Thiamine biosynthesis in Escherichia coli: in vitro reconstitution of the thiazole synthase activity.

The biosynthesis of thiamine in Escherichia coli requires the formation of an intermediate thiazole from tyrosine, 1-deoxy-d-xylulose-5-phosphate (Dxp), and cysteine using at least six structural proteins, ThiFSGH, IscS, and ThiI. We describe for the first time the reconstitution of thiazole synthase activity using cell-free extracts and proteins derived from adenosine-treated E. coli 83-1 cells. The addition of adenosine or adenine to growing cultures of Aerobacter aerogenes, Salmonella typhimurium, and E. coli has been shown previously to relieve the repression by thiamine of its own biosynthesis and increase the expression levels of the thiamine biosynthetic enzymes. By exploiting this effect, we show that the in vitro thiazole synthase activity of cleared lysates or desalted proteins from E. coli 83-1 cells is dependent upon the addition of purified ThiGH-His complex, tyrosine (but not cysteine or 1-deoxy-d-xylulose-5-phosphate), and an as yet unidentified intermediate present in the protein fraction from these cells. The activity is strongly stimulated by the addition of S-adenosylmethionine and NADPH.

Cysteine has been identified as the precursor of the thiazole sulfur atom. The sulfur atom is initially transferred to IscS, a pyridoxal phosphate-dependent cysteine desulfurase (15), in the form of a persulfide of an active site Cys residue (16). It is likely that in vivo the sulfur atom is then passed to ThiI (12) as a persulfide of residue Cys-456 (13). The sulfur atom is then transferred to the C terminus of ThiS, converting it to a thiocarboxylate (14,17). This final sulfur transfer reaction requires the activation of ThiS as its acyladenylate (17). The adenylation reaction is catalyzed by ThiF, which has been observed to form a complex with ThiS (17). In addition, an unusual acylpersulfide link between ThiS thiocarboxylate and Cys-184 of ThiF has been observed, and mutagenesis studies suggest it may be important for thiazole formation (18).
Recent studies with purified Bacillus subtilis proteins have resulted in the first demonstration of thiazole biosynthesis in a bacterial cell-free system (19). Although B. subtilis shares some metabolic components with E. coli, including some of the biosynthetic proteins (ThiFSG, ThiI, and IscS) and two metabolic precursors (Dxp and Cys), the biosynthetic pathways also differ significantly. Whereas B. subtilis depends upon ThiO, an oxygendependent flavoenzyme, to utilize glycine to provide the C-2-N-3 fragment of the thiazole (1,20), E. coli and other enteric organisms substitute ThiO with an oxygen-sensitive protein, ThiH (21)(22)(23) and use tyrosine to provide the C-2-N-3 unit.
Although genetic studies have identified ThiG and ThiH as essential for thiazole biosynthesis in E. coli, incubation of overexpressed ThiFSGH and ThiI with ThiS thiocarboxylate, [U-14 C]tyrosine, and Dxp has failed to produce any detectable amount of thiazole product (24). Under these conditions, tyrosine was cleaved to p-hydroxybenzyl alcohol, a known by-product of TPP biosynthesis (25), but sulfur transfer from 35 S-ThiS thiocarboxylate did not occur (1).
Analysis of the ThiH sequence has identified a conserved cysteine triad motif, thought to be involved in Fe-S cluster binding and characteristic of the "radical AdoMet" family (26). Recently, hexahistidine-tagged ThiH (ThiH-His) has been isolated in a complex with ThiG and shown to contain a highly oxygen-sensitive Fe-S cluster (21), which might explain the difficulty in obtaining in vitro activity under aerobic conditions. The ThiGH complex is likely to be involved in the last steps of the thiazole biosynthesis, a cyclization reaction that could conceivably require sulfur transfer to Dxp from either the ThiS thiocarboxylate or the ThiFS conjugate, condensation with Tyr (or a derivative), and cyclization to the thiazole product. * This work was supported by a University Research Fellowship from the Royal Society. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The addition of adenosine or adenine during the growth phase of enteric bacteria, such as Aerobacter aerogenes (27), Salmonella typhimurium (28,29), and E. coli (6,30,31), is known to release the repression by thiamine pyrophosphate of its own biosynthesis by decreasing the intracellular concentration of the vitamin. 5-AIR is a shared precursor of Hmp-PP and purines (32). Adenosine and adenine are negative regulators of de novo purine biosynthesis (33) and decrease the concentration of AIR, which results in a reduction in Hmp-PP biosynthesis and hence a decrease in intracellular TPP (28). Thiamine has been identified recently (34) as one of the cellular metabolites with the potential to regulate its own production by directly binding to regulatory regions upstream of thiC and thiM within polycistronic mRNAs. A decrease in intracellular thiamine might therefore be expected to result in higher levels of biosynthetic enzymes (30), thus providing a potential source of enzymes suitable for in vitro studies of thiazole biosynthesis.
As a pre-requisite to elucidating the function of the ThiGH complex and the mechanism of the thiazole cyclization in E. coli, we report here the reconstitution of thiazole biosynthetic activity in vitro using a cell-free system containing extracts from adenosine-treated E. coli 83-1 cells (6) and purified ThiGH-His under anaerobic conditions.
Bacterial Strains, Plasmids, and Proteins-E. coli 83-1 cells, which are Tyr/Trp/Phe auxotrophs, were the generous gifts from Prof. R. Azerad and Prof. M. Therisod (6). BL21(DE3) cells transformed with a pET-23b-derived plasmid encoding for a hexahistidine-tagged 1-deoxy-D-xylulose-5-phosphate synthase (Dxs-His) were kindly donated by Prof. A. Boronat (37). Dxs-His was expressed from these cells and purified to ϳ80% purity by nickel-affinity and gel filtration chromatographies. ThiG and ThiH-His were purified from E. coli BL21(DE3) pRL1020 as described previously (21). Protein concentration was routinely estimated by the method of Bradford (38); ThiG and ThiH-His concentrations in the ThiGH-His complex were determined by SDS-PAGE analysis and gel densitometry (Syngene Gene Genius imaging system) against bovine serum albumin standards. HPLC Purification of Dxp-Dxp was enzymatically synthesized as described by Taylor et al. (39), and the progress of the reaction was monitored by TLC analysis (40). When the reaction was judged to be complete, the proteins were removed by ultrafiltration (10-kDa molecular weight cut-off, Millipore), and Dxp was purified by semipreparative HPLC (PerkinElmer Life Sciences Prep-10 Octyl column, 250 ϫ 10.0 mm). Dxp was eluted with H 2 O at 2 ml/min, and the column eluate was fractionated from t ϭ 6.0 to t ϭ 8.0 min. The TPP-containing fractions were identified by TLC analysis and discarded. The TPP-free fractions were pooled and freeze-dried. The Dxp content of the resultant white solid (which also contained residual salts) was determined by 1 H NMR, and a stock solution of Dxp (50 mM) was prepared in 50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 . The absence of contaminating TPP was confirmed with the thiochrome assay (41).
Time Course of the Adenosine Derepression-E. coli 83-1 was derepressed in DM1 medium containing adenosine (300 g/ml), essentially as described previously (6). The TPP content was determined in duplicate by the thiochrome assay (41) and expressed as nanograms of TPP/mg of wet cells. To be able to use the "wet weight" of the cell pellets in the calculations, the published procedure was slightly modified. Single colonies were used to inoculate DM1 medium (10 ml) and the cells grown for 8 -10 h; the resultant culture was used as a 0.3% inoculum into fresh DM1 medium (300 ml) and grown overnight at 37°C. Cells were harvested by centrifugation (Beckman JA-14, 12,000 rpm, 10 min at 4°C), resuspended in DM medium (10 ml), transferred to pre-weighed tubes (15 ml), and centrifuged (Beckman JA-14 with adaptors, 7000 rpm, 10 min at 4°C). After having drained most of the medium, residual liquid was removed by gently wiping the walls of each tube with a clean tissue. Tubes were then re-weighed, and the wet weight was obtained by subtraction. The weighed cell pellets were resuspended in DM medium to yield 24 mg/ml suspensions, pooled, used as inoculum (1.3 ml) in DM1 medium (100 ml/flask, 12 flasks), and aerated in a rotary shaker. ␤-L-Arabinose (0.2%) was added to the DM1 medium during derepression experiments with E. coli 83-1 pRL1020, where indicated. To provide duplicate data, a pair of flasks were withdrawn at t ϭ 0 and every 30 min for 2.5 h, rapidly cooled in icy water, and the cells harvested by centrifugation. The cell pellets were then resuspended and washed with sterile 0.9% NaCl (two times, 10 ml), and the wet weight determined as described above. Cell pellets (27-180 mg) were stored at Ϫ80°C until analyzed.
The thiochrome method (41) was adapted to estimate the TPP content in cell pellets. Briefly, the frozen cell pellets were thawed on ice for 10 min and then resuspended in a 1:1 mixture of H 2 O/7.2% HClO 4 (2 ml for cell pellets of up to 80 mg of wet cells, 4 ml for heavier cell pellets); the white suspensions were sonicated on ice and the precipitated proteins removed by centrifugation. The supernatants (0.5 ml) were oxidized with a K 3 [Fe(CN) 6 ] (12 mM in 3.35 M NaOH) solution (100 l) (41) and then adjusted to pH 7.0 with phosphoric acid, and an aliquot (50 l) was analyzed by reverse phase HPLC using a Phenomenex ODS column (150 ϫ 4.6 mm, 5-m particle size) and eluted with buffer A (140 mM K 2 HPO 4 , 12% MeOH, 1.5% N,N-dimethylformamide, 0.3 mM tetrabutylammonium hydroxide, pH 7.0) and buffer B (70% MeOH, 30% H 2 O). The compounds were eluted at 0.8 ml/min with linear gradients between the following time points: t ϭ 0 min, 0% buffer B; t ϭ 2 min, 0% buffer B; t ϭ 5 min, 20% buffer B; t ϭ 10 min, 20% buffer B; t ϭ 12 min, 40% buffer B; t ϭ 15 min, 40% buffer B; and detected by fluorescence (excitation at 360 nm and emission at 454 nm). TPP, TP, and unphosphorylated thiamine solutions of known concentration were derivatized and analyzed by HPLC by the same method. The retention times of the thiochrome derivatives of TPP, TP, and unphosphorylated thiamine were determined to be 7.5, 9.0, and 15.0 min, respectively. TP and TPP calibration curves (0 -1.3 M) were constructed and used to quantify the thiamine derivatives present in samples. Unphosphorylated thiamine was not detected in any of the in vivo or in vitro enzyme assays, but the retention time was determined to exclude the possibility of unphosphorylated thiamine contributing to the total amount of thiamine derivatives present.
TPP Production by Washed Cell Suspensions-Experiments were carried out in duplicate. Cultures of E. coli 83-1 or 83-1 pRL1020 (100 ml/flask, 14 flasks) were incubated with adenosine as described above, and samples withdrawn at t ϭ 0, 1.0, and 2.5 h to monitor the derepression phase. At the end of this phase, the cultures were quickly cooled in icy water, and the cells were harvested by centrifugation (Beckman JA-14, 12,000 rpm, 10 min at 4°C). The cell pellets were then washed with sterile 0.9% NaCl (two times, 10 ml) and transferred to pre-weighed tubes to be weighed. The pellets were resuspended in DM medium to yield 30 mg/ml suspensions, pooled, and used as a 10% inoculum into DM medium (10 ml) containing Hmp and either Tyr or Thz-P or no supplement. The 10-ml cultures were shaken at 37°C, and the cells were harvested every 30 min for 2 h by centrifugation. The cell pellets were washed with NaCl solution, weighed, and stored at Ϫ80°C until analyzed. The TPP content was estimated as described above.
Large Scale Growth of E. coli 83-1 pRL1020 in the Presence of Adenosine-A single colony was used to inoculate DM1 medium (10 ml), and cells were grown at 37°C for 8 -10 h. This culture (3 ml) was used to inoculate fresh DM1 medium (800 ml). After overnight incubation at 37°C, the cells were harvested by centrifugation (Beckman JA-14, 12000 rpm, 10 min at 4°C). The cell pellets were weighed and resuspended in DM medium to yield 100 mg/ml suspensions, and then the resultant cultures were pooled and used as inocula (5 ml) in DM1 medium (four times, 1.25 liters) supplemented by the addition of adenosine (300 mg/liter). Cells were grown at 37°C for 2.5 h, then cooled down by swirling in icy water, and harvested. The typical yield of cell paste ranged between 5.5 and 7.5 g from 5 liters. The cell paste was stored at Ϫ80°C until used.
Preparation of Cell-free Extracts Under Anaerobic Conditions-Where possible cell samples were manipulated under anaerobic conditions using a Belle glove box (Ͻ2 ppm of O 2 ). Cell-free extracts from derepressed E. coli 83-1 pRL1020 or BL21(DE3) pRL1020 were prepared by introducing weighed amounts of cell paste (0.8 -1.0 g) into the glove box and resuspending it in anaerobic reaction buffer (1.6 -2.0 ml, 50 mM Tris-HCl, pH 8.0, 20 mM MgCl 2 , 5 mM DTT, 12.5% (w/v) glycerol). The suspensions were then withdrawn from the anaerobic box and rapidly lysed by sonication. The lysates were returned to the box and allowed to degas for 10 min and then cleared by centrifugation (Beckman JA-14 with appropriate adaptors, 7000 rpm, 15 min at 4°C) in gas-tight tubes. The protein concentration in these lysates was typically 35-45 mg/ml. The cleared lysate prepared from E. coli BL21(DE3) pRL1020 cells (grown in 2YT medium) was routinely gel-filtered through a NAP-10 column (AP Biotech), equilibrated in anaerobic reaction buffer to remove nearly all of the low molecular weight molecules, including TPP, and is referred to as the "ThiGH-His-enriched fraction." Cell-free lysates and desalted protein fractions were prepared by similar methods from derepressed E. coli 83-1 pRL1020 and are termed "derepressed 83-1 lysate" and "derepressed 83-1 protein fraction," respectively.
Measurement of the in Vitro Thiazole Synthase Activity by Enzymatic Coupling to Hmp-PP and Thiochrome Assay-Unless otherwise specified, the term "thiamine" is used to indicate any vitamin form, regardless of the extent of phosphorylation. A typical reaction mixture contained the following: 100 l of derepressed 83-1 lysate or protein fraction, 100 l of ThiGH-His-enriched protein fraction or purified ThiGH-His (8.0 -10.0 mg/ml, total protein concentration) in anaerobic reaction buffer, ATP (20 mM), Tyr (1.5 mM), Cys (1.5 mM), Dxp (1.0 mM), and Hmp (0.6 mM). Depending on the experiment, the total volume varied between 250 and 350 l, as required. Tyr stock solutions (40 mM) were prepared in 100 mM HCl. Cys stock solutions (80 mM) were freshly prepared by dissolving the solid in anaerobic reaction buffer. Dxp and Hmp stock solutions (50 and 14 mM, respectively) were prepared in 50 mM Tris-HCl, pH 8.0, 5 mM MgCl 2 . The ATP stock solutions (250 mM) were prepared in 300 mM Tris-HCl, pH 8.8. All solutions were prepared within the glove box using anaerobic buffer or allowed to degas in the glove box (30 min) before being added to the reaction mixtures. To determine the background level of TP or TPP present in the assays, control reactions were prepared and immediately stopped for analysis. The remaining reactions were incubated for 4 h at 37°C, then stopped by adding HClO 4 (7.2%, 500 l), and the total volume adjusted to 1 ml with H 2 O. The precipitated proteins were removed by centrifugation and the supernatant (500 l) assayed for TP and TPP. Due to the presence of DTT, the sensitivity of the thiochrome assay had to be optimized by doubling the final K 3 [Fe(CN) 6 ] concentration. For each set of conditions, the amounts of TP and TPP were estimated separately, the respective background levels subtracted, and the results expressed as pmol/mg of protein/h.
Effect of S-Adenosylmethionine and Reductants-Samples of purified ThiGH-His (100 l, 113-136 M in ThiH-His) were incubated for 2 h with a 5-fold molar excess of FeCl 3 and Na 2 S, in the presence of DTT (5 mM), under anaerobic conditions. Thus treated ThiGH-His was added to in vitro reaction mixtures containing S-adenosylmethionine (1 mM) and NADPH (1 mM) in addition to Tyr, Hmp, ATP, and derepressed 83-1 protein fraction and then incubated at 37°C for 4 h. Alternatively, the reconstituted ThiGH-His samples were anaerobically incubated with sodium dithionite (1 mM) for 30 min at room temperature (43) before being transferred to the AdoMet-containing reaction mixtures.

Effect of the Adenosine Treatment on Thiamine Biosynthesis in Washed Cell Suspensions
E. coli 83-1 cells are among those enteric bacteria that have been shown to be sensitive to the adenosine derepression of thiamine biosynthesis (6). To optimize the potential thiamine biosynthetic activity for the planned in vitro experiments, the time course of the in vivo derepression and subsequent burst in thiamine production were investigated in detail. The experimental conditions were analogous to those adopted by Newell and Tucker (28) during their investigation of the derepression of thiamine biosynthesis in S. typhimurium LT2 (mutant T), but with modifications to the growth medium to provide the auxotrophic growth requirements for E. coli 83-1 (6), including the aromatic amino acids Phe, Trp, and Tyr. As Tyr is a known thiazole precursor, if the medium is not supplemented with tyrosine, de novo thiamine biosynthesis becomes dependent on Tyr scavenged from protein turnover in E. coli 83-1.
The addition of adenosine to the medium during the growth phase caused a drop in the initial TPP content (3-4 ng/mg wet cells), with the vitamin level remaining low for about 2 h ( Fig.  2A). During this period of time, E. coli 83-1 starved of thiamine and freed from the vitamin repression could accumulate higher (derepressed) levels of biosynthetic enzymes (30). After washing, these cells were transferred to media supplemented with Tyr and Hmp and produced approximately three times more TPP than the untreated cells (negative control) (Fig. 2B). Under these conditions, the burst in the vitamin production may be ascribed to de novo synthesis of the thiazole (44), because this moiety does not accumulate in TPP-depleted cells, such as adenosine-treated cells or in mutants unable to assemble Hmp-PP (27,29).
In a positive control experiment, using media supplemented by the addition of not only Hmp but also Thz-P, both adenosinetreated and untreated cells were able to produce 60 -70 ng of TPP/mg wet cells (Table I, 3rd column). This suggests that in experiments where the medium was supplemented with Hmp and thiazole precursors, any Thz-P produced would be rapidly converted to TPP and that neither ThiD nor ThiE was limiting the TPP formation. Indeed, these results strongly suggest that, under the conditions investigated, thiazole biosynthesis was the rate-limiting step.
The observation that the pre-incubation with adenosine had lowered the intracellular TPP concentration and, as a result, may have increased the expression level of the rate-determining thiazole biosynthetic enzymes contrasted with the activities of ThiD and ThiE, which were not equally affected. This is an unexpected result as the genes encoding these two proteins belong to the thiMD and thiCEFSGH operons and therefore should be subject to coordinated repression by thiamine (30, 31, 34, 45, 46). Despite the lack of a simple explanation for these results, Newell and Tucker (29) observed a precisely analogous effect with the Sal- monella mutant used for their studies. An alternative regulatory mechanism, operating at the translational or metabolic level, can be invoked to explain these observations, but it is clear that more than one factor is controlling the balance of the different thiamine biosynthetic enzymes.
As the assembly of Thz-P was observed to be rate-limiting for TPP production in derepressed E. coli 83-1 cultures, the effect of expression plasmids encoding structural genes known to be required for Thz-P biosynthesis was investigated. For these experiments, E. coli 83-1 was transformed with pRL1020, a plasmid encoding ThiFSGH-His (21). The resultant strain, E. coli 83-1 pRL1020, was cultured with adenosine (to derepress thiamine biosynthesis) and ␤-L-arabinose (to induce the expression of ThiFSGH) for 2.5 h. However, E. coli 83-1 pRL1020 produced only marginally more TPP than E. coli 83-1, and comparative analysis of the proteins in the cleared cell lysates by SDS-PAGE did not show an appreciable difference in the protein expression levels (Fig. 3). This may be due to the short induction time or it may be a function of the high glucose content of the DM1 medium acting to repress expression from pRL1020 (47). As the presence of pRL1020 did not alter the response to the adenosine treatment, but conveniently conferred ampicillin resistance, E. coli 83-1 pRL1020 was used in the subsequent in vitro experiments.

In Vitro Assay Strategy
Several problems were overcome during the development of a functional thiazole synthase assay.
Thz-P Is Difficult to Assay Directly in E. coli Cell-free Systems-Although it is known that certain mutants impaired in thiamine biosynthesis can accumulate Hmp and also excrete it into the medium (29, 32, 46, 48), Thz-P is normally present at low levels and does not accumulate. It has been proposed that this compound may exert a feedback inhibition on one or more enzymes involved in its own biosynthesis (29). In fact, the direct detection of the thiazole has been reported to be very difficult (29,49), but this problem may be avoided by efficiently coupling any thiazole produced to Hmp-PP and then measuring the thiamine content (TP or TPP) after quantitative oxidation to the fluorescent thiochrome derivatives (41, 49 -51).
Thiamine Biosynthesis Is Repressed-The tight regulation of thiamine biosynthesis can be overcome by the addition of adenosine or adenine to the growth medium. The adenosine decreases the intracellular TPP concentration, and the cells respond by accumulating the thiamine biosynthetic enzymes (30). These derepressed cells therefore have the potential to provide a cell-free system suitable for a thiazole synthase assay. The E. coli strain 83-1 was selected as it had already been employed in the in vivo investigation of the thiazole biosynthesis using labeled precursors (6,10,44).
E. coli Accumulates Intracellular TPP When Cultured on Rich Media-When cultured on rich media (such as 2YT or LB), E. coli sequesters thiamine, resulting in very high levels of background TPP (Ͼ500 pmol/mg of protein). This makes accurate assays of de novo biosynthesis difficult. Gel filtration of the cell lysates reduced the TPP background; for example, typical TPP background levels in the 83-1 protein extract and the ThiGH-His-enriched protein extract were 60 -100 and 220 -250 pmol/mg protein, respectively.
Thiazole Biosynthesis Must Be the Rate-limiting Step in a Coupled Assay-Control experiments demonstrated that derepressed 83-1 lysates were able to produce thiamine at more than 50 pmol/mg/h when supplied with Hmp and Thz-P in the presence of ATP and Mg 2ϩ . This was far greater than the observed rate of de novo thiazole biosynthesis observed in subsequent assays and indicates that the ThiD-dependent pyro- phosphorylation of Hmp and the ThiE-dependent coupling of Thz-P and Hmp-PP to yield TP were not rate-limiting under the assay conditions.
ThiH Contains an Oxygen-sensitive Fe-S Cluster-Elegant metabolic studies with Salmonella enterica mutants (22,23,52) and the characterization of purified E. coli ThiGH-His (21) have demonstrated that ThiH contains an oxygen-sensitive Fe-S cluster required for Thz-P biosynthesis; anaerobic conditions were therefore adopted for the thiazole synthase assays.

Thiazole Synthase Activity in E. coli Cleared Cell Lysates and Desalted Protein Extracts
De novo production of both TP and TPP was measured in the derepressed 83-1 lysate, anaerobically incubated with ATP, Hmp, and the thiazole precursors, Tyr, Cys, and Dxp. This basal activity was estimated at ϳ7 pmol/mg of protein/h (Fig. 4,  1st column, or Fig. 5, sample 2) and was confirmed to be higher than the activity measured in lysates derived from repressed cells (4 -5 pmol/mg of protein/h, Fig. 5, sample 1), although the increase in the in vitro thiamine biosynthetic activity of the derepressed E. coli 83-1 lysates was less pronounced than in vivo. The ThiGH-His-enriched protein fraction (prepared by gel filtration of the E. coli BL21(DE3) pRL1020 cleared cell lysate) was also assayed under identical conditions and shown to be active, producing almost exclusively TPP (8 -11 pmol/mg of protein/h, Fig. 4, 7th column). This was a somewhat surprising result as earlier attempts to measure in vitro thiazole biosynthesis from a ThiFSGH-overexpressing extract had been unsuccessful (1,24). However, the thiazole synthase activity described here was detected in assays maintained under anaerobic conditions, using an expression system optimized for the soluble expression of the ThiGH-His complex (21).
As in vivo studies and control experiments had demonstrated that thiazole biosynthesis was likely to be the rate-limiting step for de novo thiamine biosynthesis, the effect of mixing the ThiGH-His-enriched protein extract and the derepressed 83-1 lysate was investigated. Mixtures containing different ratios of these two extracts were assayed (Fig. 4) and showed activities higher than either individual extract (to a maximum of 16.5 pmol/mg of protein/h), which suggested that the two extracts might be providing complementary components. The increasing proportion of the ThiGH-His-enriched protein extract also increased the TPP/TP ratio when compared with the derepressed 83-1 lysate (Fig. 4), revealing that the derepressed E. coli 83-1 lysate has a lower thiamine phosphate kinase (ThiL) activity than the BL21(DE3)-derived ThiGH-His-enriched protein extract.

Comparing the Effect of the ThiGH-His-enriched Protein Fraction and Purified ThiGH-His
The increased activity obtained by mixing the ThiGH-Hisenriched protein fraction and the derepressed 83-1 cell lysate suggested the possibility that the rate-enhancing components provided by the ThiGH-His-enriched protein fraction were the overexpressed proteins. Therefore, the effect of this fraction (Fig.  5, sample 3) was compared with that of purified ThiGH-His (Fig.  5, sample 4), and both showed an approximate 3-fold increase over the basal activity of the derepressed E. coli 83-1 lysate. This result may indicate that the amount of ThiGH present in the derepressed 83-1 lysate is limiting the rate of thiamine biosynthesis and provides for the first time evidence for in vitro thiazole synthase activity of the ThiGH-His complex.

Precursor Dependence of Thiazole Synthase Activity
Having established an assay for thiazole synthase activity by using purified ThiGH-His and the derepressed E. coli 83-1 lysate, the concentration dependence of the activity was determined for the known thiazole precursors Dxp, Tyr, and Cys. To remove any potential interference from endogenous amounts of these metabolites, the derepressed 83-1 cell lysate was desalted using a NAP-10 column, and the protein-containing high molecular weight fraction was isolated. Initial experiments with this derepressed 83-1 protein fraction showed that it retained at least the same level of activity of the unfractionated lysate (15-20 pmol/mg of protein/h), providing a further simplification of the system for in vitro studies. The precursor-dependent activity was investigated by varying the concentrations of each of the known precursors, otherwise maintaining the standard assay conditions. In these experiments, the Cys concentration was increased from 0 to 2 mM and the Dxp concentration from 0 to 3.7 mM, but in both cases, the de novo thiamine production was almost completely unaffected by the presence of these two putative precursors (Fig. 6). Conversely, the thiazole synthase activity appeared to be dependent on the presence and concentration of Tyr up to ϳ0.2 mM. These results raised the interesting question of the actual source of the C 5 unit and sulfur atom that are incorporated into Thz-P during these experiments. The component that provides these atoms appears to co-elute with the proteins when the derepressed E. coli 83-1 lysate is fractionated on a NAP-10 gel filtration column. In this regard, it is interesting to note that the studies of Park et al. (19) on thiazole biosynthesis in B. subtilis have provided evidence for a C-terminal derivative of ThiS that contains the atoms originating from Dxp and Cys (Fig. 7). Such an enzyme-bound intermediate may therefore be common to the thiazole-forming pathways of tyrosine/ThiH-dependent organisms (such as E. coli) and glycine/ThiO-dependent organisms (such as B. subtilis).

ThiGH-His Dosage and Time Course, Implication for the Thiazole Biosynthesis
The effect of varying the concentration of ThiGH-His in the assay was investigated (Fig. 8). The amount of thiamine produced was approximately proportional to the ThiGH-His added up to a concentration of 2 mg/ml, after which additional ThiGH-His gave no increase and in fact decreased the overall yield of thiamine.
Earlier studies had shown the marked tendency of the ThiGH-His complex to precipitate (21). Careful observations of the assays revealed that a protein precipitate formed in all the reaction mixtures containing ThiGH-His but was more pronounced when ThiGH-His concentrations exceeded 1 mg/ml. This aggregation may, at least in part, account for the decreased activity in assays containing a higher concentration of ThiGH-His. Despite this clearly observable protein precipitation, a time course experiment demonstrated that thiamine biosynthesis occurred in an approximately linear manner (rate ϭ 8 pmol/mg of protein/hour) up to 2 h and continued at a reduced rate (rate ϭ 2 pmol/mg of protein/h, Fig. 9) up to at least 6 h.

Effect of AdoMet and Reducing Agents
ThiH shares the Fe-S cluster-binding motif, Cys-XXX-Cys-XX-Cys, with members of the radical AdoMet family (26), whose activity is dependent upon the presence of AdoMet. The effects of adding AdoMet and potential reducing agents, dithionite and NADPH, on the formation of thiamine were investigated (Figs. 10 and 11). Whereas AdoMet alone gave a relatively modest increase in activity, the addition of AdoMet with a reducing agent more than doubled the observed thiazole synthase activity. Iron and sulfide have been used to reconstitute the Fe-S clusters of several "radical AdoMet" enzymes (53-55) and were added to these thiazole synthase assays in an effort to maximize activity. FIG. 10. Effect of AdoMet, reducing agents, and reconstitution on thiazole synthase activity. Sample 1, assay containing derepressed 83-1 protein fraction, ThiGH-His, Tyr, Hmp and ATP; sample 2, as assay 1 plus AdoMet; sample 3, as assay 1, but using ThiGH-His reconstituted with Fe(III) and sulfide, plus AdoMet and dithionite; sample 4, as assay 3, but substituting NADPH for sodium dithionite. Errors are estimated at less than 5%. It seems likely that the background thiazole synthase activity observed in our initial assays was dependent on endogenous AdoMet. The addition of high concentrations of ATP, to permit the pyrophosphorylation of Hmp to Hmp-PP, may have had the additional beneficial effect of allowing AdoMet to be synthesized in situ by AdoMet synthetase, using methionine presumably scavenged from protein degradation.
The stimulation of the thiazole synthase reaction by AdoMet and NADPH together with the observed low stoichiometry of turnover (0.01-0.03 mol of thiamine/mol of ThiGH-His) are similar to the properties of other proteins that belong to the radical AdoMet family of enzymes (53,56), although rigorous proof of this relationship will require a fully defined biosynthetic system. Several possible mechanisms involving AdoMet and the Fe-S cluster of ThiH can be envisaged leading to thiazole formation. One possibility is to use the AdoMet-derived deoxyadenosyl radical to initiate the formation of a tyrosinyl radical that leads to C ␣ -C ␤ bond cleavage (Fig. 12). Further experiments will be required to determine whether this side chain cleavage is directly coupled to the cyclization or leads to the formation of the glycine-derived imine that has been identified as an intermediate in B. subtilis by Park et al. (19).

Conclusions
For the first time, using lysates or partially purified protein samples and an enzyme-coupled assay, the thiazole synthase activity of E. coli has been measured. By adding purified ThiGH-His to these assays, we have demonstrated that, under the assay conditions, this complex is involved in the ratelimiting step. The reaction is completely dependent on the addition of tyrosine, a known Thz-P precursor. A further component that co-purified with the proteins derived from a lysate of adenosine-treated E. coli 83-1 was also required and is pre-sumed to provide the C 5 unit and sulfur for which Dxp and cysteine are known to be the original metabolic precursors. The thiazole synthase activity is stimulated by the addition of AdoMet and a reducing agent, with NADPH proving more effective than dithionite.
These results are consistent with our working model in which the thiazole is formed from tyrosine and an as yet uncharacterized intermediate by the ThiGH complex in a AdoMet/ NADPH-dependent reaction. Experiments to isolate and characterize the putative intermediate and to define the role of AdoMet in the thiazole synthase reaction are under way.
FIG. 12. A potential mechanism for the formation of Thz-P in E. coli. The deoxyadenosyl radical generates a tyrosinyl radical 2, which undergoes homolytic cleavage of the C ␣ -C ␤ bond to release the quinone methide 4. Hydration of 4 leads to p-hydroxybenzyl alcohol 5, a known thiamine biosynthetic by-product. The glycinyl radical 3 loses an electron to an as yet unidentified electron acceptor, to form the imminium ion 6. Reaction with the thio-sugar 7 is followed by cyclization to 9 and then dehydration and decarboxylation to Thz-P 10.