Elucidation of the anaerobic pathway for the corrin component of cobalamin (vitamin B₁₂)

Abstract

It has been known for the past 20 years that two pathways exist in nature for the de novo biosynthesis of the coenzyme form of vitamin B₁₂, adenosylcobalamin, representing aerobic and anaerobic routes. In contrast to the aerobic pathway, the anaerobic route has remained enigmatic because many of its intermediates have proven technically challenging to isolate, because of their inherent instability. However, by studying the anaerobic cobalamin biosynthetic pathway in Bacillus megaterium and using homologously overproduced enzymes, it has been possible to isolate all of the intermediates between uroporphyrinogen III and cobyrinic acid. Consequently, it has been possible to detail the activities of purified cobinamide biosynthesis (Cbi) proteins CbiF, CbiG, CbiD, CbiJ, CbiET, and CbiC, as well as show the direct in vitro conversion of 5-aminolevulinic acid into cobyrinic acid using a mixture of 14 purified enzymes. This approach has resulted in the isolation of the long sought intermediates, cobalt-precorrin-6A and -6B and cobalt-precorrin-8. EPR, in particular, has proven an effective technique in following these transformations with the cobalt(II) paramagnetic electron in the d_yz orbital, rather than the typical d_z2. This result has allowed us to speculate that the metal ion plays an unexpected role in assisting the interconversion of pathway intermediates. By determining a function for all of the pathway enzymes, we complete the tool set for cobalamin biosynthesis and pave the way for not only enhancing cobalamin production, but also design of cobalamin derivatives through their combinatorial use and modification.

Ever since the pioneering work of Dorothy Hodgkin revealed the complexity of vitamin B₁₂ (cyanocobalamin) (1), chemists and biochemists have endeavored to unravel the step-by-step biosynthesis of this essential cofactor and coenzyme. The two major biologically active forms of vitamin B₁₂ are methylcobalamin and adenosylcobalamin, which are composed of a highly modified cobalt-containing tetrapyrrole linked to a nucleotide loop that houses an unusual base called dimethylbenzimidazole. This base acts as a lower ligand to the cobalt ion, whereas the upper cobalt ligand is provided by either the methyl or 5′-deoxyadenosyl group (Fig. 1). The cobalt-containing modified tetrapyrrole component of the vitamin is referred to as the corrin ring and differs in size from other modified tetrapyrroles, because it has undergone contraction as part of the biosynthetic process.

Fig. 1.

Summary of the aerobic and anaerobic cobalamin biosynthetic pathways from precorrin-2 to cobyrinic acid a,c-diamide. The fully characterized aerobic pathway (blue) is compared with the anaerobic pathway (red), and many of the steps are predicted from sequence homology to the aerobic pathway (30–32). These unknown steps are referred to as the black box.

Research into the biosynthesis of the corrin ring has revealed the presence of two similar, although distinct, biochemical pathways, which differ on the timing of cobalt insertion and a requirement for molecular oxygen (2). The cobalt-late or aerobic pathway is the better characterized, and all of the intermediates and enzymes responsible for the de novo construction of the corrin component have now been determined (3, 4). In contrast, the cobalt-early or anaerobic pathway is poorly delineated. Many of the pathway intermediates remain hypothetical and are based on the order of events that have been determined in the aerobic pathway (4), because both pathways use a number of similar enzymes (Fig. 1) (5, 6).

Corrin ring synthesis is initiated from uroporphyrinogen III, the first macrocyclic intermediate of the tetrapyrrole pathway (7). The corrin intermediates are then defined by the number of S-adenosyl-l-methionine (SAM) -derived methyl groups that have been added to the framework (8). For instance, precorrin-2 is so called, because two methyl groups have been added to uroporphyrinogen III by CobA. Oxidized versions of these intermediates are called factors, where factor II is the oxidized version of precorrin-2. In Bacillus megaterium, which operates an anaerobic cobalamin biosynthetic pathway (6), corrin ring synthesis progresses by insertion of cobalt into factor II (9–11). This cobalt complex is further methylated at C-20 to give cobalt-factor III by CbiL, a member of the canonical methyltransferases associated with cobalamin synthesis (12). The next reaction is facilitated by CbiH₆₀ to give cobalt-precorrin-4 in a reaction that requires reducing equivalents to ensure a high yield of product (13).

Cobalt-precorrin-4 has previously been shown to be a pathway intermediate, because its incubation in a cell-free lysate of Propionibacterium shermanii resulted in its transformation into cobyrinic acid, albeit in a very low yield (∼1%) (14). It has also been shown that cobalt-precorrin-4 can be converted into cobalt-precorrin-5A by Escherichia coli extracts that contain overproduced Salmonella enterica CbiF in the presence of SAM in a yield of ∼5–10% (15). Similarly, cobalt-precorrin-4 can be transformed into cobalt-precorrin-5B by incubation with cell lysates containing CbiF and CbiG, indicating that CbiG catalyses the opening of the δ-lactone ring and the subsequent deacylation (15) to release the C2 fragment as acetalydehyde (16). However, the poor conversion rates of these reactions have prevented a detailed study of the mechanisms underpinning these transformations, and all of the intermediates have been characterized only as their derivatized methylester form rather than as free acids.

The lack of any isolated intermediates between cobalt-precorrin-5B and cobyrinic acid means that a black box exists for the latter portion of the anaerobic corrin synthesis pathway (Fig. 1). From cobalt-precorrin-5B, the C-1 position is the next expected methylation site (8). In the aerobic pathway, this step is catalyzed by a class III SAM-dependent methyltransferase, CobF. However, no equivalent enzyme is found in the anaerobic pathway. Evidence that CbiD catalyzes the C-1 methylation has been implied from indirect genetic engineering studies on whole cells (17), although this result has not been reproduced with purified enzymes; also, the expected product cobalt-precorrin-6A has not been isolated. A structure (Archaeoglobus fulgidus; Protein Data Bank ID code 1SR8) of CbiD was released, but it shares no similarity to any characterized enzyme, and the protein does not contain a Rossman fold, which is considered an essential structural feature for SAM binding. For the rest of the pathway, the intermediates and reactions have yet to be isolated. Indeed, the enzymes CbiJ, CbiE, CbiT, and CbiC share reasonable homology (∼20–40% amino acid identity) to enzymes from the aerobic pathway that have been characterized and shown to convert precorrin-6 to hydrogenobyrinic acid (3, 4). A summary of the aerobic and anaerobic pathways is shown in Fig. 1.

In this report, we describe the complete elucidation of the elusive anaerobic vitamin B₁₂ pathway, one of the last uncompleted chapters in tetrapyrrole biosynthesis. With respect to the chemical logic of this biosynthesis, the exposition of the anaerobic pathway has provided insights into the chemistry and evolution of this remarkable process, which represents one of the longest biosynthetic pathways found in nature.

Results

Homologous Production of Cobalamin Enzymes in B. megaterium.

The cobalamin biosynthetic enzymes were produced homologously in B. megaterium by cloning the relevant genes (Table S1 shows plasmids and primers) with a fusion to a His₆-tag within plasmid pC-His1622 or pN-His-TEV1622 (18). These genes included (in order of predicted enzymatic reactions) cbiF, cbiG, cbiD, cbiJ, cbiET, and cbiC. Recombinant strains were obtained by protoplast transformation of B. megaterium DSM319 with a modified minimal media method (SI Text). After growth and overproduction by xylose induction, each protein was individually purified to homogeneity (Fig. S1 and Table S2). This approach allowed significant quantities of soluble and relatively pure protein to be isolated. Furthermore, all these enzymes were stable in the standard buffers described (20 mM Tris⋅HCl, Hepes, pH 7–8, 100 mM NaCl), with the exception of CbiF and CbiH₆₀ (13), both of which required a high salt concentration (400 mM) to maintain solubility at 5–10 mg mL⁻¹. The aggregation state of the various proteins was estimated by size exclusion chromatography (Table S2).

Synthesis of Cobalt-Precorrin-5A.

Cobalt-precorrin-5A and -5B had not been characterized in their free acid forms, because the yields of these intermediates were very low when previously detected (5–10%). The activity of purified B. megaterium CbiF with cobalt-precorrin-4 was, therefore, monitored to try and improve the conversion into cobalt-precorrin-5A. Incubation of CbiF with SAM and cobalt-precorrin-4 leads to gradual changes in the color of the substrate, with a change from pale red to red-orange. Nonetheless, despite increasing the enzyme concentration, conversion into this species was not improved. Cobalt-precorrin-4 can be separated from this species on a DEAE–Sephadex column. The UV-visible spectrum of this product has a broad absorbance peak at 340 nm and a minor peak at 440 nm (Fig. 2). Liquid chomatography (LC)-MS showed that the substrate cobalt-precorrin-4 is detected at m/z 950, whereas the product of the CbiF reaction is detected with an m/z 964 (Fig. S2). This result shows an increase of 14 Da consistent with the addition of a methyl group and a molecular formula (C₄₅H₅₃CoN₄O₁₆) in agreement with cobalt-precorrin-5A.

Fig. 2.

Intermediates of the anaerobic biosynthesis of cobalamin. Panels show the UV-Vis spectrum, MS, and visual appearance of each intermediate between cobalt-factor II and cobyrinic acid. Intermediates are Cobalt-Precorrin (CoP) and Cobalt-Factor (CoF). MS data are shown in Fig. S2. Exp, expected; obs, observed.

Synthesis of Cobalt-Precorrin-5B.

Previously, CbiG had been shown to open the δ-lactone on ring A and release the methylated C-20 carbon as a C2 fragment in the form of acetaldehyde (16), generating the intermediate cobalt-precorrin-5B (15). This reaction is a prerequisite to the C-1 methylation step (8). Using cobalt-precorrin-5A, the activity of CbiG was recorded by the following changes in the UV-Vis spectrum of the substrate. The visual appearance changed from orange-red to a bright orange-colored solution. During the incubation, the Soret peak at 337 nm shifts to give a clear peak at 355 nm, with a slight shoulder at 388 nm (Fig. 2). In addition, the minor absorbance peak at 440 nm is replaced with double peaks at 469 and 500 nm. However, this intermediate is extremely sensitive to oxygen and when exposed to air, undergoes an instantaneous change to a pale brown pigment. Because of this rapid oxidation, LC-MS could not be used to determine the true mass of this intermediate, with only degraded fragments at m/z 920 (Fig. S2) and m/z 904 detected. The expected mass for cobalt-precorrin-5B is 937 Da. However, if kept under anaerobic conditions (<2 ppm O₂), the orange-colored product of the CbiG reaction, cobalt-precorrin-5B, remains stable for several months.

Enhanced Synthesis of Cobalt-Precorrin-5B from Cobalt-Factor III.

The activity of CbiF was found to be very poor on direct incubation with cobalt-precorrin-4 and SAM, with yields typically in the 10–20% range. Interestingly, higher yields could be achieved by combining the enzymes CbiH₆₀, CbiF, and CbiG with the cofactors SAM and dithionite, and the substrate cobalt-factor III in an enzyme-mixture incubation (SI Text and Fig. S3A). Typically, this approach achieves a repeatable >80% transformation of cobalt-factor III into cobalt-precorrin-5B.

Synthesis of Cobalt-Precorrin-6A.

Based on pulse-labeling experiments (8), it was known that the C-1 methylation step is the next reaction in the pathway, with CbiD most likely to catalyze this methylation (17). To show this activity conclusively, B. megaterium CbiD was incubated with cobalt-precorrin-5B and SAM, which resulted in the UV-Vis spectrum of the incubation becoming blue-shifted during the reaction (Fig. 2 and Fig. S3B). The UV-Vis spectral changes are accompanied with a subtle color change from orange to a red-bronze pigmentation. This intermediate could be detected by LC-MS (Fig. S4) with a dominate mass at m/z 951 (Fig. S2), which represents an increase of 14 Da. It is consistent with the addition of a methyl group, and it is in agreement with the predicted molecular formula (C₄₄H₅₀CoN₄O₁₆) for cobalt-precorrin-6A.

EPR has previously been used for the characterization of the early steps of the anaerobic B₁₂ pathway, including cobalt chelation into sirohydrochlorin by CbiX^L (9), C-20 methylation by CbiL (12, 19), and ring contraction reaction by CbiH₆₀ (13). For EPR, we shall now refer to each intermediate as cobalt(II)-precorrin-n, because all the pathway intermediates that we have analyzed are in the detectable paramagnetic cobalt(II) form. Thus, we do not lose or gain cobalt(II) signal in any of the reactions, indicating that cobalt(II) is not oxidized or reduced under the conditions tested.

The previous EPR studies mentioned above presented spectra of low-spin cobalt(II) ions having one unpaired electron in the d_z2 orbital. Such spectra exhibit axial or nearly axial anisotropy, having g_|| ∼ 2 and g_┴ = 2.2–3.0, and a relatively large anisotropic cobalt hyperfine interaction, giving A_||^Co ∼ 100 G. Most low-spin cobalt(II) compounds, including derivatives of coenzyme B₁₂, give rise to EPR spectra of this class. Unexpectedly, cobalt(II)-precorrin-5B and -6A do not match this class of EPR spectra and instead, most closely resemble EPR spectra of a nonbiological complex in an N₄ macrocycle (20–23). Unusually, the cobalt(II) unpaired electron resides in the ground state d_yz orbital (Fig. S5), with electron donation from two nitrogen ligands. In contrast, for most B₁₂ pathway intermediates, the four nitrogen atoms liganded to the cobalt(II) share one negative charge. The EPR spectra of these d_yz compounds are normally described as rhombic and characterized by three g values attributed as g_x, g_z, and g_y (although the difference between g_z and g_y may be small), with the z direction being perpendicular to the ligand plane, whereas x and y are mutually orthogonal within that plane. Although g_x is always the largest (lowest field) g value in such compounds, the order in which g_y and g_z occur depends on the compound in question; therefore, we shall simply refer to g tensors as g₁, g₂, and g₃ beginning at the lowest field. The theory underpinning the ground state electronic configurations of low-spin cobalt(II) compounds and the nature of the singly occupied molecular orbital have been presented (24, 25).

Cobalt(II)-precorrin-5B displays three g values (g₁ = 2.72, g₂ = 2.06, and g₃ = 1.96) and resolved hyperfine coupling at |A₃| = 31.8 G (Fig. 3 and Table S3). Cobalt(II)-precorrin-5A gives rise to a very similar spectrum (parameters are shown in Table S3); however, these data cannot be obtained without a contaminating cobalt(II)-precorrin-4 spectrum that must be removed computationally because of the failure of the CbiF-catalyzed methylation to reach completion (see above). Addition of SAM does not lead to changes in the EPR spectrum of cobalt(II)-precorrin-5B (Fig. S6). Furthermore, the EPR spectrum of a negative-control reaction with CbiD (no SAM) does not reveal any changes in the spectrum over a time period (Fig. S6). For the complete reaction (with SAM), gradual changes in the EPR spectra resulted over the time course (Fig. 3A). We attribute this change to the formation of cobalt(II)-precorrin-6A. The EPR spectrum of cobalt(II)-precorrin-6A retains similar features with g₁ = 2.63, g₂ = 2.08, g₃ = 1.98, and |A₃| = 31.5 G. Most notably, the line shape and intensity of the hyperfine splittings are enhanced. Because the d_yz orbital is retained after C-1 methylation by CbiD, the changes in line shape and enhanced splitting suggest a change in conjugation in the macrocycle and enhanced conformational stability. One plausible explanation is methylation at C-1 and consequent prototropic rearrangement of the double bond at C-1/C-19 to the C-19/C-18 position.

Fig. 3.

EPR of some of the anaerobic pathway steps. (A) Reaction of 50 µM CbiD, 200 µM cobalt(II)-precorrin-5B, and 2 mM SAM incubated at 37 °C for 24 h. Samples were removed at t = 0, 1, 4, and 24 h. (B) Reaction of 20 µM CbiET, 200 µM cobalt(II)-precorrin-6B, and 2 mM SAM at 37 °C. Samples were removed at t = 0, 2, 4, and 24 h. (C) Reaction of 10 µM CbiC and 200 µM cobalt(II)-precorrin-8 incubated at 37 °C for 24 h. Samples were removed at t = 0, 4, and 24 h. g Values and coupling constants of hyperfine splittings are shown here and in Table S3. Spectra were recorded at 20 K using a microwave power of 1 mW, a modulation frequency of 100 kHz, and a modulation amplitude of 5 G. Negative control spectra are shown in Fig. S6.

Synthesis of Cobalt-Precorrin-6B.

In the aerobic pathway, CobK drives the NADPH-dependent reduction of the C-18/C-19 double bond to convert precorrin-6A into precorrin-6B. Because of sequence identity with CobK (∼20–30%), it was assumed that CbiJ catalyzes the equivalent reaction by converting cobalt-precorrin-6A into cobalt-precorrin-6B. When cobalt-precorrin-6A was incubated with CbiJ and NADPH, there was no change observed in the UV-Vis spectrum of the substrate, suggesting that no reaction had taken place. However, when NADPH was replaced with NADH in the incubation, a change in the absorbance spectrum of the incubation was observed, with the peaks at 335 and 436 nm shifting to 318 and 419 nm, respectively (Fig. 2 and Fig. S3C). These changes were not observed in any of the controls (NADH, NADPH, or enzyme only), indicating that the reaction is dependent on the presence of CbiJ and NADH. The rate of the CbiJ-catalyzed reaction was very fast (seconds), similar to the level of activity observed with CbiG. The NADH-dependent reduction of cobalt-precorrin-6A into cobalt-precorrin-6B is expected to occur at the C-18/C-19 double bond, with a gain of two mass units. We attempted to show this change by MS but found that cobalt-precorrin-6B was unstable. Lowering the pH below 6.5 results in an instant bleaching (dark to pale red) of the color for both cobalt-precorrin-6A and cobalt-precorrin-6B. This change leads to a UV-Vis spectrum that is identical for both intermediates. Both cobalt-precorrin-6A and the expected product of the CbiJ reaction, cobalt-precorrin-6B, are detected with an identical retention time and mass at m/z 951 (Fig. S2). We were unable to solve the structure of the intermediate by NMR, possibly because of the presence of a number of tautomeric species (SI Text and Fig. S7).

Synthesis of Cobalt-Precorrin-8.

In some organisms operating the anaerobic pathway, the transformation of cobalt-precorrin-6B into cobalt-precorrin-8 is mediated by two enzymes CbiE and CbiT. It is thought that CbiE is responsible for the methylation at C-5, whereas CbiT couples the decarboxylation of the ring C acetate side chain with methylation at C-15. In the aerobic pathway, these two enzymes are always found fused together in a protein called CobL, and a similar situation occurs with the anaerobic pathway in B. megaterium with the presence of CbiET. The activity of CbiET was initially monitored by UV-Vis spectroscopy, where the spectrum of cobalt-precorrin-6B was observed to bleach, concomitant with a color change from dark red to pale yellow (Fig. 2 and Fig. S3D). LC-MS showed that cobalt-precorrin-6B (m/z 951) is converted into major and minor products at m/z 938 and m/z 936 (Fig. S4), respectively. The UV-Vis spectrum of the major peak at m/z 938 matches the spectrum described above, whereas its mass is consistent for the expected product cobalt-precorrin-8 (C₄₅H₅₉CoN₄O₁₄).

Major changes were detected in the EPR spectra during the conversion of cobalt(II)-precorrin-6B to cobalt(II)-precorrin-8 (Fig. 3B). Over the time period of the reaction, the three g values (g₁, g₂, and g₃) of cobalt(II)-precorrin-6B are replaced by three new g values: g₁ = 2.43, g₂ = 2.24, and g₃ = 2.006. Compared with cobalt-precorrin-6B, the field range narrows between g₁ and g₃ for cobalt-precorrin-8, whereas g₂ and g₃ separate, and the cobalt hyperfine splittings increase from 30.3 to 87 G. This observation suggests a change in electronic structure, in which the unpaired electron is once again mainly in the d_z2 orbital. In addition, a superhyperfine interaction with a nitrogen (¹⁴N) ligand leading to triplet formation is observed in this region, which may be caused by a contaminating N ligand, possibly from the buffer. Such a superhyperfine coupling would not be observed if the unpaired electron was located in the d_yz orbital.

Synthesis of Cobyrinic Acid.

To complete the pathway to cobyrinic acid, the final reaction in the series is the methylmutase CbiC. In the aerobic pathway, the equivalent enzyme, CobH, oversees an unusual 1,5-sigmatropic rearrangement, where the C-11 methyl group migrates to the C-12 position, creating a quaternary carbon center. Cobalt-precorrin-8 is a pale yellow intermediate. However, on incubation with CbiC, a slow but progressive change occurs as the intermediate changes to a golden orange-colored intermediate. Incubation of 5 µM CbiC and 45 µM cobalt-precorrin-8 at 37 °C for 6 h yielded a UV-Vis spectrum that resembles a typical cobalt B₁₂-like spectrum with a Soret peak at 314 nm, whereas an additional shoulder at 465 nm is also observed (Fig. 2 and Fig. S3E). The mass of this intermediate was detected at m/z 938 (Fig. S2) and as expected, is the same as the substrate, cobalt-precorrin-8. After enzyme incubation, cob(II)yrinic acid was purified to homogeneity and then oxidized to cob(III)yrinic acid by addition of potassium cyanide. Cyanide is known to coordinate B₁₂, forming a hexacoordinated cobalt(III) ion, with cyanide bound as both upper (β) and lower (α) axial ligands. This change shifted the Soret peak to 365 nm, whereas typical B₁₂-like α- and β-bands were observed at 502, 537, and 578 nm. This intermediate eluted as two separate peaks on LC-MS, although both had identical masses at m/z 964. This result is consistent for monocyanocobyrinic acid. A mass for dicyanocobyrinic acid (exp m/z 990) was not detected; however, a standard of dicyanocobinamide is also found to elute as two peaks, both as the mono form. The two peaks on LC-MS may be caused by the pentacoordinated α- or β-cyano ligand forms.

As with the UV-Vis spectroscopy, EPR provided additional evidence for our findings, because the EPR spectrum of cobalt-precorrin-8 is converted into a new B₁₂-like spectrum, which has previously been described for base-on cobinamide (26). The product, cobyrinic acid, has g values at g_┴ = 2.23, g_|| = 2.000, and A_||^Co = 110 G (Fig. 3C). These data are in agreement with published spectra of cob(II)inamides and cob(II)alamins, which like cob(II)yrinic acid, share the same UV-Vis and EPR spectra (27). The EPR spectrum of cob(II)yrinic acid seems base-on, because the superhyperfine coupling constant is 18 G. This result may be attributed to the Tris buffer or contaminating imidazole [20 G for cob(II)inamide in MeOH].

To confirm the structure of cobyrinic acid, a pure sample was synthesized directly from the starting metabolite of tetrapyrrole biosynthesis, 5-aminolevulinic acid, in a multistep mixture incubation using 14 purified enzymes HemB, HemC, HemD, CobA, SirC, CbiX, CbiL, CbiH₆₀, CbiF, CbiG, CbiD, CbiJ, CbiET, and CbiC. After purification to homogeneity by anion exchange and reverse-phase chromatography, 1.8 mg pure cob(II)yrinic acid were isolated [estimated at 40% yield from 5-aminolevulinic acid (ALA)]. Solid potassium cyanide was added to provide upper and lower cobalt ligands and oxidization to the cobalt(III) form. Datasets of ¹³C-¹H heteronuclear single quantum coherence (Fig. S8 and Table S4), total correlation, ¹H-¹H homonuclear correlation, heteronuclear multiple bond correlation, and rotating frame Overhauser effect were collected measuring resonance of naturally abundant ¹³C and ¹H. In the ¹³C-¹H heteronuclear single quantum coherence spectrum, one major species (>75%) is detected representing dicyanocob(III)yrinic acid along with two possible derivatives. Eight methyl groups at the C-1, C-2, C-5, C-7, C-12α, C-12β, C-15, and C-17 positions were detected. The C-5, C-10, and C-15 carbon positions are sp², which suggests the correct formation of the corrin ring system. A methylene bridge is present at C-10, whereas C-18 and C-19 are protonated. The C-1 methyl group shows nuclear Overhauser effect contacts to the C-19 proton and the ring D acetate CH₂ group, whereas the heteronuclear multiple bond correlation also reveals contacts with pyrrole ring carbons in rings A and D. The methyl group at C-1 and protons at C-18 and C-19 provide evidence for the earlier actions of CbiD and CbiJ, respectively. These findings are similar to previously published spectra of cobyrinic acid heptamethyl ester, although these data were obtained in d₆-hexane (28).

Discussion

In an attempt to elucidate completely the pathway associated with the anaerobic synthesis of the corrin ring, we took advantage of the tool box of enzymes found in B. megaterium. This approach proved successful, because high yields of soluble enzymes were obtained through homologous protein production. The starting point was the production of both cobalt-precorrin-5A and -5B from cobalt-precorrin-4. Although these syntheses validated the activities of CbiF and CbiG, the yields were low (10–20%) as previously noted (15). In contrast, the direct synthesis of cobalt-precorrin-5B from cobalt-factor III with an enzyme mixture incubation results in yields greater than 80%. The poor transformation from cobalt-precorrin-4 likely reflects the observation that isolated cobalt-factor IV and cobalt-precorrin-4 exist in multiple tautomeric forms, some of which may act as inhibitors. The improvement in yield of cobalt-precorrin-5B by use of an enzyme mixture suggests that the pathway enzymes may act to facilitate substrate channeling. For the next step, we have been able to show that CbiD is active as an SAM-dependent methyltransferase, because it transforms cobalt-precorrin-5B into cobalt-precorrin-6A. This result shows that CbiD is able to catalyze the reaction in isolation and does not function only as part of a multienzyme complex (17).

The EPR spectroscopic studies provide a metallocentric view of the black box stages of adenosylcobalamin biosynthesis. The results presented here show that the cobalt ion in the early metal insertion pathway is an active participant in the biosynthesis and not merely an observer. There is no evidence that the ion undergoes oxidation/reduction reactions during the course of the biosynthesis. The cobalt ion remains in the +2 oxidation state throughout, and thus, it seems that the cobalt ion does not play the role of an electron sink, except possibly transiently. However, it is clear that the orbital structure of the ion is altered in cobalt(II)-precorrin-5A/B and cobalt(II)-precorrin-6A/B relative to other intermediates in the pathway, with the singly occupied molecular orbital (SOMO) becoming the d_yz orbital rather than the d_z2 orbital. Such an arrangement of the d electrons is very unusual and has not previously been reported for compounds isolated from biological systems, only being observed in a few synthetic compounds (29). Such synthetic compounds show that the nature of the ligand is the driving force for the rearrangement, with an increase in Lewis base character favoring the d_yz orbital as the SOMO (29). Thus, the cobalt ion participates in the biosynthetic pathway by contributing to the stabilization of the structures of cobalt(II)-precorrin-5A/B and cobalt(II)-precorrin-6A/B that are more electron donating and show more Lewis base character. All other intermediates in the pathway contribute only one covalent bond to cobalt and therefore, exhibit less Lewis basicity and favor the d_z2 orbital as the SOMO. Therefore, the anaerobic biosynthesis of vitamin B₁₂ provides an example of a metal ion with several accessible redox states playing an unexpected role in Lewis acid/base chemistry while apparently having no redox role.

One of the major differences between the aerobic and anaerobic corrin biosynthetic pathways relates to the enzymes used to methylate the C1 position. In the aerobic pathway, this reaction is mediated by CobF, which is similar to the majority of other cobalamin biosynthetic methyltransferases. CobF aids in the removal of the extruded methylated C20 position as acetic acid and then methylates the C-1 position. The anaerobic pathway instead uses two unique enzymes: CbiG and CbiD. CbiG opens the δ-lactone ring and extrudes the methylated C20 position as acetaldehyde (16), forming cobalt-precorrin-5B. This change results in a double bond between C-1 and C-19 (15). Because the cobalt(II) unpaired electron occupies a d_yz orbital in cobalt-precorrin-5B, the metal behaves like a Lewis-base, allowing it to aid in the activation of the C-1 position for methyl addition by electrophilic substitution. Methylation at C-1 would also result in a prototropic rearrangement of the C-1 double bond to C-19 in ring D. This double bond, consequently, has to be reduced by CbiJ, thereby providing a chemical rationale for this sequence of reactions. The dramatic changes in UV-Vis and EPR spectra during these stages are also highly indicative of double bond rearrangement (CbiD) and reduction (CbiJ). It seems that these key differences in chemistry between the early and late cobalt insertion pathways at the C-1 methylation site have crafted two separate methyltransferase enzymes (CbiD/CobF).

The full step-by-step in vitro synthesis of cobyrinic acid from ALA requires 14 enzymes, and we have shown that this transformation can be accomplished by an enzyme mixture approach. This report also provides an accurate description of the anaerobic pathway (Fig. 4) and essentially defines the activities of enzymes involved in the black box region of the biosynthesis. In a broader context, this understanding provides an important prerequisite within the area of synthetic biology if we are to use biology to assist in the production of new chemicals and drugs through natural synthetic pathways. With the enzymes for corrin ring synthesis now defined for both the aerobic and anaerobic pathways, the potential exists to manipulate the pathway for the construction of vitamin analogs and derivatives.

Fig. 4.

The anaerobic biosynthesis of cobyrinic acid from cobalt-precorrin-4. The complete step-by-step pathway with enzymes and cofactors is indicated. Cobalt-precorrin-7 was not isolated as part of this study, because the enzymes CbiE and -T are fused into a single protein in B. megaterium, allowing the direct conversion of cobalt-precorrin-6B into cobalt-precorrin-8. The tautomeric forms shown are consistent with the EPR and MS data but have not been rigorously confirmed. Indeed, the intermediates may exist as mixtures of tautomers.

Materials and Methods

Cloning and Genetic Manipulation.

B. megaterium DSM509 cbiF, cbiG, cbiD, cbiJ, cbiE, and cbiC were PCR amplified from genomic DNA. The PCR fragment and plasmid were digested with the appropriate restriction enzyme followed by purification, ligation, and transformation of E. coli DH10B with 100 µg/mL ampicillin for selection. Table S1 shows plasmids and primers. B. megaterium DSM319 was transformed by a modified minimal media protoplast transformation protocol. Full details are in SI Text. Enzymes were overproduced and purified as previously described (13) but with modifications (SI Text).

Synthesis of Anaerobic Pathway Intermediates.

Individual synthesis of intermediates is provided in SI Text. As an example, for the enzyme mixture synthesis of cobalt-precorrin-5B, the following protocol is provided. Protein lysates derived from recombinant strains of CbiH₆₀ (4 L), CbiF (2 L), and CbiG (0.5 L) were pooled and transferred to the glove box. The pooled lysate was applied to a single immobilized metal affinity chromatography column (7 mL resin) using the standard buffers described. The proteins CbiH₆₀, CbiF, and CbiG were purified to homogeneity; 7.5 mL concentrated protein were buffer exchanged into 10.5 mL Buffer H. The incubation contained ∼40 mg CbiH₆₀, 20 mg CbiF, 15 mg CbiG, 3 mg cobalt-factor III, 21.6 mg SAM, and 43.5 mg sodium dithionite in 10 mL Buffer H. The pH of SAM was adjusted to 8.0 before addition to the reaction. The incubation was left at 37 °C overnight to reach completion (∼80% yield) of cobalt-precorrin-5B.

Analytical Analysis.

For LC-MS analysis, tetrapyrroles were injected onto an Ace 5 AQ column (2.1 × 150 mm, 5 μm; Advanced Chromatography Technologies) that was attached to an Agilent 1100 series HPLC coupled to a micrOTOF-Q II (Bruker) mass spectrometer equipped with an online diode array run at a flow rate of 0.2 mL min⁻¹. Tetrapyrroles were routinely separated with a linear gradient of acetonitrile in 0.1% (vol/vol) trifluoroacetic acid. For EPR, samples were prepared and then flash-frozen in liquid nitrogen. EPR experiments were performed on a Bruker ELEXSYS E500 Spectrometer operating at X-band using a Super High Q Cylindrical Cavity (Q factor ∼ 16,000) equipped with an Oxford Instruments ESR900 Liquid Helium Cryostat linked to an ITC503 Temperature Controller. The experimental parameters are given in the figures. For NMR, all experiments were carried out using a 14.1 T Bruker Avance III Spectrometer (600 MHz ¹H resonance frequency) equipped with a QCI (+ ¹⁹F) cryoprobe. Resonance assignment was completed using a range of 2D NMR experiments.