Sacrificial Cobalt−Carbon Bond Homolysis in Coenzyme B₁₂ as a Cofactor Conservation Strategy

Abstract

A sophisticated intracellular trafficking pathway in humans is used to tailor vitamin B₁₂ into its active cofactor forms, and to deliver it to two known B_12-dependent enzymes. Herein, we report an unexpected strategy for cellular retention of B₁₂, an essential and reactive cofactor. If methylmalonyl-CoA mutase is unavailable to accept the coenzyme B₁₂ product of adenosyltransferase, the latter catalyzes homolytic scission of the cobalt−carbon bond in an unconventional reversal of the nucleophilic displacement reaction that was used to make it. The resulting homolysis product binds more tightly to adenosyltransferase than does coenzyme B₁₂, facilitating cofactor retention. We have trapped, and characterized spectroscopically, an intermediate in which the cobalt−carbon bond is weakened prior to being broken. The physiological relevance of this sacrificial catalytic activity for cofactor retention is supported by the significantly lower coenzyme B₁₂ concentration in patients with dysfunctional methylmalonyl-CoA mutase but normal adenosyltransferase activity.

The landmark accomplishment of elucidating the structure of vitamin B₁₂¹ was followed by clinical genetics studies on patients with inborn errors of B₁₂ metabolism, which hinted at a complex pathway for its intracellular trafficking.² These studies revealed that at least seven protein handlers assimilate and escort this reactive cofactor³ to the two enzymes that use it: methylmalonyl-CoA mutase (MCM)⁴ and methionine synthase (MS).⁵ MCM, a mitochondrial enzyme, uses 5′-deoxyadenosylcobalamin (AdoCbl or coenzyme B₁₂) and isomerizes (R)-methylmalonyl-CoA to succinyl-CoA. AdoCbl is synthesized by adenosyltransferase (ATR), which catalyzes the adenosylation of cob(I)alamin by ATP in a direct nucleophilic displacement reaction that leads to cobalt−carbon (Co−C) bond formation yielding AdoCbl and inorganic triphosphate (PPP_i)⁶ (Figure 1a). ATR also delivers the AdoCbl product directly to MCM, thereby averting cofactor loss by release into solution.⁶ Herein, we report the unusual chemical versatility of ATR, which catalyzes homolytic cleavage of the newly formed Co−C bond of AdoCbl in the presence of PPP_i if MCM is unavailable as an acceptor. Unlike AdoCbl, the resulting cob(II)alamin product is tightly bound to human ATR, reducing loss of the rare but essential cofactor into solution. Patients with mutations in MCM but with normal ATR exhibit significantly reduced AdoCbl levels, supporting the relevance of this cofactor sequestration mechanism in a cellular milieu.

Figure 1.

PPP_i induces Co−C homolysis in ATR-bound AdoCbl. (a) Scheme showing the role of ATR (green) in AdoCbl synthesis and transfer to MCM. (b) Transfer of AdoCbl from ATR (black) to MCM (red). (c) Addition of PPP_i to AdoCbl bound to ATR (black) results in the formation of intermediate X (red) under anaerobic conditions. Inset shows the spectrum of base-off NpCbl. (d) Under aerobic conditions, X (red) converts to 4C cob(II)alamin (black). (e) Scheme showing conversion of ATR-bound AdoCbl to X and to 4C cob(II)alamin and 5′-hydroperoxyadenosine.

ATR is a homotrimer that binds AdoCbl in a five-coordinate (5C) “base-off” state⁷ in which the lower axial ligand, dimethylbenzimidazole (DMB), is displaced from cobalt. In contrast, MCM, which also binds AdoCbl in a base-off state, fills the lower axial position with a histidine residue.⁸ The change from 5- to 6C AdoCbl is accompanied by large spectral changes, which allow facile monitoring of the cofactor’s movement between the active sites. Complete cofactor transfer from ATR·AdoCbl to MCM is observed upon mixing the two proteins as indicated by a shift in the absorption maximum from 455 to 530 nm (Figure 1b). Transfer of AdoCbl from ATR (K_d = 0.96 ± 0.31 μM) to MCM (K_d = 0.27 ± 0.11 μM) is thermodynamically favored (Table S1). The reverse transfer was not observed even in the presence of excess ATR (Figure S1), which contrasts with the bidirectional transfer of AdoCbl observed with the homologous Methylobacterium extorquens proteins.⁶

The fate of the newly synthesized and precious AdoCbl product was examined in the absence of MCM. Addition of PPP_i to human ATR·AdoCbl to mimic the initially formed ternary product complex, elicited an unexpected spectral change (Figure 1c) forming intermediate “X” with absorption maxima at 388 and 439 nm. Whereas X was stable under anaerobic conditions, it converted under aerobic conditions to a new species with a 465 nm absorption maximum (Figure 1d), which is spectroscopically identical to 4C cob(II)alamin bound to ATR·ATP (Figure S2). We hypothesized that PPP_i binding to ATR·AdoCbl weakened the Co−C bond forming X, which then underwent homolytic bond cleavage yielding cob(II)-alamin and Ado (Figure 1e). Whereas geminate recombination of the radical pair predominated under anaerobic conditions, irreversible interception of Ado by O₂ shifted the equilibrium away from X to cob(II)alamin.

To test this hypothesis, we characterized the products obtained upon addition of PPP_i to ATR·AdoCbl. The spectrum of X is very similar to that of base-off neopentylcobalamin (NpCbl) (Figure 1c, inset). Steric strain from the bulky neopentyl group weakens the Co−C bond and leads to a 10⁶-fold enhancement of the homolysis rate with NpCbl compared to AdoCbl.^9,10 Although the precise nature of X is unclear, the spectral similarity between it and NpCbl suggests stabilization of an intermediate in which the Co−C bond is weakened prior to being homolytically cleaved. Evidence that X is diamagnetic and that 4C cob(II)alamin is formed under aerobic conditions was obtained by EPR spectroscopy. The large g spread and substantial Co hyperfine splittings are diagnostic of a square planar Co(II) coordination environment (Figures 2a and S3). Consistent with these results, the corresponding MCD spectrum showed an intense positively signed feature at 12 500 cm⁻¹ that is characteristic of 4C cob(II)alamin. This feature is considerably red-shifted from its counterpart in cob(II)inamide, which has an axial water ligand in solution (Figure 2b and S4).¹¹ The fate of Ado· was investigated by NMR spectroscopy using [U−¹³C-Ado]-Cbl. The ¹H and ¹³C chemical shifts of the product (Figure S5 and Tables S2 and S3) were nearly identical to those reported previously for 5′-hydroperoxyadenosine formed during aerobic photolysis of AdoCbl.¹² However, while 5′-hydroperoxyadenosine arises in aqueous solution from hydrolysis of a 5′-peroxy-AdoCbl intermediate and yields cob(III)alamin, 4C cob(II)alamin is produced in the ATR active site. Hence, 5′-hydroperoxyadenosine formed in the ATR reaction likely results from quenching of the initially formed alkylperoxy (Ado-OO·) radical. The source of the hydrogen atom is not known.

Figure 2.

Spectroscopic and structural characterization of B₁₂ bound to ATR. (a) EPR spectrum of 4C cob(II)alamin formed after addition of PPP_i under aerobic conditions (black). The red line corresponds to the simulated spectrum. (b) MCD spectra of 4C cob(II)alamin (solid line) generated under the same conditions as (a) and of 5C cob(II)inamide (dashed lines) as a reference. (c) Structure of human ATR showing two subunits occupied by AdoCbl and a third by ATP.(d) Superposition of the adenosine rings in ATP (cyan, PDB: 21DX) and AdoCbl (yellow, PDB: 6D5K, this study) bound to ATR. (e) Close-up of the ATR·AdoCbl·PPP_i active site structure (PDB: 6D5X, this study). (f) Close-up of human ATR showing that the DMB tail, tucked below the corrin ring, is kept from coordinating to the cobalt by Phe170 on a hydrophobic loop (pink). Adjacent subunits that form the active site are in blue and green. The gray spheres in panels d and e represent magnesium.

We solved the structure of human ATR·ATP crystals soaked with AdoCbl (Table S4). Electron density for AdoCbl was observed in two active sites (with occupancies of 1.00 and 0.86) and for ATP in the third site (Figures 2c and S6). The adenosine moieties in the sites containing ATP¹³ and AdoCbl are virtually superimposable (Figure 2d), held via electrostatic and hydrophobic interactions. Fortuitously, we also obtained a structure with AdoCbl and PPP_i bound (Figure 2e), which we ascribe to contaminating PPP_i in the ATP sample and to its high affinity for ATR·AdoCbl (Table S1). The structures in the absence and presence of ATP are very similar, and at the current resolution, do not provide insights into how PPP_i labilizes the Co−C bond. AdoCbl is bound in a base-off state with Phe170 on a hydrophobic loop wedging between the DMB and the corrin ring, precluding water coordination (Figure 2f). The DMB tail in this structure is fully resolved in contrast to the Lactobacillus reuteri ATR structure where it is disordered.¹⁴ While AdoCbl transfer to MCM (k = 0.63 ± 0.06 min⁻¹) and sacrificial formation of 5′-hydroperoxyadenosine and cob(II)alamin (k = 0.082 ± 0.01 min⁻¹) both proceed slowly, the former is favored ~8-fold over the latter. We note that CblC, a B₁₂ chaperone with decyanase and dealkylase activities, also exhibits low reaction rates (~0.003−0.2 min⁻¹) that nevertheless support cellular B processing.^15,16 Apparently, the low reaction rates are sufficient to support low flux through the B₁₂ trafficking pathway in which fidelity is prioritized over speed.¹⁷

What could be the physiological significance of AdoCbl synthesis by ATR at the expense of three ATP equivalents only to be followed by its cleavage in the absence of MCM? The kinetic choices facing the newly formed ternary ATR·AdoCbl·PPP_i complex in the absence of AdoCbl transfer to MCM are to form cob(II)alamin, as a strategy to sequester the cofactor or to release AdoCbl into solution. From a thermodynamic standpoint, binding of ATP and cob(II)alamin (K_d = 0.08 ± 0.01 μM) to ATR is favored over AdoCbl (K_d = 0.96 ± 0.31 μM) (Table S1). Our model predicts that AdoCbl levels should be reduced in MCM deficiency even though AdoCbl synthesis by ATR is unimpaired. To test this model, we examined AdoCbl levels in normal versus patient fibroblasts. In control fibroblasts (n = 192), 15.1 ± 3.4% of the intracellular B₁₂ pool is AdoCbl (Figure 3a); the majority exists as methylcobalamin. As expected for fibroblasts with ATR deficiency (due to mutations in the cblB locus), the AdoCbl pool size is reduced (3.0 ± 2.1%, n = 45, p < 0.0005). Consistent with our model, the AdoCbl pool is smaller in patients with MCM (mut) deficiency, which is differentiated into groups containing either low (8.3 ± 6.9% in mut⁻, n = 53, p < 0.0005) or undetectable (8.1 ± 4.4% in mut⁰, n = 195, p <0.0005) mutase activity.

Figure 3.

Impact of ATR deficiency on AdoCbl synthesis. (a) AdoCbl as a percent of total cobalamin in fibroblasts from ATR (cblB) or MCM (mut) deficiency versus control individuals. The mut⁻ and mut⁰ subgroups represent fibroblasts with diminished or undetectable MCM activity, respectively. (b) Comparison of the EPR spectra of cob(II)alamin bound to R186Q versus wild-type (WT) ATR. The presence of the triplet superhyperfine splittings in the high-field lines of the R186Q spectrum is caused by the presence of axial nitrogen coordination (absent in WT). (c) Absorbance spectrum of 6C AdoCbl bound to R186Q ATR (black). Addition of PPP_i under anaerobic conditions gives rise to a spectrum (red) that resembles “base-on” NpCbl (inset). (d) Intermediate X partitions between transfer to MCM, release into solution or undergoes sacrificial Co−C bond homolysis leading to retention of tightly bound cob(II)alamin.

Because of its low intracellular concentration, it is unlikely that free PPP_i binds to the binary ATR·AdoCbl complex (K_d =0.42 ± 0.08 μM, Figure S7) to trigger Co−C bond homolysis. Instead, we predict that the fate of AdoCbl is dictated by the kinetics of its transfer to MCM from the newly formed product complex versus Co−C bond homolysis. This model is indirectly supported by the properties of the R186Q patient mutation.¹⁸ The in vitro activity of R186Q (11 ± 3.6 min⁻¹) and wild-type (14.8 ± 3.5 min⁻¹) ATR are comparable. Arg186 forms an intersubunit salt bridge with Asp90 only when B₁₂ is bound (Figure 2d). The mutant protein binds B₁₂ in the base-on state as monitored by EPR and absorption spectroscopy (Figure 3b,c), revealing the tenuous hold of ATR on the base-off state. The homologous mutation in L. reuteri ATR also led to bound base-on B₁₂.¹⁹ The R186Q mutation destabilizes the ternary ATR·AdoCbl·PPP_i complex but does not affect ATP binding (Figure S8 and Tables S1 and S5). Furthermore, while the affinity for cob(II)alamin is unaffected (≤0.2 μM), the affinity for AdoCbl is significantly diminished. The product of Co−C bond homolysis in the R186Q mutant resembles base-on 6C Np-Cbl (Figure 2c, inset) but is only seen when ATR active sites are present in vast excess over AdoCbl. Hence, the inability to sequester AdoCbl via direct transfer to MCM predisposes the R186Q mutant to cofactor release into solution (Figure 3d). We speculate that the twin challenges of reducing base-on cob(II)alamin (which has a lower redox potential than the base-off form²⁰), and weakened affinity for AdoCbl, contribute to methylmalonic aciduria in patients with the biallelic R186Q mutation. We recently reported that itaconyl-CoA inactivates MCM by rapid cob(II)alamin accumulation on the enzyme.²¹ Human MCM binds cob(II)alamin with a lower affinity than AdoCbl (Table S1), which begs the question as to how cob(II)alamin loss is averted from MCM into solution.

A cytoplasmic strategy for cofactor sequestration is exemplified by MS, whose synthesis is translationally up-regulated by B₁₂.²² Prioritization of the B₁₂ supply for MS, an essential enzyme, over MCM, which serves an anaplerotic function, is suggested by the prevalence of holo-MS (~100%)²³ versus fractional B₁₂ saturation of MCM (~25%)²⁴ in liver. Furthermore, under conditions of B₁₂ deficiency, the magnitude of cofactor loss from MCM is greater than from MS.²⁴ We now report a cellular strategy for B₁₂ retention that exploits an unprecedented chemical versatility whereby AdoCbl is synthesized via nucleophilic chemistry but cleaved via radical chemistry within the same active site. The mechanism for mitochondrial import of B₁₂ is unknown and the possibility that ATR-induced Co−C bond homolysis serves to return cob(II)alamin to the cytoplasm remains open.