Identification of the [Fe-S] cluster-binding residues of Escherichia coli biotin synthase.

The gene encoding Escherichia coli biotin synthase (bioB) has been expressed as a histidine fusion protein, and the protein was purified in a single step using immobilized metal affinity chromatography. The His(6)-tagged protein was fully functional in in vitro and in vivo biotin production assays. Analysis of all the published bioB sequences identified a number of conserved residues. Single point mutations, to either serine or threonine, were carried out on the four conserved (Cys-53, Cys-57, Cys-60, and Cys-188) and one non-conserved (Cys-288) cysteine residues, and the purified mutant proteins were tested both for ability to reconstitute the [2Fe-2S] clusters of the native (oxidized) dimer and enzymatic activity. The C188S mutant was insoluble. The wild-type and four of the mutant proteins were characterized by UV-visible spectroscopy, metal and sulfide analysis, and both in vitro and in vivo biotin production assays. The molecular masses of all proteins were verified using electrospray mass spectrometry. The results indicate that the His(6) tag and the C288T mutation have no effect on the activity of biotin synthase when compared with the wild-type protein. The C53S, C57S, and C60S mutant proteins, both as prepared and reconstituted, were unable to covert dethiobiotin to biotin in vitro and in vivo. We conclude that three of the conserved cysteine residues (Cys-53, Cys-57, and Cys-60), all of which lie in the highly conserved "cysteine box" motif, are crucial for [Fe-S] cluster binding, whereas Cys-188 plays a hitherto unknown structural role in biotin synthase.

The biotin operon of Escherichia coli contains six open reading frames that encode at least four of the proteins essential for the conversion of pimeloyl-CoA to biotin (1). The recent determination of the structure of three of these enzymes, 8-amino-7-oxononanoate synthase (2), 7,8-diaminononanoate synthase (3), and dethiobiotin synthase (4), encoded by the bioF, bioA, and bioD genes respectively, has shed light on the mechanism of the steps involved in the conversion of alanine to dethiobi-otin. However, the mechanism of the final step, the conversion of dethiobiotin to biotin catalyzed by biotin synthase (see Scheme 1), remains unresolved.
Numerous studies have shown that a fully functional in vitro system for conversion of dethiobiotin to biotin requires not only the product of the bioB gene but also several other proteins and low molecular weight molecules. Although the exact components of this system vary between different laboratories, Sadenosylmethionine (AdoMet), 1 NADPH, cysteine, DTT, Fe 2ϩ , and a reducing system (5) are common to all. The reducing system we and others use consists of flavodoxin (FLD) (6) and flavodoxin (ferredoxin) NADP ϩ oxidoreductase (FLDR) (7) (which we believe to be the physiological system) which Marquet and co-workers (8) have found can be replaced by a photoreduced deazaflavin. Both systems appear to be equally effective in providing reducing equivalents for the reaction. There is, however, conflicting evidence that other proteins, cofactors, and allosteric activators may also be required for competence of the biotin synthase complex. There is no doubt that addition of extracts derived from E. coli bioB Ϫ strains can enhance activity of biotin synthase, but whether this is due to the presence of a thiamine pyrophosphate-stabilized protein (7), high levels of constitutive low molecular weight cellular components such as fructose 1,6-biphosphate and the labile (and as yet uncharacterized) AdoMet-derived product of the 7,8-diaminononanoate synthase reaction (5), or indeed to a combination of several of these factors is still open to question.
Recent results indicate that the sulfur incorporated in the tetrahydrothiophene ring of biotin is ultimately derived from cysteine (7) and that 9-mercaptodethiobiotin (but not 6-mercaptodethiobiotin) can act as an intermediate (9). The intermediacy of 9-mercaptodethiobiotin, racemization at C-9 (10) (but not at C-6 (11)), and the requirement for at least two molecules of AdoMet for each ring formation step provides clues to the mechanism of C-S bond formation (12). Taken together these results suggest that the mechanism involves the following: (a) initial AdoMet-dependent radical formation at C-9; (b) capture of sulfur by the C-9 radical to form a 9-mercaptodethiobiotin derivative; and (c) a second AdoMet-initiated radical formation by abstraction of the pro-S hydrogen at C-6 to generate a conformationally restrained radical that subsequently captures the C-9 thiol resulting in tetrahydrothiophene ring clo- 1 The abbreviations used are: AdoMet, S-adenosylmethionine; CV, column volume; DTT, dithiothreitol; FLD, E. coli Flavodoxin; FLDR, E. coli flavodoxin (ferredoxin) NADP ϩ oxidoreductase; His 6 , six consecutive histidine residues; ICP-AES, inductively coupled plasma-atomic emission spectroscopy; IMAC, immobilized metal affinity chromatography; IPTG, isopropyl-␤,D-thiogalactopyranoside; LAM, lysine 2,3-aminomutase; LIPA, lipoic acid synthase; PFL-AE, pyruvate formate lyaseactivating enzyme; ARR-AE, ribonucleotide reductase-activating enzyme; PAGE, polyacrylamide gel electrophoresis; Bis-Tris, 2-[bis(2hydroxyethyl)amino]-2-(hydroxymethyl)propane-1,3-diol; PCR, polymerase chain reaction. sure (13). Recent studies indicate that a sulfur atom of the [Fe-S] cluster of biotin synthase, rather than cysteine itself, may act as the actual sulfur donor, suggesting that the [Fe-S] cluster of the enzyme must be regenerated prior to the next reaction (14). Thus biotin synthase could be regarded as both a reagent and a catalyst. It has been suggested that a NifS-like enzyme, similar to the Azotobacter vinelandii NifS (15), may be required for construction of the [Fe-S] cluster in vivo and regeneration in vitro but such an activity has still to be identified. NifS is a pyridoxal phosphate-dependent enzyme that catalyzes the desulfuration of L-cysteine to yield L-alanine and sulfide that has been shown to catalyze [Fe-S] cluster formation in vitro and in vivo.
The fact that E. coli biotin synthase is based on a core [2Fe-2S] protein with a requirement for a FLD/FLDR redox couple and AdoMet as a radical generator suggests that it belongs to the family of enzymes that include anaerobic ribonucleotide reductase-activating enzyme (ARR-AE) (16), pyruvate formate-lyase-activating enzyme (PFL-AE) (17), lysine 2,3-aminomutase (LAM) (18), and lipoic acid synthase (LIPA) (19). Even within this small group there already appears to be differences with regard to subunit composition and chemical function; biotin synthase, LIPA, and ARR-AE are dimers; LAM is a hexamer, and PFL-AE is a monomer. Biotin synthase, LIPA, and LAM generate a radical on a small molecule, whereas ARR-AE and PFL-AE generate a protein glycyl radical.
Characterization of the E. coli biotin synthase has been complicated by the fact that most preparations of the protein contain variable amounts of polymeric forms (mainly dimers and tetramers) with variable [Fe-S] content. The major native form of the core biotin synthase, however, appears to be a 76-kDa dimer, each monomer containing an [2Fe-2S] cluster (20). Spectroscopic studies suggest that reduction of the two [2Fe-2S] clusters in the oxidized protein dimer, an obligatory step in the mechanism, results in formation of a biotin synthase dimer containing a single [4Fe-4S] cluster in which each of the iron centers is coordinated to a thiol ligand(s) of the protein (21)(22)(23).
However, almost nothing is known about the protein residues involved in coordination to the [Fe-S] clusters of biotin synthase. Whereas the active sites have been proposed to involve the cysteine thiols of the common GXCXXXCXXCXQ motif as a "cysteine ([Fe-S] coordination) box" motif, there has been as yet no experimental evidence to support this contention (20). In this paper we describe studies on mutants of the biotin synthase protein that uniquely identify the key protein cysteine residues involved in formation of the [Fe-S] cluster.

EXPERIMENTAL PROCEDURES
Materials-Electrophoresis was carried out using a Bio-Rad Protean II minigel system (protein) and a Life Technologies, Inc., H5 system (DNA). PCR and sequencing reactions were done on a Perkin-Elmer 480 thermal cycler, and automated DNA sequencing was carried out using an ABI prism 377 DNA sequencer. An Amersham Pharmacia Biotech FPLC system and columns were used for chromatographic separations of proteins. Electrospray mass spectrometry was performed on a Micromass Platform II quadrupole mass spectrometer. UV-visible spectrophotometry was done on a Unicam UV4. Determination of metal ion concentrations were carried out using a Thermo Jarrell Ash IRIS inductively coupled plasma atomic emission spectrometer. Primers were purchased from Life Technologies, Inc., and Ready to Go PCR TM beads from Amersham Pharmacia Biotech. Pre-cast SDS-PAGE gels (10% Bis-Tris) were purchased from Novex and were used according to the manufacturer's instructions.
Mutagenesis-Site-directed mutagenesis was performed using the "mega-primer" PCR method using vector forward (BIOB PCR) and reverse (M13) oligonucleotides and the mutagenic primers (C53S, C57S, C60S, C188S and C288T) (25). The typical PCR reaction volume was 50 l and consisted of template DNA (1 l), forward primer (1 M), reverse primer (1 M), water (39 l), and two Taq beads. The primers for the cysteine mutations used were C53S (5Ј AAG ACC GGA GCT TCC CCG GAA GAT 3Ј), C57S (5Ј CCG GAA GAT TCT AAA TAC TGC CCG 3Ј), C60S (5Ј AAA TAC TCT CCG CAA ACG TCG CGC 3Ј), C188S (5Ј GGG ATC AAA GTC TCT TCT GGC 3Ј), and C288T (5Ј GGT CAG CAG TTT GGT ACC GTA 3Ј). The bases underlined in each primer indicate the position where the cysteine codon is replaced by one encoding a serine or in the case of C288T a threonine. The bases in bold indicate a KpnI restriction site. All clones were sequenced, and the data were analyzed using ABI Prism editview software.
Preparation of Biotin Synthase Cell-free Extracts-Wild-type biotin synthase, His 6 -tagged biotin synthase (from HMS174 (DE3)/pET16b/ bioB and pET6H/bioB), and mutant proteins (from HMS174 (DE3)/ pHC53S, pHC57S, pHC60S, pHC188S, and pHC288T) were prepared from cultures grown in 2-10 liters of 2YT media (tryptone (16 g/liter), yeast extract (10 g/liter), sodium chloride (5 g/liter), pH 7.5) containing ampicillin (100 g/ml). Transformed cells were grown at 37°C (shaking 250 rpm) until the A 600 nm ϭ 1.0 and biotin synthase was produced by addition of IPTG to a final concentration of 1 mM. Cells (ϳ2.5 g/liter wet weight) were harvested (5,000 ϫ g for 20 min, 4°C) after a further 4 h of growth. The cells pellets were washed by resuspension in ice-cold buffer A (20 mM Tris-HCl, pH 7.9, 0.5 M sodium chloride, 5 mM imidazole). All other cell pellets were washed in buffer B (100 mM sodium phosphate, pH 7.5). The cell pellets were lysed by intermittent sonication (30 s on, 30 s off) 4°C for 15 min. Cellular debris was removed by centrifugation (15,000 ϫ g for 30 min, 4°C), and the supernatants were stored frozen (Ϫ20°C) in glycerol (15%). The PCOi (bioB Ϫ ) strain was grown in 2YT, and the cells washed with buffer A and cell extracts, used for augmentation in the in vitro assay, were prepared in a similar fashion to that described above.
Purification of Wild-type Biotin Synthase-Protamine sulfate (1%, 0.5 ml per 10 ml) was added to the cell-free extract to remove nucleic acids. After ammonium sulfate (45%) precipitation, precipitated proteins were collected by centrifugation, resuspended in buffer B, and filtered (0.45 m) prior to being loaded onto a Q-Sepharose 26/10 high load anion exchange column (2.5 ϫ 10 cm) which had been equilibrated with buffer B. The proteins were eluted with an increasing gradient of potassium chloride (0 -1 M, 20 CV) in buffer B. The deep red fractions containing biotin synthase eluted at 300 mM salt. These were combined, brought to a final concentration of 10% ammonium sulfate, and loaded onto a phenyl-Sepharose hydrophobic interaction column (1 ϫ 30 cm) that had been equilibrated with buffer B containing ammonium sulfate (10%). Biotin synthase was eluted with a decreasing gradient of buffer B containing ammonium sulfate (10 -0%, 20 CV). Biotin synthase eluted at the end of the gradient, and these fractions were combined and filtration concentrated (Amicon PM10 membrane). The protein was purified by gel filtration on Sephacryl S-200 HR eluting with buffer B. Protein samples were diluted with glycerol (15%) and stored (Ϫ80°C).
Purification of His 6 -tagged Biotin Synthase and Mutants-Cell-free extracts were loaded directly onto a Hi-Trap TM -chelating column (5 ml, Amersham Pharmacia Biotech) that had been charged with NiSO 4 (100 mM) and prewashed with buffer A. Proteins were eluted with a linear gradient of (5 mM to 1 M) imidazole in buffer A. Biotin synthasecontaining fractions typically eluted between 120 and 150 mM imidazole. These were combined and exhaustively dialyzed against buffer B to remove imidazole and nickel salts and concentrated by ultrafiltration. SCHEME 1.
Mass Spectrometry-Prior to mass spectrometry all protein samples were passed through an Aquapure reverse phase C-4 column at a constant trifluoroacetic acid concentration of 0.1% using a linear gradient of 10 -100% acetonitrile in water over 40 min at a flow rate of 1 ml/min and the separation monitored at 280 nm. The total ion count of all the ions in the m/z range 500 -2,000 was recorded. The mass spectrometer was scanned at intervals of 0.1 s, the scans accumulated, spectra combined, and the average molecular mass determined using MaxEnt and Transform algorithms of MassLynx software.
Sequence Alignments-Swiss-Prot and the Genomes Representation Organization data bases were used to search for amino acid sequences that were then aligned using Expasy, Clustal W, and Alscript (27).
Preparation of Apo and Holo Wild-type Biotin Synthase and Mutants-This was carried out using methods previously described (14,20).
In Vitro Assay for Biotin Synthase Activity-In vitro biotin production was determined using the Lactobacillus assay (24). Incubations were carried out both aerobically and anaerobically in buffer B. A typical assay mixture contained biotin synthase or mutant protein (5 M), potassium chloride (10 mM The final reaction volume was typically 1 ml. The reaction was incubated at 37°C for 2 h and then stopped by adding 100 l of 10% trichloroacetic acid. The resulting precipitate was centrifuged, and 5 l of the supernatant was used for bioassay. In Vivo Assays for Biotin Synthase Activity-Transformants of HMS174 (DE3) with pET16b, pET6H, pET16b/bioB, pET6H/bioB, pHC53S, pHC57S, pHC60S, pHC188S, and pHC288T were grown in M9CA media containing dethiobiotin (5 g/ml), glucose (0.4%), thiamine (0.8 g/ml), MgSO 4 (2 mM), CaCl 2 (0.1 mM), and ampicillin (100 g/ml). At A 600 nm ϭ 1.0 the cells were equally divided, and one sample was kept as a control and the other was induced with IPTG (1 mM). Aliquots (1 ml) of cells were removed from both control and induced samples at various time intervals. The cells were pelleted by centrifugation, and 5 l of the supernatants were used in the bioassay as above. Supernatant from HMS174 (DE3) cells were grown in the same media minus ampicillin and used as the control.
Elemental Analyses-Inductively coupled plasma atomic emission spectroscopy (ICP-AES) was used for metal analyses. All data were interpreted using thermospec/CID software. The radiofrequency power was 1150 watts, the nebulizer flow was 30 pounds/square inch, the pump rate was 100 rpm, and the purge time was 90 s. Fe(NO 3 ) 3 (0.01-100 ppm; 180 nM to 1.8 mM) and Ni(NO 3 ) 2 (0.0001-1.0 ppm; 1.7 nM to 17 mM) solutions were used as standards. All biotin synthase samples (27 M) and standard solutions were made up in buffer B (28). Chemical analysis for iron was carried out using the ferrozine assay as described by Stookey (29) using the standard described above. Labile sulfide analysis was determined as previously reported (20).

RESULTS
Overexpression of Biotin Synthase-To study the molecular properties of wild-type and mutant forms of biotin synthase, we overproduced the protein using the pET system. In our hands we found the purification procedure described by Sanyal et al. (20) both technically difficult and time-consuming, although the procedure provided ϳ80% pure biotin synthase. We found it impossible to reproduce the purification of biotin synthase using a cobalt column, as described by Marquet and co-workers (30). To overcome these problems, we decided to express the protein with an N-terminal polyhistidine tag. The bioB gene was cloned into pET6H which fused the 5Ј terminus of the gene to a sequence encoding a MHHHHHHA tag. Overexpression of His 6 -tagged biotin synthase was carried out in HMS174 (DE3) and gave comparable expression levels when compared with wild-type biotin synthase. Optimal yields of both proteins were obtained by aerobic cell growth in 2YT at 37°C. By using the IMAC purification system, we were able to routinely prepare 100-mg quantities of Ͼ95% pure His 6 -tagged biotin synthase from 5 liters of transformed cell culture.
Essential Cysteine Residues of Biotin Synthase-Spectroscopic studies on biotin synthase, PFL activase, and LIPA (17,19,31) suggest that three cysteine residues from each monomer are required to coordinate the iron in the [2Fe-2S] clusters of the oxidized dimer and that two of these are required ligands in the reduced catalytically active form. Five cysteine residues of biotin synthase (Cys-53, Cys-57, Cys-60, Cys-188, and Cys-288) were replaced separately with serine or threonine residues by site-directed mutagenesis (threonine was chosen as a replacement for Cys-288 since it enabled us to engineer a KpnI restriction site into the bioB gene to facilitate mutant selection). Each of the mutant proteins was overproduced in E. coli HMS174 (DE3), and its catalytic activity was determined. Since the His 6 -tagged biotin synthase showed identical in vitro spectroscopic parameters and catalytic activity to wild-type protein, the mutant bioB genes were also expressed as histidine fusion proteins. Each of the mutant proteins was purified using IMAC under identical conditions to the histidine-tagged wild-type biotin synthase except for mutant C188S, which despite numerous attempts (low temperature induction and IPTG concentrations Ͻ0.1 mM), remained intractably insoluble. Protein purity was estimated by SDS-PAGE (Fig. 1).

Preparation of Wild-type and His 6 -tagged Apoenzymes and the Reconstitution of [2Fe-2S] Cluster Containing Holoenzymes-
The apoenzymes of each of wild-type biotin synthase, His 6 -tagged biotin synthase, and mutants were prepared anaerobically using sodium dithionite and EDTA. The samples were then desalted using a G10 gel filtration column. ICP-AES analysis and UV-visible spectroscopy confirmed the absence of iron in the proteins. The samples were then reconstituted by incubation with FeCl 3 and Na 2 S under similar conditions to those described previously (14,20).
Spectral Characteristics-Solutions of wild-type, His 6tagged, and C288T biotin synthase were all reddish brown in color, and they showed virtually identical UV-visible spectra with a maximum at 274 nm, a shoulder at 330 nm, and a distinctive peak at 420 nm (see Fig. 2). The spectra of the apo forms of each of these proteins were colorless and showed no peaks at 330 or 420 nm. Upon incubation with FeCl 3 and Na 2 S for 2 h, their spectra were identical to that of the wild-type isolate. The mutants (C53S, C57S, and C60S) were colorless, and their spectra showed no peaks indicative of the presence of an [Fe-S] cluster. Moreover, their spectra did not show any significant changes even after prolonged incubation with FeCl 3 and Na 2 S.
All the mass spectrometry results were within 0.1% of the expected values for the polypeptide chains (Table I)

indicating that the [Fe-S] cluster is lost under the MS conditions.
Elemental Analyses-Analyses for iron, sulfur, and nickel were carried out on all proteins using ICP-AES, the Beinert method, and the ferrozine method (29). The protein concentrations were adjusted to ϳ27 M (based on a monomeric mass of 39,673 Da for His 6 -tagged proteins and 38,648 Da for wild-type biotin synthase). The amount of sulfur (as S 2-) in His 6 -tagged biotin synthase was virtually identical to that in the untagged protein and similar to the value previously reported by Sanyal et al. (20) for wild type. Analysis for nickel content proved that there was negligible nickel binding. Iron analysis (Table II) on the purified proteins revealed a 1:1 ratio between iron and sulfide in both unmodified and His 6 -tagged biotin synthase in agreement with previously reported values (14), although it should be noted that Sanyal (20) has reported a 2:1 ratio. In order to try to increase the iron/protein ratio, the proteins were converted to their apoenzymes by incubation with sodium dithionite and EDTA, and the [Fe-S] clusters were reconstituted by incubation with FeCl 3 and Na 2 S. In the case of the wildtype, the His 6 -tagged, and C288T proteins, the protein/iron ratios increased to values comparable with those previously reported (20). The ferrozine assay gave a slightly lower iron/ protein ratio than that obtained by ICP-AES (Table III).
In Vitro Assay for Biotin Production-The in vitro activities of all proteins were determined using the Lactobacillus assay. Biotin was produced by the wild-type, the His 6 -tagged, and the C288T mutant but not by the three other cysteine mutants (Table IV). There was no difference in biotin production between aerobic and anaerobic incubations.
In Vivo Assay for Biotin Production-HMS174 (DE3) cells did not produce detectable levels of biotin either on their own or when transformed with pET16b and pET6H and were used as a control. To examine whether or not the His 6 tag affected biotin synthase activity, an in vivo assay was carried out using HMS174 (DE3) cells transformed with either pET16b/BioB or pET6H/bioB. All the cells grew in a similar fashion, and production of biotin was found to increase with cell growth. There was no difference in the cells transformed with pET16b/BioB when compared with those that were producing His 6 -tagged protein indicating that the tag does not seem to affect the activity of wild-type biotin synthase. Similar experiments were carried out with the five mutant plasmids, all the clones grew as vigorously to those of the wild-type, but only the non-conserved C288T mutant produced significant amounts of biotin. Cells expressing the C53S, C57S, C60S, or C188S mutants did not produce any detectable quantities of biotin (Table V). DISCUSSION Biotin synthase has two dimeric forms that contain different types of [Fe-S] clusters. In its oxidized form the protein contains a single [2Fe-2S] cluster in each monomer unit, but on reduction this converts to a single [4Fe-4S] cluster in the dimer. EPR and Raman spectroscopy studies have provided evidence that in the oxidized [2Fe-2S] form biotin synthase has one Fe 3ϩ that is coordinated to two cysteines and one Fe 3ϩ that is coordinated to a cysteine and an oxygenic ligand (21,22). In contrast, after dithionite reduction a [4Fe-4S] cluster is formed in which each iron is coordinated to a single cysteine sulfur. A possible mechanism for this interconversion has been proposed by Johnson et al. (22) (Fig. 3). It has been speculated that the cysteine residues of the highly conserved "cysteine box" GX-CXXXCXXCXQ motif of these proteins are involved in [Fe-S] cluster binding. This is supported by the fact that an identical motif is found in other [Fe-S] cluster proteins such as LIPA (Swiss-Prot accession number P25845) (19), ARR-AE (P39329) (32), LAM (33), the nitrogen fixation B gene product (P11067) (34), pyrrolo-quinoline-quinone synthase (Q01060) (35), PFL-AE (P09374) (17), NARA protein (P39757) (36), FNR (P03019) (37), benzylsuccinate synthase-activating enzyme (CAA05050) (38), spore photoproduct lyase (P37956) (39), molybdenum cofactor synthase (O27593) (40), and the thiamine synthase H gene product (P30140) (39). Mutational studies on PFL-AE have shown that the three cysteine residues of the cysteine box are vital for its activity (41).
Our attempts to purify wild-type biotin synthase from overexpressing clones using previously reported methods afforded protein that was ϳ90% pure contaminated with several high molecular weight contaminants. To circumvent this problem we elected to clone the bioB gene with an N-terminal His 6 tag and to purify the fusion using nickel-bound affinity chromatography. This appeared to be a safe approach since E. coli biotin synthase has a relatively long N-terminal sequence, which is either not present or unconserved in the ORFs of genes from other species, and the tag was unlikely to affect enzymatic activity.
A further significant result was that the activity of the N terminus His 6 -tagged biotin synthase proved indistinguishable from that of the wild-type protein. The His 6 tag allows purification of much larger quantities of homogeneous E. coli biotin synthase than previously available. Typically, protein of Ͼ95%  homogeneity can be isolated in a single chromatographic operation. Metal analyses on both the wild-type and the His 6 -tagged protein revealed that they contained the same amount of iron as that reported by Marquet and co-workers (14) but less than that reported by Sanyal et al. (20). This is possibly a result of the high protein expression levels. Shortage of free intracellular Fe 2ϩ and/or the fact that the enzyme involved in cluster formation in the cell is expressed at wild-type levels would lead to a proportion of the biotin synthase population being formed with incomplete clusters. In order to increase the iron content of biotin synthase, the apo-enzyme was prepared anaerobically by treatment of the proteins with sodium dithionite and EDTA, and the [2Fe-2S] cluster was reconstituted by incubation with FeCl 3 and Na 2 S. Although the concentration of biotin synthase with intact [2Fe-2S] clusters was significantly increased by this procedure, both forms produced equal amounts of biotin (0.2 mol biotin/mol protein/h) which is in agreement with reported values (8,12,20). This suggests that there is sufficient iron and sulfide in the assay mixture to reconstitute the cluster in situ. Clearly, biotin synthase is not acting as a true catalyst, and one reasonable explanation for this is that an unknown cofactor(s) (possible sulfur donor) is missing from the reaction mixture.
Sequence similarity searches reveal that there at least 20 known species of bacteria, yeasts, and plants that contain a bioB gene sequence similar to that from E. coli. To date, biotin synthase from Bacillus sphaericus (42), Arabidopsis thaliana (43), Saccharomyces cerevisiae (44), and E. coli (20) have been cloned, overexpressed, and purified. Sufficient complementation analysis has been carried out on Methylobacillus, Bacillus flavum (45), Bacillus subtilis (46), Erwinia herbicola (47), and Serratia marcescens (48) to confirm that they also express active biotin synthases. Sequence alignment reveals a number of conserved regions (Fig. 4). There are 25 completely conserved residues, four of which are cysteine. Three of these (Cys-53, Cys-57, and Cys-60) lie in the 15-residue cysteine box motif (K(S/T)GXCXE(D/N)CX(Y/F)CXQS), and the fourth (Cys-188 in the E. coli sequence) is remote.    Wild-type ϩ ϩ ϩ

TABLE V In vivo production of biotin
Biotin production was measured using the Lactobacillus assay as described under "Experimental Procedures." ϩ, biotin produced, Ϫ, no biotin production.

Protein
Biotin produced

(in vivo)
Wild-type ϩ To determine which of the cysteine residues are involved in stabilizing the [2Fe-2S] and [4Fe-4S] forms of biotin synthase, single point mutations were made of all five cysteine residues (Cys-53, Cys-57, Cys-60, Cys-188, and Cys-288) to either serine or threonine. The choice of cysteine substitution by hydroxylated residues was made to preserve as much as possible the H-bonding character and the side chain geometry of the original cysteine residues. All the mutant proteins were expressed as His 6 tag fusions and purified using IMAC. The molecular weights of all the polypeptides were verified by mass spectrometry (Table I), and the purified proteins were characterized by UV-visible spectrophotometry (Fig. 2), nickel, iron, and sulfur analyses (Tables II and III), and in vivo and in vitro assays for biotin production (Tables IV and V).
Native biotin synthase, His 6 -tagged biotin synthase, and the His 6 -tagged non-conserved mutant C288T all had similar UVvisible spectrophotometric characteristics that were indicative of intact [2Fe-2S] cluster formation. All of these proteins, along with their apo-proteins and reconstituted holo-proteins, were active both in vitro (Table IV) and in vivo (Table V). In contrast, C53S, C57S, and C60S were all colorless, and their absorption spectra showed no indication of the presence of [2Fe-2S] clusters (Fig. 2) indicating that these residues are absolutely vital for [Fe-S] cluster formation which is in turn crucial for activity. C188S proved to be insoluble and could only be characterized in vivo. The fact that only three cysteine residues appear to be needed for [2Fe-2S] cluster binding (22) and that the C53S, C57S, and C60S mutants are inactive suggest that Cys-188 may be required for correct folding of the protein. Attempts to induce [Fe-S] cluster formation by incubation of these mutants with excess reagents did not result in any significant increase of iron binding.
The only cysteine mutant of E. coli biotin synthase that was catalytically active and could support formation of a [2Fe-2S] cluster was C288S, the mutant of the non-conserved cysteine residue, showing that the presence of C288 is not essential for folding, [2Fe-2S] and [4Fe-4S] cluster formation, or for activity. In contrast, all the conserved residues Cys-53, Cys-57, and Cys-60 of the cysteine box are proven to be absolutely crucial for both the formation of the [2Fe-2S] cluster and the activity of the enzyme. These results are consistent with recent mutational studies on PFL-AE where it was also found that mutations of cysteine residues in the cysteine box abolish metal binding (41).
In the absence of crystallographic evidence, this work is a step toward the definition of the essential protein residues involved in the mechanism of [Fe-S] cluster formation in biotin synthase and the relevance of these to the catalytic action of the enzyme. The inference of spectroscopic studies on E. coli biotin synthase is that a hydroxylated residue is also involved in the initial [2Fe-2S] cluster formation (22) and here the conserved cysteine box residues Thr-50 and Ser-63 are obvious candidates. Another region in the protein that is completely conserved is the sequence Tyr-150, Asn-151, His-152, Asn-153, and Leu-154 which is also found in lipoic acid synthase (13) (see Fig. 4). Biotin synthase mutants at sites within these sequences are the subject of current studies. We have recently described the overexpression, purification, and redox characteristics of the flavoprotein components of the E. coli biotin synthase system, FLD, and FLDR (49) and the fact that it is now possible to obtain relatively large quantities of these, and functional, homogeneous BioB protein should facilitate exam- FIG. 4. Sequence alignment of 17 known biotin synthases. The conserved residues are boxed in gray and the two extended conserved motifs outlined in black. The black arrows show the residues mutated in this study, and the residue numbering is that of the E. coli sequence. (The Swiss-Prot accession numbers from the top are P12996, O67104, P54967, P19206, P53557, P46396, Q9Z6L5, Q47862, P44987, O25956, P94966, P46715, O06601, P32451, O60050, P36569, and P73538). The sequence alignment was carried out using Expasy and ClustalW, and the data were formatted using Alscript. ination of the protein-protein and protein-cofactor interactions involved in the redox chemistry of the biotin synthase complex.