Switch I-dependent allosteric signaling in a G-protein chaperone–B12 enzyme complex

G-proteins regulate various processes ranging from DNA replication and protein synthesis to cytoskeletal dynamics and cofactor assimilation and serve as models for uncovering strategies deployed for allosteric signal transduction. MeaB is a multifunctional G-protein chaperone, which gates loading of the active 5′-deoxyadenosylcobalamin cofactor onto methylmalonyl-CoA mutase (MCM) and precludes loading of inactive cofactor forms. MeaB also safeguards MCM, which uses radical chemistry, against inactivation and rescues MCM inactivated during catalytic turnover by using the GTP-binding energy to offload inactive cofactor. The conserved switch I and II signaling motifs used by G-proteins are predicted to mediate allosteric regulation in response to nucleotide binding and hydrolysis in MeaB. Herein, we targeted conserved residues in the MeaB switch I motif to interrogate the function of this loop. Unexpectedly, the switch I mutations had only modest effects on GTP binding and on GTPase activity and did not perturb stability of the MCM–MeaB complex. However, these mutations disrupted multiple MeaB chaperone functions, including cofactor editing, loading, and offloading. Hence, although residues in the switch I motif are not essential for catalysis, they are important for allosteric regulation. Furthermore, single-particle EM analysis revealed, for the first time, the overall architecture of the MCM–MeaB complex, which exhibits a 2:1 stoichiometry. These EM studies also demonstrate that the complex exhibits considerable conformational flexibility. In conclusion, the switch I element does not significantly stabilize the MCM–MeaB complex or influence the affinity of MeaB for GTP but is required for transducing signals between MeaB and MCM.

loading and offloading in a nucleotide-dependent manner (2)(3)(4). 5Ј-Deoxyadenosylcobalamin (AdoCbl) serves as a cofactor for MCM, which is a B 12 -dependent isomerase (5,6). In mammals, the tissue concentration of cobalamins is low, and formation of the active MCM holoenzyme relies on a B 12 trafficking pathway that assimilates and inserts AdoCbl posttranslationally (7)(8)(9). Mutations in the B 12 trafficking pathway that lead to a functional deficiency of AdoCbl or affect MCM itself result in methylmalonic aciduria, an inborn error of metabolism that is inherited as an autosomal recessive disorder (10 -14).
MeaB gates the transfer of AdoCbl to MCM from adenosyltransferase, the enzyme that synthesizes this active cofactor form ( Fig. 1) (15,16). GTP hydrolysis by MeaB controls the AdoCbl transfer process and precludes loading by the inactive cob(II)alamin form (3). Once loaded with AdoCbl, MCM catalyzes the radical-based isomerization of methylmalonyl-CoA to succinyl-CoA. Inadvertent dissociation of the 5Ј-deoxyadenosine moiety from the active site precludes reformation of AdoCbl at the end of the catalytic cycle and results in inactivation of MCM (3). MeaB plays additional roles following cofactor loading. It protects MCM from inactivation, slowing escape from the turnover cycle ϳ30-fold (2,17), and promotes ejection of inactive cob(II)alamin when it forms during the catalytic cycle (3). Although MeaB has been studied quite extensively (2)(3)(4)18), the human ortholog, CblA, is poorly characterized (19,20). Nearly 40 pathogenic mutations have been identified in CblA that are associated with methylmalonic aciduria (12,21).
MeaB exists as a homodimer and binds MCM with high affinity (K D ϭ 34 -525 nM), which varies depending on the presence or absence and the identity of the ligands bound to the two proteins (4). In contrast, the metallochaperones CooC and HypB, which catalyze the insertion of nickel into carbon monoxide dehydrogenase and [Ni-Fe]-hydrogenase, respectively, undergo NTP-dependent dimerization (22)(23)(24).
Like most G-proteins, MeaB exhibits low intrinsic GTPase activity (4). In complex with MCM, the GTPase activity of MeaB is enhanced ϳ100-fold. Hence, MCM functions as a GTPase-activating protein (GAP) for MeaB (4). Like other G-proteins, MeaB contains the signature switch I/II elements ( Fig. 2A), which are important for catalysis and signaling to downstream effector proteins (25). Residues in the switch I/II This work was supported by National Institutes of Health Grants DK45776 (to to R. B.), 5 F32 GM113405 (to G. C. C.), and GM110001 (to D. S.). The authors declare that they have no conflicts of interest with the contents of this article. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health. This article contains supplemental Fig. S1. 1 Both authors contributed equally to this work. 2 To whom correspondence should be addressed. cro ARTICLE motifs provide some of the ligands to the Mg 2ϩ ion and undergo conformational rearrangements in response to nucleotide binding, exchange, and hydrolysis. In the so-called "loaded spring" mechanism, direct H 2 O-mediated or Mg 2ϩmediated contacts form between the switch motif residues and the ␥-phosphate of GTP and are disrupted upon hydrolysis (26).
Besides the switch I/II elements, MeaB and its homologs have an additional mobile loop called switch III (residues 178 -188), which is distant from the GTP-binding site ( Fig. 2A). Switch III has been captured crystallographically in multiple conformations (17). Mutations localizing in the switch III loop have been identified in methylmalonic aciduria patients (27) and compromise the fidelity of AdoCbl loading/offloading processes and alleviate the protective effect on MCM during turnover, making it more susceptible to inactivation (17). GTP hydrolysis triggers conformational changes in the switch I/II loops that are propagated to switch III via switch II (17).
The switch II element (residues 154 -162) plays an additional role in suppressing the intrinsic GTPase activity of MeaB (28). Val 156 and Val 158 in switch II are postulated to impede access of the water-activating general base (Asp 92 as discussed below). Glu 154 in switch II interacts with Arg 108 in switch I (Fig. 2B), and its mutation to alanine increases the intrinsic GTPase activity 10-fold and renders MeaB insensitive to GAP activation in the presence of MCM.
The role of the switch I sequence (residues 92-108) in the chaperone function of MeaB is not known. The switch I sequence, D 92 ,D 105 KTR 108 , is not strongly conserved and resides on an extended loop (Fig. 2B). An identical D 180 ,D 193 KTR 196 sequence is found in the human homolog, CblA (Fig. 2C). The crystal structures of both MeaB and CblA indicate that they do not represent catalytically active conformations. Despite the presence of nucleotides in some MeaB structures, Mg 2ϩ is absent from all of them, and the switch I residue Asp 92 , which is predicted to activate the nucleophilic water for attack on the ␥-phosphorus, is displaced 11-14 Å from the active site (17,28,29). We have postulated that the active site of MeaB is brought into catalytic register in the MCM-MeaB complex either via donation of critical residues from MCM and/or via a conformational rearrangement of residues in the MeaB active site (28). Interestingly, most of the switch I residues are not resolved in the CblA structure ( Fig.   Figure 1. Nucleotide-gated transfer of AdoCbl from ATR to MCM. AdoCbl transfer is driven by the binding of ATP to a vacant site in the ATR homotrimer and gated by GTP hydrolysis catalyzed by MeaB. AdoCbl bound to ATR is in a five-coordinate "base-off" conformation ( max ϭ 458 nm) and switches to a six-coordinate "base-off/His-on" ( max ϭ 527 nm) conformation in the MCM active site.  2C), although Asp 180 (corresponding to Asp 92 in MeaB) also appears to be blocked from the active site by a bulky hydrophobic residue (Val 244 ) in this inactive conformation (19).

Allosteric signaling in the MeaB-MCM complex
The crystal structure of IcmF (30,31), which represents a fusion between isobutyryl-CoA mutase and its G-protein (32), provides a useful structural model for the possible interface between MeaB and MCM. The structure of IcmF with AdoCbl shows two Mg 2ϩ ions and a GDP bound in each G-domain (Fig.  2D). Asp 249 in IcmF (corresponding to Asp 92 in MeaB), hydrogen bonds to a water molecule while Asp 262 (equivalent to MeaB Asp 105 ) coordinates a second Mg 2ϩ and Arg 265 (MeaB Arg 108 ) hydrogen bonds to an oxygen atom on the ␣-phosphate group. Interestingly, IcmF lacks a bulky, hydrophobic residue, which would restrict access of the catalytic Asp 249 residue to the nucleophilic water and exhibits higher intrinsic GTPase activity (32) than either MeaB (4) or CblA (19).
In this study, we have interrogated the role of the switch I loop in the chaperone functions of MeaB. Surprisingly, our data reveal that switch I residues do not play essential roles in GTP hydrolysis but, like the previously characterized switch II and III loops, play a critical role in gating the transfer of AdoCbl from ATR to MCM and in evicting inactive cob(II)alamin from the MCM active site. Furthermore, our single-particle EM studies provide the first structural insights into the organization of a stand-alone mutase, MCM, in complex with its G-protein chaperone and reveal a 2:1 stoichiometry in a conformationally dynamic MCM-MeaB complex.

Effect of switch I mutations on GDP binding
We used isothermal titration calorimetry to determine the K D values for GDP binding to MeaB, which is a homodimeric protein. MeaB exhibited negative cooperativity for nucleotide binding with the affinity at site 1 (K D ϭ 0.20 Ϯ 0.01 M) being ϳ10-fold stronger than for site 2 (K D of 2.1 Ϯ 0.2 M) ( Table 1). All switch I mutants similarly exhibited negative cooperativity for GDP binding. Perturbations in the binding affinities for site 1 were relatively modest with K D values ranging from 0.07 Ϯ 0.02 M for D105A MeaB to 1.3 Ϯ 0.6 M for K106A MeaB. Perturbations at site 2 followed the same pattern with the D105A mutant showing the highest affinity for GDP (K D ϭ 0.8 Ϯ 0.2 M) and the K106A mutant showing the weakest affinity (K D ϭ 7.2 Ϯ 1.9 M). Unfortunately, the R108M mutant could not be characterized due to its insolubility. Importantly, the binding isotherms for the switch I mutants were similar to that of wild-type MeaB (supplemental Fig. S1), suggesting that the mutations do not significantly affect nucleotide binding.

GTPase activity of switch I mutants
Modest changes in the intrinsic GTPase activity were observed with all the mutants compared with wild-type MeaB ( Table 2). Wild-type MeaB, in the presence of MCM, exhibits an ϳ100-fold increase in the k cat (4.1 Ϯ 0.2 min Ϫ1 ) for GTP hydrolysis (4). The GAP effect of MCM ranged from 16-(D92A) to 40-fold (K106A) activation for the switch I mutants. The magnitude of GAP activation for the D105A and D92N mutants was dampened by their higher intrinsic GTPase activity.

Switch I mutations protect MCM from inactivation
Inactivation of MCM during catalysis results in the formation of OH 2 Cbl, which can be distinguished from the active AdoCbl cofactor by its absorption peak at 351 nm (3). MCM is inactivated to a similar extent in the presence or absence of MeaB ( Fig. 3A, B, and D). However, in the presence of GMP-PNP, MeaB affords some protection to MCM against inactivation ( Fig. 3, C and D). The switch I mutants D92A, D92N, D105A, and K106A also protect MCM from inactivation ( Fig.  3D). Hence, the switch I loop does not appear to be important in the allosteric protection of MCM against inactivation during turnover.

Switch I mutations uncouple GTPase activity from AdoCbl loading to MCM from ATR
Two nucleotide-dependent switches, ATP binding to ATR and GTP hydrolysis by MeaB, trigger AdoCbl transfer from ATR to the MCM-MeaB complex ( Fig. 1) (15,16). GTP hydrolysis by MeaB is essential for cofactor transfer, and in the presence of GMPPNP, ϳ0.06 eq of AdoCbl is transferred to MCM-MeaB, whereas the rest is released into solution (Fig. 3E). Mutations in the switch I element uncouple GTP hydrolysis from AdoCbl transfer. Thus, in the presence of GMPPNP, ϳ0.4 eq of AdoCbl is transferred from ATR to the MCM-MeaB complex containing the D92N or K106A mutant with the rest of the cofactor appearing in solution. An even greater fraction of AdoCbl transfer (ϳ0.8 eq) is seen in the presence of the D92A or D105A mutant of MeaB. Hence, the switch I loop is important for interprotein signaling during GTPase-gated AdoCbl transfer from ATR to the MCM-MeaB complex.

Reactivation of MCM by switch I mutants
Accumulation of cob(II)alamin in the active site of MCM from which the 5Ј-deoxyadenosine moiety has been lost triggers the repair function of MeaB (3). This rescue function of MeaB specifically evicts cob(II)alamin but not the oxidized product, OH 2 Cbl. Addition of GMPPNP to the wild-type Table 1 Dissociation constants for GDP binding to MeaB switch I mutants ND, not determined.

Allosteric signaling in the MeaB-MCM complex
cob(II)alamin-containing MCM-MeaB complex under anaerobic conditions triggers ejection of ϳ97% of the cofactor (Fig.  3F). The released cofactor can be quantified as OH 2 Cbl (at 351 nm) following separation from the protein using a Centricon filter under aerobic conditions. Addition of GMPPNP to the MCM-MeaB complex loaded with cob(II)alamin and MeaB harboring switch I mutations led to ejection of 51-66% of the cofactor (Fig. 3F). Hence, the switch I mutations impair GTPdependent displacement of cob(II)alamin from MCM.

EM studies on the MCM-MeaB complex
The stoichiometry of the MCM-MeaB complex was estimated by size exclusion chromatography coupled to multiangle light scattering (SEC-MALS). In the presence of GMPPNP, the wild-type complex elutes as a single monodisperse species that has an average molecular mass of ϳ359 kDa (Fig. 4A). This matches a 2:1 stoichiometry for MCM:MeaB, which has a predicted molecular mass of 352 kDa (MCM, 142 kDa per ␣␤ heterodimer; MeaB, 71 kDa per ␣ 2 homodimer). The SEC-MALS chromatograms for complexes containing switch I MeaB mutants were identical to the wild-type complex. Hence, the mutations in the switch I loop do not cause significant perturbations in the gross architecture or stability of the MCM-MeaB complex. Therefore, perturbations in allosteric communication between MeaB and MCM, which impair its chaperone functions, are not due to destabilization of the complex.
Next, we analyzed the MCM-MeaB complex by negativestain EM (Fig. 4B). Reference-free 2D class averages showed three distinct globular densities that appear flexibly connected. The outside densities are elongated and larger than the central density, indicating that these domains might be MCM dimers, which flank a central MeaB dimer. To test this assignment, crystal structures of the related bacterial MCM from Propionibacterium shermanii (34) and of MeaB (17) were compared by manual alignment based on the 2D averages without symmetry imposed and back-projected. The resulting 2D projections of these structures indeed match the arrangement in the 2D averages of the MCM-MeaB complex, supporting the proposed architecture (Fig. 4C).
The complex is dynamic and exists in a large array of conformations ranging from an "open" conformation where each MCM heterodimer is extended out from MeaB to one that is more "closed." Some class averages also display a mixture of states with one MCM open and the other closed. However, the effect of this complex heterogeneity is not yet understood by these data. Overall, the EM analysis is consistent with a model of MCM-MeaB complex in which each GTP-binding site in the MeaB homodimer interacts with the active ␣-subunit in the ␣␤ MCM heterodimer (Fig. 5A), giving rise to the observed 2:1 stoichiometry.

Discussion
The crystal structure of IcmF (30,31) provides the best molecular level model for understanding allosteric signaling between MCM and MeaB. However, the organization of the IcmF structure is inherently different from that of the standalone MeaB and mutase proteins, highlighting the limitations

Allosteric signaling in the MeaB-MCM complex
of using IcmF as a predictive model for understanding the interaction between MeaB and MCM. A major organizational difference is that MeaB is a homodimer and forms a hub for interaction of two MCM heterodimers as indicated by our negative-stain EM data (Fig. 4B). In contrast, the two monomeric G-domain units in the IcmF homodimer are far apart and interact with the B 12 -domain and substrate-binding domain of the mutase but not with each other (Fig. 5, A versus B). Also unlike the M. extorquens MCM, which is an ␣␤ homodimer containing a single B 12 -and substrate-binding site in the ␣-subunit (Fig. 5), each subunit in the IcmF dimer binds B 12 and is active for loading. Hence, although the stoichiometry of the MCM-MeaB complex is 2:1, the stoichiometry of the mutase:G-domain in IcmF is 1:1. We predict that the human MCM-CblA complex will exhibit yet another organizational structure in which the ␣ 2 homodimer of MCM is organized around an ␣ 2 CblA core (Fig. 5C).
The IcmF structure reveals an extensive network of hydrogen bonds and salt bridges between the B 12 -and G-domains. The switch I loop (D 249 ,D 262 RIR 263 ) in the IcmF structure is posi-

Allosteric signaling in the MeaB-MCM complex
tioned directly at the interface between the B 12 -and G-domains. Based on sequence homology to IcmF and HypB (D 69 ,D 75 AER 78 ), the switch I motif in MeaB (D 92 ,D 105 KTR 108 ) is predicted to supply the catalytic residues for activating a water molecule (Asp 92 ), ligating Mg 2ϩ (Asp 105 ), and binding the ␤and ␥-phosphate oxygens of GTP (Arg 108 ) ( Fig. 2A) (22,29). The GTPase activity of G-proteins is highly sensitive to amino acid substitutions at conserved positions in the switch I and switch II loops (35)(36)(37). Mutation of the glutamine in the EXXGQ switch II motif of Ras-related proteins abolishes GTPase activity and activates Ras proteins toward malignant transformation (38). Similarly, substitution of the putative catalytic aspartate residue in the switch I motif of the ATPase-dependent iron protein involved in nitrogenase function leads to a drastic decrease in the rate of nucleotide hydrolysis (39). Mutation of the residue in NifH corresponding to Asp 92 in MeaB severely impairs its GTPase activity and maturation of the MoFe cofactor for nitrogenase maturation (39).
Given the predicted role of Asp 92 in catalysis, it is surprising that the D92A mutation had no effect on the intrinsic GTPase activity, whereas the conservative D92N substitution increased it ϳ3-fold (Table 2). Both substitutions at the Asp 92 position decreased MCM-dependent GAP activation of MeaB, albeit by a modest 2-4-fold. SEC-MALS analyses ruled out that the modest effects on GAP activation were due to weakened complexation (Fig. 4A); nor was the effect due to decreased nucleotide-binding affinity (Table 1). These results indicate that Asp 92 does not play an essential catalytic role in MeaB in the absence or presence of MCM. In light of these results, we conclude that the general base used for activating water is provided by MCM as observed with other GAPs (40). In IcmF, Asp 249 , the presumed water-activating general base, is derived from the G-domain rather than from the mutase domain (Fig. 2D). Although the role of Asp 249 in catalysis has not been evaluated, its positioning is consistent with its postulated role (31) and suggests that the geometry of the GTPase active site in the MCM-MeaB complex might be different in this regard.
The switch I residue, Lys 106 in MeaB, is equivalent to Arg 263 in IcmF that forms a salt bridge with Glu 76 derived from the B 12 -domain (Fig. 2D). The functional significance of this interdomain interaction is not known. In the various structures of MeaB, Lys 106 is either positioned on a disordered loop, solventexposed, or hydrogen-bonded to the carbonyl groups of Leu 103 in switch I or Ile 120 on the ␤3 sheet (17,28,29). The K106A mutant exhibits the largest GAP activation of all switch I mutants studied, indicating that it is the least impacted (Table 2).
In contrast to their modest effects on the GTPase activity, the switch I mutants exhibit significant effects on the following chaperone functions: GTPase-gated AdoCbl loading onto MCM from ATR and cob(II)alamin offloading from MCM. For example, although the K106A mutation has very modest effects on the intrinsic GTPase activity of MeaB and on GAP activation by MCM, it has a drastic effect on the chaperoning functions of MeaB (Fig. 3, E and F). The surface exposure of Lys 106 and the adverse impact of the K106A mutation on gating AdoCbl loading and supporting cob(II)alamin off-loading suggests that this residue, like the others in the switch I domain, is important for communicating between MeaB and the B 12 -binding domain of MCM.
In summary, our results lead us to the unexpected conclusion that the conserved switch I residues are not essential for the GTPase activity of MeaB or for transmitting the GAP activity of MCM. They do, however, profoundly impair communication between ATR and the MCM-MeaB complex during AdoCbl loading and between MCM and MeaB during cob(II)alamin off-loading. Unlike IcmF, ATR is not required during off-loading inactive cofactor from MCM (41). Together with our previous studies, we conclude that the switch I, II, and III elements do not contribute significantly to stabilizing the MCM-MeaB complex or influence the affinity of MeaB for GTP. Furthermore, the switch elements are not significantly involved in protecting MCM against inactivation. Instead, they are collectively important for transducing signals between MeaB and MCM, and mutations in any of the switch elements impact the same chaperone functions that ensure the fidelity of loading active cofactor and allow off-loading of inactive cofactor.
G-proteins and ATPases are evolutionarily diverse regulators of varied functions ranging from DNA replication and protein synthesis and localization to cytoskeletal dynamics and coenzyme trafficking and assimilation (35,42,43). Studies on these proteins are uncovering multiple strategies by which they transduce allosteric signals. MeaB, a prototype of the G3E metallochaperone subfamily, with its atypical active site and an auxiliary switch III element amplifies the diversity of the catalytic and signaling strategies used by G-proteins. A high-resolution structure of MeaB in complex with MCM will provide needed insights into how a catalytically competent MeaB active site is constructed.

Materials
AdoCbl, GMPPNP, ATP, GTP, GDP, methylmalonic acid, coenzyme A, and other reagent grade chemicals were purchased from Sigma. Trifluoroacetic acid was purchased from MeaB is an ␣ 2 homodimer with two GTP-binding sites. B, IcmF is a fusion protein with AdoCbl, MeaI (G-domain), and substrate-binding domains fused in the direction of N to C terminus as shown. The crystal structure reveals that IcmF exists as a homodimer with the G-domains split into monomers, each interacting with an active ␣-subunit. C, human MCM is an ␣ 2 homodimer with two active subunits. The organization of the human complex is proposed to be two MCMs to one CblA and is presently unknown.

Thermodynamic analysis of GDP binding
Isothermal titration calorimetry experiments were performed at 10°C in Buffer A using a 300-l injection syringe and a 1.43-ml injection cell. Samples were prepared by filtration through a 0.2-m filter and then degassed under vacuum at 4°C using a ThermoVac sample degasser. Each titration was performed at least in duplicate. GDP (10-l injections of 150 -400 M) was added to 15-20 M MeaB. The data were analyzed for a two-site binding model using the MicroCal Origin program.

Protection of MCM from inactivation by MeaB
Inactivation of MCM under steady-state turnover conditions was examined by enzyme-monitored conversion of AdoCbl to OH 2 Cbl (formed via oxidation of the cob(II)alamin intermediate) at 20°C in 0.1 M potassium phosphate, pH 7.5, containing 10 mM MgCl 2 . The reactions and sample preparations were performed in the dark to avert spurious formation of OH 2 Cbl by photolysis of AdoCbl. Samples were prepared by the addition of reaction components in the order described below. MCM (25-30 M) was reconstituted using an equimolar concentration of AdoCbl. A molar excess (35-40 M) of MeaB was added to the MCM holoenzyme to obtain the MCM-MeaB complex. GMPPNP was then added to the reaction mixture to a final concentration of 1-2 mM, and the reaction was initiated by the addition of methylmalonyl-CoA to a final concentration of 4.5-5 mM. The change in absorbance at 351 nm was determined between t ϭ 70 and 0 min and was used as a measure of the degree of cofactor inactivation.

Transfer of AdoCbl from ATR to MCM in complex with MeaB
The ATP-dependent transfer of AdoCbl from ATR to the MCM-MeaB⅐GMPPNP complex was performed in the dark at 20°C and monitored by UV/visible spectroscopy. Two equivalents of AdoCbl were added to 1 eq of ATR to generate holo-ATR in Buffer A. The apo-MCM-MeaB⅐GMPPNP complex was reconstituted in Buffer A by mixing 40 -50 M MCM with 50 -60 M MeaB. GMPPNP was added to a final concentration of 1 mM. Holo-ATR (2:1 AdoCbl:ATR trimer) and the MCM-MeaB⅐GMPPNP complex (40 -50 M) were mixed and incubated for 10 min at 20°C before addition of ATP to a final concentration of 5 mM. Release/transfer of AdoCbl from ATR was calculated using ⌬⑀ 525 ϭ 6.69 mM Ϫ1 cm Ϫ1 . Bound versus free cofactor was separated using an Amicon centrifuge filter (10-kDa cutoff; 20 min at 4°C at 16,000 ϫ g). The concentration of free AdoCbl in the filtrate was calculated using ⑀ 525 ϭ 8.0 mM Ϫ1 cm Ϫ1 .

Ejection of Cob(II)alamin by switch I mutants
MCM (30 -40 M) was mixed with 45-60 M MeaB (wild type, D92A, D92N, D105A, or K106A) in Buffer A at 20°C under strictly anaerobic conditions such that the MCM-MeaB ratio was 1:1.5. Cob(II)alamin was generated by reduction of OH 2 Cbl with tris(2-carboxyethyl)phosphine hydrochloride and was added to a final concentration equal to that of MCM. The reaction mixture was incubated for 10 min at 20°C. GMP-PNP in anaerobic Buffer A was added to a final concentration of 2 mM. The mixture was then incubated for 20 min at 20°C. Subsequently, the sample was air-oxidized for 2 h and then applied to a Centricon YM10 filter (10-kDa cutoff) to separate free from bound OH 2 Cbl. Cob(II)alamin (but not OH 2 Cbl) is released from MCM in the presence of MeaB⅐GTP and is subsequently oxidized to OH 2 Cbl, which was quantified using ⑀ 535 ϭ 9.3 mM Ϫ1 cm Ϫ1 .

GTPase activity of MeaB
The GTPase activity of MeaB was determined using an HPLC assay as described previously (2). The activity assays were performed at two GTP concentrations (2.5 and 5 mM), which represent a saturating concentration. The k cat values were determined using 20 M wild-type or mutant MeaB. The k cat values for GTPase activity in MCM-MeaB complexes were determined using 2.5 M MeaB (wild type or mutant) and 10 M MCM.

SEC-MALS analysis of MCM-MeaB complexes
Purified MeaB (25 M; mutant or wild type) and MCM (25 M) proteins in Buffer A were mixed with GMPPNP (250 M) and allowed to equilibrate for 15 min on ice. Next, 50 l of this sample was injected onto a silica-based size exclusion column (model 050S5, Wyatt Technology, Santa Barbara, CA) preequilibrated with 50 mM HEPES, 0.3 M KCl, 3 mM MgCl 2 , pH 8.0, connected to an FPLC system and coupled to a MALS detector (Dawn Heleos II, Wyatt Technology).

Negative-stain electron microscopy
The MCM-MeaB complex sample was taken directly from the SEC-MALS column, diluted 1:30 with the column buffer Allosteric signaling in the MeaB-MCM complex onto carbon-coated grids, and stained with uranyl formate (1%, w/v). Negative-stain micrographs were taken at 52,000ϫ magnification with 2.16 Å per pixel using a 4000 ϫ 4000 chargecoupled device camera (Gatan, Pleasanton, CA) on a G2 Spirit transmission electron microscope (FEI, Hillsboro, OR) operated at 120 kV. From 233 micrographs, 18,747 single particles were selected manually using E2boxer (EMAN2 (33)). The total particle set was classified into reference-free 2D class averages using iterative stable alignment and classification using 20 initial iterations and three-way matching. This process generated 200 stable classes composed of 14,884 single particles, which represented ϳ80% of all particles. Models of 2D projections were generated using UCSF Chimera to position MCM (Protein Data Bank code 1REQ) and MeaB (Protein Data Bank code 4JYB) relative to iterative stable alignment and classification class averages.