry is 1:0.66, and that during purification 34% of ThiS dissociates from the ThiF. It was not possible to determine whether ThiF on its own would catalyze the hydrolysis of ATP because overexpression of ThiF in the absence of ThiS gave only inclusion bodies. The overexpression of ThiS using the pET-16B vector was also unsuccessful.
Mass Spectrometric Detection of a Covalent ThiS-AMP Adduct-The ESI/FTMS of ATP-treated ThiFS isolated from E. coli VJS2890(DE3) (thiI Ϫ ) is shown in Fig. 6. In addition to showing ions that can be assigned to [ThiS] 5ϩ (m/z ϭ 1463.15), the spectrum shows ions at m/z ϭ 1528.97 that can be assigned to [ThiS-COAMP] 5ϩ .
Mass Spectrometric Characterization of ThiS Isolated from E. coli BL21(DE3) (thiI ϩ ) and from E. coli VJS2890(DE3) (thiI Ϫ )-ESI/FTMS on ThiFS isolated from E. coli BL21(DE3) (thiI ϩ ) demonstrated that the predominant ThiS ion had an additional mass of 16 Da compared with ThiS isolated from E. coli VJS2890(DE3) (thiI Ϫ ) (Fig. 7, A and B, respectively). This is consistent with the replacement of one oxygen with sulfur on the protein. The ratio of ThiS to ThiSϩ16 in the (thiI ϩ ) sample was approximately 1:9. SWIFT isolation and SORI activation of the [ThiSϩ16] 6ϩ ions gave 10 fragment ions; two of these had masses of 7235.64 and 7178.69 and were assigned as b 65 and b 64 ions (fragments that contain the NH 2 terminus of the protein) (33). These represent the loss of 91 Da (COOH-terminal glycine ϩ 16 Da) and 148 Da (COOH-terminal glycine ϩ 16 Da ϩ internal glycine), respectively, from the COOH terminus of the ThiSϩ16 adduct and strongly suggested that the COOHterminal glycine carboxylate of ThiS has been converted to a thiocarboxylate in the (thiI ϩ ) strain (Fig. 8).
To confirm this, ThiFS from E. coli BL21(DE3) (thiI ϩ ) was alkylated with iodoacetic acid. Analysis   (Fig. 9A). SWIFT isolation of the 6ϩ charge state of the sodium adduct at m ϭ 7406.7 (formed in a separate sample ionization), followed by SORI activation gave a series of 5ϩ fragments (Fig. 9B). All of these fragments were assigned either as b ions or y ions (fragments that contain the COOH terminus of the protein) (33). All of the observed b ions through b 63 had a mass identical to that predicted from the unmodified ThiS sequence, and the y 62 , y 63 , and y 64 ions (sodium adducts) all contained the modified amino acid. This localized the site of modification to the carboxyl-terminal Ala 64 , Gly 65 , and Gly 66 residues (Fig. 9C). Because the only functional group in these three residues is the Gly66 carboxylate, ThiS isolated from E. coli BL21(DE3) (thiI ϩ ) must contain a thiocarboxylate at the COOH terminus. DISCUSSION The sequence similarity between ThiF/ThiS and the ubiquitin-activating enzyme E1/ubiquitin suggested that ThiF might catalyze the adenylation of ThiS and that ThiS might be the sulfur carrier in thiamin biosynthesis. In addition, the dual role of ThiI in both thiazole biosynthesis and in sulfur transfer in 4-thiouridine biosynthesis in tRNA suggested that ThiI may play a role in the sulfur transfer to ThiS. The high level overexpression of soluble ThiFS from E. coli BL21(DE3) (thiI ϩ ) and from E. coli VJS2890(DE3) (thiI Ϫ ) and the partial purification of these proteins (Ͼ90%) has made it possible to test these hypotheses (Fig. 3).
When ThiFS isolated from E. coli BL21(DE3) (thiI ϩ ) was incubated with ATP, a rapid initial burst of pyrophosphate formation was observed (Fig. 5). The reaction then stops presumably because of the complete conversion of ThiS to ThiS-COAMP. It was possible to directly detect this species using ESI/FTMS (Fig. 6). This demonstrates that ThiF catalyzes the adenylation of ThiS on an oxygen nucleophile.
Analysis of ThiFS isolated from E. coli BL21(DE3) (thiI ϩ ) by ESI/FTMS demonstrated the presence of a ThiSϩ16 adduct as the major form of isolated ThiS (Fig. 7A). This could be because of methionine oxidation during isolation or to the replacement of an oxygen by sulfur. Partial sequencing of ThiSϩ16 (Fig. 8) and the product resulting from alkylation of ThiSϩ16 by iodoacetate (Fig. 9A) unambiguously identified the ϩ16 adduct as an oxygen to sulfur replacement and uniquely localized the site of modification to the carboxylic acid group of the carboxylterminal glycine residue.
MS analysis of ThiS isolated from E. coli VJS2890(DE3) (thiI Ϫ ) demonstrated that the ThiS-COSH was not formed in this strain (Fig. 7B). This suggests that ThiI plays an essential role in the conversion of ThiS-COAMP to ThiS-COSH.
The detection of ThiS-COSH was surprising. None of the previous mechanistic proposals for the formation of the thiazole moiety of thiamin predicted the existence of such an intermediate (1).
The mechanistic enzymology of the sulfur transfer chemistry involved in the biosynthesis of thiamin, molybdopterin, biotin, and lipoic acid is still poorly understood. The immediate sulfur source for the formation of biotin and lipoic acid is unknown, and the mechanism of the sulfur insertion chemistry is not yet clear (34 -38). The sulfur transfer chemistry involved in thiamin and molybdopterin biosynthesis is likely to be similar. A three-enzyme system (MoeB, MoaD, and MoaE) for the sulfur transfer to precursor Z in molybdopterin biosynthesis has been reconstituted (39). ThiF and MoeB (GenBank™ accession number 127233) show high sequence similarity, and both MoaD (Gen-Bank™ accession number 2507065) and ThiS contain the carboxyl-terminal GG residues and bind sulfur as a thiocarboxylate (20,40). With the identification of the immediate sulfur donor in thiamin biosynthesis and the development of an overexpression strain yielding milligram quantities of ThiS-COSH, 6 it should now be possible to reconstitute the biosynthesis of the thiazole moiety. Efforts along these lines are in progress.