|Home | About | Journals | Submit | Contact Us | Français|
Derivatives of vitamin B12 are used in methyl group transfer in biological processes as diverse as methionine synthesis in humans and CO2 fixation in acetogenic bacteria1–3. This seemingly straightforward reaction requires large, multimodular enzyme complexes that adopt multiple conformations to alternately activate, protect, and perform catalysis on the reactive B12 cofactor. Crystal structures determined thus far have provided structural information for only fragments of these complexes4–12, inspiring speculation regarding the overall protein assembly and conformational movements inherent to activity. Here we present X-ray crystal structures of a complete ~220 kDa complex that contains all enzymes responsible for B12-dependent methyltransfer, namely the corrinoid iron-sulfur protein (CFeSP) and its methyltransferase (MeTr) from the model acetogen Moorella thermoacetica. These structures provide the first three-dimensional depiction of all protein modules required for the activation, protection, and catalytic steps of B12-dependent methyltransfer. In addition, the structures capture B12 at multiple locations between its “resting” and catalytic positions, allowing visualisation of the dramatic protein rearrangements that enable methyltransfer and identification of the trajectory for B12 movement within the large enzyme scaffold. The structures are also presented alongside in crystallo UV-vis spectroscopic data, which confirm enzymatic activity within crystals and demonstrate the largest known conformational movements of proteins in a crystalline state. Taken together, this work provides a model for the molecular juggling that accompanies turnover and helps explain why such an elaborate protein framework is required for such a simple, yet biologically essential reaction.
B12-dependent methyltransfer lies at the heart of methylation biochemistry and is an essential reaction in human health and microbial CO2 sequestration2,3. In humans, methionine synthase (MetH) methylates homocysteine to form methionine to maintain cellular pools of folate (vitamin B9) and S-adenosylmethionine (AdoMet), the universal methyl donor. MetH mutation or vitamin B12 deficiency can cause serious health consequences, including megaloblastic anemia and birth abnormalities such as neural tube defects13. Acetogenic bacteria, including M. thermoacetica, use CFeSP and MeTr together to catalyze methyltransfer in the Wood-Ljungdahl carbon fixation pathway for growth on CO2 as the sole carbon source14.
For both MetH and CFeSP/MeTr, methyltetrahydrofolate (CH3-H4folate) is the methyl donor, and a protein-bound B12 derivative (cobalamin for MetH and 5′-methoxybenzimidazolyl cobamide for CFeSP) is the methyl carrier. In acetogenic bacteria, the CH3-H4folate methyl group is derived from enzymatic reduction of CO2, whereas in humans, CH3-H4folate is the predominant circulating form of the vitamin. Although CH3-H4folate is the common methyl source, methyl removal from the N5 tertiary amine is chemically challenging because the product, tetrahydrofolate (H4folate), is a poor leaving group1. Therefore, a particularly powerful nucleophile is required, and B12 with cobalt in the +1 oxidation state, a Co(I) species dubbed a “supernucleophile”15, is recruited. Such strong reactivity comes at a price: reducing the inactive Co(II) state to active Co(I) is thermodynamically challenging, as the Co(II/I) reduction potential is one of the lowest in nature, −504 mV in CFeSP and −526 mV in MetH16,17. In CFeSP, an electron is first delivered from a partner protein to an Fe4S4 cluster harboured by an activation domain18,19. The electron is then passed to Co(II) to yield Co(I) (Eq. 1), which attacks CH3-H4folate to form CH3-Co(III) (Eq. 2). CFeSP then delivers the methyl group to the Ni2Fe4S4 active site metallocluster (A-cluster) of acetyl-CoA synthase (ACS), where it becomes the methyl of acetyl CoA, and B12 returns to its nucleophilic Co(I) state.
During the catalytic cycle of both MetH and CFeSP/MeTr (Supplementary Fig. 1), a series of “molecular juggling” acts must be carried out in which domains rearrange to contact the B12 cofactor. Crystal structures of a MetH B12-binding fragment4 and CFeSP from Carboxydothermus hydrogenoformans (ChCFeSP)9 both depict a “resting” state, where B12 is buried by a protective “capping” domain, shielded from unwanted chemistry but inaccessible to substrate. Because methyltransfer employs SN2 substitution20, large conformational changes must “uncap” B12 before chemistry can occur. B12 is “uncapped” in structures of MetH fragments that depict B12 activation7,11,12, but no structure has been solved that shows B12- and CH3-H4folate-binding domains together to illustrate methyltransfer.
To visualise this elusive methyltransfer complex, we determined a 2.38 Å resolution structure of folate-free CFeSP/MeTr from M. thermoacetica (Fig. 1, Supplementary Table 1). The homodimeric MeTr component (58 kDa) is virtually identical to prior structures of both MeTr6,10 (Supplementary Fig. 2a), rms deviation for Cα atoms (rmsd): 0.39 Å, and the analogous MetH domain that binds CH3-H4folate8, rmsd: 1.03–1.08 Å. MeTr and MetH both use (β/α)8 triosephosphate isomerase (TIM) barrels to bind and activate CH3-H4folate for nucleophilic attack. Two CFeSPs are present in the complex, each containing two subunits. The small subunit (35 kDa) is a TIM barrel which acts as the B12 “cap” in the ChCFeSP structure9, while the large subunit (48 kDa) has three domains joined by linkers: an N-terminal Fe4S4 activation domain (residues 1–57), a TIM barrel domain (residues 93–312), and a C-terminal B12-binding domain (residues 325–446). With the exception of Fe4S4- and B12- domains, discussed below, both CFeSP copies align well to the ChCFeSP structure, rmsd: 0.81–0.85 Å (Supplementary Fig. 2b).
In the ~220 kDa CFeSP/MeTr assembly (Fig. 1), the MeTr homodimer lies in the centre, with one CFeSP bound on either side. Each MeTr monomer has a C-terminal α-helix (residues 255–262) protruding from the TIM barrel rim. Contacts between this helix and its preceding loop with a CFeSP small subunit helix (residues 191–204) form the primary interactions between MeTr and CFeSP (Supplementary Fig. 3). Weak interactions between MeTr and CFeSP Fe4S4-domains have stabilized these highly flexible9 domains responsible for B12 activation, allowing their visualisation as bundles of short α-helices connected by long loops that coordinate the Fe4S4 cubane (Fig. 1 and Supplementary Figs. 4–5). The Fe4S4-domains are observed to adopt a variety of positions that are all too far from the B12 to afford reductive activation9. However, the long and primarily unstructured protein linkers that connect both the Fe4S4- and B12- domains to the central TIM barrel must allow for the requisite flexibility for B12 activation (Supplementary Figs. 2b, 6).
B12-domains of both MetH and CFeSP adopt Rossmann-like architectures that bind B12 in the base-off conformation (Fig. 1)4,9. High B-factors support the notion of flexibility mentioned above (Supplementary Fig. 7, Supplementary Table 2), where electron density for both B12-domains represents a highest occupancy position within an ensemble, rather than a sole conformation. In both CFeSPs, the average B12-domain position resides between the “capping” small subunit TIM barrel and the TIM barrel of a MeTr monomer, which are adjacent and nearly perpendicular to each other (Fig. 1). On average, the B12 Co has shifted ~6.5 Å away from its “resting” location towards the MeTr folate-binding site. B12 in this structure is thus positioned “en route” towards catalysis, with ~18 Å remaining to the methyl group of folate modeled into the MeTr active site, based on an alignment with the folate-bound MeTr structure10. In transitioning between “resting” and “en route” positions, the B12 corrin ring breaks three interactions with the “capping” domain and forms new contacts, including an H bond with Asn203 of MeTr (Fig. 2a-b).
Given the flexibility suggested by this structural analysis, we explored whether the B12 domain can sample the full ~25 Å to afford turnover within intact crystals, using anaerobic in crystallo UV-vis spectroscopy to monitor the state of B12. In crystallo and analogous solution spectra were collected in parallel (Fig. 3) for the as-isolated Co(II) form of B12. Reduction to Co(I) and methylation to CH3-Co(III) were then achieved in crystallo and in solution, with all spectra matching well-established CFeSP UV-vis features18,19,21–23. Importantly, absorption features disappear when light is passed through the solution surrounding the crystals, indicating that spectra represent protein in crystals and not protein that may have been liberated into the solution. Collectively, these data demonstrate enzymatic transfer of the CH3-H4folate methyl group to CFeSP-bound B12, evidence that the B12-domain is able to move at least ~18 Å to trigger methyltransfer within the crystal. To our knowledge, this conformational movement represents the largest observed in a crystallized protein (Supplementary Discussion). Such dramatic B12-domain movement is likely facilitated by the fact that the majority of CFeSP/MeTr is composed of rigid TIM barrels that provide all the lattice contacts (Supplementary Fig. 8). Although their biosynthesis is energetically expensive, these high molecular weight TIM barrel scaffolds may be important for B12-dependent methyltransferases to maintain structural integrity during the conformational gymnastics that alternately enable activation, protection, and catalysis of the highly reactive B12 cofactor. Thus, despite the small size of the transferred methyl moiety, these large conformational changes appear to necessitate large enzyme sizes.
While the folate-free CFeSP/MeTr structure describes large B12-domain movements that “uncap” B12 from the small subunit, it is interesting to consider why binding of CFeSP to MeTr does not simply position the B12-domain directly over the MeTr active site. One explanation is that the structure represents an inactive complex; however, in crystallo results clearly demonstrate that CFeSP/MeTr crystals are active. Another explanation posits that an ensemble of “en route” conformations exists when CH3-H4folate is absent, and that CH3-H4folate binding would shift the conformational equilibrium, moving B12 closer to the folate-binding site. To obtain experimental support for this hypothesis, we solved additional CFeSP/MeTr structures co crystallized with CH3-H4folate, with and without Ti(III) citrate as a reductant at 3.03 Å and 3.50 Å resolution, respectively. UV-vis spectroscopy performed on these crystals shows that these structures represent a substrate form (CH3–H4folate bound, B12 in the Co(II) state) and a product form (H4folate bound, B12 in the CH3 Co(III) state) of the complex (Supplementary Fig. 9). Compared with the folate-free structure, B12 in both folate-bound structures has indeed moved even closer to the MeTr folate-binding site (by an average of ~7.7 Å) and exhibits new H bonding features (Fig. 2c). In these folate-bound structures, the B12 corrin ring has severed all interactions with the “capping” CFeSP small subunit and contacts only MeTr residues. Here, asparagine and glutamine residues that line the MeTr surface appear to participate in an “amide hand off”, sequentially passing B12 along its trajectory as it progresses towards folate (Fig. 2b-c and Supplementary Fig. 10).
Interestingly, the terminal amide in this “hand off”, Asn199, is strictly conserved in both MeTr and MetH and was previously shown to switch conformations between folate-free and folate-bound forms10, a feature also observed in the CFeSP/MeTr structures presented here (Fig. 2b-c,e). In apo-MeTr structures, Asn199 points upwards and out of the active site, while in folate-bound MeTr structures Asn199 turns down to H-bond with the N5 of folate. Since N199A mutation moderately affects folate binding (20-fold in Kd) but dramatically compromises catalytic efficiency (25,000-fold in kcat/Km), Asn199 is thought to be important for formation of the transition state10. In our CFeSP/MeTr structures, we observe a new role for Asn199 in B12-domain conformational switching: when folate is absent, Asn199 points out of the active site, blocking a closer B12 position. However, when folate binds and Asn199 reorients to H-bond with folate, space becomes available for B12 to move closer to the MeTr folate-binding site. Therefore, the position of Asn199 itself could help shift the conformational equilibrium of the B12-domain, signalling that substrate has bound to MeTr. Asn199 is an ideal signal for substrate binding, as it is the only MeTr residue known to reposition upon folate binding10.
Displacement of the B12-domain from its resting position to the position nearest the folate-binding site can be attributed to two independent conformational changes within the complex, best described as “swinging” and “clamping” motions (Fig. 2d). The B12-domain can “swing” relative to the rest of CFeSP (Supplementary Fig. 11), and CFeSP can “clamp” the B12-domain towards the MeTr active site (Fig. 2d, Supplementary Fig. 12). Despite varying degrees of “clamping” over a range of ~14° across the structures, the interface between the CFeSP small subunit and MeTr is preserved (Supplementary Fig. 13).
Although folate binding shifts the average position of the B12-domain closer to the MeTr folate-binding site, the B12 Co is still too far for SN2 methyl transfer (Figure 2e). Intriguingly, a large, continuous electron density peak is present in 2Fo-Fc, Fo-Fc, and composite omit maps, emanating from the corrin ring and stretching directly over the folate-binding site, suggestive of an alternative, low-occupancy corrin conformation (Fig. 2f, Supplementary Fig. 14). A trial refinement of a putative corrin ring at 40% occupancy satisfies the Fo-Fc difference maps (Supplementary Fig. 15), positioning B12 over the folate-binding site.
The multiple positions of B12 captured here (Fig. 2g) highlight the conformational flexibility of the CFeSP/MeTr scaffold and provide a framework to understand the molecular juggling of domains during B12-dependent methyltransfer (Fig. 4). Prior to MeTr binding, the CFeSP B12-domain rests against the “capping” small subunit, as in the structure of ChCFeSP9, with reactive Co(I) of B12 protected (“resting” state). From this conformation, either the “cap” or the B12-domain must move to allow substrate access. Our folate-free CFeSP/MeTr structure indicates that upon MeTr binding, the B12-domain becomes “loosened” and flexible, adopting an ensemble of conformations that lie en route towards the MeTr active site. Here, reactive B12 species would be protected by the CFeSP small subunit and MeTr TIM barrels. CH3-H4folate binding to MeTr accompanied by movement of Asn199 shifts the equilibrium of B12-domain conformers, placing B12 closer to folate, as in our folate-bound CFeSP/MeTr structures, with B12 protection afforded by MeTr. It is notable that even after CH3-H4folate binds, the major B12 position is still not directly over the folate methyl group, as such a position is expected to be transient. After methyltransfer, the B12-domain can return to the small subunit to “re-cap” the methylated B12 product, protected by the small subunit TIM barrel.
Overall, our data indicate that B12-domain movement is not a simple, two-state switch between “resting” and “catalytic” conformations. Instead, a flexible B12-domain samples an ensemble of conformations, where subtle shifts of the conformational equilibrium place B12 progressively closer to the active site, thereby increasing the population of conformers capable of methyltransfer without obstructing substrate access or hindering domain movement. This model is consistent with MetH studies where ligation, alkylation, and redox state of the B12 cobalt can favour/disfavour various binding modes, alternately shifting the equilibrium of conformers for ordered domain rearrangements during the reaction cycle11,12,24,25. We further identify MeTr residues that contact B12 along its trajectory, ending with Asn199. In MetH the B12 ligating residue His759 has been shown to play a dual role in catalysis and in signalling conformational shifts11. The strictly conserved, folate binding Asn199 of MeTr could similarly play a dual role in both catalysis and conformational signalling. We thus expect this model for dynamic domain juggling, communicated by residues involved in substrate and cofactor binding, to be a common theme in methyltransfer between the B vitamins folate and B12.
CFeSP and MeTr were expressed and purified anaerobically from M. thermoacetica ATCC 39073 and from recombinant Escherichia coli, respectively. Crystals were grown anaerobically by hanging drop vapour diffusion. Diffraction data were collected at 24ID-C at the Advanced Photon Source, Argonne National Laboratory, and 5.0.2 and 8.2.2 at the Advanced Light Source, Lawrence Berkeley National Laboratory. Structures were solved by molecular replacement. Data collection and refinement statistics are presented in Supplementary Table 1, and representative electron density for protein domains and for cofactors/substrate are shown in Supplementary Figs. 16–21. Solution and in crystallo UV-vis absorption spectra were collected as described in the text and in the Full Methods.
Full Methods and any associated references are available in the online version of the paper at www.nature.com/nature.
CFeSP was expressed and purified anaerobically from Moorella thermoacetica ATCC 39073 as described21, except for the following modifications. CFeSP was purified from cell extracts using DEAE-cellulose and high resolution Q-Sepharose anion exchange chromatography followed by phenyl-Sepharose hydrophobic interaction chromatography. Fractions containing CFeSP were concentrated and buffer exchanged using Amicon ultracentrifuge concentrators in the anaerobic chamber. MeTr was expressed and purified anaerobically from recombinant Escherichia coli as described6. Concentrations of CFeSP and MeTr protein samples were determined using the Rose-Bengal method28 and kept in storage buffer: 50 mM Tris-HCl, pH 7.6, 100 mM NaCl, 2 mM dithiothreitol.
Crystals of the folate-free CFeSP/MeTr complex were grown by hanging drop vapour diffusion in an anaerobic chamber (Coy Laboratories) at room temperature by adding 1 µL of precipitant (100 mM Bis-Tris, pH 6.5, 100 mM calcium acetate, 9% PEG 5000 monomethyl ether, 20% glycerol) to 2 µL of an equimolar mixture of CFeSP and MeTr (~250 µM monomer for each), over a 0.5 mL reservoir solution of precipitant. Large, brown, rod-shaped crystals appeared overnight. Crystals were looped and cryo-cooled in liquid nitrogen anaerobically prior to collection of X-ray diffraction data at 100 K. Crystals of CFeSP/MeTr co crystallized with the CH3-H4folate substrate were obtained in the same manner as above, except the protein solution also contained CH3-H4folate at 1 mM concentration. Crystals of CFeSP/MeTr co-crystallized with both CH3-H4folate and Ti(III) citrate as a reductant were obtained in the same manner, except the precipitant solution also contained Ti(III) citrate at 3 mM concentration.
Two X-ray diffraction data sets were collected for the folate-free CFeSP/MeTr structure. A lower-resolution data set (3.3 Å) was collected at the Advanced Light Source (ALS) beam line 5.0.2 (λ=1.1000 Å), and a higher-resolution data set (2.38 Å) was later collected at the Advanced Photon Source (APS) beam line 24ID-C (λ=0.9792 Å).
The initial data set to 3.3 Å resolution was processed in HKL2000 and Scalepack29. The structure was solved by molecular replacement in Phaser30, using individual structures of MeTr10 (PDB ID: 2E7F) and ChCFeSP9 (PDB ID: 2H9A) lacking its B12-domain as independent search models. Two CFeSP/MeTr complexes (~220 kDa each) were found in the asymmetric unit, and crystals belonged to the space group P212121 with unit cell dimensions (Å): a=137.42, b=159.87, and c=241.92. Iterative rounds of refinement with residue-grouped B-factors were carried out in CNS31 and PHENIX32, with model building in Coot33. The four B12-domains present in the complex were kept as a polyalanine model. Final Rwork and Rfree values for the model were 29.2% and 33.7%, respectively, when refinement of the structure to higher resolution began. Data collection and refinement statistics for this data set are shown in Supplementary Table 1. Ramachandran analysis was carried out in PROCHECK34: 74.0% of residues resided in the most favoured region, with 21.2% additionally allowed, 3.1% generously allowed, and 1.7% disfavoured.
The data set to 2.38 Å resolution was processed in HKL2000 and Scalepack29. Although this crystal formed in similar conditions as the crystal which gave the 3.3 Å resolution data set, the space group was now P21212, with unit cell dimensions (Å): a=125.71, b=242.84, and c=79.67. The structure for this crystal was thus solved by molecular replacement in Phaser30 using the MeTr homodimer and CFeSP heterodimers lacking B12-domains from the previously refined model of the 3.3 Å resolution structure as independent search models. Only one CFeSP/MeTr complex was present in the asymmetric unit. Iterative rounds of refinement were carried out in CNS31 and PHENIX32, with model building in Coot33. Translation/libration/screw (TLS) refinement was carried out in latter refinement rounds with seven TLS groups: the MeTr homodimer (chains A and B), the Fe4S4 domain of one CFeSP large subunit (chain C), the TIM domain of chain C with one small subunit (chain D), the B12 domain of chain C, the Fe4S4 domain of the second large subunit (chain E), the TIM domain of chain E with the second small subunit (chain F), and the B12 domain of chain E. Data collection and refinement statistics are shown in Supplementary Table 1, and average B-factors for each domain of the final model are given in Supplementary Table 2. Ramachandran analysis was carried out in PROCHECK34: 90.1% residues reside in the most favoured region, with 9.5% additionally allowed, 0.3% generously allowed, and 0.2% disfavoured. The final model contains residues 1–262 (of 262) for both MeTr chains (A and B), residues 2–442 (of 446) for both CFeSP large subunit chains (C and E), and residues 1–323 (of 323) for both CFeSP small subunit chains (D and F).
Except for the Fe4S4 and B12-domains, the entire structure is composed of TIM barrels for which the electron density is well-defined (Supplementary Fig. 16). Electron density is weaker for the Fe4S4 domains (Supplementary Figs. 5, 17, and 19), consistent with the fact that these domains exhibit higher B-factors (Supplementary Table 2, Supplementary Fig. 7). However, reasonable electron density is present for the main chain and most side chains, allowing us to build a model for this domain. Still, several side chains of the Fe4S4 domains lack clear electron density, and thus for these residues, atoms were truncated past the Cβ atom (chain C: 12 residues truncated and chain E: 16 residues truncated, out of 56 total residues).
Although B-factors are high and electron density is weak for the B12-domains in general, electron density for the B12 cofactors is unambiguous (Supplementary Fig. 20), and density is also clear in several helical regions, including those near B12 (Supplementary Fig. 18). Because the structure of a CFeSP B12-domain bound with B12 was already known9, we used the clear electron density of the B12 cofactor and the resolvable helices to position the B12-domain during model building. Still, many side chains of the CFeSP B12-domains lacked clear electron density, and thus for these residues, atoms were truncated past the Cβ atom (chain C: 50 residues truncated and chain E: 59 residues truncated, out of 118 total residues).
The B12 cofactor in the final model contains 5,6-dimethylbenzimidazole as the lower ligand moiety, as in cobalamin. Although active with cobalamin35, previous studies have shown that CFeSP isolated from M. thermoacetica harbours an unusual B12 derivative that contains 5′ methoxybenzimidazole as the lower ligand instead36. However, disorder of the B12 cofactor and B12-domain due to thermal motion of these regions in the CFeSP/MeTr crystal resulted in weak electron density for substituents of the benzimidazolyl ring (Supplementary Figs. 18 and 20). Therefore, we cannot confirm the presence of this unusual B12 derivative from our crystallographic studies, and we have thus modeled cobalamin as the form of B12 in the structure.
Previous spectroscopic studies27 in addition to the crystal structure of ChCFeSP9 have indicated that a water molecule coordinates the central cobalt of B12 in the as-isolated CFeSP. Here, Co(II) is the major species and is expected to be five-coordinate. However, because of disorder we do not observe electron density to suggest a water molecule bound to cobalt (Supplementary Figs. 18 and 20). Accordingly, we have not modeled a water molecule.
For crystals grown with CH3-H4folate, X-ray diffraction data were collected at APS beam line 24ID-C to 3.50 Å resolution at λ=1.6039 Å to optimize the cobalt peak anomalous signal. For crystals grown with both CH3-H4folate and Ti(III) citrate, X-ray diffraction data were collected at ALS beam line 8.2.2 to 3.03 Å resolution at λ=1.0000 Å. The structures were solved by molecular replacement using the MeTr homodimer and CFeSP heterodimers lacking Fe4S4- and B12-domains from the folate-free 2.38 Å CFeSP/MeTr structure as independent search models. Refinement of the folate-free CFeSP/MeTr structure against either folate bound X-ray data set was not sufficient to solve the structure, as the unit cell dimensions were markedly different (Supplementary Table 1). Following molecular replacement, one CFeSP/MeTr complex was present in the asymmetric unit, and omit electron density clearly indicated the presence of bound folate (Supplementary Fig. 21). Iterative rounds of refinement were carried out in CNS31 and PHENIX32, with model building in Coot33. The same test set of reflections for Rfree calculations was used for both folate-bound data sets. Data collection and refinement statistics are shown in Supplementary Table 1. Ramachandran analysis was carried out in PROCHECK34: for the CH3-H4folate-only structure, 89.6% of residues reside in the most favoured region, with 9.8% additionally allowed, 0.3% generously allowed, and 0.2% disfavoured. For the CH3-H4folate with Ti(III) citrate structure, 89.5% of residues reside in the most favoured region, with 10.0% additionally allowed, 0.2% generously allowed, and 0.3% disfavoured. The final models both contain folate, B12 and residues 1–262 (of 262) for MeTr chains (A and B), residues 2–442 (of 446) for CFeSP large subunit chains (C and E), and residues 1–323 (of 323) for CFeSP small subunit chains (D and F). As with the folate-free structure, several side chains of the Fe4S4 domains for both folate-bound structures lacked clear electron density, and thus for these residues, atoms were truncated past the Cβ atom (chain C: 15 residues truncated and chain E: 18 residues truncated, out of 52 total residues). Similarly, many side chains of the B12-domains lacked electron density, and thus for these residues, atoms were truncated past the Cβ atom (chain C: 51 residues truncated and chain E: 78 residues truncated, out of 118 total residues). The liganded/oxidation states of folate and B12 in these structures were determined by use of a UV-vis microspectrophotometer (see below).
Titanium(III) citrate (100 mM in 50 mM Tris, pH 7.6) was prepared37, and (6S)-5-methyl-5,6,7,8-tetrahydrofolate (CH3-H4folate) containing one glutamate tail was purchased from Schircks Laboratories. As-isolated, reduced, and methylated CFeSP samples in solution were prepared in a room temperature anaerobic chamber (MBraun) following similar procedures to those previously described18,19,22,23,26,27. Briefly, purified CFeSP (20 µM) was used for the as-isolated sample, CFeSP mixed with Ti(III) citrate (1 mM) was used for the reduced sample, and CFeSP mixed with equimolar MeTr, Ti(III) citrate (1 mM), and CH3-H4folate (1 mM) was used for the methylated sample. Spectra were taken using a Nanodrop 2000c (Thermo Scientific) in a quartz cuvette or on the sample stage in the anaerobic chamber directly after mixing; identical solutions lacking CFeSP were used as blanks.
To obtain in crystallo UV-vis spectra, CFeSP/MeTr crystals in as-isolated, reduced, and methylated forms were prepared in a similar fashion. In a room temperature anaerobic chamber (Coy Laboratories), crystals were looped into a 2 µL drop, which was placed on a cover slide and contained one of the following three solutions for as-isolated, reduced, and methylated samples, respectively: well solution, well solution with Ti(III) citrate (10 mM), and well solution with Ti(III) citrate (10 mM) and CH3-H4folate (1 mM). A ring of epoxy surrounding each drop was applied to the cover side, and a second cover slide was placed on top, sandwiching the drops within a uniform distance separation and sealing the crystals within an anaerobic environment. Upon curing of the epoxy, crystals were brought out of the anaerobic chamber and mounted on a an XZ translation stage (Newport, UMR8.25 & SM-13) in a fibre optic coupled microspectrophotometer (Ocean Optics, Jaz) with 40 mm diameter reflective objectives (Optique Peter, France) and a deuterium-halogen lamp (DH2000-BAL, Ocean Optics), similar to that previously described38,39. Stray light was blocked with black-out material. The light focus was coarsely aligned to the crystals by visual inspection and then finely aligned by monitoring light transmission in real time. Data were acquired at room temperature with the SpectraSuite software (Ocean Optics). The background transmission was measured through the solution immediately surrounding the crystals. The dark current was measured with the light shuttered off. Sample, reference, and dark current spectra were acquired by averaging 10–50 scans with total exposure times of 90–1000 ms. Experiments were completed within 60 min of sample preparation, and crystals remained intact over the course of the experiment, as observed using a microscope following data collection.
To generate Fig. 3, absorbance spectra were scaled relative to each other to account for variable crystal sizes and pathlengths, where absolute peak absorbances did not exceed 1 absorbance unit.
UV-vis absorption spectra were collected on a microspectrophotometer at 100 K for crystals of folate-free CFeSP/MeTr, crystals that were grown in the presence of CH3-H4folate only, and for crystals grown in the presence of both CH3 H4folate and Ti(III) citrate (Supplementary Fig. 9). The spectra were compared to the analogous solution spectra (Fig. 3). The spectrum for the folate-free crystal is similar the spectrum of CFeSP alone in solution, with broad features at ~400 nm and 470 nm indicative of the Fe4S4 cluster and B12 primarily in the Co(II) state. The spectrum for the crystal grown with CH3 H4folate matches the spectrum of folate-free crystals, indicating that B12 has remained primarily in the Co(II) state, and turnover has not occurred. However, the spectrum for the crystal grown with both CH3-H4folate and Ti(III) citrate is markedly different and contains a peak at 450 nm, indicating that B12 in these crystals is methylated to the CH3-Co(III) state. Based on these data, we have modeled the methyl group on folate in the structure co-crystallized with CH3-H4folate only, while we have modeled the methyl group bound to Co of B12 for the CH3-H4folate/Ti(III) citrate structure. Without these spectroscopic data, assignment of the location of the methyl group would have otherwise been prevented by the resolution limits of the data.
We thank Joseph E. Darty for his assistance to the purification of CFeSP. This work was supported by NIH grants GM69857 (C.L.D.) and GM39451 (S.W.R.) and the MIT Energy Initiative (C.L.D.). C.L.D. is a Howard Hughes Medical Institute Investigator. This work is based upon research conducted at the Advanced Photon Source on the Northeastern Collaborative Access Team beamlines, which are supported by award RR-15301 from the National Center for Research Resources at the National Institutes of Health. Use of the Advanced Photon Source is supported by the U.S. Department of Energy, Office of Basic Energy Sciences, under Contract No. DE-AC02-06CH11357. The Advanced Light Source is supported by the Director, Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under Contract No. DE-AC02-05CH11231.
Supplementary Information is linked to the online version of the paper at www.nature.com/nature.
Author Contributions Y.K. performed crystallization and data collection, processing, and refinement that gave the folate-free and folate-bound CFeSP/MeTr structures. N.A. built the microspectrophotometer and performed in crystallo UV-vis experiments with the aid of Y.K., who performed the parallel solution UV-vis experiments. T.I.D. determined initial crystallization conditions and performed initial data collection, and L.C.B. processed and refined these data. G.B. and J.S. expressed and purified protein samples, and S.W.R and C.L.D. were involved in study design. Y.K. and C.L.D. wrote the manuscript.
Author Information Atomic coordinates are deposited in the Protein Data Bank under accession codes XXXX, XXXX, and XXXX. Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests.