Reconstitution and characterization of the polynuclear iron-sulfur cluster in pyruvate formate-lyase-activating enzyme. Molecular properties of the holoenzyme form.

The glycyl radical (Gly-734) contained in the active form of pyruvate formate-lyase (PFL) of Escherichia coli is generated by the S-adenosylmethionine-dependent pyruvate formate-lyase-activating enzyme (PFL activase). A 5'-deoxyadenosyl radical intermediate produced by the activase has been suggested as the species that abstracts the pro-S hydrogen of the glycine 734 residue in PFL (Frey, M., Rothe, M., Wagner, A. F. V., and Knappe, J. (1994) J. Biol. Chem. 269, 12432-12437). To enable mechanistic investigations of this system we have worked out a convenient large scale preparation of functionally competent PFL activase from its apoform. The previously inferred metallic cofactor was identified as redox-interconvertible polynuclear iron-sulfur cluster, most probably of the [4Fe-4S] type, according to UV-visible and EPR spectroscopic information. Cys --> Ser replacements by site-directed mutagenesis determined Cys-29, Cys-33, and Cys-36 to be essential to yield active holoenzyme. Gel filtration chromatography showed a monomeric structure (28 kDa) for both the apoenzyme and holoenzyme form. The iron-sulfur cluster complement proved to be a prerequisite for effective binding of adenosylmethionine, which induces a characteristic shift of the EPR signal shape of the reduced enzyme form ([4Fe-4S]+) from axial to rhombic symmetry.

Pyruvate formate-lyase (PFL) 1 is a key enzyme of the anaerobic glucose fermentation in Escherichia coli and other microorganisms, catalyzing the CoA-dependent cleavage of pyruvate to acetyl-CoA and formate (1). In its active form, this enzyme contains a stable glycyl radical (Gly-734) required for catalysis (2). Studies of mutants and substrate analogs propose that on substrate binding, the spin is transferred from Gly-734 to the reaction center (Cys-418/Cys-419), where a thiyl radical initiates the homolytic cleavage of the pyruvate C-C bond (3,4).
The PFL radical is produced post-translationally by abstraction of the H si atom from the Gly-734 methylene group (5). This occurs by the action of pyruvate formate-lyase-activating en-zyme (PFL activase), which employs adenosylmethionine (AdoMet) and dihydroflavodoxin (or artificial 1e Ϫ donors) as co-substrates, yielding 5Ј-deoxyadenosine and methionine as stoichiometric co-products (Equation 1).
͓-NH-CH 2 -CO-͔ ϩ AdoMet ϩ e Ϫ 3 ͓-NH-Ċ H-CO-͔ ϩ 5Ј-deoxyadenosine ϩ Met (Eq. 1) Because the abstracted H atom is recovered in the 5Ј-CH 3 of deoxyadenosine, a 5Ј-deoxyadenosyl radical intermediate has been proposed as the actual H abstracting species in this system. How this nucleoside radical is generated from AdoMet is an intriguing problem, requiring in the first place PFL activase to be fully characterized at the molecular level.
Previous work has identified PFL activase as a monomer of 28 kDa, and its primary structure, as deduced from the DNAnucleotide sequence (246 amino acids), has been established by N-and C-terminal amino acid sequencing (6,7). The enzyme has long since been recognized to require Fe 2ϩ for catalytic activity (6), and iron contents in the order of one iron/polypeptide chain were determined in purified preparations (3,8). Only more recently, an iron-sulfur cluster has been reported in anaerobically purified enzyme (9). Previously, a [4Fe-4S] cluster had been identified in the analogous activating protein of anaerobic ribonucleotide reductase (RNR) (10).
We report here a new procedure for the isolation of larger quantities of PFL activase from overproducing E. coli cells. The enzyme was carried through the apoform and reconstituted in vitro with Fe 2ϩ and sulfide. The holoenzyme form obtained has full catalytic activity under modified assay conditions that omit the previous free Fe 2ϩ supplement. UV-visible and EPR spectroscopic data indicate that the metal center comprises a redoxinterconvertible [4Fe-4S] cluster, which is essential for the interaction of the enzyme with AdoMet. In addition, single Cys 3 Ser mutations were performed that corroborate that the cysteinyl residues in the Cys-Xaa 3 -Cys-Xaa 2 -Cys motif of the N-terminal polypeptide section, proposed previously as metal interaction site (7), are essential for reconstituting enzyme activity. Our data place PFL activase, together with the AdoMetdependent activating protein of anaerobic RNR (11), into a new subclass of iron-sulfur proteins that generate glycyl radicals in free radical enzymes.

EXPERIMENTAL PROCEDURES
Materials-Pyruvate formate-lyase (nonradical form) was prepared by previous procedures (2,6). High potential iron-sulfur protein from Ectothiorhodospira halophila was available in the laboratory. [8][9][10][11][12][13][14] C]Adenosylmethionine was prepared enzymatically as described in Ref. 5 using [8-14 C]ATP from Amersham-Buchler. AdoMet, S-adenosylhomocysteine, catalase, DNase I, and RNase were from Boehringer Mannheim; Q-Sepharose FF, Sephadex G-25 fine, Superdex 75 HR, and Ultrogel AcA44 were from Pharmacia Biotech Inc.; Polymin G35 was from BASF AG (Ludwigshafen, Germany). * This work was supported by the Deutsche Forschungsgemeinschaft and by the Fonds der Chemischen Industrie. 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.
Bacterial Strain, Plasmids, and Site-directed Mutagenesis-E. coli strain 234M1 (2), a MC4100 derivative carrying the Cm r gene inserted into the chromosomal act gene, was used as host for the overproduction of activase proteins (wild type or mutants). The parental expression plasmid pXA-1 contained the act gene encompassing the 1-kilobase pair DNA fragment comprising nucleotides 2529 -3541 in the sequence (7) linked to the Tac promotor of pKK223-3 (Pharmacia). This construct differs from the earlier pKE-1 (5) by deleting 0.4 kilobase pairs of the act downstream region.
Site-directed mutagenesis was performed with the phosphorothioate method (12) using the Sculptor in vitro mutagenesis system (RPN 1526, Amersham) according to the supplier's manual. This comprised single Ser substitutions of the six Cys residues: TGT codons (for Cys in amino acid positions 12, 36, and 102) and TGC codons (for Cys in amino acid positions 29, 33, and 94) were mutated to TCT and TCC codons, respectively, by using 24 -30-mer antisense mutagenic oligonucleotides that were synthesized by R. Frank (Zentrum fü r Molekulave Biologie, Heidelberg, Germany). All mutant genes were proof sequenced.
Activase Assays-The reaction mixtures (0.5 ml) for PFL conversion to the radical form contained in general: 0.15 M Tris-HCl, pH 7.5, 0.1 M KCl, 5 mM DTT, 10 mM potassium oxamate, 0.2 mM AdoMet, 50 M 5-deazariboflavin, 1 M riboflavin, 10 g of catalase, 60 g (340 pmol) of PFL, and 0.1-2 g of the activase sample. A slide projector light was used for continuous photoreduction of deazaflavin, and reactions were run at 30°C under argon, usually for 10 min. In the assay version termed "holoactivase assay" (see "Results"), the reaction mixture contained additionally 10 M EDTA, and the reaction was started with the activase sample. In the "standard assay," which was used to determine intact activase together with iron center defective enzyme forms, the reaction mixture contained additionally 0.2 mM Fe(NH 4 ) 2 (SO 4 ) 2 (and optionally 0.1 mM Na 2 S) and was preincubated, in the presence of activase, in the dark for 20 min; the reaction was started by illumination. The produced radical form of PFL was determined by activity measurement (coupled optical assay with a 5-10-l aliquot); 35 IU correspond to 1 nmol of PFL dimer containing one Gly radical.
One unit of activase activity was defined to produce 1 nmol of the PFL radical/min. (Activity data reported here are lower by a factor of 700 when compared with those in previous work (6), which used a different unit definition.
Chemical Analyses-Protein concentrations of purified apoenzyme solutions were determined by the 280-nm absorption using a 1 mg/ml value of 1.38 (as calculated from the Trp and Tyr residues in the polypeptide and verified by amino acid analysis). On reconstituted enzyme and other enzyme fractions, the Bradford assay (13) or a biuret procedure, preceded by trichloroacetic acid precipitation, was employed; these assays were standardized against apoenzyme. For quantitation of polypeptides or enzyme concentrations we used the protein values and the molecular mass of 28.1 kDa. Iron was determined according to Ref. 14, and sulfide was determined according to Ref. 15.
Bacterial Growth and Isolation of the Apoenzyme-E. coli 234M1/ pXA-1 cells were grown in LB medium with 50 g/ml ampicillin and 30 g/ml chloramphenicol at 30°C under aeration, maintaining pH 7 by addition of 5 N KOH. Cells were harvested at the transition to the stationary growth phase (A 550 ϭ 4.8) and frozen in liquid nitrogen (yield ϭ 5 g/liter).
Purification of the enzyme was made at 0 -6°C, except for the Q-Sepharose step. Buffer solutions were previously degassed and flushed with argon, but column operations were not strictly anoxic. Enzyme fractions collected were stored under argon.
A 50-g portion of cell paste was suspended in 100 ml of 50 mM Mops/KOH, pH 7.5, 5 mM DTT, 0.1 mM EDTA, and 0.2 mM phenylmethanesulfonyl fluoride. After sonication, DNase I (1 mg) and RNase (1 mg) were added, and the sample was centrifuged at 100,000 ϫ g for 60 min. The supernatant (138 ml, 45 mg of protein/ml) was adjusted to pH 7.5 and 31 ml of 1% (w/v) PolyminG35 (neutralized with H 2 SO 4 ) were added during 60 min of stirring; the precipitate was discarded. The enzyme solution was then chromatographed on a column of Ultrogel AcA44 (20 cm 2 ϫ 85 cm) with 50 mM Mops/KOH, pH 7.5, 0.1 M KCl, 5 mM DTT, and 0.1 mM EDTA (flow rate ϭ 5 cm/h). The brown-red fractions containing activase activity, which eluted between 0.75 and 0.85 column volumes, were concentrated by ultrafiltration on a Amicon PM-10 membrane (32 ml, 25 mg/ml). For final chromatographic purification with simultaneous removal of protein-bound iron, we used Q-Sepharose FF (20 cm 2 ϫ 9 cm). The column was equilibrated with 50 mM Tris-phosphate (KH 2 PO 4 titrated with Tris base), pH 8.0, containing 5 mM DTT and 1 mM EDTA and was operated at 25°C. A 490-mg portion of the AcA44 fraction, previously gel filtered into the same buffer and warmed to 25°C, was applied at a flow rate of 12 cm/h. After a wash with 450 ml of the starting buffer, 90 mM Tris-phosphate pH 8.0 (with DTT and EDTA as before) was applied, which eluted the apoactivase as a symmetric protein peak (275 ml), whereas the iron-based color remained column-bound. The colorless solution was concentrated (in the cold) over a PM-10 membrane, then gel filtered into 50 mM Mops/KOH, pH 7.5, containing 0.1 M KCl, 5 mM DTT, and 0.1 mM ETDA (18 ml, 8.7 mg/ml). Long term storage was at Ϫ70°C. Analysis by SDS-polyacrylamide gel electrophoresis (12% gel) showed the 28-kDa band of PFL activase (6,7) in addition to a few minor bands comprising Յ5% of the total protein.
The same protocols (but down-scaled) of bacterial growth and isolation as apoenzyme were applied for C12S, C29S, C33S, C36S, and C102S mutant enzymes. The C94S mutant was carried only to the Ultrogel AcA44 stage (protein purity ϭ 85%).
Reconstitution of Holoenzyme-Samples of holoenzyme were prepared in batches up to 20 ml composed of 50 mM Mops/KOH, pH 7.5, 0.1 M KCl, 0.1 M mercaptoethanol, 1 mg (36 nmol) of apoprotein/ml, 0.2 mM Na 2 S, and 0.25 mM Fe(NH 4 ) 2 (SO 4 ) 2 . Prior to the addition of iron, the mixture was purged with argon and cooled to 0°C; Fe 2ϩ was admixed from an anaerobic 10 mM stock solution. Incubation was performed routinely for 10 h. With protein concentrations of Յ0.1 mg/ml (see Fig.  1), reconstitution was complete within 5 min.
For analyses of the iron center, reconstitution mixtures were gel filtered anaerobically at 2°C through Sephadex G-25 columns into 50 mM Mops/KOH, pH 7.5, containing 0.1 M KCl, 10 mM mercaptoethanol, and 10 g/ml of catalase. The sample size was 0.1 column vol. and the protein fraction collected between 0.38 and 0.51 column vol.; excessive free iron appeared at Ͼ0.65 column vol. Concentrations up to 15 mg/ml (for EPR spectra) were made by ultrafiltration. The greenish brown colored solutions were stored in closed tubes under argon at 0 or Ϫ70°C, conditions under which the holoactivase activity was stable for days and several weeks, respectively.
Spectroscopy-UV-visible spectra were recorded anoxically, using septum-sealed cells (type 117, Hellma, Mü llheim, Germany). EPR first derivative spectra (5-10 accumulations) were recorded with a Bruker ESP 300 X-band spectrometer (9.50 GHz), using a continuous flow helium cryostat (ESR 900, Oxford Instruments) in a temperature range from 4 to 80 K. The microwave frequency and the magnetic field were measured with a frequency counter and a Gaussmeter, respectively. Recording conditions are indicated in the legends to Figs. 6 and 8. Samples (0.2 ml) with 0.5 mM enzyme were filled into 0.36-cm quartz tubes under argon and stored in liquid nitrogen. Spin quantification was performed by comparing the double integrals of the activase sample and the [4Fe-4S] 3ϩ cluster of E. halophila of known concentration determined for nonsaturating conditions. Molecular Size Determination-To determine the apparent molecular weight of holoactivase, a 100-l sample (0.1 mg) of the reconstitution mixture was gel chromatographed on a Superdex 75 HR 10/30 that was preequilibrated and operated (at ambient temperature) with anoxic buffer composed of 50 mM Mops/KOH, pH 7.5, 0.15 M KCl, 10 mM mercaptoethanol, 0.2 mM Fe 2ϩ , and 0.05 mM sulfide (flow rate ϭ 0.4 ml/min). The eluate was monitored for protein (280 nm) and holoactivase activity (see Fig. 3). For the apoenzyme sample monitored at 280 nm, the column was run with anoxic buffer that lacked Fe 2ϩ and sulfide. Proteins for column calibration were bovine serum albumin, ovalbumin, carbonic anhydrase, soybean trypsin inhibitor, and cytochrome c.
AdoMet Binding Assays with Penefsky Columns-The centrifuged column procedure (16) for separating enzyme-bound and free small molecules was used to monitor binding of [ 14 C]AdoMet to various activase forms. Bio-Spin columns (Bio-Rad) in which the plastic support was replaced by a glass fiber disc (GF-C, Whatman) were employed with a 1-ml filling of Sephadex G-25 fine (Pharmacia), which was equilibrated with anaerobic buffer composed of 50 mM Mops/KOH, pH 7.5, 0.15 M KCl, and 30 mM mercaptoethanol. The experiments were carried out in the cold room (4°C), using a swing bucket table top centrifuge. After precentrifugation at 380 ϫ g for 30 s, which displaced 400 l of buffer, a 150-l sample of the enzyme-[ 14 C]AdoMet mixture was applied, and the centrifugation was run as before. The effluent (155-160 l) was analyzed for radioactivity and protein content. The total amount of enzyme-bound AdoMet (see Fig. 7) was estimated from the effluent radioactivity with accountancy of the protein recovery, which was typically 42%.

RESULTS
Overproduction and Functional Assay of PFL Activase-To study the molecular properties of wild-type and mutant forms of PFL activase, we overproduced the protein in the E. coli act null strain 234M1 carrying the expression vector pXA-1, wherein the act gene is under control of the Tac promoter. Optimal yields of activase protein, amounting to 15% of the soluble protein fraction, were obtained by aerobic cell growth in LB medium at 30°C. Glucose/minimal medium, at aerobic or anaerobic conditions, proved unsuitable because this produced the enzyme chiefly deposited in inclusion bodies.
Two protocols were used to assay PFL activase activity via the production of the radical form of PFL. In the first protocol, which has been used in previous work (6,8) and will be termed now standard assay, free Fe 2ϩ (0.2 mM) is contained in the DTT-based anaerobic reaction mixture, which is preincubated (20 min) before initiating the reaction by photoreduced 5-deazaflavin. We recognized during our present work that both the intact enzyme and iron center-defective or -deficient enzyme forms are measured by this procedure. The native iron center is readily restored during preincubation, with the DTT reagent furnishing the sulfide source. To assay specifically the intact enzyme, we turned to a modified protocol, termed holoactivase assay, that omits free Fe 2ϩ from the reaction mixture and starts the reaction by the activase sample.
In both assay versions, oxamate, replacing the physiological pyruvate, was present as the effector compound required for the PFL activation process (17). It should be noted that oxamate or pyruvate are allosteric ligands to the nonradical form of PFL, not to the activating enzyme.
Isolation of Homogenous Apoenzyme and Reconstitution of the Fe-S Center- Table I summarizes our experimental strategy that led finally to 20-mg quantities of homogeneous PFL activase with a virtually full complement of the native iron center. As in previous work (5,6), the high speed centrifuged cell extract was first subjected to molecular sieve chromatography on Ultrogel AcA44, where the enzyme elutes, after 0.8 column volume, in fractions displaying a red-brown color. At this stage, the protein purity, estimated by SDS-polyacrylamide gel electrophoresis analysis, is about 50%. The fraction of intact enzyme molecules, determined by the two activase assays, is only about 0.1, compared with a value of 0.5 at the cell extract stage. EDTA (0.1 mM) was routinely included in the column buffer, which was not rigorously made anoxic; this probably caused substantial degradation of the native iron center. The spectral properties (see below) suggest that the iron content of the AcA44 prep is largely due to inactive [3Fe-4S] clusters.
Various standard anion-exchange columns, which we tested for further purification, proved unsatisfactory because of low recoveries and separation into subfractions. Therefore we turned to removal of any protein-bound iron as a first step. Although standard metal chelators proved generally suitable, the optimal method was finally found in using a Q-Sepharose column that is operated with 50 mM phosphate buffer, pH 8, at a temperature of 25°C. The iron center is disintegrated upon enzyme adsorption to this column, and the virtually homogeneous apoenzyme can be eluted subsequently by a single elution step with 90 mM phosphate. The isolation protocol is described under "Experimental Procedures," and the purification analysis is shown in Fig. 1.
Through monitoring the appearance of holoactivase activity, the native iron center was then found to become rapidly reconstructed by anaerobic incubation of the apoenzyme, at pH 7.5 and 0°C, with Fe 2ϩ and S 2Ϫ . With 2-mercaptoethanol in place of dithiothreitol as protective thiol reagent in the reconstitution mixture, the sulfide requirement was total (Fig. 2), thus immediately demonstrating a polynuclear Fe-S cluster in the catalytically competent enzyme form.
The specific activity values of reconstituted enzyme were in the range of 60 Ϯ 5 units/mg. This calculates the turnover number to be 1.7 min Ϫ1 at assay conditions with 0.7 M PFL (K m ϭ 1.4 M) or 5 min Ϫ1 at a saturating PFL concentration. It is interesting to note that the presence of up to 5 mM EDTA in the holoactivase assay mixture does not impair the catalytic efficiency within the usual reaction period of 10 min. Reconstituted enzyme alone, however, is quite labile against EDTA (see below).
Molecular Size, Iron-Sulfur Content, and Stability Properties of Reconstituted Holoenzyme-To structurally characterize the functional competent enzyme form, we determined in the first place its molecular size using a calibrated Superdex-75 column. Holoactivase activity migrated as a single band, together with the principal protein band, according to 25 Ϯ 1 kDa (Fig. 3). For the apoenzyme (data not shown) the molecular size was found to be 23 Ϯ 1 kDa. These values compare with the sequencededuced molecular mass of the polypeptide of 28.1 kDa (7), indicating that activase is monomeric with or without the Fe-S complement.
On Superdex-75 chromatography, which was operated with anaerobic reconstitution medium containing iron and sulfide, the recovery of holoactivase activity was routinely Ն95%. However, yields dropped to Յ10% when free Fe 2ϩ and sulfide was omitted from the running buffer, even at 4°C column operation. The residual activity was again found exclusively in the fractions corresponding to a monomeric size.
Removal of excessive ligand components contained in the reconstitution mixture without damaging the Fe-S center proved technically difficult and has as yet been accomplished only partially. With Sephadex G25 columns run at 4°C with anaerobic Mops buffer, pH 7.5 (see "Experimental Procedures"), activase preparations were obtained that showed activity values from 40 to 50 units/mg in the holoactivase assay and about 60 units/mg in the standard assay. Iron and sulfide contents were determined in such samples to be 2.6 Ϯ 0.2 iron/polypeptide chain and 2.5 Ϯ 0.2 sulfur/polypeptide chain. Taken that the deficiency of holoactivase activity is due to partial loss of the iron-sulfur center, a content of 3.3 Ϯ 0.3 iron Iron-Sulfur Cluster of PFL Activase and 3.2 Ϯ 0.3 sulfide/polypeptide would result for the fully complemented enzyme. In SG25 gel filtered enzyme samples stored under argon at 0°C, the functionally competent Fe-S cluster, as monitored by the holoactivase activity, proved to be totally stable for at least several days. However, exposure to air under slight agitation destroyed the activity with a half-time of about 2 min. Likewise, incubation with metal chelators under anaerobic conditions caused rapid inactivation, the half-life with 0.3 mM EDTA being 5 min.
Spectral Characteristics-Gel filtered holoenzyme samples, which are greenish brown in color, showed a UV-visible spectrum (Fig. 4) comprising a prominent 280 nm peak followed by a shoulder at 320 nm and a broad peak at 370 -420 nm, with extensions of the absorbance to wavelengths Ͼ600 nm, which is typical for Fe-S proteins containing [4Fe-4S] 2ϩ clusters. The difference spectrum of holo versus apoenzyme calculates absorption coefficients for iron at maximum wavelengths of 310 and 420 nm to be 5.3 and 3.8 mM Ϫ1 cm Ϫ1 , respectively. These spectral data closely match those reported for [Fe 4 S 4 (SEt) 4 ] 2Ϫ (18) but are different from those for [Fe 2 S 2 (SEt) 4 ] 2Ϫ (19). Addition of Ն0.1 mM dithionite (Ն5-fold excess) reduced the 420-nm absorption by 45% (Fig. 4, inset), and a subsequent short exposure (1-2 min) to air restored the original spectrum, which, however, became bleached upon prolonged admission of air.
For comparison, Fig. 5 shows the optical spectrum of the Ultrogel AcA44 fraction of the enzyme purification scheme, which is largely defective as to holoenzyme activity (Table I). The prominent feature is a pronounced peak at 416 nm (estimated absorption coefficient for iron ϭ 6 mM Ϫ1 cm Ϫ1 ), followed by shoulders at 470 and 580 nm, giving such samples a characteristic red-brown color appearance. These spectral data suggest the presence of 3Fe-4S clusters, possibly a mixture of linear and cubic forms, when the spectrum is compared with the optical spectra of the 3Fe-4S forms of aconitase (20). This is supported by the EPR spectrum of the Ultrogel fraction (not  Table I shown), which displays a strong nearly isotropic signal at g av ϭ 2.026 similar to the g ϭ 2.01 signal of the (cubic) [3Fe-4S] ϩ cluster form of aconitase (20,21). It should be noted here that the functionally defective Fe-S core is usually retained on carrying the Ultrogel fraction through conventional protein purification protocols. This occurred most probably also with the first isolation of chromosomally encoded PFL activase from wild-type E. coli, where an associated color had been ascribed to an organic factor (6). However, this possibility was dismissed later on by studies of overproduced or heterologically expressed enzyme (3,8).
EPR spectroscopy of reconstituted holoenzyme samples as obtained after gel filtration showed a very weak, nearly isotropic signal at g ϭ 2.01, which we ascribe to a small content of [3Fe-4S] ϩ species. On reduction with dithionite, an intense EPR signal of axial symmetry, g ϭ 2.029, and g Ќ ϭ 1.925, was obtained suggestive of an abundant S ϭ 1/2 [4Fe-4S] ϩ cluster (Fig. 6). (A small shoulder at g ϭ 1.95, registered on various samples, could arise from a minor component, the origin of which is presently unknown). Signal intensity was maximal in the temperature range 15-20 K and virtually undetectable above 45 K. Spectral saturation effects were observed only for microwave powers above 20 mW (at 20 K). Using a HiPIP (high potential iron-sulfur protein) reference sample for spin quantification, we estimated that the axial resonance accounts for ϳ0.4 spins/four-iron atoms in the enzyme preparations.
From these electronic and EPR spectroscopic characteristics we conclude that the metal content in reconstituted holoenzyme preparations comprises predominantly a [4Fe-4S] 2ϩ cluster that is readily redox-interconvertible into the [4Fe-4S] ϩ state.
Essential Cysteine Residues of PFL Activase-Because cysteine residues are the prominent ligands of Fe-S cores in ironsulfur proteins, we investigated the effect of Cys mutations as a first, simple approach toward the coordination chemistry of the metal cluster in PFL activase. Each of the six cysteine residues (Cys-12, Cys-29, Cys-33, Cys-36, Cys-94, and Cys-102) in the polypeptide chain was replaced separately by a serine residue via site-directed mutagenesis. The mutant proteins, which were overproduced in the E. coli act Ϫ host strain in soluble form to about the same extent as the wild-type enzyme, were purified, and their catalytic activities were determined after subjection to Fe-S cluster reconstitution conditions. Results are compiled in Table II. C12S, C94S, and C102S were reddish brown colored proteins on purification to the Ultrogel AcA44 stage, with spectral features resembling that of the wild-type enzyme at this stage (Fig. 5). Although C12S and C102S were readily convertible to the apoenzyme form, the C94S mutant protein became insoluble upon removing the iron center by the Q-Sepharose step and therefore was analyzed with the Ultrogel sample. All three mutants displayed full holoactivase activity (as compared with standard assay activities), albeit absolute values were slightly lower, by a factor of Յ2, than the wild-type value. In contrast, C29S, C33S, and C36S mutants proved catalytically incompetent. These proteins, which were colorless at the Ultrogel purification stage, behaved normally throughout the isolation procedure, suggesting that they are unimpaired with respect to global protein-folding. These observations suggest that C29, C33, and C36 constitute cysteine sulfur ligands of the Fe-S core in PFL activase.

Studies of Adenosylmethionine Interaction with PFL Activase-The
Fe-S cluster in PFL activase is a prerequisite for effective binding of AdoMet, the crucial co-substrate of this enzyme. This was demonstrated by assays using 14 C-labeled AdoMet and the convenient column centrifugation technique (16) as shown in Fig. 7. No interaction was detected with the apoenzyme sample or with the critical Cys 3 Ser mutants (C29S, C33S, and C36S) previously subjected to Fe-S cluster reconstitution conditions. The data for the holoenzyme sample yielded a tentative K d value of 3 Ϯ 1 M, which compares to reported K m values of 2. 8 -7 M (8, 22). (The maximal quantity of 0.55 AdoMet bound per polypeptide, as calculated for this experiment, is lower than expected, but our analytical technique is insufficient to give accurate data.) The holoenzyme-AdoMet complex was also detectable by conventional gel filtration at 4°C on Sephadex G-25 run with anoxic Mops buffer, pH 7.5. It should be noted that the 14 C label associated with the protein fraction in these binding experiments was verified by high pressure liquid chromatography analysis to be due to intact AdoMet.
Spectroscopic examinations found that binding of AdoMet slightly decreased the holoenzyme absorbance between 320 and 440 nm (maximally by 5% at 375 nm). Drastic effects, however, were found with the EPR spectrum as shown in Fig. 8A. On incubation with AdoMet, the axial signal of dithionite-reduced enzyme (Fig. 6) changed to a distinctive rhombic symmetry (g 1 ϭ 2.009, g 2 ϭ 1.921, and g 3 ϭ 1.886). A similar but subtly different signal change to rhombicity, with g-factors of 2.038, 1.930, and 1.899, was obtained with S-adenosylhomocysteine (Fig. 8B). In the activity assays, this compound is a competitive inhibitor of PFL activase with respect to AdoMet. 2

DISCUSSION
Through isolation as apoenzyme and chemical reconstitution of the iron-sulfur cluster, substantial quantities of PFL activase in its catalytically competent form have now been made available enabling closer structure-function analyses of this radical-generating enzyme. The reconstituted enzyme has been determined to be monomeric (28 kDa), and data reported here are strongly indicative of a 4Fe-4S cluster content (rather than a 2Fe-2S type). The electronic spectrum of the greenish brown colored solution shows features (shoulder at 320 nm and broad absorption in the 420 nm region) typical of a [4Fe-4S] 2ϩ cluster type, which is reducible by dithionite. Accordingly, enzyme samples as obtained are largely EPR silent but on reduction yield an axial S ϭ 1/2 signal (g ϭ 2.03, 1.92) indicative of a [4Fe-4S] ϩ cluster. Integration of the fast relaxing EPR signal, observed optimally at 20 K, estimated Ϸ0.4 spin/four-iron cluster as calculated from the protein-bound iron. Iron and sulfide analyses regularly found 2.6 Ϯ 0.2 iron/polypeptide and 2.5 Ϯ 0.2 sulfur/polypeptide, values that are too low for a 4Fe-4S stoichiometry/monomer. However, allowance has to be made for the fact that the samples of reconstituted enzyme, on which all cluster characterizations were made, were previously gel filtered, an operation that looses between 15 and 30% of the catalytic holoenzyme activity.
For PFL activase samples prepared by anaerobic enzyme isolation from E. coli extracts, Broderick et al. (9)   presence of AdoMet, yielding an almost axial EPR resonance (g ϭ 2.01, 1.89, and 1.88). This behavior is strikingly different from what we observe for reconstituted PFL activase (ready reduction to the 1ϩ state without AdoMet and shift of the axial S ϭ 1 ⁄2 EPR signal to rhombicity (g ϭ 2.01, 1.92, and 1.88) in the presence of AdoMet). The difference remains to be resolved.
It should be noted that the anaerobically isolated enzyme had been obtained via protein expression in E. coli from a different vector (pMG-AE) at the increased temperature of 42°C (9). The sample under study might have contained functionally damaged iron center(s). It is reported as red-brown colored (not greenish brown as with the reconstituted holoenzyme), showing a specific activity of only 5.5 units/mg (recalculated for our unit definition) under assay conditions equivalent to our standard assay (an assay version that is unable to report the actual amount of catalytically competent enzyme form contained in activase samples).
A close analog of PFL activase is the activating protein of anaerobic ribonucleotide reductase, which also uses AdoMet and reduced flavodoxin for generating a protein-based glycyl radical (11,23,24). Its 4Fe-4S center (10) has a spectroscopic signature (UV-visible and EPR) resembling closely the signature that we have found here for PFL activase. The two activases are different in molecular size and operational terms, however. RNR activase (␤, 17 kDa) is dimeric, with the [4Fe-4S] cluster placed between the two polypeptides and is permanently associated with the reductase enzyme proper in a ␣ 2 ␤ 2 complex (10,11). In contrast, PFL activase is a monomeric protein (28 kDa) interacting with its target protein pyruvate formate-lyase only transiently, i.e. as a catalytically operating converter enzyme. This property is reflected also in the separate transcriptional control of PFL and PFL activase genes (1). At the amino acid sequence level, the two activases are strongly homologous in the N-terminal section harboring the characteristic Cys-Xaa 3 -Cys-Xaa 2 -Cys pattern that we had previously proposed as iron interaction site (7). Mutagenesis studies reported here have now demonstrated that each of these three Cys residues is essential, whereas single Ser replacements of other cysteines in the polypeptide chain do not affect the reconstitution of active PFL activase. The sequence similarity between the two activases is strongest for a stretch of 26 amino acid residues as shown in Fig. 9, wherein alignments with additional sequences for organisms other than E. coli have been integrated. The consensus comprises 11 amino acid residues. (Other homologous sequences in the current data bases that refer to proteins of different or nonestablished functions were not rated. It should be noted that biotin synthase, another Fe-S protein employing AdoMet for a radical reaction (25,26), shares the Cys-Xaa 3 -Cys-Xaa 2 -Cys pattern but no other significant homology.) The unique utilization of AdoMet as a radical generating cofactor had been recognized at first in 1984 by studies of the pyruvate formate-lyase system, when we identified 5Ј-deoxyadenosine and methionine as co-products of the post-translational radical introduction into PFL (27). The chemical mechanism of this process, proposed to involve intermediary formation of a H abstracting 5Ј-deoxyadenosine radical (3,5), should be approachable now more directly because the metal center of PFL activase has been characterized, and functionally correct enzyme samples are available in abundance. The ligands of the Fe-S core must still be identified. It is tempting to speculate that the metal cluster is not completely cysteinyl coordinated, thus enabling close interaction with the AdoMet co-substrate. Its effective binding to PFL activase has been demonstrated to require the Fe-S complement. Reductive cleavage of AdoMet occurs presumably through the Fe-S cluster; however, this process is largely suppressed when the PFL substrate or Gly-734 site analogous peptide substrates are not present simultaneously (5). The essential role of the metal center for AdoMet reduction has recently been shown also for the parallel case of anaerobic RNR (28).