Riboflavin Synthase of Escherichia coli

Conserved amino acid residues of riboflavin synthase from Escherichia coli were modified by site-directed mutagenesis. Replacement or deletion of phenylalanine 2 afforded catalytically inactive proteins. S41A and H102Q mutants had substantially reduced reaction velocities. Replacements of various other conserved polar residues had little impact on catalytic activity.19F NMR protein perturbation experiments using a fluorinated intermediate analog suggest that the N-terminal sequence motif MFTG is part of one of the substrate-binding sites of the protein.

Conserved amino acid residues of riboflavin synthase from Escherichia coli were modified by site-directed mutagenesis. Replacement or deletion of phenylalanine 2 afforded catalytically inactive proteins. S41A and H102Q mutants had substantially reduced reaction velocities. Replacements of various other conserved polar residues had little impact on catalytic activity. 19

F NMR protein perturbation experiments using a fluorinated intermediate analog suggest that the N-terminal sequence motif MFTG is part of one of the substrate-binding sites of the protein.
Riboflavin synthase catalyzes a mechanistically complex dismutation of 6,7-dimethyl-8-ribityllumazine (Compound 1) affording riboflavin (Compound 3) and 5-amino-6-ribitylamino-2,4(1H,3H)-pyrimidinedione (Compound 4) ( Fig. 1) (1-3). The enzyme is a potential target for the chemotherapy of infections by Gram-negative bacteria which are unable to absorb riboflavin from the environment due to the absence of a transport system and are therefore absolutely dependent on its endogenous synthesis. Specifically, it has been shown that riboflavindeficient mutants of Escherichia coli and Salmonella typhimurium require riboflavin concentrations well above 10 mg/ liter of culture medium (7,8). Similar levels of the vitamin are required for the growth of riboflavin-deficient mutants of the yeasts Saccharomyces cerevisiae and Candida guilliermondii (9 -11). Inhibitors of riboflavin biosynthesis may therefore also qualify as potential agents for the therapy of yeast infections, which are becoming progressively more important in immunocompromised patients.
The riboflavin synthases of eubacteria and yeasts are homotrimeric proteins with an approximate subunit mass of 24 kDa. The sequence similarity between the N-terminal and Cterminal part of the enzyme suggested that each subunit folds into two domains with similar folding topology (12). In line with this hypothesis, preliminary crystallographic analysis of ribo-flavin synthase from E. coli provided evidence for pseudo d 3 symmetry of the homotrimeric molecule (13).
The reaction mechanism of the enzyme is incompletely understood despite the efforts of several research groups extending over a period of 4 decades (for review, see Refs. 6,and 14). The dismutation reaction involves the transfer of a 4-carbon unit between the two identical substrate molecules and requires the simultaneous presence of two 6,7-dimethyl-8-ribityllumazine molecules at the active site of the enzyme (15)(16)(17). This is well in line with the hypothesis of two topologically similar folding domains and with the experimentally observed binding of two substrate molecules per enzyme subunit.
Recently, the two-domain hypothesis received considerable support by the recombinant expression of N-terminal and Cterminal segments of riboflavin synthase from E. coli. Both recombinant proteins were shown to bind the intermediate analog, Compound 5, but not its diastereomer, Compound 6 ( Fig. 2) 1 (19). Moreover, the N-terminal domain could bind riboflavin and 6,7-dimethyl-8-ribityllumazine with high affinity. Surprisingly, the N-terminal domain was found to form a homodimer in contrast to the heterotrimeric riboflavin synthase.
The three-dimensional structure of riboflavin synthase is still unknown despite considerable efforts. Therefore, we decided to use systematic mutation analysis to identify catalytically relevant amino acid residues.
Construction of a ribC Hyperexpression Strain-The ribC gene coding for riboflavin synthase of E. coli was amplified by PCR using the plasmid pERS (22) as template and oligonucleotides 1 and 2 as primers ( Table I). The amplificate was purified, digested with XbaI and PstI, * This work was supported by grants from the Deutsche Forschungsgemeinschaft, the Fonds der Chemischen Industrie, the Hans Fischer Gesellschaft, the Alexander von Humboldt-Stiftung (to B. I.), and National Institutes of Health Grant GM51469. 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.
‡ On leave of absence from the Institute for Biophysics, Krasnoyarsk, Russia.
Bacterial Culture-Recombinant E. coli strains were grown in LB medium at 37°C in shaking flasks to an optical density of 0.7. Isopropyl-␤-D-thiogalactopyranoside was added to a concentration of 0.5 mM, and incubation was continued at 37°C for 4 h. The cells were harvested by centrifugation and stored at Ϫ75°C.
Site-directed Mutagenesis-A procedure modified after Marini et al. (24) was used for site-directed mutagenesis. PCR was performed with Vent DNA Polymerase (New England Biolabs, Schwalbach, Germany) to minimize the error rate. Plasmid pERS was used as template. The internal mismatch primers are shown in the Table I.
The general scheme of mutagenic PCR involved two rounds of amplification cycles using one mismatch and two flanking primers (primers 1 and 2, Table I). During the first round, 5 amplification cycles were carried out with the respective mismatch primer and with one of the flanking primers. Then the second flanking primer was added and the reaction was continued for an 10 additional cycles. The amplificate was subjected to agarose gel electrophoresis, digested with XbaI and PstI, purified using the QIAquick PCR Purification Kit, and ligated into plasmid pT7-7 that had been digested with the same restriction enzymes. The ligation mixture was transformed into E. coli XL1-blue cells (Stratagene, Heidelberg, Germany). Mutant clones were identified by the presence of the respective new restriction site. All plasmid constructs were sequenced by the automated dideoxynucleotide method (Sanger) using a 377 Prism DNA sequencer from Applied Biosystems (Weiterstadt, Germany).
Protein Purification-Purification procedures were performed at 4°C, unless otherwise stated. Frozen cell mass (5 g) was thawed in 25 ml of 50 mM Tris hydrochloride, pH 7.2, containing 0.5 mM EDTA and 0.5 mM dithiothreitol (buffer A). The suspension was subjected to ultrasonic treatment and was then centrifuged. The supernatant was dialyzed against 10 volumes of buffer A and centrifuged. The supernatant was passed through a column of Q-Sepharose Fast Flow (2 ϫ 18 cm) pre-equilibrated with buffer A (flow rate, 1 ml min Ϫ1 ). The column was washed with 100 ml of buffer A and developed with a linear gradient of 0 -0.5 M NaCl in buffer A (total volume, 280 ml). Riboflavin synthase was eluted from 200 to 240 ml. The enzyme fraction was brought to a concentration of 1 M (NH 4 ) 2 SO 4 by the slow addition of an equal volume of 2 M (NH 4 ) 2 SO 4 in buffer A. The solution was passed through a column of phenyl-Sepharose CL-4B (1.5 ϫ 12 cm) pre-equilibrated with 1 M (NH 4 ) 2 SO 4 in buffer A (flow rate, 1 ml min Ϫ1 ). The column was washed with 40 ml of 1 M (NH 4 ) 2 SO 4 in buffer A and developed with a linear gradient of 1.0 -0 M (NH 4 ) 2 SO 4 in buffer A (total volume, 120 ml). The enzyme was eluted from 80 to 95 ml. Fractions were combined, concentrated by ultrafiltration, and dialyzed against 20 mM sodium/potassium phosphate, pH 6.9, containing 0.2 mM EDTA and 0.5 mM dithiothreitol (buffer B). The dialyzed solution was stored at Ϫ75°C. According to SDSϪpolyacrylamide gel electrophoresis (25), the protein samples contained less than 5% impurities. The oligomeric structure of mutant proteins was determined using a 6% polyacrylamide nondenaturing gel electrophoresis system as previously described (26). The protein concentration was determined photometrically (⑀ 280 ϭ 47, 700 M Ϫ1 cm Ϫ1 ).
Monitoring of Riboflavin Formation-Reaction rates were measured using a stopped-flow module SFM-4 and Diode Array Spectrometer (Bio-Logic, Claix, France) at 20°C. All mixing solutions were prepared in buffer B. The reaction was initiated by rapid mixing of equal volumes of 8 M enzyme solution and 25-160 M 6,7-dimethyl-8-ribityllumazine solution. The reference absorbance spectrum of 4 M protein solution in reaction buffer was automatically subtracted from the experimental data. Absorbance spectra were collected and analyzed using Spectralys software from J&M Analytische Meß-und Regeltechnik GmbH (Aalen, Protein Perturbation Experiments-Samples contained 20 mM sodium/potassium phosphate, pH 6.9, 0.2 mM EDTA, 0.5 mM dithiothreitol, 10% D 2 O and protein (10.5Ϫ35.0 mg ml Ϫ1 ). Aliquots (5-30 l) of 6,7bis(trifluoromethyl)-8-ribityllumazine hydrate (epimer A) were added. After each addition, 19 F NMR spectra were recorded at 338 MHz using an AM 360 NMR spectrometer from Bruker Instruments (Karlsruhe, Germany) at 24°C. Experimental parameters were as follows: pulse angle, 30°(2 s); repetition rate, 1.6 s; 32 K data set; 800 to 3000 scans. Chemical shifts were referenced to an external standard containing sodium trifluoroacetate, pH 7.0. The concentrations of bound and free ligands were calculated from 19 F NMR signal integrals.

Construction of Riboflavin Synthase
Mutants-Plaut, Wood, and co-workers (4,15,27) had proposed a hypothetical mechanism for riboflavin synthase involving the initial addition of a nucleophile to the lumazine molecule serving as donor of the 4-C unit which is transferred between the two substrate molecules in the enzyme-catalyzed reaction (Fig. 1). More specifically, this hypothetical nucleophile could be either a water molecule or a polar amino acid residue. To check this hypothesis, we decided to mutate polar amino acid residues in search of the hypothetical nucleophile thought to be involved in covalent catalysis.
Sequence comparison of all known or presumed riboflavin synthases of the fungal/eubacterial type showed that the similarity between the proteins from different species was relatively low (data not shown). More specifically, the enzymes from E. coli and the closely related Hemeophilus influenzae share 63% identical amino acid residues. Otherwise, the fraction of identical amino acids was in the range of 29 to 38%.
In light of the relatively low overall conservation of the riboflavin synthase sequence, it was most notable that all enzymes invariably had the N-terminal sequence MFXG. A functional role of this sequence segment appeared likely, and it was therefore included in the mutagenesis study.
Site-specific mutagenesis was performed by PCR and generated mutant genes were checked for sequence accuracy by restriction analysis of novel restriction sites and by full-length sequence analysis. Five mutated genes (C48S, T50R, T67R, T148R, and T165R) could not be expressed in recombinant E. coli strains. It was tentatively assumed that the mutated amino acid residues affected the folding and overall conformation rather than the active site of the enzyme, and the respec-  tive mutant genes were not studied in more detail. Four mutant genes (Y133A, D143G, D143N, and N181G) ( Table I) afforded soluble proteins that were too unstable to be purified. One mutant gene (T3R) could be expressed to a level around 3% of cellular protein in recombinant E. coli host strains. All other mutant genes listed in Table I could be expressed to a level of ϳ20% of cellular protein and were purified to a level of at least 95%. The migration properties of all mutants as well as the wild type enzyme, in gel electrophoresis under nondenaturing conditions were very similar (data not shown). Riboflavin Formation-The catalytic activity of the mutant proteins was studied by single turnover experiments using a stopped-flow photometer. The respective mutant protein and the substrate, 6,7-dimethyl-8-ribityllumazine, were rapidly mixed at a molar ratio of 6:1 to 20:1 and the absorbency at 470 nm was recorded. At this wavelength, the absorbency contribution of 6,7-dimethyl-8-ribityllumazine is low, and the recorded signal is dominated by the concentration of riboflavin formed. The maximal rate of riboflavin formation (V max ) in the case of the wild type enzyme was 21 nmol mg Ϫ1 min Ϫ1 (Table  II), which is equivalent to a turnover number of 0.5 per min and subunit.
A substantial number of absolutely or highly conserved, polar amino acid residues could be exchanged without significant impact on the catalytic rate. Thus, the replacement of residues Thr 3 , Asn 45 , Glu 66 , Thr 71 , Asn 83 , Glu 85 , His 97 , Lys 137 , Ser 146 , and Glu 183 decreased the reaction velocity by less than 10-fold (Table II).
The replacement of serine 41 by alanine was accompanied by a substantial reduction in catalytic capacity. V max decreased to 3 pmol mg Ϫ1 min Ϫ1 . Similarly, the replacement of histidine 102 by glutamine afforded a mutant protein with a decreased V max of 83 pmol mg Ϫ1 min Ϫ1 .
The replacement of the absolutely conserved phenylalanine 2 by alanine afforded soluble protein that had no detectable catalytic activity. No riboflavin whatsoever could be observed even after long-term incubation with substrate. Similarly, the deletion of phenylalanine 2 resulted in the complete loss of catalytic activity. The more conservative replacement of phenylalanine by tyrosine decreased V max of the enzyme 50-fold, to 0.42 nmol mg Ϫ1 min Ϫ1 .
Protein Perturbation Experiments-In the absence of protein structural data, we decided to perform protein perturbation studies to characterize the interaction of the mutant proteins with a ligand that binds to the catalytic site of riboflavin synthase. The fluorinated Compounds 5 and 6 have been used extensively in ligand binding studies with riboflavin synthase and lumazine synthase of Bacillus subtilis (17,20,28,29). Fluorine substitution favors the covalent hydration of the pteridine ring system to such an extent that the diasteromeric Compounds 5 and 6 are not subject to epimerization. The covalent hydrate structure mimics the hypothetical intermediate, Compound 2, in the reaction mechanism proposed by Plaut, Wood, and co-workers ( Fig. 1) (4, 15, 27).
Riboflavin synthase of B. subtilis has been shown to bind Compound 5 at a molar ratio of 1:1 per subunit (17,20). Binding of Compound 5 to the enzyme was accompanied by downfield shifts of the 19 F NMR signals of both trifluoromethyl groups. Compound 6 (epimer B) could not be bound even at high concentration (20).
A titration experiment performed with E. coli riboflavin synthase and Compound 5 is shown in Fig. 4. Aliquots of the ligand were added, and 19 F NMR spectra were recorded after each addition. At the early stages of the titration experiment (Fig. 4,  a and b), the 19 F NMR signals of the free ligand are very low in intensity, and the spectra are dominated by the signals of the enzyme-bound ligand. With progressive saturation of the enzyme's binding sites, the signals of the free ligand increase. The position 7 trifluoromethyl group of the enzyme-bound ligand appears as a relatively broad signal with a line width of 180 Hz which is 1.4 ppm downfield shifted by comparison with the signal of the free ligand. The position 6 trifluoromethyl group of the enzyme-bound ligand affords a more complex signal pattern. At low ligand concentrations (Fig. 4a), a relatively intense signal with a line width of 160 Hz is observed at 14.2 ppm and a signal of lower intensity with a line width of 90 Hz at 15.7 ppm. The latter signal increases with progressive addition of ligand, thus suggesting that the enzyme has two different types of binding sites whose affinities to Compound 5 differ significantly (Table III).
The quantitative evaluation of the signal integrals afforded the nonlinear Scatchard plot shown in NMR signal integrals affords a K D value of 30 M for the low affinity site (Fig. 5B).
These data are well in line with our recent finding that the recombinant N-terminal and C-terminal domains of E. coli riboflavin synthase can both bind Compound 5 (Fig. 4, Table  III). 1 The comparison between the spectra of the wild type enzyme and the domains suggests tentatively that the ligand signal at 15.7 ppm is associated with the N-terminal domain of the wild type enzyme, whereas the signal at 14.2 ppm is associated with the C-terminal domain.
Titration experiments with Compound 5 were performed with all mutants listed in Table II. In most cases, the spectra were qualitatively similar to that obtained with the wild type enzyme. K D values obtained from 19 F NMR titration differ from that of the wild type enzyme by less than a factor of 10 (Table  II) and the number of ligand molecules bound per protein molecule (trimer) varied from 5.2 to 6.3 (relative standard error, 6 -15%). The few exceptional cases are discussed in detail below.
An NMR titration experiment with the catalytically inactive F2⌬ mutant with a deletion of phenylalanine 2 is shown in Fig.  6. The position 7 trifluoromethyl group of the bound ligand affords two signals with a distance of 0.9 ppm, respectively (Table III). On the other hand, the signals of the position 6 trifluoromethyl group are located very close to each other with a distance of 0.4 ppm (Fig. 6). The signal pattern suggests that the mutant retains the capacity to bind one ligand molecule each to the N-terminal and C-terminal domain. The signals of   (Table III). Titration experiments with the F2⌬ mutant indicate that the different binding sites bind the ligand with different affinity (Table III). The low affinity site is characterized by a dissociation constant of 57 M. The dissociation constant for the high affinity site cannot be measured accurately for experimental reasons (the NMR signals of the free ligand are too small to be integrated in the early titration steps; the value of the dissociation constant is estimated to be less than 1 M).
The F2A mutant is very similar to the F2⌬ mutant. It is catalytically inactive, and it modulates the NMR spectrum of Compound 5 in the same way as the deletion mutant (Table  III). However, the conservative replacement of phenylalanine 2 by tyrosine affords a catalytically active protein (Table II) whose interactions with Compound 5 are similar to that of the wild type protein.
Apart from the phenylalanine 2 modifications, only a small number of mutations analyzed in this study were conducive to substantial modification of the interaction with Compound 5 (Table III, Fig. 7). In two cases (K137A and S146G), the catalytic activity was reduced by approximately 1 order of magnitude, whereas the E183G mutant had almost normal catalytic activity. In all three mutants, two signals are clearly resolved for the position 7 CF 3 group of the enzyme bound ligand. In the case of S146G mutant, the signal of the position 6 CF 3 group assumed to represent the ligand in association with the Cterminal domain is shifted upfield by comparison with the wild type enzyme. In the K137A and E183G mutants, there is tentative evidence for three enzyme bound species giving rise to additional signals for the position 6 CF 3 group.
Comparison of the NMR chemical shift values shown in Table III indicates that, except of Glu 183 , replacements of other amino acids in N-or C-teminal halves of riboflavin synthase affected exclusively (Lys 137 and Ser 146 ) or largely (Phe 2 ) the NMR pattern afforded by the ligand molecule bound to the same half of the protein. The fact that all mutations listed in Table III have a visible effect on the appearance of the 6-and 7-trifluoromethyl groups of the ligand, relative to the wild type enzyme, suggests that residues 2, 137, 146, and 183 may be in close proximity to the ligand and to each other. DISCUSSION The reaction catalyzed by riboflavin synthase involves the cleavage of two CN bonds and the formation of two CC bonds. It appears obvious that the reaction trajectory must involve a complex series of intermediates. Surprisingly, however, the reaction can proceed without enzyme catalysis under relatively mild conditions in aqueous solution (30). Notably, the catalyzed and the uncatalyzed reactions share the same regiospecificity (4,15,18).
The hypothetical reaction mechanisms suggested by Wood, Plaut, and co-workers (3)(4)(5)(6) implicate the addition of a nucleophile to one of the substrate molecules as an early reaction step. In the uncatalyzed reaction, a water molecule could serve this purpose. Either a water molecule or a polar amino acid side chain could serve as a nucleophile in the enzyme-catalyzed reaction.
The protein perturbation data provide evidence for the existence of two different types of binding sites whose affinities for Compound 5 differ by approximately 1 order of magnitude. Evidence for two different types of binding sites had been obtained earlier by studies with riboflavin synthase from B. subtilis and 6-trifluoromethyl-7-oxo-8-ribityllumazine (16,28), but the 19 F NMR spectra produced by Compound 5 in this case not only differed substantially from those shown in Fig. 4, but also gave the molar ratio of 1:1 of bound ligand per subunit, indicating that only one type of binding site could bind the ligand. E. coli and B. subtilis proteins share only 33% identical amino acids (data not shown), so the structural difference between the ligand-binding sites of these two enzymes that leads to the different NMR signal pattern seems plausible.
We have shown recently that the N-terminal and C-terminal parts of the riboflavin synthase of E. coli can be expressed as recombinant proteins. 1 Specifically, the N-terminal domain was shown to form a relatively stable, soluble homodimer whereas the full-length riboflavin synthase forms a homotrimer. The recombinant C-terminal domain has comparatively low stability and could not be obtained in pure form. Both recombinant domains were shown to bind Compound 5 by protein perturbation experiments similar to those reported in the present study. A comparison of the chemical shift values indicates that binding of Compound 5 to the N-terminal domain is associated with a relatively large downfield shift of the position 6 trifluoromethyl signal whereas the binding to the C-terminal domain has a more pronounced effect on the signals from the position 7 trifluoromethyl group.
The replacement of serine 41 or histidine 102 reduced the catalytic activity of the enzyme by several orders of magnitude. The mutations did not decrease substantially the affinity of the enzyme for the putative intermediate analog, Compound 5. Moreover, the mutations appeared to have little impact on the environment of enzyme-bound Compound 5 as gleaned by 19 F NMR spectrometry. If either of these amino acids were involved in covalent catalysis, the residual activity of the mutant proteins could be due to a water molecule replacing the amino acid nucleophile. Whereas the data are insufficient to prove a direct catalytic role of serine 41 or histidine 102, they rule out covalent catalysis by any of the other polar amino acid residues addressed in the present mutagenesis study.
Whereas this study was primarily focused on conserved polar amino acid residues, the inclusion of the lipophilic N terminus appeared mandatory in light of the absolute conservation of the terminal MFXG motif. Deletion or nonconservative replacement of phenylalanine 2 afforded completely inactive proteins. Although these mutants could still bind two lumazine-type molecules, the chemical shifts of the ligand bound at the putative N-terminal-binding site were modulated significantly. This finding suggests that the N terminus of the polypeptide forms an essential part of the active site.