Cysteine-mediated decyanation of vitamin B12 by the predicted membrane transporter BtuM

Uptake of vitamin B12 is essential for many prokaryotes, but in most cases the membrane proteins involved are yet to be identified. We present the biochemical characterization and high-resolution crystal structure of BtuM, a predicted bacterial vitamin B12 uptake system. BtuM binds vitamin B12 in its base-off conformation, with a cysteine residue as axial ligand of the corrin cobalt ion. Spectroscopic analysis indicates that the unusual thiolate coordination allows for decyanation of vitamin B12. Chemical modification of the substrate is a property other characterized vitamin B12-transport proteins do not exhibit.

C obalamin (Cbl) is one of the most complex cofactors (Supplementary Figure 1a) known, and used by enzymes catalyzing for instance methyl-group transfer and ribonucleotide reduction reactions 1,2 . For example, in the methionine synthase MetH, the cofactor is used to transfer a methyl moiety onto L-homocysteine to produce L-methionine 1,3 . Many bacteria require Cbl for survival 1,2,4,5 , but only a small subset of prokaryotic species can produce this molecule de novo, via either an aerobic or anaerobic pathway 4,5 . Roughly two thirds of archaea and eubacteria are Cbl-auxotrophs that rely on uptake of either Cbl or its precursor cobinamide 2,5,6 (Cbi, Fig. 1a, Supplementary  Figure 1b). Dependence on uptake has probably evolved, because synthesis of Cbl involves roughly 30 different enzymes and is energetically costly. Gram-negative bacteria require the TonBdependent transporter BtuB 1 to transport Cbl across the outer membrane (Supplementary Figure 1c). For subsequent transport of vitamin B12 across the cytoplasmic membrane, the only characterized bacterial uptake system is the ABC transporter BtuCDF, which is predicted to be present in approximately 50% of Cbl-auxotrophic bacteria 5,7 . Many Cbl-auxotrophic Gramnegative bacteria do not encode BtuCDF, whereas they do contain BtuB. Metabolic reconstruction and chromosomal context analyses, e.g. co-localization with the gene for BtuB, have identified potential alternative inner membrane vitamin B12 transporters, one of which is BtuM 5 . BtuM homologues are small membrane proteins of~22 kDa, and found predominantly in Gram-negative species, distributed mostly over α-, β-, and γ-proteobacteria (Supplementary Data 1).
Here, we sought to characterize the predicted vitamin B12 transporter BtuM from Thiobacillus denitrificans (BtuM Td ). We show that BtuM Td is involved in transport of Cbl in vivo and we solved its structure to 2 Å resolution. A cobalt-cysteine interaction allows for chemical modification of the substrate prior to translocation, which is a rare feature among uptake systems.

Results
BtuM Td supports vitamin B12-dependent growth. To test experimentally whether BtuM Td is a potential Cbl-transporter, we constructed an Escherichia coli triple knockout strain, E. coli ΔFEC, based on Cadieux et al. 8 . In this strain, the gene encoding the Cblindependent methionine synthase, MetE 9 , is deleted. The metE deletion makes it impossible for E. coli ΔFEC to synthesize methionine, unless it can import Cbl 8,9 . In that case, L-methionine can be synthesized using the Cbl-dependent methionine synthase, MetH 3,8 . E. coli ΔFEC has additional deletions in btuF and btuC, encoding subunits of the endogenous Cbl-transporter BtuCDF 7,8 . Therefore, E. coli ΔFEC cannot import Cbl, prohibiting MetHmediated L-methionine synthesis. Consequently, E. coli ΔFEC can grow only if L-methionine is present or if vitamin B12 import is restored by (heterologous) expression of a Cbl-transport system 8 . The phenotype of E. coli ΔFEC was confirmed in growth assays (Supplementary Figure 2a). Cells that are not expressing any Cbltransporter did not exhibit substantial growth in methionine-free medium, whereas cells complemented with an expression plasmid for BtuCDF grew readily (Fig. 1b). Cells expressing BtuM Td had a similar growth phenotype, indicating that BtuM Td is a potential transporter for vitamin B12 (Fig. 1b).
Crystal structure of BtuM Td bound to vitamin B12. The BtuM family contains an invariably conserved cysteine residue (Supplementary Figure 3a). In BtuM Td , this cysteine is located at position 80, and mutation to serine abolishes the ability of the protein to complement the E. coli ΔFEC strain (Fig. 1b). To investigate the role of the cysteine, we solved a crystal structure at 2 Å resolution of BtuM Td in complex with Cbl. Data collection as well as refinement statistics are summarized in Table 1. BtuM Td consists of six transmembrane helices with both termini located on the predicted cytosolic side (Fig. 1c). The amino acid sequences of BtuM proteins are not related to any other protein 5 but, surprisingly, BtuM Td resembles the structure of Scomponents from energy-coupling factor (ECF)-type ABC transporters 10 (Supplementary Figure 4 and Supplementary  Table 1). In contrast to BtuM proteins, ECF-type ABC transporters are predominantly found in Gram-positive bacteria. They are multi-subunit complexes consisting of two peripheral ATPases and two transmembrane components (EcfT and Scomponent) 10,11 . EcfT and the ATPases together form the socalled ECF-module. S-components bind the transported substrate, and dynamically associate with the ECF-module to allow substrate translocation 10,[12][13][14] . Intriguingly, no homologues of EcfT could be found in T. denitrificans. In addition, all ABC-type ATPases encoded by the organism are predicted to be part of classical ABC transporters, and not ECF transporters. Therefore, we conclude that the organism does not encode an ECF-module, and hypothesize that solitary BtuM Td may be responsible for Cbl uptake. This hypothesis is supported by the ability of BtuM Td to transport vitamin B12 when expressed heterologously in E. coli ΔFEC. Importantly, E.coli also does not encode an ECFmodule 11 , hence BtuM Td cannot interact with a module from the host, and BtuM Td must be able to support Cbl uptake using a different mechanism than that of ECF transporters 10,11 . In a few cases, the biotin-specific S-component BioY 15 has also been found in organisms lacking an ECF-module and was shown to mediate transport without the need for an ECF-module 15 . However, organisms encoding only BioY without an ECF-module are rare 15 , and in the large majority of organisms BioY is associated with an ECF-module 11 . In contrast, BtuM homologues (apart from one exception) are found exclusively in organisms lacking an ECF-module (Supplementary Data 1).
Further experiments, for instance using purified protein reconstituted in proteoliposomes, are required to test whether BtuM Td also catalyses transport in vitro without any additional component involved. However, the in vivo assay gives a very strong indication that BtuM Td is a transporter itself, as the protein was expressed in a heterologous host that does not contain any ECF-module or S-component. Similar in vivo experiments have been used extensively in the past to identify other transporters (for instance ref. 16 ) and have the advantage over in vitro assays that physiologically relevant conditions are used.
BtuM Td binds cobalamin using cysteine ligation. Close to the predicted periplasmic surface of BtuM Td , we found well-defined electron density (Supplementary Figure 5) representing a bound Cbl molecule. The binding mode of Cbl in the crystal structure ( Fig. 2a) is striking for two reasons. First, the essential Cys80 is the α-axial ligand of the cobalt ion. To our knowledge, cobalt coordination by cysteine has not been observed in any other Cblbinding protein of known structure. Binding of cysteine to cobalt in a corrinoid has been hypothesized for the mercury methylating enzyme HgcA 17 and observed in a synthetic cyclo-decapeptide, but in the latter case the residue replaced the β-ligand 18 .
Second, Cbl is bound to BtuM Td in the base-off conformation in which the 5,6-dimethylbenzimidazole moiety does not bind to the cobalt ion (Fig. 2a). In contrast, at physiological pH the conformation of free Cbl in aqueous solution is base-on with the 5,6-dimethylbenzimidazole moiety coordinated to the cobalt ion in the α-axial position 1 (Fig. 1a). The base-off conformation has been found only in a subset of Cbl-containing enzymes, but not in Cbl-binding proteins without enzymatic activity 1 , such as the periplasmic substrate-binding protein BtuF 19 , the outer membrane transporter BtuB 20 , and human Cbl-carriers intrinsic factor 21 , haptocorrin 22 and transcobalamin 23 . Enzymes that bind Cbl with the base-off conformation usually use a histidine residue as the α-axial ligand. In this way, the reactivity of the cobalt at the β-axial position is altered, allowing among others a variety of methyl-group transfer reactions 1 . Therefore, the base-off binding mode by BtuM Td could indicate that the protein may exhibit enzymatic activity.  Figure 1a, b). The ligand at the α-axial position (base-on conformation) is the 5,6-dimethylbenzimidazole base, which is covalently linked to the corrinoid ring. When this coordination is lost, Cbl is termed base-off. Cbi lacks the 5,6-dimethylbenzimidazole base (indicated by the zigzagged red line). b Growth assays with E. coli ΔFEC was conducted in the presence of 50 μg ml −1 L-methionine or 1 nM Cbl. Additional experiments in the presence of different Cbl concentration are shown in Supplementary  Figure 2f and g. All growth curves are averages of nine experiments (three biological triplicates, each with three technical replicates). Top panel: cells containing the empty expression vector (pBAD24) in the presence of methionine (blue line) or Cbl (grey line) and cells expressing the BtuCDF system (black and red lines, respectively). Bottom panel: cells expressing BtuM Td (black and red lines) or mutant BtuM Td _C80S (blue and grey) in the presence of methionine and Cbl, respectively. The inset displays a western blot showing that the mutant is expressed to wild-type levels (the full-length western blot can be found in Supplementary Figure 2h). c The structure of BtuM Td in cartoon representation, coloured from blue (N terminus) to red (C terminus) and viewed from the membrane plane. α-helices (H1-6) and connecting loops (L1-5) are indicated. Cbl is shown in stick representation with carbon atoms coloured wheat, the oxygen and nitrogen atoms in red and blue, respectively, the cobalt ion in pink. Four n-nonyl-β-D-glucopyranoside detergent molecules are also shown in stick representation (carbons in light grey) BtuM Td catalyses decyanation of vitamin B12. Indeed, the structure of BtuM Td suggests that the protein can catalyse chemical modification of the substrate. We co-crystallized BtuM Td with cyano-Cbl, which contains a cyano-group as the β-ligand 1,4 . Cyano-Cbl is the most stable form of vitamin B12 4 but, despite the tight binding of the β-ligand, in the crystal structure the cyano-group is absent indicating protein-mediated decyanation. Consistent with decyanation and the presence of a cysteine ligand in BtuM Td , the absorbance spectrum of Cbl-bound BtuM Td showed pronounced differences compared to that of free Cbl 18,24 (Fig. 2b). The characteristic absorption peak at 361 nm of Cbl is absent and two peaks with lower absorption appear around 330 and 370 nm. The absorption between 500 and 580 nm is lower than in free Cbl, and a new peak at 430 nm is present.
In place of the cyano-group, the imidazole group of His207 from a neighbouring BtuM Td molecule in the crystal is located at the β-axial position. His207 is the last histidine residue of the His 8 affinity-tag (His-tag) engineered at the C terminus of the protein (Supplementary Figure 6). Because crystal contacts may be nonphysiological and the His-tag is a non-natural addition to the protein, we performed control experiments to exclude the possibility that decyanation is an artefact. First, we showed by mass spectrometry (MS) that the loss of the cyanide does not require crystal formation (Supplementary Figure 7a). Second, we showed that decyanation also occurred by BtuM Td with a C-terminal Glu-Pro-Glu-Ala (EPEA)-tag instead of a His-tag (Supplementary Figure 7b). Notably, the EPEA-tagged protein was active in the growth assay and also removal of the His-tag did not affect activity (Supplementary Figure 2b, c). Finally, binding of Cbl to BtuM Td with His-tag or EPEA-tag was accompanied by the same changes in absorption spectrum (Figs. 2b, 3b). Therefore, we conclude that decyanation takes place regardless of crystal formation or presence of a His-tag.
Kinetics of the BtuM Td catalysed decyanation reaction. To study the kinetics of BtuM Td -catalysed decyanation we used cobinamide (Cbi) instead of Cbl as substrate. Because Cbi does not contain the 5,6-dimethylbenzimidazole moiety (Fig. 1a), it mimics the base-off conformation of cobalamin, which makes the compound suitable to study decyanation without interference from the slow conversion 25 of base-on to base-off Cbl. The absorption spectra of Cbl-bound and Cbi-bound BtuM Td are almost identical (Fig. 2c), indicating identical coordination of the cobalt ion of Cbi at the α-axial and β-axial positions. MS analysis showed that binding of Cbi to BtuM Td also results in decyanation (Supplementary Figure 8a). To probe Cbi binding by BtuM Td , we used isothermal titration calorimetry (ITC), which revealed dissociation constants for the His-tagged and EPEAtagged protein of 0.65 ± 0.27 and 0.58 ± 0.13 μM (s.d. of the mean of technical triplicates), respectively (Fig. 3a). It is noteworthy that we were unable to assay for Cbl-binding by ITC. We speculate that the conversion from base-on to base-off Cbl is so slow 25 that it may prevent detection of Cbl-binding by ITC. Additionally, the absence of the membrane environment also appears to preclude Cbl binding to purified BtuM Td , as binding was observed only when the substrate was added before solubilisation (Fig. 2b, c, Supplementary Figure 9).
Because binding of cyanide to cobinamide causes a decrease in absorbance at 330 nm and an increase at 369 nm 24 , we expected the opposite spectral changes upon decyanation. Addition of excess of apo-BtuM Td (Supplementary Figure 10a and b) to a solution of Cbi indeed revealed time-dependent changes in absorbance consistent with a decyanation reaction (Fig. 3b, c). Decyanation occurred with an apparent time constant of τ = 12 ± 0.7 min (s.d. from technical triplicates, Fig. 3c), which is comparable to the rate observed in the human decyanating enzyme CblC 25,26 . We also tested Cbi binding and decyanation using mutant proteins C80A and C80S. While these mutants were unable to bind Cbl, they remained capable of binding Cbi as demonstrated by co-purification of the molecule with the protein (Supplementary Figure 10c, d). We measured the affinity of BtuM Td _C80S to Cbi with ITC and found a dissociation constant of 5.6 ± 2.8 μM (s.d. of the mean of technical triplicates), which is an order of magnitude weaker than the wild-type (WT) protein.
The absorbance spectra of Cbi bound to the mutant proteins showed the characteristic features for cyano-Cbi, indicating that decyanation was abolished (Supplementary Figure 10c, d). Consistently, the decyanation assay with BtuM Td _C80S did not reveal the slow spectral changes observed for the WT protein (Fig. 3b, c). These results show that Cys80 is required for decyanation of Cbi and that binding and modification of the substrate are separate events: fast binding (detected by ITC) is followed by slow modification. The lack of detectable binding of Cbl to BtuM Td _C80S (measured by lack of co-purification, Supplementary Figure 10c, d) may indicate that the cysteine is also required for base-on to base-off conversion, and that the base-on conformer binds with too low affinity for detection by co-purification. To understand BtuM Td -catalysed decyanation of Cbl and Cbi in more detail, we mutated conserved amino acids H28, D67, Y85, and R153 located in the binding pocket (Supplementary Figure 3b). Mutant D67A could not be purified, and was not analysed further. Cbl-bound mutants H28A, Y85L, and R153A displayed the same spectral properties as the WT protein ( Supplementary Figure 11a), and MS analysis showed that the binding of Cbl was accompanied by decyanation, indicating that the conserved residues are not essential for the reaction (Supplementary Figure 11b-d). Finally, to exclude that BtuM Td is merely a decyanating enzyme, and that the potential reaction product hydroxyl-Cbl is subsequently transported by another protein, we show that BtuM Td also mediates uptake of hydroxyl-Cbl in the growth assay (Supplementary Figure 2d, e).

Discussion
We showed in vivo that BtuM Td is a vitamin B12 transporter, which is consistent with the predictions based on bioinformatics analysis 5 . Our work sheds light on the diversity of transport systems used for the uptake of vitamin B12. The outer membrane transporter BtuB is a TonB-dependent active transporter, which uses a different mechanism of transport than inner membrane proteins 1,20 . The well-studied inner membrane type II ABC transporter BtuCDF uses hydrolysis of ATP to pump Cbl into the cell like the ECF-transporter, ECF-CbrT 27 . Both systems require a substrate-binding protein and are multiprotein complexes 7,8,27,28 . BtuM Td on the other hand, must operate by a different mechanism because the protein lacks accessory components and the expected ATPase motifs of ABC transporters 10 . BtuM Td structurally resembles the S-components of ECF transporters. In ECF transporters, the S-components bind the transported substrate with high affinity and then associate with an ECF-module for energizing transport. During the transport cycle, the Scomponents rotate (topple over) in the membrane to bring the substrate from the outside to the cytoplasm. We hypothesize that BtuM Td mediates the translocation of Cbl through the membrane by a similar toppling mechanism. Because BtuM Td does not require an ECF-module, the transport mode may be facilitated diffusion along the concentration gradient of the substrate. In T. denitrificans and most other BtuM hosts, the BtuM Td gene colocalizes with btuR, which encodes for the cobalamin adenosyltransferase BtuR. This enzyme catalyses the synthesis of 5′-deoxyadenosyl-cobalamin and would offer a mechanism of metabolic trapping, similar to what has been proposed for other vitamin transporters in bacteria 29 .
Our work provides experimental evidence for a binding mode of Cbl, in which cysteine ligation and the base-off conformation are linked. This binding mode leads to decyanation of cyano-Cbl, for which we propose a reductive decyanation mechanism, which depends on Cys80 17,18 (Supplementary Figure 8d). The proposed decyanation mechanism differs from the mechanism used by CblC, where a flavin acts as reducing agent. In CblC, the flavin donates two electrons resulting in the reductive decyanation (CN − ) and the reduction of the Coion 25,26 . For BtuM Td , cysteine-catalysed reductive decyanation would only result in the release of CN − , but not in the reduction of the Co-ion. Finally, BtuM Td likely combines two functions: transport of the substrate into the bacterial cell, and chemical modification of the substrate. Such combined functionality rarely occurs in transporters, and has been observed only in phosphotransferase systems 30 . However, in that case the modification (phosphorylation) takes place on the cytoplasmic side of the membrane 30 , whereas BtuM Td appears to modify on the periplasmic side of the membrane. Internalisation of decyanated vitamin B12 may be relevant because environmental cyano-Cbl exists 26 . A combination of decyanation and transport activity would make cyano-Cbl directly accessible for conversion into physiological forms, for example, by BtuR.

Methods
Bioinformatic identification of BtuM homologues and ECF-modules. The amino acid sequence of BtuM Td was used as a search query using the iterative jackHMMer algorithm (default settings) with the reference proteome database 31 until the search converged leading to 131 hits. Within the genomes of the identified 131 organisms, we screened for the presence of an ECF-module using the pHMMer algorithm (default settings) 31 with the amino acid sequence of the transmembrane component (ECF-T) from Lactobacillus delbrueckii 14 as a search query. Additionally, we used the SEED viewer (http://pseed.theseed.org) to verify the absence of any ECFtransporter in a subset of organisms (46 present in the SEED database) and also used this tool to find all ABC transporters in T. denitrificans to verify that none of these are an ECF-transporter.
Molecular methods. For expression in E. coli MC1061 32 a codon optimized version (Invitrogen) of btuM (Tbd_2719) from Thiobacillus denitrificans ATCC25259 with a C-terminal eight histidine affinity-tag or EPEA-tag was used and introduced into pBAD24 33 with NcoI and HindIII restriction sites. A single glycine (Gly2) was introduced to be in-frame with the start-codon of the NcoI restriction site. Single amino acid substitutions and removal of the affinity tag were conducted using sitedirected mutagenesis. The complementation plasmid for expression of BtuC and BtuF was constructed using Gibson Assembly following the standard procedure (NEB). All constructs were checked for correct sequences by DNA sequencing. All primers are listed in Supplementary Table 2.
Construction of the ΔFEC strain. E. coli ΔFEC, was constructed by P1-mediated generalized transduction 27,34,35 . In short: E. coli JW0154 (ΔbtuF::Km R ) was used as the basis for construction of E. coli ΔFEC. The kanamycin resistance cassette was removed using the FLP-recombinase 36 . The metE::Km R locus from E. coli JW3805 and the ΔbtuC::Km R locus of E. coli JW1701 was introduced 34,35 , resulting in E. coli ΔFEC (ΔbtuF, ΔmetE, ΔbtuC::Km R ). Colony PCRs based on three primer pairs 27 were used to verify Km R -insertions, FLP-recombinase-mediated removal of Km Rmarkers and absence of genomic duplications (Supplementary Figure 13).
Growth assays. The strains carrying various expression vectors were grown overnight at 37°C on LB-agar plates supplemented with 25 μg ml −1 kanamycin and 100 μg ml −1 ampicillin. M9 minimal medium (47.7 mM Na 2 HPO 4 × 12H 2 O, 17.2 mM KH 2 PO 4 , 18.7 mM NH 4 Cl, 8.6 mM NaCl) was supplemented with 0.4% glycerol, 2 mM MgSO 4 , 0.1 mM CaCl 2 , 100 μg ml −1 L-arginine, 25 μg ml −1 kanamycin and 100 μg ml −1 ampicillin. A single colony was picked and used to inoculate an M9-medium pre-culture supplemented with 50 μg ml −1 L-methionine (Sigma-Aldrich). The pre-culture was grown~24 h at 37°C, shaking in tubes with gas-permeable lids (Cellstar), and then used to inoculate the assay medium in a 1:500 ratio. The assay medium was supplemented with 0.00001% L-arabinose (Sigma-Aldrich) and either 50 μg ml −1 L-methionine, 0.01 nM, 1 nM and 5 nM cyano-cobalamin (Acros Organics), or 0.1 nM hydroxy-cobalamin (Sigma-Aldrich). Overall, 200 μl medium was added per well of a sterile 96 well plate (Cellstar). Plates were sealed with a sterile and gas-permeable foil (BreatheEasy, Diversified Biotech). The cultures were grown for 1000 min (1250 min for Cbi) in a BioTek Power Wave 340 plate reader at 37°C, shaking. The OD 600 was measured every 5 min at 600 nm. All experiments were conducted as technical triplicates  Western blotting. Cells grown in LB-medium were broken in 50 mM K-P i pH 7.5 supplemented with 10% glycerol, 1 mM MgSO 4 , 1 mM phenylmethylsulfonyl fluoride (PMSF) and DNaseI with glass beads in a tissue lyser at 50 hertz. The lysate was centrifuged for 10 min at 20,000×g and 4°C and the supernatant was used for further analysis. The samples analysed by SDS-polyacrylamide gel electrophoresis followed by semi-dry western blotting. The primary antibody was mouse anti-Tetra·His Antibody, BSA-free from Qiagen (Cat.No. 34670) and the secondary antibody was anti-mouse IgG (whole molecule)-alkaline phosphatase conjugate antibody from Sigma-Aldrich ( Purification of His-tagged BtuM Td . Purification of His-tagged protein for biochemical analyses was essentially performed as described above with the following adaptations. HEPES was replaced with 50 mM K-P i pH 7 or 7.5 (for ITC and spectral analyses, respectively), NG was replaced with 0.04% DDM, and 100 mM NaCl was used throughout. For purification of the apo protein, substrate was omitted from all buffers. For spectral analyses of substrate-bound proteins, substrate was omitted from buffer B.
Purification of EPEA-tagged BtuM Td . EPEA-tagged protein was purified as described above with the following adaptations. CaptureSelect TM C-tagXL Affnity Matrix (Thermo Fisher Scientific) was used. DTT and imidazole were omitted in all steps and 50 mM Tris/HCl pH 7.5 was used instead of K-P i . The column was washed once with 10 CV buffer supplemented with 500 mM MgCl 2 . Elution was done in four fractions of 0.5 (first) ml-0.8 ml (others) in buffer containing 2 M MgCl 2 .
Crystallisation and phasing and structure determination. BtuM Td purified for crystallisation was concentrated to between 1.1 and 1.6 mg ml −1 with a 10,000 kDa cut-off Vivaspin concentrator (Sartorius) at 4000×g at 2°C. The initial screening was done using a Mosquito robot (TTP Labtech), and a hit was found after 1 month in the H1 condition (50 mM Tris pH 8.5, 28% (v/v) PEG400) of the MemGold2 screen (Molecular Dimensions) at 4°C. Larger and better diffracting pyramid-shaped crystals were obtained at 8°C after 3 to 4 weeks in a crystallization buffer containing 25 mM Tris pH 8.5 and 25 to 30% (v/v) PEG400, 50 mM Tris pH 8.5 and 27 to 30% (v/v) PEG400 or 75 mM Tris pH 8.5 and 29 to 30% (v/v) PEG400, using the sitting drop vapour diffusion method (in MRC Maxi 48-well plate) and a 1:1 mixing ratio (2 μl final drop volume) of mother liquor and protein solution. Phases were obtained from crystals that were soaked for 1 min with 100 mM Tb-Xo4 37 (Molecular Dimensions) mother liquor solution (0.5 μl added directly to the drop). Diffraction data of the native crystals were collected at the Swiss Light Source (SLS) at PXI (X06SA) beamline (λ = 1.000 Å, T = 100 K) and two anomalous diffraction datasets were collected at the European Synchrotron Radiation Facility (ESRF) at beamline ID23-1 (λ = 1.400 Å and 1.476 Å, T = 100 K). Data were processed with XDS 38 and the two datasets containing anomalous information were merged and subsequently used to solve the structure with ShelX 39 . Autobuild 40 was used to obtain a starting model, which was refined further with Phenix refine 41 with manual adjustments done in Coot 42 . The model was used as an input to solve the phase problem for the native dataset, which was carried out with Phaser-MR 43 . The model of the native data was refined iteratively with Phenix refine 41 and manual adjustments were done in Coot 42 . The Ramachandran statistics for the final model are 99.47% for favoured regions, 0.53% for allowed regions and 0.00% for outliers. A stereo view of 2F o -F c electron density of the entire structure including the backbone trace molecule, the binding pocket and the Cbl-ligand is provided in Supplementary Figure 5a-c, respectively. All structural figures were prepared with an open-source version of pymol (https:// sourceforge.net/projects/pymol/).
UV-Vis assay to determine decyanation of vitamin B12. All measurements were carried out in a Cary100Bio spectrophotometer (Varian) at room temperature and baseline corrected for buffer B in a quartz cuvette. To monitor the binding of dicyano-Cbi or cyano-Cbl by BtuM Td over time, every minute a spectrum was recorded between 260 and 640 nm for 40 min (Cbi, n = 3) or every 20 min for 12 h (Cbl, n = 1) at room temperature. For this measurement, a molar protein to substrate ratio of 5:1 (Cbi) or 1:1 (Cbl) was used. To obtain the apparent time constant, τ, the absorbance ratio of 369/330 nm was plotted against the time and fitted with a single exponential decay function in Origin 8. Decyanation assays with Cbi were conducted as technical triplicates and errors are standard deviations of the averaged ratios (if not specified otherwise).
ITC measurement with Cbi. Binding of dicyano-Cbi to purified BtuM Td was measured on a microcal200 ITC (GE healthcare) in high feedback mode. The cell temperature was set to 25°C with a reference power of 9.5 μcal s −1 . During the measurement, the sample was stirred at 750 rpm and a 15-fold excess (WT) or 32.5-fold excess (C80S) of Cbi in the syringe was used over the protein concentration in the cell. The data was analysed in Origin and experiments were done as technical triplicates (n = 3). The obtained dissociation constants were averaged and the error is the standard deviation of the replicates.
Mass spectrometry. BtuM Td variants and mutant proteins were purified as described above. BtuM Td proteins were diluted in a 1:1 (v/v) ratio with 0.1% formic acid and 5 μl were injected into an Ultimate 3000-UPLC system (Dionex), connected to a Q-Exactive mass spectrometer (Thermo Fisher Scientific) and separated on a 2.1 mm × 50 mm Acquity UPLC BEHC18, 1.7 μm (Waters). Solvent A was H 2 O with 0.1% formic acid and solvent B was acetonitrile with 0.1% formic acid. The following mobile phase gradient was delivered at a flow rate of 0.6 ml min −1 starting with a mixture of 60% solvent B for 1 min. Solvent B was increased to 90% over 5 min with a linear gradient and kept at this concentration for 5 min. Solvent B was reduced to 60% in 0.1 min and kept for 3.9 min resulting in a total elution time of 15 min. The column temperature was kept constant at 40°C. The mass spectrometer was operated in positive mode. Full scan MS spectra were acquired for 10 min from m/z 1000 to 2000 at a target value of 1 × 10 6 and a max IT of 500 ms with a resolution of 140,000 at m/z 200. Scans were averaged using Xcalibur 4.0.27.42 Qualbrowser and the isotopically resolved MS spectrum was deconvoluted using the built-in Xtract algorithm.
Data availability. Data supporting the findings of this manuscript are available from the corresponding author upon reasonable request. Atomic coordinates and structure factors for the crystal structure of BtuM Td have been deposited in the Protein Data Bank under the accession code 6FFV. The mass spectrometry data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD010024.