Coordination chemistry controls the thiol oxidase activity of the B12-trafficking protein CblC

The cobalamin or B12 cofactor supports sulfur and one-carbon metabolism and the catabolism of odd-chain fatty acids, branched-chain amino acids, and cholesterol. CblC is a B12-processing enzyme involved in an early cytoplasmic step in the cofactor-trafficking pathway. It catalyzes the glutathione (GSH)-dependent dealkylation of alkylcobalamins and the reductive decyanation of cyanocobalamin. CblC from Caenorhabditis elegans (ceCblC) also exhibits a robust thiol oxidase activity, converting reduced GSH to oxidized GSSG with concomitant scrubbing of ambient dissolved O2. The mechanism of thiol oxidation catalyzed by ceCblC is not known. In this study, we demonstrate that novel coordination chemistry accessible to ceCblC-bound cobalamin supports its thiol oxidase activity via a glutathionyl-cobalamin intermediate. Deglutathionylation of glutathionyl-cobalamin by a second molecule of GSH yields GSSG. The crystal structure of ceCblC provides insights into how architectural differences at the α- and β-faces of cobalamin promote the thiol oxidase activity of ceCblC but mute it in wild-type human CblC. The R161G and R161Q mutations in human CblC unmask its latent thiol oxidase activity and are correlated with increased cellular oxidative stress disease. In summary, we have uncovered key architectural features in the cobalamin-binding pocket that support unusual cob(II)alamin coordination chemistry and enable the thiol oxidase activity of ceCblC.

are the active cofactor derivatives found in mammals and are required by methionine synthase and methylmalonyl-CoA mutase, respectively (1). Mammals are unable to synthesize this essential cofactor and house an elaborate pathway for converting dietary cobalamins to the active cofactor forms and for delivering them to the target proteins (2)(3)(4). Defects in this pathway result in a functional B 12 deficiency and lead to cobalamin disorders that are classified into 10 genetic complementation groups, cblA-G, cblJ, cblX, and mut (5)(6)(7)(8).
CblC (also known as MMACHC for methylmalonic aciduria type C and homocystinuria) is a B 12 -processing protein that converts cobalamin derivatives entering the cytoplasm to a common intermediate that is subsequently partitioned for synthesis of the two biologically active alkylcobalamins, MeCbl and AdoCbl (3,4). Mutations in human CblC (hCblC) represent the most common inborn error of cobalamin metabolism (9). CblC exhibits unusual chemical versatility with demonstrated glutathione transferase, reductive decyanase, and aquocobalamin (OH 2 Cbl) reductase activities (10 -12). The Caenorhabdiitis elegans CblC (ceCblC) catalyzes essentially the same chemical reactions as hCblC but also harbors an additional oxygenscrubbing and glutathione (GSH) oxidation activity, which was proposed to occur via Reactions 1-4 (13). The thiol oxidase activity associated with ceCblC is dependent on its dealkylation or decyanation activity and gives rise to an apparently futile redox cycle, wasting the GSH antioxidant pool and consuming oxygen. The thiol oxidase activity of hCblC is muted in contrast to that of ceCblC. Interestingly, two pathological mutants of hCblC, R161Q and R161G, exhibit elevated thiol oxidase activity, which, in turn, is correlated with increased oxidative stress that is associated with the CblC disorder (14).

Reactions 1-4
In this study, a combination of kinetic, spectroscopic, computational, and crystallographic approaches has revealed the involvement of unusual cobalt coordination chemistry that is supported by the active-site architecture of ceCblC but not hCblC. These studies support a chemical mechanism of thiol oxidation in which glutathionyl-cobalamin (GSCbl) is formed as an intermediate and GSH-dependent deglutathionylation of GSCbl yields GSSG.

Kinetics of thiol oxidase activity in the presence of cobalamin derivatives
The thiol oxidase activity of ceCblC was initially monitored during the dealkylation of AdoCbl by GSH by monitoring O 2 consumption as well as GSSG and H 2 O 2 generation (Fig. 1A). The corresponding k cat values for O 2 consumption and GSSG and H 2 O 2 production were estimated to be 17 Ϯ 1, 19 Ϯ 2, and 15.8 Ϯ 0.4 min Ϫ1 , respectively, at 20°C. Neither GSSG nor H 2 O 2 synthesis was supported by apo-ceCblC or ceCblC loaded with ethylphenylcobalamin (EtPhCbl), an analog that cannot be dealkylated (15) (Fig. 1B), indicating that the thiol oxidase activity requires dealkylation of the alkylcobalamin that is initially bound to CblC. Furthermore, the thiol oxidase activity is dependent on the presence of the initially formed cob(I)alamin or the subsequently oxidized cob(II)alamin state. The K m for GSH was estimated to be 181 Ϯ 12 M from the O 2 consumption assay and 134 Ϯ 7 M from the GSSG production assay (Fig.  1C). By comparison, the K act for GSH in the AdoCbl dealkylation assay was estimated to be 183 Ϯ 11 M (13).
Next, we compared the thiol oxidase activity of ceCblC loaded with different cobalamin derivatives (supplemental Fig.   S1). A pronounced lag phase was observed with some derivatives (e.g. AdoCbl and CNCbl). Comparing the rates of GSHdependent cobalamin processing activities (i.e. dealkylation or reductive decyanation) and GSSG production, a correlation was observed between the rates of cobalamin processing and GSSG production ( Table 1). When cofactor processing rates are high as with MeCbl and GSCbl, GSSG production rates are correspondingly high, reaching a maximal value of ϳ100 min Ϫ1 under our experimental conditions. This result suggests that either cob(I)alamin (the initial product of dealkylation) or cob(II)alamin (product of reductive decyanation or formed by oxidation of cob(I)alamin) initiates the futile cycle.

Redox cycling is not dependent on OH 2 Cbl reduction by GSH
Previously, we had proposed a mechanism of futile redox cycling involving OH 2 Cbl as an intermediate (Reactions 1-4) (12). Due to the technical challenge of loading ceCblC with "base-off" OH 2 Cbl (in which the endogenous dimethylbenzimadazole ligand is not coordinated to cobalt; see supplemental . . The resulting spectrum exhibited absorption maxima at 357 and 513 nm, in good agreement with the spectrum of ceCblC-aquocobinamide (OH 2 Cbi), which, lacking the dimethylbenzimadazole ligand, is naturally base-off (supplemental Fig. S2A). Next, we tested the reactivity of ceCblC-OH 2 Cbl toward GSH under anaerobic conditions. As seen previously under aerobic conditions (12), cob(II)alamin ( max ϭ 473 nm) accumulation was observed (supplemental Fig. S3). Formation of GSCbl from OH 2 Cbl (by ligand exchange with GSH) was not observed, although GSCbl forms readily in solution. This result is consistent with ligand exchange being disfavored in the baseoff state (16), which facilitates OH 2 Cbl reduction instead (17). The observed rate of cob(II)alamin formation from OH 2 Cbl and GSH was estimated to be 0.46 Ϯ 0.02 min Ϫ1 , which is significantly smaller than the rate of GSSG production under comparable conditions (ranging from ϳ3 to 100 min Ϫ1 ( Table  1). The discrepancy between the k obs for OH 2 Cbl reduction by GSH and the k obs for GSSG synthesis by ceCblC rules out OH 2 Cbl reduction as a kinetically competent step during redox cycling (Reaction 1).

An alternative mechanism for redox cycling
In light of the above results, we considered an alternative mechanism in which GSCbl is formed as a reaction intermediate and is deglutathionylated by GSH to form GSSG in a redox cycle in which O 2 undergoes a net two-electron oxidation to H 2 O 2 (Reactions 5-7). In this model, cob(I)alamin formed by dealkylation of MeCbl or AdoCbl enters the cycle via Reaction 6.

Reactions 5-7
To test this mechanism, GSH-dependent deglutathionylation of GSCbl by ceCblC (Reaction 5) was monitored by stopped-flow spectrophotometry under anaerobic conditions. GSCbl binds to ceCblC in the base-off state and exhibits absorption maxima (417 and 506 nm) that are distinct from the base-on cofactor in solution (535 nm) and identical to ceCblCbound glutathionylcobinamide (GSCbi) (supplemental Fig. S4). The addition of GSH initiates rapid deglutathionylation and yields cob(I)alamin (389 nm) with a k obs of 108 Ϯ 12 min Ϫ1 ( Fig.  2A, top and inset). An ϳ0.2-s lag precedes formation of cob (I)alamin, during which time the spectrum of ceCblC-GSCbl shifts from 417 to 507 nm with isosbestic points at 453 and 532 nm ( Fig. 2A, bottom). We tentatively assign the 507-nm species as the [ceCblC-GSCbl-GS Ϫ ] intermediate.
Next, we examined deglutathionylation of GSCbl under aerobic conditions. Due to its air sensitivity, cob(I)alamin accumulation was observed ϳ7 s after initiation of the reaction (Fig. 2B, top), presumably when O 2 in the reaction mixture had been scrubbed via redox cycling. The accumulation of cob(I)alamin under aerobic conditions is remarkable and was previously ascribed to thiol oxidase activity-mediated depletion of O 2 during dealkylation of alklycobalamins (13). Before cob(I)alamin accumulation, formation of the 507-nm absorbing [ceCblC-GSCbl-GS Ϫ ] intermediate was observed within 0.2 s (Fig. 2B, bottom). Between 0.2 and 7 s, a mixture of [ceCblC-GSCbl-GS Ϫ ] and cob(II)alamin accumulated, as indicated by the absorption increase at 475 nm (Fig. 2B, middle). These observations are consistent with the intermediacy of GSCbl and cob(II)alamin (Reactions 7 and 6, respectively), during redox cycling. After 7 s, the absorbance associated with the 507-nm species decreased sharply and was accompanied by an increase in absorbance at 389 nm, consistent with cob(I)alamin accumulation. The final products of the reaction were a mixture of cob(I)alamin and cob(II)alamin.

GSCbl is an intermediate during dealkylation of MeCbl
Under anaerobic conditions, GSH-dependent dealkylation of MeCbl by ceCblC results in the full conversion to cob(I)alamin within 3 s (13). In contrast, under aerobic conditions, cob(I)alamin accumulation is observed only after 7.5 s (Fig. 2C,  top). Between 0 and 4 s, an increase in absorbance between 500 and 550 nm is observed with a concomitant decrease in absor- bance at 454 nm corresponding to base-off MeCbl (Fig. 2C, bottom).
To assess whether the spectral change in the 500 -550 nm region was due to formation of the [ceCblC-GSCbl-GS Ϫ ] intermediate, the spectral changes during 0 -1 s were analyzed (supplemental Fig. S5). A reasonably good correspondence was seen between the experimental and predicted difference spectra for the conversion of ceCblC-MeCbl to [ceCblC-GSCbl-GS Ϫ ], supporting the conclusion that GSCbl is also formed during aerobic dealkylation of ceCblC-MeCbl, where redox cycling occurs. Between 4 -7.5 s, the spectral changes (Fig. 2C, middle) were similar to those observed during aerobic deglutathionylation of GSCbl (Fig. 2B, middle), consistent with the formation of both cob(II)alamin and GSCbl (Reactions 6 and 7).

Redox activity of cob(II)alamin bound to ceCblC
To further dissect the mechanism of redox cycling, we examined the mechanism of oxidation of cob(I)alamin and cob (II)alamin bound to ceCblC. Cob(I)alamin is the product of both the dealkylation and deglutathionylation reactions and is highly prone to oxidation. Single electron oxidation of cob(I) alamin by O 2 generates cob(II)alamin and O 2 . (Reaction 6). The redox potential for the O 2 /O 2 . at pH 7.0 is Ϫ330 mV (18), and for base-off cob(II)alamin/cob(I)alamin, it is Ϫ500 mV (19); hence, cob(I)alamin oxidation by O 2 is highly favorable. Although free cob(II)alamin is also readily oxidized in air (the redox potential of base-on OH 2 Cbl/cob(II)alamin is ϩ190 mV) (19), cob(II)alamin bound to ceCblC is stabilized against oxidation (supplemental Fig. S6). The stabilization against air oxidation is consistent with OH 2 Cbl not being an intermediate in redox cycling (Reaction 4) and necessitates consideration of an alternative mechanism (Reaction 7). ceCblC-cob(II)alamin exposed to O 2 . (generated by xanthine/ xanthine oxidase) resulted in OH 2 Cbl formation (supplemental Fig. S7A and Fig. 3A). In the presence of superoxide dismutase, the rate of OH 2 Cbl formation was significantly diminished (supplemental Fig. S7B). The rates of O 2 . generation and ceCblC-cob(II)alamin oxidation were virtually identical (supplemental Fig. S7C), as reported previously with free cob(II)alamin (20). H 2 O 2 (which could form via dismutation of O 2 . ) did not contribute to ceCblC-cob(II)alamin oxidation because catalase had no significant effect on the oxidation rate (supplemental Fig. S7, B and C).
In the presence of GSH, the absorption spectrum of ceCblCcob(II)alamin was unaffected by exposure to O 2 . (Fig. 3A), although GSSG production, a hallmark of the thiol oxidase activity, was observed (Fig. 3B). GSSG production was diminished in the presence of SOD or when the dealkylation-inert EtPhCbl derivative was used. SOD inhibited GSSG formation when the thiol oxidase activity was monitored during dealkylation of AdoCbl or decyanation of CNCbl (Fig. 3C). These data are consistent with the involvement of O 2 . and cob(II)alamin in redox cycling (Reaction 7).

EPR and magnetic CD (MCD) analysis of ceCblC-cob(II)alamin
GSCbl formation during the redox cycle requires ligation of the GS Ϫ anion to the cobalt atom. However, ceCblC-OH 2 Cbl does not undergo ligand exchange to form GSCbl (supplemen-  Binding of GSH leads to expulsion of O 2 and to the appearance of additional features in the EPR spectrum at g ϳ2. 3, which resembles the spectrum of cob(II)alamin bound to the methyltransferase, TsrM (22) (Fig. 4, C and D). The increase in g ani-sotropy and 59 Co hyperfine coupling constants observed upon GSH binding to ceCblC-cob(II)alamin (Table 2) is consistent with a weakening of the axial Co(II)-ligand bonding interaction.
To evaluate the possibility that GSH directly coordinates to the Co(II) ion in ceCblC-cob(II)alamin, density functional theory (DFT) geometry optimizations and EPR parameter calculations were performed for 5-coordinate cysteine-cob(II)inamide and H 2 O-cob(II)inamide (Table 2). Although DFT fails to accurately predict g values and hyperfine coupling constants for cob(II)alamin species, it reliably reproduces experimental trends in these parameters upon axial ligand perturbations (23). Thus, based on the computational prediction that replacement of the water ligand in H 2 O-cob(II)inamide by cysteine causes a substantial decrease in the g values and 59 Co hyperfine coupling constants, direct coordination by GSH to cob(II)alamin was excluded.
The MCD spectrum of ceCblC-cob(II)alamin revealed the presence of an elongated Co-OH 2 bond, as judged by the ϳ320 cm Ϫ1 red shift in the lowest-energy, positively signed feature at 16,134 cm Ϫ1 compared with free cob(II)inamide (23) (Fig. 5A). GSH caused an additional red shift by ϳ520 cm Ϫ1 (Fig. 5B), indicating a further weakening of the Co-OH 2 bond. We speculate that lengthening of the Co-O bond facilitates ligand exchange, promoting formation of a transient GS Ϫ -cob(II)alamin intermediate. EPR and MCD spectra of samples frozen during redox cycling (Figs. 4 (E and F) and 5C) are similar to those of ceCblC-cob(II)alamin plus GSH, consistent with the presence of the same species in both sets of samples.

Crystal structure of ceCblC
The structure of ceCblC(⌬C8) with MeCbl bound was solved at 1.35 Å resolution ( Table 3). The overall fold of ceCblC is identical to that of hCblC, with which it shares 35% sequence identity (Fig. 6A) (24,25). Based on structural superposition, the root mean square deviations for C␣ atoms and for all atoms were 0.78 and 1.04 Å, respectively. Clear electron density for MeCbl and tartrate (present in the crystallization buffer) was observed in the active site. The ceCblC form used for crystallization had 6 additional residues (ENLYFQ) that remained following cleavage of the C-terminal His 6 tag. Surprisingly, the electron density for this C-terminal extension was observed with the tripeptide YFQ, protruding into the active site of an adjacent monomer ( Fig. 6B and supplemental Fig. S8A). Using the interactions observed between ceCblC, tartrate, and Gln-270 from the tripeptide, GSH was modeled in the active site, which placed its sulfur atom at a 3.1-Å distance from the methyl group of MeCbl, oriented for methyl group transfer (Fig. 6C).
The position of the modeled GSH is supported by the newly obtained structure of hCblC with GSH and a cobalamin analog bound. 3 The modeled GSH interacts with Arg-187, Arg-232, and Arg-256, which form an "arginine-rich" pocket, which was previously implicated in GSH binding (25). The active site is . ) (55). Samples C and D were generated by adding 4 mM GSH to samples A and B, respectively, prior to freezing. The spectra in C and D are identical, and the absence of Co(III)-O 2 . in D indicates that GSH binding excludes oxygen. The blue trace in A and the red trace in C represent the simulated EPR spectra obtained using the fit parameters reported in Table 2. Sample E was prepared by incubating ceCblC (300 commodious, and in fact, a second molecule of GSH can be accommodated in it (supplemental Fig. S8, B and C), supporting the plausibility of the postulated [ceCblC-GSCbl-GS Ϫ ] intermediate.
Two major structural differences exist between ceCblC and hCblC, which could be important for differential tuning of their thiol oxidase activity (Fig. 6, D and E). The first is the conformation of a ␤-hairpin loop (residues 146 -175) that is folded over the cofactor pocket in ceCblC but is flayed away from it in hCblC (Fig. 6A). In this "closed" conformation in ceCblC, electrostatic interactions between the backbone amido and carbonyl groups of Lys-170 and two propionamide chains from the corrin ring push the cofactor closer toward GSH (Fig. 6D). In hCblC, the two propionamide chains bend in toward the corrin ring and avoid contact with Arg-144 (Fig. 6E). The second feature is the presence in ceCblC of the hydrophobic and bulky Ile-172 directly beneath the cobalt on the ␣-face of the corin ring. In hCblC, the ␣-face is hydrophilic and has an ordered water molecule that bridges a propionamide side chain with Ser-146.

Discussion
The thiol oxidase activity involves unusual cobalamin chemistry, and differences in accessible coordination chemistry are correlated with promoting this reactivity in ceCblC but suppressing it in hCblC. Both cob(I)alamin generated from dealkylation and cob(II)alamin formed from GSH-dependent reduc- a Obtained from simulation of the experimental spectrum shown in Fig. 4A. b Obtained from simulation of the experimental spectrum shown in Fig. 4C. c Obtained from DFT calculations.  tion of CNCbl or OH 2 Cbl can enter the redox cycle (Fig. 7). Formation of GSCbl, a key intermediate in the cycle, is supported by the spectroscopic detection of a [ceCblC-GSCbl-GS Ϫ ] species (Fig. 2) and by kinetic studies, which established that GSCbl is deglutathionylated (k obs ϭ 108 Ϯ 12 min Ϫ1 ) at a rate that is comparable with GSSG production (k GSSG ϭ 102 Ϯ 9 min Ϫ1 ). Thus, deglutathionylation of GSCbl is rate-limiting (step 6) within the redox cycle. Because GSCbl does not form via ligand exchange of ceCblC-OH 2 Cbl (supplemental Fig. S4), we propose that ligand exchange occurs at the cob(II)alamin oxidation level instead (step 3). Why is the thiol oxidase chemistry muted in hCblC? The crystal structure of ceCblC reveals a more compact active site than in hCblC (Fig. 6, D and E) and suggests that GSH might be positioned closer to cobalamin, favoring Co-S bond formation. Additionally, we posit that in hCblC, the equilibrium between O 2 . -cob(II)alamin and O 2 . -cob(III)alamin favors the latter via ␣-face H 2 O ligation, stabilizing the preferred 6-coordinate geometry and taking it out of the redox cycle (step 11). Alternatively, in hCblC, H 2 O coordination at the ␣-face of cob(II) alamin, which prefers 5-coordinate geometry, would stabilize it against further chemistry on the ␤-face and curtail redox cycling. DFT geometry optimizations performed for the 5-coordinate cysteine-cob(II)inamide model furnished a 2.51-Å bond distance for Co-S(cysteine) and support the feasibility of a baseoff GS Ϫ -cob(II)alamin intermediate ( Table 2 and supplemental  Table S1). However, the GS Ϫ -cob(II)alamin intermediate was not observed during the redox cycle. EPR and MCD analyses revealed that although the H 2 O ligand is retained in ceCblC-cob(II)alamin in the presence of GSH, the Co-O bond is elongated, indicating bond weakening. It is possible that ligand exchange (step 8) is kinetically coupled to GS Ϫ -cob(II)alamin oxidation (step 4), leading to the lack of GS Ϫ -cob(II)alamin accumulation.
Entering the redox cycle from OH 2 Cbl or CNCbl (steps 9 and 10) poses the following challenge. The redox potential for GS ⅐ / GSH is ϩ920 mV at pH 7.4 (27), which is more positive than the potentials for base-off OH 2 Cbl/cob(II)alamin (ϩ510 mV) or CNCbl/cob(II)alamin (-110 mV) (28). Based on the crystal structure of ceCblC, the ␣-side of cobalamin faces a hydrophobic environment, which disfavors binding of water and, therefore, formation of 6-coordinate OH 2 Cbl. Consequently, the redox potential of the OH 2 Cbl/cob(II)alamin or CNCbl/cob (II)alamin couples in the ceCblC active site could be higher, rendering reduction more favorable. In addition, kinetic coupling of the unfavorable B 12 reduction step to the highly favorable quenching of the thiyl radical (GS ⅐ ) could provide the driving force (Reactions 2 and 3).
Thiol oxidation by free cobalamins and cobinamides has been reported previously (29). Cobinamides, lacking the lower axial base, are ϳ10 3 times more active than the corresponding free cobalamins in oxidizing thiols. CblC binds base-off cobalamins, which not only facilitates reduction but also has the potential to promote adventitious redox cycling, as observed with ceCblC. Although futile redox cycling is suppressed in wild-type hCblC, it is unleashed by two disease-causing mutations (R161Q and R161G) (14). We posit that architectural constraints lead to a larger separation between the cobalt and thiolate (of GSH) ions on the ␤-face in hCblC than in ceCblC, inhibiting ligand exchange, and that loss of the interaction between Arg-161 and GSH allows greater proximity and promotes thiol oxidation. The increased thiol oxidase activity of the Arg-161 mutants is correlated with elevated reactive oxygen species in cell lines harboring CblC mutations (30).
In summary, key architectural differences in the cobalaminbinding pocket between the C. elegans and human CblCs support a role for stereoelectronic and coordination chemistry Tartarate (TAR shown in blue) forms hydrogen bonds with Arg-187, Tyr-241, and Arg-232. The C-terminal extension from the adjacent monomer (shown in gray) protrudes into the active site. The carboxyl group of the C-terminal residue, Gln-270, forms hydrogen bonds with the backbone of Ile-94 and Gln-95, and its side chain interacts with Arg-256. C, modeled structure of GSH (orange) in ceCblC, which was generated by aligning the glutamate and glycine moieties in GSH with Gln-270 and tartrate, respectively. The distance between the sulfur atom of the modeled GSH and the methyl group of MeCbl is 3.1 Å. The predicted hydrogen bonds between GSH and Arg-187, Arg-232, and Arg-256 are shown. Selected residues are shown in a stick representation. D, due to the extended ␤-hairpin loop (pink) in ceCblC being in the closed conformation, Lys-170 is pushed in toward MeCbl and forms electrostatic interactions with a propionamide chain. The hydrophobic side chain of Ile-172 is directly under the cobalt. The position of GSH (orange) is modeled as in C. E, the propionamide side chains in hCblC (yellow) curve toward the corrin ring and do not interact with Arg-144. In contrast to ceCblC, the ␣-face of MeCbl is hydrophilic and solvent-accessible, as evidenced by the presence of a water molecule, bridging between a propionamide and Ser-146. The position of GSH (orange) in hCblC⅐MeCbl was predicted by overlaying the structure of hCblC⅐MeCbl (PDB entry 3SC0) and hCblC⅐AntiB 12 ⅐GSH (PDB entry 5UOS). 3 control of the thiol oxidase activity. Pathogenic mutations that corrupt the precise positioning of GSH in hCblC on the ␤-face and/or water coordination on the ␣-face can potentially unmask its latent thiol oxidase activity.

Expression and purification of ceCblC
ceCblC was expressed and purified as described previously (12) and was dialyzed into Buffer A containing 100 mM HEPES, pH 7.0, 150 mM KCl, and 10% glycerol. All assays were performed in Buffer A unless otherwise specified.

Synthesis of OH 2 Cbi
OH 2 Cbi was prepared by cereous hydroxide hydrolysis of OH 2 Cbl by modification of the published procedure (31). In brief, NaOH (110 mg, 2.75 mmol) was dissolved in 13 ml of H 2 O followed by the addition of cerium nitrate hexahydrate (585 mg, 1.35 mmol). A solution of OHCbl⅐HCl (50 mg, 36.1 mol in 2 ml of H 2 O was added to the suspension, and the reaction mixture was heated to 100°C with vigorous stirring. After 1.5 h, the reaction was cooled to room temperature, and the pH was adjusted to 8.5 with concentrated ammonium hydroxide. The mixture was centrifuged for 15 min at 2236 ϫ g. The superna-tant was decanted, and the residue was washed twice with 10 ml of water. The supernatants were pooled, and the solution was desalted using an RP-18 cartridge (Waters). The crude OH 2 Cbi product was purified using a carboxymethyl cellulose (2.5 ϫ 10-cm) column (Sigma-Aldrich). Unreacted OH 2 Cbl was removed by washing the column with 100 ml of 10 mM phosphate buffer, pH 8.5. OH 2 Cbi was eluted with 50 ml of 0.1 M NaCl in 10 mM phosphate buffer, pH 8.5, and desalted using an RP-18 cartridge. The cartridge was washed with 10 ml of water, 1 ml of 0.1 M NaCl followed by 10 ml of water. The product was eluted with 50% aqueous MeOH and lyophilized to obtain 9.9 mg of OH 2 Cbi⅐Cl (28% yield) as a red powder. Further purification was achieved by preparative HPLC on an RP-18 column

Synthesis of GSCbl and GSCbi
GSCbl was prepared as previously described (32). GSCbi was prepared in an analogous way as described for N-acetylcysteinylcobalamin (32). OH 2 Cbi⅐Cl (3.8 mg, 3.7 mol) was dissolved in 250 l of 100 mM MES buffer, pH 6.0, in a black 1.5-ml sample tube. The solution was cooled to 0°C, and GSH (65 l of a 100 mM stock solution in MES buffer, pH 6.0) was added dropwise. The reaction mixture was stirred for 1.5 h, after which time only one major peak was observed by analytical HPLC. The reaction mixture was then loaded onto an RP-18 cartridge (Waters). The column was washed sequentially with 10 ml of water, 1 ml of 0.1 M NaCl, and 5 ml of water, and the product was eluted with 10 ml of 50% aqueous MeOH. Following lyophilization, an orange product was obtained, which was further purified by preparative HPLC on an RP-18 column (Phenomenex Luna C-18 250 ϫ 10 mm) with the following solvent system: solvent A: 10 mM phosphate buffer, pH 6.

GSSG quantification by a coupled glutathione reductase assay
Dealkylation of ceCblC (10 -50 M)-bound AdoCbl (5-25 M) in the presence of 4 mM GSH was carried out in Buffer A under aerobic conditions. The reactions were stopped at the desired time points by precipitating the protein with an equal volume of metaphosphoric acid solution (16.8 mg/ml metaphosphoric acid, 2 mg/ml EDTA, and 9 mg/ml NaCl). The supernatant (40 -80 l) was mixed with 0.7 units/ml glutathione reductase (from bakers' yeast; Sigma-Aldrich) and 0.2 mM NADPH in 50 mM Tris-HCl, pH 7.5, to a total volume of 400 l and incubated for 20 min at room temperature to allow for the complete reduction of GSSG and stoichiometric oxidation of NADPH. The amount of GSSG produced in the dealkylation reaction was quantified spectrophotometrically using a ⌬⑀ 340 nm ϭ 6.2 mM Ϫ1 cm Ϫ1 for the oxidation of NADPH. Alternatively, GSSG production rates were measured in a continuous assay in reactions containing ceCblC (1-4 M)-cobalamin

Oxidation of ceCblC-cob(II)alamin by superoxide
Cob(II)alamin was generated by photolysis of 60 M AdoCbl in Buffer A and mixed with 100 M ceCblC under anaerobic conditions. Subsequently, ceCblC-cob(II)alamin was mixed with aerobic Buffer A in a 1:2 (v/v) ratio. A superoxide flux was generated by the addition of 1.4 mM xanthine and varying concentrations (0 -40 g/ml) of bovine milk xanthine oxidase (Sigma-Aldrich) to the ceCblC-cob(II)alamin mixture. The superoxide flux generated at a given concentration of xanthine oxidase was estimated by cytochrome c oxidation (⌬⑀ 550 nm ϭ 21 mM Ϫ1 cm Ϫ1 ) separately, as described previously (20). Oxidation of cob(II)alamin to OH 2 Cbl was monitored by UV-visible spectroscopy and quantified by the change in absorption at 525 nm (⌬⑀ 525 nm ϭ 5.5 mM Ϫ1 cm Ϫ1 ).

Estimation of the binding affinity of ceCblC for cobalamins
Cobalamins (20 M) and ceCblC (0 -200 M) were mixed in a total volume of 200 l of Buffer A. Unbound cobalamin was separated from protein by filtration using Nanosep centrifugal devices (cutoff molecular weight ϭ 10 kDa; Pall Life Sciences) at 4°C to estimate the free cofactor concentration. The ratio of bound cofactor to total cofactor was estimated using the following equation, where E 0 and L 0 are the total concentration of ceCblC and cofactor, respectively.
Binding of OH 2 Cbi was followed by the change in absorbance at 550 nm (⌬A 550 nm ), and the fractional change in absor-Thiol oxidase activity of ceCblC bance (⌬A 550 nm /⌬A 550 nm max ) was estimated using the following equation.

Deglutathionylation of GSCbl by ceCblC under anaerobic conditions
The reaction mixtures contained 50 M ceCblC and 30 M GSCbl in Buffer A. The reactions were initiated by the addition of 4 mM GSH and carried out at 20°C under anaerobic conditions. Formation of cob(I)alamin was followed for 5 s by the increase in absorbance at 389 nm (⌬⑀ ϭ 19 mM Ϫ1 cm Ϫ1 ) using stopped-flow spectrophotometry and fitted to the single exponential equation below to obtain k obs .

Deglutathionylation of GSCbl and dealkylation of MeCbl by ceCblC under aerobic conditions
The reaction mixtures contained 50 M ceCblC and 30 M GSCbl or MeCbl in Buffer A. The reactions were initiated by the addition of 4 mM GSH and carried out at 20°C under aerobic conditions. The reactions were followed for 28 s using stoppedflow spectrophotometry.

EPR Spectroscopy
One set of samples was anaerobic and contained 100 M ceCblC and 80 M cob(II)alamin Ϯ 4 mM GSH in Buffer A. A second set of samples was prepared to monitor the EPR spectrum of ceCblC-cob(II)alamin under aerobic conditions. For this, anaerobic ceCblC (300 M reconstituted with 240 M cob(II)alamin) was mixed with aerobic Buffer A (1:2, v/v) with or without 4 mM GSH and incubated on ice for 20 min before freezing in liquid nitrogen. A third set of samples was prepared to test whether cob(II)alamin is formed during redox cycling. For this, ceCblC (300 M), AdoCbl or CNCbl (200 M), and 4 mM GSH were mixed aerobically and incubated in the dark for 2-10 min at 15°C before freezing in liquid nitrogen. The settings for the EPR spectrometer were as follows: temperature, 70 K; microwave power, 20 milliwatts; microwave frequency, 9.32 GHz; receiver gain, 0.5 ϫ 10 5 ; modulation amplitude, 10.0 G; modulation frequency, 100 kHz. EPR spectral simulations were performed using Dr. Mark Nilges' SIMPOW program (33).

DFT calculations
Models of H 2 O-and Cys-cob(II)inamide were derived from the X-ray crystal structure of cob(II)alamin (34). In each case, the entire nucleotide loop and all other corrin ring substituents were replaced by hydrogen atoms separated by 1.08 Å from the neighboring carbon atoms. Full geometry optimizations were performed in ORCA version 3.0 (35), using DFT in conjunction with the BP86 functional (36,37). The SV(P) (Ahlrichs polarized split valence) basis (38) in conjunction with the SV/C auxiliary basis (39) were used for all atoms except for the cobalt ion and the atoms directly bound to it, for which the TZVP (Ahlrichs polarized triplevalence) basis (40) was used instead.
Molecular g values and 59 Co hyperfine coupling parameters were also calculated with ORCA version 3.0 using DFT by solving the coupled-perturbed self-consistent field (SCF) equations (41) and employing the B3LYP hybrid functional (42,43). The "core properties" with extended polarization (CP(PPP)) basis (44) and Kutzelnigg's NMR/EPR (IGLOIII) basis (45) were used to treat the cobalt ion and all ligating atoms (nitrogen and oxygen or sulfur), respectively, whereas the same basis sets described above were used for all other atoms. A high-resolution radial grid with an integration accuracy of 7 was used for the cobalt ion, and spin-orbit coupling contributions were included in the calculation of the 59 Co hyperfine tensor (46).

MCD spectroscopy
Samples containing 250 M ceCblC and 200 M cob(II)alamin with or without 4 mM GSH were prepared anaerobically in buffer containing 0.1 M HEPES, pH 7.0, 150 mM KCl, and 55% glycerol. A third sample was prepared by mixing 500 M ceCblC, 400 M AdoCbl, and 4 mM GSH aerobically in the same buffer and incubating in the dark for 2 min at room temperature before freezing in liquid nitrogen. MCD and low-temperature absorption spectra were collected on a Jasco J-715 spectropolarimeter in conjunction with an Oxford Instruments SM-4000 8T magnetocryostat. All MCD spectra presented herein were obtained by taking the difference between spectra collected with the magnetic field oriented parallel and antiparallel to the light propagation axis to remove contributions from the natural CD and glass strain.

Expression and purification of ceCblC for crystallization
A construct missing the last eight C-terminal residues (ceCblC(⌬C8)) and fused to a C-terminal cleavable His 6 tag was generated for crystallization. The deletion of the last 8 residues was needed to circumvent the insolubility of the full-length construct. ceCblC(⌬C8)-His 6 was expressed and purified as described previously (12). The His 6 tag was removed following purification on a nickel-nitrilotriacetic acid-agarose column using tobacco etch virus protease and dialyzed overnight against 1 liter of dialysis buffer containing 50 mM Tris-HCl, pH 8.0, 150 mM KCl, 1 mM DTT, and 10% glycerol at 4°C. The dialyzed protein was passed through a second nickel-nitrilotriacetic acid-agarose column (2.5 ϫ 2 cm), and unbound protein containing ceCblC(⌬C8) was collected, concentrated, and loaded on to a Superdex S200 column (120 ml) pre-equilibrated with buffer containing 50 mM Tris-HCl, pH 8.0. ceCblC(⌬C8) eluted as a monomer and was reconstituted with 1.2 eq of MeCbl in the presence of 0.5 mM tris(2-carboxyethyl)phosphine and was concentrated to at least 30 mg/ml for crystallization.

Data collection and structure determination
Diffraction data were collected at GM/CA-CAT 23-IDB (Advanced Photon Source, Argonne National Laboratory) on a Mar 300 detector and processed with XDS (47) to 1.35 Å resolution. Initial phases were obtained by the molecular replacement method with Phaser (48) using the apo-hCblC structure (PDB entry 3SBZ) as a search model (24). The resulting Phaser solution was used for refinement with PHENIX (49), including rigid body refinement of the individual domains followed by simulated annealing in torsional and Cartesian space, coordination minimization, and restrained individual B-factor adjustment with maximum-likelihood targets. Refmac5 (50) in the CCP4i suite (51) was subsequently employed for restrained refinement by using a mixture of anisotropic individual B-factors for protein and cofactor atoms and isotropic individual B-factors for water molecules with maximum-likelihood targets and also using the Babinet model for bulk solvent scaling. Refinement was followed by model building and modification with Coot (52). Several iterative rounds of refinement followed by model building/modifications were performed before the cobalamin cofactor was modeled in the electron density and further refined with Refmac5. All residues corresponding to the ceCblC(⌬C8) construct with the exception of residues 1 and 2 were traced and accounted for in the electron density of the final model, including residues (ENLYFQ) left from the cleavage of the C-terminal His tag. Crystallographic information as well as refinement statistics are provided in Table 3. The geometric quality of the model and its agreement with the structure factors were assessed with MolProbity (53). For ceCblC, MolProbity reported a clash and a molprobity score of 2.35 (99th percentile) and 1.01 (99th percentile), respectively, whereas 98.92% of the residues were in the favored Ramachandran plot regions with no residues in outlier regions. Crystal structure figures were generated with PyMOL (54).