Computer-assisted Docking of Flavodoxin with the ATP:Co(I)rrinoid Adenosyltransferase (CobA) Enzyme Reveals Residues Critical for Protein-Protein Interactions but Not for Catalysis*

The activity of the housekeeping ATP:co(I)rrinoid adenosyltransferase (CobA) enzyme of Salmonella enterica sv. Typhimurium is required to adenosylate de novo biosynthetic intermediates of adenosylcobalamin and to salvage incomplete and complete corrinoids from the environment of this bacterium. In vitro, reduced flavodoxin (FldA) provides an electron to generate the co(I)rrinoid substrate in the CobA active site. To understand how CobA and FldA interact, a computer model of a CobA·FldA complex was generated. This model was used to guide the introduction of mutations into CobA using site-directed mutagenesis and the synthesis of a peptide mimic of FldA. Residues Arg-9 and Arg-165 of CobA were critical for FldA-dependent adenosylation but were catalytically as competent as the wild-type protein when cob(I)alamin was provided as substrate. These results indicate that Arg-9 and Arg-165 are important for CobA·FldA docking but not to catalysis. A truncation of the 9-amino acid N-terminal helix of CobA reduced its FldA-dependent cobalamin adenosyltransferase activity by 97.4%. The same protein, however, had a 4-fold higher specific activity than the native enzyme when cob(I)alamin was generated chemically in situ.

The Co-C bond in coenzyme B 12 is the result of a nucleophilic attack by the Co(I) ion on the C5Ј carbon of ATP. In vitro the Co(I) ion is generated by the transfer of an electron from reduced flavodoxin A (FldA) to Co(II), an event that is currently thought to occur in the active site of CobA (14).
FldA is an electron transfer protein essential to cell survival (20). In addition to its involvement in the corrinoid adenosylation pathway (14), FldA provides reducing equivalents for the reactivation of MetH, the B 12 -dependent methionine synthase (21). Elegant NMR spectroscopy studies identified the interacting surfaces of a fragment of MetH with FldA (22).
One unanswered question regarding the mechanism of catalysis by CobA is how the redox potential of the Co(II) to Co(I) transition (Ϫ610 mV) (23) is increased enough so the electron transfer from reduced FldA (Ϫ450 mV; semiquinonine/hydroquinone) (21,24) can occur. As mentioned above, we previously hypothesized that FldA reduces Co(II) to Co(I) in the CobA active site, triggering the attack of the Co(I) nucleophile on the 5Ј carbon of the ribosyl moiety of ATP (14). Stich et al. (25) recently reported that binding of the corrinoid substrate to the CobA/MgATP complex increases the Co(II)/Co(I) redox potential to within the range for FldA reduction (25), leading to the reduction of Co(II) to Co(I) in the active site of CobA. At present, the interactions of FldA with CobA are not understood. Here we report results of studies aimed at advancing our understanding of CobA⅐FldA interactions highlighting the use of computer modeling to generate experimentally testable models.

EXPERIMENTAL PROCEDURES
Strains and Plasmids-TABLE ONE lists all strains and plasmids used in this work. TABLE TWO lists all the mutagenic PCR oligonucleotides used to generate the plasmids used in these studies.
PCR Methods-An Eppendorf Mastercycler thermocycler was used to amplify templates listed in TABLE TWO using primers listed in  TABLE TWO. Amplification products were cloned in vector pET15b (Novagen) to fuse an N-terminal hexahistidine tag to the gene product.
Site-directed Mutagenesis-The Stratagene QuikChange site-directed mutagenesis kit was used to create point mutants in plasmids pCOBA17 and pFLDA4 (TABLE TWO). Nonradioactive BigDye (Amersham Biosciences) protocols for DNA sequencing were used to verify the presence of mutations. Plasmids were reconstructed and then sequenced at the DNA sequencing facility at the University of Wisconsin-Madison Biotechnology Center.
Purification of His-tagged CobA Proteins-Plasmids were transformed into Escherichia coli strain BL21(DE3) for overexpression (JE3892). Fresh transformants were used to inoculate 2 ml of fresh Luria-Burtani broth (LB)/ampicillin medium. A starter culture (1 ml) was used to inoculate 100-ml cultures of LB/ampicillin medium. After overnight growth at 37°C with shaking, cells were harvested by centrifugation at 15,000 ϫ g at 4°C in a Beckman/Coulter Avanti J-25I centrifuge equipped with a JLA-16.250 rotor. Cell paste was frozen at Ϫ80°C until use. To purify proteins 5 ml of lysis buffer (1ϫ Bug Buster reagent (Novagen) in 1ϫ binding buffer, 0.1 mM phenylmethanesulfonyl fluoride) was used to resuspend the cell pellet. Cells were lysed at room temperature for 20 min and centrifuged for 30 min at 43,667 ϫ g. Cellfree extract was passed through a 0.45-m syringe filter before loading onto HisBind 900 cartridges (Novagen). Protein was purified as per the manufacturer's instructions. Protein was eluted into 1 mM EDTA. Fractions containing highly purified protein were pooled and dialyzed once against 50 mM Tris-Cl buffer, pH 8, (at 4°C) containing 10 mM EDTA. Subsequent dialyses were performed against buffer without EDTA. In some cases protein was concentrated using a YM10 Centricon unit (molecular weight cutoff ϭ 10,000; Amicon). Glycerol was added to 10% (v/v) and dithiothreitol to 1 mM; 100-l drops of protein solution were flash-frozen using liquid N 2 , and pellets were stored at Ϫ80°C until used. Native Wild Type (WT) and CobA ⌬N2-26 Protein Purification-Native WT CobA protein was overexpressed on strain JE2886 and purified as described (15). Allele cobA1328 encoding the mutant CobA protein with the N-terminal 2-26 amino acids deleted (CobA ⌬N2-26 ) was cloned into plasmid pT7-7 (27) and overproduced in E. coli strain BL21(DE3); all chromatographic steps were identical to those used to isolate native WT CobA protein.
FldA Overexpression and Purification-Plasmids encoding WT or mutant His-tagged FldA proteins were freshly transformed into strain JE4182. Four 2-liter flasks containing 750 ml of LB medium containing kanamycin (50 l/ml) and ampicillin (100 l/ml) were each inoculated with 10 ml of an overnight starter culture grown in the same medium. After the culture reached mid-log phase at 30°C and 160 rpm shaking, riboflavin was added to a final concentration of 10 M, and cultures were shifted to 42°C for 1 h and incubated at 37°C overnight. Cells were harvested at 4°C and 15,000 ϫ g for 10 min; cell paste was frozen at Ϫ80°C until used.
Frozen cell paste was resuspended in 20 mM sodium phosphate buffer, pH 7.5, (at 4°C) containing 0.5 M NaCl and 1 mM phenylmethanesulfonyl fluoride. The suspension was passed through a French press twice, 10 mg of DNase was added, and the lysate was incubated on ice for 10 min. The lysate was centrifuged at 4°C at 43,667 ϫ g for 30 min, and soluble protein was passed through a 0.45-m syringe filter. Proteins were isolated from crude cell extracts using a Ä KTA fast protein liquid chromatograph equipped with a 5-ml HisTrap column (Amersham Biosciences). After column equilibration, cell-free extract was loaded onto the column, which was washed with 5 column volumes (CV) of 20 mM sodium phosphate buffer, pH 7.5, (at 4°C) containing 0.5 M NaCl. Mutant FldA proteins were eluted with a 0 -50% linear gradient of 20 mM sodium phosphate, pH 7.5, (at 4°C), 0.5 M NaCl, 0.5 M imidazole over 15 column volumes. Fractions containing pure protein were pooled, concentrated in an YM10 Centricon unit, and dialyzed against 50 mM Tris-Cl, pH 7.5, (at 4°C) containing 1 mM EDTA. Protein concentration was determined by A 466 to measure flavin mononucleotide (FMN) bound to FldA (⑀ 466 ϭ 8250 M Ϫ1 ) (28). Glycerol was added to 10% (v/v), and dithiothreitol was added to 1 mM before protein was flash-frozen and stored at Ϫ80°C until used. Fpr protein (ferredoxin (flavodoxin):NADP ϩ reductase) was purified as described (14).
Fre Protein Overexpression and Purification-H 6 Fre (flavin reductase; EC 1.5.1.29) protein was overexpressed using E. coli BL21(DE3)/ pFRE3 (13). The latter was re-constructed before overexpression by transforming strain JE3892 with plasmid pFRE3. A single colony was used to inoculate 5 ml of fresh LB/ampicillin medium and incubated to stationary phase at 37°C. 2.5 ml of the starter culture was used to inoculate four 1-liter flasks containing 250 ml of LB/ampicillin medium. All cultures were incubated at 37°C to late log phase and induced with 0.5 mM isopropyl-␤-D-thiogalactopyranoside with shaking for an additional 4 h. Cells were harvested at 15,000 ϫ g for 10 min; the cell pellet was frozen at Ϫ80°C until used. To purify H 6 Fre (flavin reductase) protein, 20 ml of lysis buffer (1ϫ Bug Buster reagent (Novagen) in 1ϫ binding buffer, 1 mM phenylmethanesulfonyl fluoride) was used to resuspend the cell pellet. Cells were lysed at room temperature for 20 min and centrifuged for 30 min at 43,667 ϫ g. Cell-free extract was filtered through a 0.45-m and then a 0.2-m syringe filter before loading onto HisBind 900® cartridges (Novagen). H 6 -Fre Protein was purified as per the manufacturer's instructions. Protein was eluted directly into 5 mM EDTA. Fractions containing homogeneous protein were pooled, concentrated in an YM10 Centricon unit, and dialyzed once against 50 mM Tris-Cl buffer, pH 7.5, (at 4°C) containing 10 mM EDTA. Subsequent dialyses were against buffer without EDTA. Glycerol was added to 10% (v/v) and dithiothreitol to 1 mM before aliquots were flash-frozen and stored at Ϫ80°C.
Growth Behavior Analysis-Ten l of overnight cultures of all strains was used to inoculate 190 l of fresh medium in each well of a 96-well microtiter dish; all strains were cultured in triplicate. Growth media contained 1ϫ no-carbon E medium (29), 1 mM Mg ϩ2 , 0.5 mM methionine, 1ϫ trace minerals (30), 50 mM NH 4 Cl, 50 mM ethanolamine, pH 7.0. When added, ampicillin was used at 100 g/ml, kanamycin was used at 50 g/ml, and AdoCbl, hydroxycobalamin, or dicyano-cobinamide was used at 200 nM. For growth on glycerol as the sole carbon and energy source, ethanolamine and methionine were omitted, and glycerol was added to 30 mM. Cultures were incubated with shaking for up to 96 h at 37°C in a SpectraMAX Plus automatic plate reader (Molecular Devices). Western blots were performed to determine whether complementation defects were due to protein expression or stability issues. However, because of the sensitivity of the growth assay, even undetectable levels of CobA WT allowed full complementation.
Cobalamin Adenosyltransferase Assays- Fig. 1 shows three different ways to generate the co(I)rrinoid substrate for CobA. Reactions contained 4 mg of potassium borohydride (KBH 4  enzyme. In this hypothetical CobA⅐FldA complex, the FMN cofactor of FldA was within 10 Å of the Cbl substrate in the CobA active site, and hydrophobic patches on both proteins were brought together (Fig. 2D).
Competitive Inhibition of FldA-dependent AdoCbl Synthesis by a FldA Peptide-Based on our computer model, residues TWYY-GEAQCDWDD 68 of FldA were predicted to be part of the surface interacting with CobA. We hypothesized that a peptide mimic of this region should block CobA⅐FldA interactions. Because the same region of FldA is known to also interact with Fpr, the latter was present in the reaction mixture in large excess and was preincubated with FldA before peptide, ATP, and CobA were added to start the cobalamin adenosyltransferase assay reaction. The control peptide TWYYGAAQCDWDA 68 contained E61A and D68A mutations to determine whether, as predicted, these two charged residues contributed to any interactions of FldA with CobA. A third FldA peptide of similar size and hydrophobicity (WPTAGYHFEASKG 132 ) was used as negative control. The control peptide did not contain any residues known to interact with MetH ( Fig.  3) and, hence, were predicted to have no effect of the interactions of FldA with CobA.
The effect of all three peptides on FldA-dependent AdoCbl synthesis was assayed in vitro. In these assays, AdoCbl formation relied on productive CobA⅐FldA interactions for the generation of the cob(I)alamin nucleophile. To optimize the assay conditions, various concentrations of FldA were used in assays where substrates and CobA were held constant. The K m for FldA was 5.65 M and V max ϭ 31.8 nmol min Ϫ1 mg Ϫ1 (Fig. 4). For Lineweaver-Burk analysis, subsaturating levels of FldA were used such that the reaction proceeded at a rate between 4 and 50% V max when no peptides were added. Each of the three peptides was then varied from 2-6 M for each concentration of FldA.
Peptide TWYYGEAQCDWDD 68 inhibited AdoCbl formation by ϳ30% when present at a 1.5:1 peptide:FldA ratio. At 59% V max , the specific activities were 19 and 13, respectively. This level of inhibition was significant considering the small size of the peptide. According to Lineweaver-Burk analysis, this inhibition was competitive (K m peptide ϭ 29 M and K i ϭ 1.3 mM). A 13-amino acid peptide mimic of FldA had only a 5-fold decrease in CobA binding efficiency. This decreased affinity might be due to several reasons. The peptide could have greater conformational flexibility than the FldA protein or lack a secondary structure needed for interactions with CobA. In addition, the peptide may not span all the residues involved in the CobA⅐FldA interaction.
In contrast, the mutagenized and control peptides (TWYYGAAQC-DWDA 68 and WPTAGYHFEASKG 132 , respectively) failed to inhibit AdoCbl production even when present at 10-fold increase over FldA. These results suggested that the TWYYGEAQCDWDD 68 peptide did bind to CobA and interfered with the FldA docking. Results with the mutagenized peptide indicated that this inhibition depended on residues Glu-61 and Asp-68.
In Vivo Assessment of CobA Activity-To test the effect of mutations in residues Arg-9, Arg-98, and Arg-165, we site-directed mutagenized each one of these residues to Ala or Glu. Permutations of mutated residues were also constructed, and their effects were tested in vitro and in vivo.
Plasmids containing mutant cobA alleles were transformed into a metE cobA strain of S. enterica (JE7180) to assess their level of function in vivo. Strains were grown under conditions that demanded low or high levels of cobalamin synthesis. When grown on glycerol as sole carbon source, a metE cobA strain cannot convert dicyano-cobinamide into Cbl because enzymes that assemble the nucleotide loop of Cbl require AdoCbi (4,31). Therefore, in the absence of Cbl, the Cbl-dependent methionine synthase (MetH) cannot methylate homocysteine to produce methionine. Hence, the metE cobA strain is an AdoCbi, Cbl, or methionine auxotroph. When grown on ethanolamine as the carbon and energy source, AdoCbl synthesized by CobA is needed to activate transcription of the ethanolamine utilization (eut) operon and to provide AdoCbl for ethanolamine ammonia-lyase function (18,32).
Of the CobA mutant proteins tested, only CobA R98A , CobA R98E , and CobA R165A supported growth on ethanolamine as carbon and energy source, albeit to a reduced level relative to the wild-type strain (Fig. 5A). It was surprising to learn, however, that protein CobA R98A with Ͻ20% of the CobA WT activity (TABLE THREE, column C) supported growth on ethanolamine at ϳ80% that of the growth rate observed with a strain synthesizing CobA WT enzyme (Fig. 5). Surprisingly, all cobA alleles tested complemented the Cbl auxotrophy of the cobA mutant strain during growth on glycerol (data not shown). We interpret these data to mean that all mutant CobA proteins sufficiently interacted with FldA to produce enough Cbl to satisfy the methionine requirement of the cell. These results indicated that the wild-type CobA enzyme synthesizes an excess of AdoCbl. The inability of some CobA variants to complement a cobA strain during growth on ethanolamine (demands more AdoCbl) reflected the negative effects of the mutations on CobA⅐FldA interactions rather than being the result of expression problems or unstable CobA mutant proteins.
In Vitro Assessment of CobA Activity-Having two ways of generating the Co(I) nucleophile (Fig. 1, A and B) allowed us to determine whether the CobA residues identified by the model were relevant to docking, catalysis, or both.
Activity of CobA Variants When the Cob(I)alamin Substrate Was Generated Using a Chemical Reductant-To distinguish between docking defects and loss of catalytic activity, we measured the activity of mutant CobA enzymes independent of CobA⅐FldA docking. For these experiments, we reduced cob(III)alamin to cob(I)alamin with potassium borohydride (KBH 4 ) before adding CobA (Fig. 1A). Regardless of the change at residues Arg-9 or Arg-165 (i.e. R9A, R9E, R165A, or R165E), the specific activity of the resulting mutant CobA proteins was  equal or better than that of the wild-type enzyme (TABLE THREE,  column B; lines 2, 3, 6, 7 versus line 1). Although it is a formal possibility that Arg-9 and Arg-165 mutations negatively affect binding of cob(II)alamin but not cob(I)alamin to CobA, we believe such a defect would have to be pronounced since saturating levels of cob(II)alamin were used in the FldA-dependent assay.
Activity of CobA Variants When the Cob(I)alamin Substrate Was Generated Using the Fpr/FldA Reduction System-To determine whether CobA residues Arg-9 and Arg-165 were involved in docking, the specific activity of each mutant CobA protein was determined in assays where CobA⅐FldA interaction was required for the generation of the cob(I)alamin substrate (Fig. 1B). The results from these assays (TABLE THREE, column C) were consistent with the results from in vivo experiments using these plasmids (Fig. 5). In reactions where CobA R9A , CobA R9E , and CobA R165E were used, we measured very low levels of adenosyltransferase specific activity (TABLE THREE, column C; lines 2, 3, 7 versus line 1). In contrast, when CobA R98A , CobA R98E , and CobA R165A proteins were used in the assay, the adenosyltransferase specific activity we measured was 7-40-fold higher (TABLE THREE, column C; lines 4, 5, 6 versus line 1). Combinations of the above mutations resulted in proteins with undetectable activity (TABLE THREE, column C; lines 8 -11). On the basis of the data in Fig. 5 and TABLE THREE, we conclude that residues Arg-9 and Arg-165 are important to docking but not to catalysis.
The interpretation of the data regarding residue Arg-98 is less straightforward. In isolation, the effect of a mutation at Arg-98 depends on the nature of the substitution. A drastic change in charge and size of the side chain (R98A) results in a substantial loss of specific activity relative to the wild-type enzyme (143 versus 10; TABLE THREE, column B; lines 1 versus 4). A drastic change in charge but not in the size of the side chain (R98E) reduces the specific activity of the enzyme less than 2-fold relative to the wild type (143 versus 97; TABLE THREE, column B; lines 1 versus 5). CobA R98A was more active in vivo than in either in vitro assay. Interestingly, the combination of R98E with R165E mimicked the effect of the R98A change (CobA R98E/R165E specific activity ϭ 20; TABLE THREE, column B; lines 1 versus 10), an effect that was reversed by the introduction of the R9E mutation (CobA R9E/R98E/R165E specific activity ϭ 108; TABLE THREE, column B; lines 1 versus 11). With the data in hand we suggest that Arg-98 is necessary for structural integrity of CobA.
Effect of Compensatory Mutations in FldA-Based on the model of the CobA⅐FldA complex (Fig. 2), we hypothesized that changing FldA residues Asp-68 and Asp-93 to Arg would compensate for the negative  effects of CobA mutations R9E and R165E on FldA-dependent synthesis of AdoCbl. A problem with these experiments arose when the Fpr enzyme failed to reduce FldA D68R and FldA D93R mutant proteins (data not shown). In an attempt to circumvent this problem, we used the NADH:FMN reductase (Fre) enzyme (13,33) to reduce free FMN, which would indirectly reduce the flavin cofactor in mutant FldA enzymes (Fig. 1C). If the Fre system worked, CobA R9E was predicted to preferentially interact with either FldA D68R or FldA D93R proteins. Or else CobA R9E would interact equally well with the FldA D68R and FldA D93R proteins if the orientation for docking were not important. However, CobA R9E was not expected to interact with wild-type FldA or the FldA D68R/D93R proteins. The first observation we made was that when we used the Fre enzyme, the specific activity of CobA R9E increased 12-fold over the level measured when Fpr was used to reduce FldA (4.5 versus 0.4). CobA R9E was not as active when either FldA D68R (specific activity ϭ 1.7) or FldA D93R (specific activity ϭ 2) was used in the assay. When FldA D68R/D93R protein was used, the specific activity of CobA R9E was as high as when wild-type FldA was used (both 4.5). Results from assays using permutations of CobA and FldA mutant proteins did not reveal any additional insights into CobA⅐FldA interactions.
Effect of Changes in the Hydrophobic Topology of the CobA Docking Surface-The hydrophobic patch of CobA that is proposed to be involved in CobA⅐FldA docking has a concave surface. To investigate the contributions of hydrophobic packing to the interactions of CobA with FldA, residue Ala-134 was mutated to Leu. Ala-134 was chosen for mutagenesis because it is the only Ala in the hydrophobic patch. Hence, a change to Leu would be tolerated by the polypeptide. In contrast, mutating other residues in the patch (Phe or Trp) to a less bulky side chain like Ala or Gly would be expected to affect the structural integrity of the protein. The CobA A134L protein supported growth on glycerol or on ethanolamine as well as did CobA WT (Fig. 5), and its specific activity under all assay conditions (KBH 4 or Fpr/FldA) was as high as that of the CobA WT protein (TABLE THREE columns  Deletion of the CobA N-terminal Helix Increases the Specific Activity of the Enzyme but Decreases CobA⅐FldA Docking-A striking feature of the CobA crystal structure is an N-terminal helix of 26 amino acids that remains disordered unless Mg/ATP and the corrinoid substrate are bound to the active site (12). The N-terminal helix of CobA may have several functions. It may help secure the corrinoid substrate in the active site, it may exclude water from the active site, thereby helping to stabilize the co(I)rrinoid nucleophile, or it may be important for FldA recognition or a combination of the above. To gain insights into the role of the N-terminal helix, we deleted it and tested the ability of the resulting CobA ⌬N2-26 variant to function in vivo and in vitro. Deletion of this helix did not affect the ability of the mutant CobA protein to support growth of a metE cobA strain on glycerol (data not shown). However, the CobA ⌬N2-26 protein failed to support growth on ethanolamine as sole carbon source (Fig. 5).
Results from in vitro assays showed that the specific activity of the CobA ⌬N2-26 enzyme was only 3% of the CobA WT enzyme when FldA was used as the reductant. The same protein has 4-fold higher specific activity than the wild-type when cob(I)alamin was generated with KBH 4 . Eliminating the N-terminal helix resulted in a more efficient enzyme as long as cob(I)alamin is in excess. When the concentration of cob(I)alamin in the active site depends on interaction with FldA, the absence of the N-terminal helix of CobA affects AdoCbl production.
Deletion of the N-terminal helix of CobA may affect electron transfer, catalysis, or both.

Computer Modeling as a Tool to Investigate Protein-Protein
Interactions-At present, there is a great deal of interest in improving our molecular understanding of how protein complexes form. Fortunately, the accumulation of data in protein databases is rapidly increasing as a result of structural genomics initiatives. Here we report one example of how this valuable information can be used to generate experimentally testable models of protein docking. Sophisticated, userfriendly software for the manipulation of protein structure data is available to researchers interested in studying enzyme complexes. As shown here, a computer-assisted model facilitates the design of peptide mimics predicted to strongly inhibit the interactions between partner proteins. Structure models and peptide mimics can be used to investigate the contribution of specific amino acid residues to the formation of biologically relevant protein-protein contacts.
Insights into CobA⅐FldA Interactions-We have identified part of the regions of the CobA and FldA proteins that allow electron transfer from the flavin in FldA to the co(II)rrinoid in the active site of CobA. Four pieces of evidence support the conclusion that FldA residues TWYY-GEAQCDWDD 68 are part of the FldA structure that interacts with CobA. First, a 30% inhibition of CobA activity when the TWYY-GEAQCDWDD 68 peptide is present in the assay is significant given the small size of the peptide (13 amino acids). Second, there was a complete lack of an inhibitory effect by another FldA-derived peptide of similar size and hydrophobicity. Third, mutagenized peptide TWYYGAAQC-DWDA 68 failed to inhibit adenosylation, demonstrating that residues Glu-61 and Asp-68 are key components of the FldA surface interacting with CobA. Fourth, this peptide is part of the surface of FldA that interacts with MetH and Fpr (28).
Residues of CobA Critical for Docking-We suggest that residue Arg-9 is essential for docking of CobA with FldA (TABLE THREE). Regardless of the nature of the change at this position, CobA cannot dock with FldA if Arg-9 is altered. A strong case for docking can also be made for residue Arg-165. However, there is more tolerance for change at this position. A CobA R165A variant still supported growth on ethanolamine, indicating that FldA-dependent reduction of the substrate was sufficient to meet the high demand of AdoCbl required to grow on ethanolamine. Only a drastic change in the charge of the side chain (e.g. Arg to Glu) prevented docking of CobA with FldA. The role of residues Arg-9 and Arg-165 appears to be limited to CobA⅐FldA interaction, since the catalytic competence of mutant enzymes unable to dock with FldA was equal to that of the wild-type protein (TABLE THREE,  It is not surprising that a truncation of the N-terminal helix results in poor interactions of CobA with FldA. CobA ⌬N2-26 protein is missing residue Arg-9 as well as other residues that may be involved in docking. What is surprising, however, is the substantial increase in the specific activity of the enzyme (4-fold higher than CobA WT enzyme). We speculate that this increase in specific activity may be due to an increase in the rate of substrate binding, product release, or both.
Results from recent EPR and MCD studies by Stich et al. (25) have provided evidence for the existence of a four-coordinate co(II)rrinoid in the active site of the wild-type CobA enzyme. The effect that FldA docking to CobA may have on electron transfer or stability of the co(II)rrinoid substrate is under investigation. Use of the CobA ⌬N2-26 protein might be especially useful in probing the role of residues in the N-terminal helix in activation and stabilization of the corrinoid substrate in the CobA active site.
Differences in the Sensitivity of the in Vivo Assays for AdoCbl Reveal the Coenzyme B 12 Requirement of S. enterica-We typically use low Cbl levels (1 nM) in the culture medium to satisfy the methionine requirement of S. enterica. In contrast, optimal growth rates on ethanolamine as carbon and energy source require a Ͼ2 orders of magnitude higher concentration of Cbl in the medium (150 nM (34)). It was surprising to learn that a mutant CobA protein that is one-tenth as active as the wild-type enzyme can meet the demand for AdoCbl during growth on ethanolamine. This result explains why it has been difficult to isolate cobA alleles that do not contain nonsense or missense mutations that severely destabilize the protein. The low requirement for endogenous AdoCbl when growing on ethanolamine means that Salmonella strains carrying cobA alleles encoding defective CobA variants appear indistinguishable from cobA ϩ strains on solid media.