An in vitro reducing system for the enzymic conversion of cobalamin to adenosylcobalamin.

Homogeneous ferredoxin (flavodoxin):NADP(+) reductase and flavodoxin A proteins served as electron donors for the reduction of co(III)rrinoids to co(I)rrinoids in vitro. The resulting co(I)rrinoids served as substrates for the ATP:co(I)rrinoid adenosyltransferase (CobA) enzyme of Salmonella enterica serovar Typhimurium LT2 and were converted to their respective adenosylated derivatives. The reaction products were isolated by reverse phase high performance liquid chromatography, and their identities were confirmed by UV-visible spectroscopy, mass spectrometry, and in vivo biological activity assays. Adenosylcobalamin generated by this system supported the activity of 1,2-propanediol dehydratase as effectively as authentic adenosylcobalamin. This is the first report of a protein system that can be coupled to the adenosyltransferase CobA enzyme for the conversion of co(III)rrinoids to their adenosylated derivatives.

The reducing system involved in the corrinoid adenosylation pathway is not well understood in any of the organisms in which Cbl biosynthesis has been studied. The corrinoid reducing activities described to date in procaryotes were reported to require pyridine nucleotides and free flavin nucleotides for activity (7,8,19), and thus activity has been attributed to easily dissociable flavoproteins. However, it was recently shown that co(III)rrinoid reduction can be chemically achieved by dihydroflavin nucleotides in the absence of any enzymes (20). The identity of the enzyme catalyzing the reduction of Co(II) to Co(I) remained unknown. Although a cob(II)yrinic acid a,c-diamide reductase was isolated from Pseudomonas denitrificans, neither the enzyme nor the gene responsible for this activity was identified (19).
At present, the only well documented cob(II)alamin reduction reaction is the one required for reductive activation of the E. coli methionine synthase enzyme (MetH). The MetH enzyme becomes inactive once every 2,000 turnovers when cob(I)alamin oxidizes to cob(II)alamin (21). The electrons for the reduction of cob(II)alamin to cob(I)alamin in MetH are provided by reduced flavodoxin (FldA) (22)(23)(24). It has been demonstrated that very specific interactions between FldA and MetH are required for cob(II)alamin reduction to occur (21,24,25). The data presented herein show that the NADP ϩ :ferredoxin (flavodoxin) reductase (Fpr) and FldA proteins can reduce co(III)rrinoids to co(I)rrinoids, the substrates of the CobA. The results from spectroscopic analyses and enzymic and biological activity assays are reported to show that the product of the in vitro adenosylation reaction was the adenosylated derivative of the corrinoid substrate used. It is proposed that the Fpr and FldA proteins are part of the corrinoid adenosylation pathway in S. enterica.

Protein Purification
Purification of E. coli Fpr Step 1: Mass Culturing of the Overexpressing Strain and Generation of Cell-free Extracts-Fpr protein was overproduced from strain JE4937 (E. coli C-1a/pEE1010, a gift from Elisabeth Haggård-Ljungquist, Stockholm University, Stockholm, Sweden) grown overnight at 37°C on Luria Bertani broth containing 100 g/ml of ampicillin and 30 mM glucose. The cells were harvested by centrifugation at 10,000 ϫ g for 10 * This work was supported in part by National Institutes of Health Grant GM40313 and by a DuPont Aid-To-Education grant (to J. C. E.-S.). 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.
Use Approximately 20 g of cells were resuspended in 50 mM Tris-Cl buffer pH 7.5 at 4°C (buffer A) containing 16 g/ml of protein inhibitor phenylmethanesulfonyl fluoride. The cell suspension was broken by sonication using a Sonic Dismembrator (model 550; Fisher) for 13 min (large tip, setting of 6, 50% duty). Cell-free extracts were generated by centrifugation at 40,000 ϫ g using a KOMP Spin rotor (model KA-21.50; Composite Rotor TM , Mountain View, CA) for 1.5 h.
Step 2: Anion Exchange Chromatography-Cell-free extract (0.84 g of protein) was applied to a column containing DEAE 650M TSK TOYO-PEARL® anion exchange resin (2.5 ϫ 20 cm, 41.7-ml bed volume; equilibrated with buffer A; Tosohaas, Montgomeryville, PA). Twenty mg of protein was applied per ml of resin. The column was developed at a flow rate of 60 ml h Ϫ1 . After the sample was applied, the column was washed with 170 ml of the buffer A, followed by a 65-ml wash with buffer A containing 1 M NaCl. Purification of the Fpr protein was monitored by SDS-PAGE.
Step 3: Dye Ligand Chromatography-Fractions from the previous step containing Fpr (254 mg of total protein) were applied to a Reactive Red 120 column (2.5 ϫ 20 cm, 76-ml bed volume; Sigma) equilibrated with buffer A. Approximately 4 mg of protein was loaded per ml of resin. The column was developed at a flow rate of 79 ml h Ϫ1 . After washing the column with 120 ml of buffer A, bound proteins were eluted with a 310-ml linear gradient of NaCl (0 -1.0 M) in buffer A. Following the gradient, the column was washed with 240 ml of buffer A containing 1.3 M NaCl to elute tightly bound proteins. Fractions containing Fpr protein, as judged by SDS-PAGE and UV-visible spectrum of the enzymebound flavin, were pooled and concentrated using a Centriprep 10 TM concentrator (Millipore, Bedford, MA).
Step 4: Anion Exchange FPLC-Purified Fpr protein was applied to a Resource Q FPLC column (Amersha Pharmacia Biotech) equilibrated with buffer A at 41 mg/ml of resin. Proteins bound to this resin were eluted with a linear gradient of NaCl (0 -1.0 M) in buffer A. After this purification step, the Fpr protein was assessed by SDS-PAGE to be Ն98% homogeneous. Eighteen mg of homogeneous Fpr protein was obtained from the starting 20 g of cell material.

Purification of E. coli FldA
Step 1: Overproduction of FldA Protein and Generation of Cell-free Extracts-The flavodoxin protein was overproduced from four 2-liter cultures of E. coli strain DH1/pDHO2 (R. G. Matthews, University of Michigan) grown on LB broth containing 100 g/ml of ampicillin. The cells were grown at 37°C to an A 650 of about 0.7. Isopropyl-␤-D-thiogalactoside was added to a final concentration of 400 M, and the incubation was continued overnight. After incubation, the cells were harvested by centrifugation at 10,000 ϫ g for 10 min using a Sorvall GSA rotor (DuPont), the cells (ϳ25 g) were resuspended in 150 ml of 50 mM Tris-Cl buffer, pH 7.5, at 4°C containing 16 g of phenylmethanesulfonyl fluoride/ml of extract. The cells were broken by sonication using a sonic dismembrator for 7.5 min (large tip, setting of 6, 50% duty). The cellfree extracts were generated by centrifugation at 40,000 ϫ g for 1.5 h using a KOMP Spin rotor.
Step 2: Anion Exchange Chromatography on DEAE-650M-Cell-free extract (ϳ1.5 g of protein) was applied to a column containing DEAE 650M resin (2.5 ϫ 20 cm, 70-ml bed volume) equilibrated with buffer A. Approximately 22 mg of protein was applied per ml of resin. The column was washed with 70 ml of loading buffer followed by 140 ml of 100 mM sodium acetate buffer, pH 5.0. Bound proteins were eluted with a linear gradient 100 -500 mM NaCl in 100 mM sodium acetate buffer, pH 5.0. The pH of the fractions containing the FldA protein was brought up to 7.0 with 2 M Tris-Cl buffer, pH 8.0, at 4°C. Fractions containing the FldA protein were pooled and dialyzed against buffer A.
Step 3: Anion Exchange FPLC-Fractions containing dialyzed FldA protein (ϳ315 mg of total protein) were applied to a POROS HQ (Per-Septive Biosystems, Framingham, MA) anion exchange FPLC column (1.0 ϫ 10 cm, 7-ml bed volume) equilibrated with buffer A. Forty-five mg of protein was applied per ml of resin. The column was washed with 14 ml of buffer A followed by a linear gradient of 0 -1.5 M NaCl in buffer A. Fractions containing FldA were pooled, dialyzed against buffer A, and stored at Ϫ90°C in buffer A ϩ 20% (v/v) glycerol. After this purification step, the FldA protein was assessed to be Ͼ95% homogeneous by SDS-PAGE. A total of 85 mg of FldA was obtained from 25 g of starting material.

Purification of the CobA Enzyme of S. enterica
Purification of the CobA enzyme was performed as described in Bauer et al. (14) without modifications.

Protein Techniques
Total protein concentration was determined by the Bradford method using the Bio-Rad protein reagent (26). Protein analysis was performed by SDS-PAGE (27) stained with Coomassie Blue (28). The concentration of the Fpr and FldA proteins was determined using the reported extinction coefficients for the protein-bound flavins (29).
AdoCbl formation was monitored by the increase in the absorbance of the solution at 525 nm as a function of time. The amount of product generated was calculated using the difference in molar extinction coefficient between the substrate, cob(II)alamin, and the product, AdoCbl, at 525 nm (⑀ 525 ϭ 4.8 ϫ 10 Ϫ3 M Ϫ1 cm Ϫ1 ). One unit of CobA activity was defined as the amount of enzyme required to generate 1 nmol of AdoCbl/ min (13).

Cytochrome c Reductase Assays
Activity of the Fpr protein was monitored using the cytochrome c reductase assay as reported by Fujii et al. (22,23), except the assays were performed under anoxic conditions to mimic the conditions used in the corrinoid adenosylation assays.

1,2-Propanediol Dehydratase Assays
In vitro assays for AdoCbl-dependent 1,2-propanediol dehydratase (a gift from Perry A. Frey) were performed as described (30). The conver- Enzymic Reduction and Adenosylation of Corrinoids sion of 1,2-propanediol to propionaldehyde was measured by derivatization of the product with 3-methyl-2-benzothiazoline hydrazone yielding an azine product that can be measured spectrophotometrically at 305 nm (⑀ 305 ϭ 13,300 M Ϫ1 cm Ϫ1 ). This assay was used as a measure of coenzymic activity of enzymically generated AdoCbl. One unit of 1,2propanediol dehydratase is defined as the amount of enzyme required to generate 1 mol of propionaldehyde/min under the conditions of the assay.

Synthesis of Aquocobinamide (H 2 OCbi)
H 2 OCbi was synthesized in vitro from (CN) 2 Cbi. First, AdoCbi was synthesized from (CN) 2 Cbi using the potassium borohydride-dependent corrinoid adenosylation assays. The assays were performed as described (13), except (CN) 2 Cbi (50 M) was substituted for HOCbl. After substrate conversion was complete, the reaction mixture was exposed to air and filtered through a 45-m (25-mm diameter) syringe filter (Nalgene, Rochester, NY). The corrinoid product was isolated using a C 18 SepPak cartridge (Waters, Milford, MA) and dried using a SpeedVac concentrator (Savant Instruments, Farmingdale, NY). Concentrated AdoCbi was resuspended in 400 l of water and photolyzed aerobically using a tungsten/halogen lamp placed ϳ20 cm away from the tube containing the sample. Using this procedure, AdoCbi was converted by heterolysis to H 2 OCbi. The sample was passed through a C 18 cartridge as described above. H 2 OCbi was converted to (CN) 2 Cbi by photolysis in 0.1 M KCN, pH 10.0, using a tungsten/halogen lamp and quantitated using the molar extinction coefficient for (CN) 2 Cbi at 367 nm (⑀ 367 ϭ 30,800 M Ϫ1 cm Ϫ1 ) (31).

Reverse Phase HPLC (RP-HPLC) Analysis of Reaction Products
The products of the corrinoid adenosylation assays were isolated and analyzed by RP-HPLC on a Prodigy ODS (8) column (Phenomenex, Torrance, CA) attached to an HPLC system as described (32). A photodiode array detector (Waters) was used to identify corrinoids by their UV-visible spectra. Authentic HOCbl and AdoCbl were used as standards. The isolated corrinoids were desalted using a C 18 SepPak cartridge and dried in a SpeedVac concentrator.

Mass Spectrometry
Mass spectrometry of purified reaction products was performed at the Mass Spectrometry Facility at the University of Wisconsin-Madison Biotechnology Center.

In Vivo Assessment of Adenosylcorrinoid Synthesis
In vitro synthesis of adenosylcorrinoids was assessed in vivo by using strain JE1096 (metE205 ara-9 cobA343::MudJ) as an indicator strain in a biological activity assays. The cells of the indicator strain grown overnight in 2 ml of nutrient broth were concentrated by centrifugation and washed with 14.5 mM sterile saline; 0.1 ml of the cell suspension (ϳ10 8 cells) was added to 3 ml of molten soft agar (0.7% w/v) and used to overlay a plate of Vogel-Bonner minimal medium (33) containing glucose (11 mM) as carbon and energy source. The corrinoid standards AdoCbi, (CN) 2 Cbi, and Cbl (ϳ50 pmol each) and reaction products were spotted onto the overlay and incubated overnight at 37°C.

RESULTS
Homogeneous Fpr, FldA, and CobA Proteins Are Sufficient for in Vitro Conversion of Cob(III)alamin to AdoCbl-NADPHdependent synthesis of AdoCbl was conducted when cell-free extracts of strains overexpressing Fpr (ferredoxin(flavodoxin): NADP ϩ oxidoreductase) and FldA (flavodoxin A) proteins were added to corrinoid adenosylation assay mixtures in place of potassium borohydride (data not shown). Control experiments in which either cell-free extract was omitted failed to yield a product, suggesting a requirement for both Fpr and FldA proteins. These proteins were isolated to homogeneity to further investigate their involvement the corrinoid adenosylation reaction.

Enzymic Reduction and Adenosylation of Corrinoids
Homogeneous Fpr protein (Fig. 2, lane A) used in this work had a cytochrome c reductase specific activity of 0.36 mol cyt c min Ϫ1 mg Ϫ1 of protein, and a specific content of 42 nmol of Fpr/mg of protein. Initially, the cob(II)alamin substrate was generated using NAD(P)H:flavin oxidoreductase H 6 Fre enzyme, NADH, and FMN (20). Once the Fpr protein was isolated, it was used in lieu of H 6 Fre to reduce HOCbl to cob(II)alamin. Elimination of the H 6 Fre protein simplified the reaction conditions. The FldA protein was purified to Ͼ95% homogeneity (Fig. 2,  lane B). Approximately 85 mg of protein was obtained from 25 g of wet mass. The specific flavoprotein content for this preparation was 22 nmol of FldA/mg of protein. FldA protein was reduced by Fpr, and reduced FldA was coupled to CobA enzyme in the corrinoid adenosylation assays. This system converted cob(III)alamin or cob(III)inamide to their corresponding adeno-sylated derivatives. NADPH was required, but unlike other systems, no further addition of FAD or FMN to the reaction mixture was necessary (7,8,19,20).
When Fpr and FldA proteins were coupled to the CobA enzyme, AdoCbl synthesis was observed (Fig. 3A). The characteristic changes associated with conversion of cob(II)alamin to the final product, AdoCbl (Fig. 3B), were not observed when either one of the enzymes was omitted from the reaction mixture. These spectral changes were in good agreement with those observed for this conversion in other systems (7). The identity of the products of these reactions was further confirmed by the analyses shown below.
Linearity of the Reaction-The rate of product formation was linear when the concentration of the FldA protein in the reaction mixture ranged between 0.5 and 2.0 M (0.2-0.4 nmol of AdoCbl min Ϫ1 ). The rate of the reaction was also linear when CobA enzyme was present in the reaction mixture in the range of 0.5-1.5 M (0.2-0.5 nmol AdoCbl min Ϫ1 ). These results indicated a 1:1 CobA:FldA stoichiometry.
Identification of the Product of the Reaction-The product of the corrinoid adenosylation reaction was isolated by RP-HPLC. When HOCbl was used as substrate, the elution time (Fig. 4A, 21.9 min) and UV-visible spectrum of the isolated product was identical to an AdoCbl standard isolated by the same procedure (Fig. 4B). When the reducing system (Fpr, FldA) was omitted from the reaction mixture, a product with an elution time (18.7 min) and UV-visible spectrum identical to that of a HOCbl standard was obtained (data not shown).
The isolated product was analyzed by electrospray ionization (ESI) mass spectrometry (Fig. 5). The positive-ion ESI mass spectrum showed signals with m/z values of 829.2 and 1582.2, which corresponded to masses of KAdoCbl (z ϭ ϩ2) and AdoCbl (z ϭ ϩ1), respectively. The negative-ion ESI mass spectrum of the product showed a parent peak with an m/z value 1367.8 (z ϭ ϩ1), in agreement with the molecular mass for NaHOCbl and suggesting that the Ado group was removed during analysis. This was confirmed by the negative-ion ESI mass spectrum of authentic AdoCbl isolated by HPLC (Fig. 5B). The positive ion ESI mass spectrum of authentic AdoCbl showed a  7. In vivo assessment of adenosylcorrinoid synthesis. A biological activity assay was used to detect AdoCbi in the reaction mixture. The product of the reaction was provided to strain JE1096 (met205 ara-9 cobA343::MudJ). This strain is a methionine auxotroph unless AdoCbi or Cbl is provided in the medium. A growth response was observed when the product of the reaction is provided to the indicator strain. Authentic cyanocobalamin, (CN) 2 Cbi, and AdoCbi were provided as controls for the growth response. signal at m/z ϭ 829.4, consistent with the molecular masses of KAdoCbl (z ϭ ϩ2).
AdoCbl Generated by the Enzymic System Has Coenzymic Activity-Approximately 0.1 nmol of enzymically generated AdoCbl was added to 1,2-propanediol dehydratase reaction mixtures. The amount of propionaldehyde generated from the dehydration of 1,2-propanediol during the reaction was identical to that of a reaction performed using 0.1 nmol of commercially available AdoCbl (specific activity, 0.02 unit/mg), indicating that enzymically generated AdoCbl was biologically active.
In Vitro Reduction and Adenosylation of Cobinamide Derivatives-Corrinoid adenosylation assays were performed using either (CN) 2 Cbi or H 2 OCbi as substrate. When (CN) 2 Cbi was used, the rate conversion of this substrate to AdoCbi was slow, as judged by the changes in the UV-visible spectrum associated with this conversion (Fig. 6A). High concentrations of the flavoprotein system and of the CobA enzyme were required to drive the reaction to near completion (Fig. 6A, spectrum 5). The UV-visible spectrum of the product obtained under these conditions was in good agreement with that reported for AdoCbi (13,34). The absorbance peak at 579 nm (Fig. 6A, spectrum 5) was indicative of unreacted substrate. These results indicated that the cyanide ligands were removed upon reduction of the cobalt ion to generate the four-coordinate cob(I)inamide substrate for CobA. When H 2 OCbi was used as substrate (Fig. 6B), the apparent rate for adenosylation was approximately twice as fast, indicating that removal of the cyanide ligands was the rate-limiting step when (CN) 2 Cbi was used as substrate.
To confirm that the product of the in vitro reaction was AdoCbi, strain JE1096 was used to test for biological activity (Fig. 7). Strain JE1096 is a methionine auxotroph correctable by AdoCbi or Cbl (11). A strong growth response was observed when the product of the reaction was provided to the indicator strain. Growth of the cobA mutant strain was also observed with AdoCbi generated in KBH 4 -dependent adenosylation assays (Fig. 7). As expected, a growth response was not observed when (CN) 2 Cbi was provided in the assay. These results confirmed that the product was AdoCbi.

DISCUSSION
The data reported herein show that the reducing system comprised of the Fpr and FldA proteins is sufficient for the generation of the co(I)rrinoid substrate of the ATP:co(I)rrinoid adenosyltransferase CobA enzyme of S. enterica. To the best of our knowledge, this is the first report of a reducing system that can be coupled to the CobA enzyme for the generation of adenosylated corrinoids.
The involvement of the FldA protein in the corrinoid adenosylation reaction is consistent with the ability of this protein to work at low redox potentials as a one-electron carrier (21,22,35,36) and with its involvement in the reduction of cob(II)alamin to cob(I)alamin on the Cbl-dependent methionine synthase MetH enzyme (21)(22)(23)(24)37). The validity of the conclusions drawn from this work is not affected by the use of E. coli FldA and Fpr proteins given the high degree of identity shared by the proteins in these bacteria (Fig. 8), strongly suggesting that these proteins perform equivalent functions in these procaryotes. Thus it is reasonable to think that reduced FldA might represent a natural reducing system for corrinoid adenosylation in enterobacteria (e.g. E. coli and S. enterica).
The Site of Corrinoid Reduction-The very low redox potential of the free cob(II)alamin/cob(I)alamin couple (E°Ј ϭ Ϫ610 mV) (38) and the extreme reactivity of the cob(I)alamin nucleophile (39) strongly suggest that reduction of cob(II)alamin occurs on the CobA enzyme. It is unlikely that cob(II)alamin reduction occurs in solution because the midpoint redox potentials for the semiquinone and hydroquinone forms of free FldA (Ϫ260 and Ϫ440 mV, respectively) (21) are too high to account for the generation of the cob(I)alamin nucleophile via a second order reaction. The redox potential for this reaction is expected to be close to the redox potential for free cob(II)alamin and thus considerably lower than that of the cob(II)alamin form of free methionine synthase (21,38). In addition, the lower axial ligand of HOCbl has been shown to be coordinated to the cobalt in the CobA crystal structure (Fig. 8)  ligand dissociation from the cobalt ion. This reaction is known to raise the redox potential of the cob(II)alamin/cob(I)alamin couple (21,40). Interactions between FldA and CobA coupled to product formation are therefore expected to raise the redox potential to facilitate the reaction. The documented mechanisms for MetH reactivation as well as the FldA-dependent anaerobic ribonucleotide reductase activation reaction support this hypothesis (24,41). Interactions between the FldA and CobA proteins are an attractive possibility, when one also considers the location of the corrinoid-binding site in CobA (14). The ternary complex structure between HOCbl, ATP, and CobA show that the corrinoid binding site is located near the surface of the enzyme (Fig. 9). The interactions between FldA and CobA and/or the reduction of the corrinoid may be responsible for the hypothesized conformational changes CobA must undergo to bring the target C5Ј of the ribose closer to the cobalt atom for the nucleophilic attack to occur (14). Work addressing these important aspects of the reaction is currently being performed.
Is FldA Involved in Corrinoid Adenosylation in Vivo?-The answer to this question is complicated by the fact that mutations in the fldA gene have been found to be lethal under aerobic and anaerobic growth conditions, suggesting that FldA is an essential function (42). Thus a genetic approach to this question is not possible. Because of this, we are not able to determine whether FldA is one redundant function in co(II)rrinoid reduction for adenosylation or whether it is solely responsible for this process. If FldA were an in vivo corrinoid reductase involved in corrinoid adenosylation, it would explain why cob(II)alamin reductase mutants have not been isolated. The chemistry of the reaction together with the data presented herein suggest that a true in vivo reductant for this process must be a low redox one-electron carrier and that, if the proposed interactions with the CobA enzyme are required, at least some specificity would be expected.
The Corrinoid Adenosylation Pathway in S. enterica-The schematic of the corrinoid adenosylation pathway shown in Fig. 10 incorporates the findings reported in this paper. In this model, reduction of co(III)rrinoid to co(II)rrinoid is shown catalyzed by endogenous dihydroflavin nucleotides or by the Fpr enzyme. The Fpr enzyme uses NADPH to reduce FldA, which in turn transfers an electron to Co(II) to yield Co(I). Electrons for co(II)rrinoid reduction are proposed to be transferred by reduced FldA when the corrinoid is bound to the CobA enzyme. The nucleophilic attack by Co(I) on the C5Ј of the ribosyl moiety of ATP results in the release of inorganic tripolyphosphate. 2 Putative interactions between the FldA and CobA proteins are currently under investigation. FIG. 9. Stereo view of the ternary complex between CobA, HOCbl and ATP (14). Shown is a ribbon representation of the complex between CobA and its substrates. Only one HOCbl molecule is observed in this representation, bound at the surface of subunit A. The N-terminal ␣-helix from subunit B is shown interacting with the nucleotide loop of the HOCbl molecule bound to subunit A. The lower ligand, 5,6-dimethylbenzimidazole, is shown coordinated to the cobalt ion of the corrin ring.