Structures of the Human GTPase MMAA and Vitamin B12-dependent Methylmalonyl-CoA Mutase and Insight into Their Complex Formation*

Vitamin B12 (cobalamin, Cbl) is essential to the function of two human enzymes, methionine synthase (MS) and methylmalonyl-CoA mutase (MUT). The conversion of dietary Cbl to its cofactor forms, methyl-Cbl (MeCbl) for MS and adenosyl-Cbl (AdoCbl) for MUT, located in the cytosol and mitochondria, respectively, requires a complex pathway of intracellular processing and trafficking. One of the processing proteins, MMAA (methylmalonic aciduria type A), is implicated in the mitochondrial assembly of AdoCbl into MUT and is defective in children from the cblA complementation group of cobalamin disorders. To characterize the functional interplay between MMAA and MUT, we have crystallized human MMAA in the GDP-bound form and human MUT in the apo, holo, and substrate-bound ternary forms. Structures of both proteins reveal highly conserved domain architecture and catalytic machinery for ligand binding, yet they show substantially different dimeric assembly and interaction, compared with their bacterial counterparts. We show that MMAA exhibits GTPase activity that is modulated by MUT and that the two proteins interact in vitro and in vivo. Formation of a stable MMAA-MUT complex is nucleotide-selective for MMAA (GMPPNP over GDP) and apoenzyme-dependent for MUT. The physiological importance of this interaction is highlighted by a recently identified homoallelic patient mutation of MMAA, G188R, which, we show, retains basal GTPase activity but has abrogated interaction. Together, our data point to a gatekeeping role for MMAA by favoring complex formation with MUT apoenzyme for AdoCbl assembly and releasing the AdoCbl-loaded holoenzyme from the complex, in a GTP-dependent manner.

data point to a gatekeeping role for MMAA by favoring complex formation with MUT apoenzyme for AdoCbl assembly and releasing the AdoCbl-loaded holoenzyme from the complex, in a GTP-dependent manner.
In humans, vitamin B 12 (cobalamin, Cbl) is the cofactor for two enzymes: cytosolic methionine synthase and mitochondrial methylmalonyl-CoA mutase (MUT, 3 also known as MCM), utilizing methylcobalamin (MeCbl) and adenosylcobalamin (AdoCbl) as cofactor forms, respectively. Because of the different subcellular locations of the two enzymes and distinct modifications required to produce usable forms of Cbl, an exquisite subcellular pathway has evolved for the uptake, processing and delivery of Cbl to its destination enzymes. The essential nature of intracellular Cbl processing is underscored by the presence of inherited defects in intermediate metabolic steps, giving rise to the rare metabolic disorders of methylmalonic aciduria and homocystinuria (1). These defects, designated complementation groups cblA-G and mut, represent mutations in genes encoding the Cbl processing and utilization enzymes (2)(3)(4). Among them, three complementation groups, cblA, cblB, and mut, correspond to defects of the mitochondrial enzymes MMAA, MMAB, and MUT, respectively (5)(6)(7). MMAB is an ATP:cob(I)alamin adenosyltransferase (ATR) that transfers 5Ј-deoxyadenosyl from ATP to Cbl forming AdoCbl and delivers it to MUT (5,8). MMAA, whose function is not clearly defined, has been shown to be integral to MUT function and AdoCbl synthesis (9).
MUT catalyzes the AdoCbl-dependent rearrangement of methylmalonyl-CoA (mmCoA) to succinyl-CoA (sCoA). The structure of MUT from Propionibacterium shermanii (psMUT) has been determined (10 -12). It contains an N-terminal TIM barrel accommodating the substrate binding site, a C-terminal AdoCbl binding domain, and an interdomain linker (10). Whereas this work has enhanced enormously our understanding of bacterial MUT structure and function, its application to human MUT (hMUT; 60% sequence identity) is hindered by the different structural configuration of the human enzyme. Unlike psMUT, which is a heterodimer of one catalytic (␣) and one acatalytic subunit (␤) (13), hMUT is a homodimer with two catalytic (␣) subunits per dimer (14).
Likewise, MeaB, the Methylobacterium extorquens ortholog of human MMAA (hMMAA), has also been crystallized (15). MeaB is a homodimer, with the C-terminal ϳ70 amino acids (aa) critical for dimerization. Each subunit contains a central G-domain encompassing the GTPase active site, typical of the G3E family of P-loop GTPases to which it belongs (16). The N-terminal ϳ50 aa extension, whose function is not readily apparent, was predicted to interact with MUT (15). Indeed, in vitro interaction has been demonstrated between M. extorquens MeaB and MUT by native-PAGE and isothermal titration calorimetry (17,18), and between the Escherichia coli orthologs YgfD and Sbm by size exclusion chromatography (19). Additionally, the energetics of the binding interaction were dependent on the liganded state of MUT (apo versus holo) and MeaB (apo or GDP/GTP-bound) (17,18). A role for MeaB has been recently proposed in gating AdoCbl transfer to MUT in order to discriminate against the binding of inactive Cbl forms, in a manner dependent upon GTP binding and hydrolysis (20). These results suggest that the human proteins hMMAA and hMUT might functionally interact in a similar manner.
In this study we determine, by x-ray crystallography, the structures of hMMAA and hMUT. We show that both proteins are homodimeric in native state albeit with different modes of dimeric assembly compared with their bacterial counterparts. We show further that hMMAA and hMUT interact in vitro and in vivo and that this interaction is nucleotide-selective for hMMAA and apoenzyme-dependent for hMUT and is associated with stimulation of hMMAA GTPase activity by hMUT. Finally, we show that complex formation and GTPase stimulation are abrogated in hMMAA with the G188R patient mutation.
Crystallization and Structure Determination-Crystals were grown by vapor diffusion at 20°C in sitting drops containing 50 nl of protein and 100 nl mother liquor for hMMAA, or 100 nl protein and 50 nl mother liquor for hMUT. The mother liquor conditions for hMMAA and hMUT crystals are summarized in Table 1. To crystallize hMMAA, protein (10 mg/ml) was pre-incubated with 1.5 mM GMPPNP. To crystallize hMUT-AdoCbl binary complex, protein (19.6 mg/ml) was preincubated with 250 M AdoCbl. To yield a ternary complex, protein (19.6 mg/ml) was preincubated with 250 M AdoCbl and 2.5 mM mCoA. Crystals were cryo-protected in mother liquor containing 25% (v/v) ethylene glycol and flash-cooled in liquid nitrogen. Diffraction data were collected at the Diamond Light Source beamlines I02 and I03, and processed using the CCP4 program suite (22) ( Table 1). Initial phases were calculated by molecular replacement with PHASER (23) using as search models the structure of M. extorquens MeaB for hMMAA (PDB code 2QM7) (15) or the structure of psMUT for hMUT (PDB code 4REQ) (12). ARP/ wARP (24) was used for automated model building, followed by iterative cycles of REFMAC5 (25) refinement and model building with COOT (26). Calculation of difference Fourier maps revealed unambiguous electron density for AdoCbl and mCoA in the corresponding hMUT complexes (supplemental Fig. S1). For hMMAA, an attempt to co-crystallize GMPPNP resulted in GDP bound in the crystal (supplemental Fig. S1), likely due to the absence of Mg 2ϩ in the crystallization condition, which is required to coordinate with the ␥-phosphate (27). As observed previously (28), GDP may have originated as a contaminant from commercial sources of GMPPNP, reported as only 80% pure. Statistics for data collection and refinement are summarized in Table 1.
Kinetics of hMMAA GTP Hydrolysis-The GTPase activity of hMMAA (20 M) was determined in the presence of various concentrations of GTP (50 -5000 M) by the Enzchek Phosphate Assay (Invitrogen), following manufacturer's instructions and using a POLARstar Omega plate reader (BMG Labtech). Initial velocities for each substrate concentration were calculated and GraphPad Prism 5.0 (GraphPad Software Inc.) was used to calculate the K m and V max by nonlinear regres-sion analysis. k cat was calculated with the formula k cat ϭ V max / [E total ]. The effect of hMUT (apo or holoenzyme) was determined by preincubating the complex (20 M each protein) for 5 min on ice before initiating the GTPase assay.
Immunoprecipitation and Immunoblotting-HepG2 cells were transfected with pCMV-hMMAA-Flag, pCMV-Flag or pCMV-hMMAA G188R -Flag using a nucleofection system (Amaxa), according to the manufacturer's instructions. After 48 h of incubation, cells were treated with 1% paraformaldehyde (PFA) at 37°C for 20 min to cross-link protein complexes in situ, following which the cells were resuspended in lysis buffer (1% Nonidet P-40, 0.5% deoxycholine, 150 mM NaCl, 50 mM Tris-HCl, pH 7.5) for 1 h and then centrifuged at 10,000 ϫ g for 10 min. The supernatant was incubated with anti-Flag antibody for 20 min then incubated with protein-G-agarose (Qiagen) overnight. To recover the pellet, the mixture was centrifuged at 1500 ϫ g for 2 min. The antibody-bound protein was resuspended, heated at 95°C for 20 min to release the protein from cross-linking, and immunoblotted with anti-Flag or anti-hMUT antibodies.

Homodimeric Assembly of hMUT and Comparison to psMUT-
The structure of hMUT has been determined in the apoenzyme at 2.60 Å resolution (hMUT apo ), in binary complex with AdoCbl at 1.95 Å resolution (hMUT holo ), and in ternary complex with AdoCbl and the substrate analog malonyl-CoA (mCoA; lacking a branching methyl group) at 1.95 Å resolution (hMUT ter ) ( Table 1). hMUT crystallized as a homodimer (Fig.  1a), consistent with its solution state (supplemental Fig. S2). Each hMUT monomer features a two-domain structure as observed for the ␣-subunit of psMUT (60% sequence identity), namely a large substrate-binding TIM barrel (N-domain; Fig.  1a, blue) connected to a small AdoCbl-binding domain (C-domain; Fig. 1a, pink) via a ϳ100 aa inter-domain linker (Belt; Fig.  1a, yellow), with the active site situated at the N/C-domain interface. Although hMUT and psMUT monomers superimpose well (rmsd Ͻ 2.0 Å), the two proteins differ substantially in their mode of dimeric assembly. The hMUT homodimer adopts a square cuboidal shape ( Fig. 1, a and c), where the two AdoCbl-binding domains are positioned on the same square face (Fig. 1c, white arrows) with the entrances of the two substrate channels at opposing sides, close to the central grooves at the dimer interface ( Fig. 1c, black arrows). psMUT, by contrast, forms an ␣␤ heterodimer where cofactor and substrate binding regions are found only in the catalytic ␣ subunit but not the acatalytic ␤-subunit ( Fig. 1, b and d). As a result, hMUT has two active sites per dimer, whereas psMUT has only one (10). Furthermore, the psMUT heterodimer is more compact (shorter dimensions at the square face) and has less extensive surface grooves at the subunit interface than the hMUT homodimer (compare Fig. 1, c and d). These observations suggest that hMUT may interact with its functional partner(s) in a different manner than psMUT and other heterodimeric MUT orthologs e.g. M. extorquens MUT.
hMUT Adopts an Induced-fit Mechanism for Binding Cofactor and Substrate-Previously, comparison of the psMUT structures in the holo and ternary states revealed a substrateinduced conformational rearrangement of the protein (11). However, in the absence of an apo structure, it could not be assessed whether cofactor-induced structural rearrangement occurs. Here, the structures of hMUT apo , hMUT holo , and hMUT ter provide snapshots of both the substrate and AdoCbl binding events in the human enzyme. Structural rearrangements upon cofactor binding can be identified by comparison of the hMUT apo and hMUT holo structures, which reveal significant outward displacement (rmsd ϳ2.5 Å) in several C-domain helices (Fig. 2a), while the N-domain remains largely stationary (rmsd ϳ0.6 Å). This creates an induced-fit pocket at the domain interface for binding AdoCbl, by positioning the His-627 imidazole to coordinate with the Cbl cobalt atom, rearranging the Leu-674 and Ala-675 side chains that would otherwise sterically clash with one of the cofactor propionamide tails, and burying the Cbl dimethylbenzimidazole (DMB) tail in a glycine-rich hydrophobic cavity (Fig. 2a). These conformational changes enable the "base-off/His-on" cofactor configuration, where the DMB tail, usually coordinated to the Cbl cobalt, is displaced by His-627 from hMUT, an important criterion for catalysis (10). In addition, the cofactor-binding pocket is shielded from the solvent exterior due to a reordering of the linker belt in hMUT holo (aa 500 -505) that was flexible in hMUT apo (Fig. 2c, top and middle), while the substrate binding channel remains exposed via a long crevice (Fig.  2d, top and middle).
Comparison of the hMUT holo and hMUT ter structures reveal major C␣ differences in the N-domain TIM barrel (Fig. 2b) upon substrate binding, similar to those observed in psMUT. In contrast to the open substrate channel in hMUT apo and hMUT holo , the N-domain TIM barrel tightens and constricts the channel pore in the hMUT ter structure (arrows in Fig. 2b),  thereby sealing off the long crevice from the solvent exterior (Fig. 2, c, bottom and d, bottom). Consequently, this brings the highly-conserved residues lining the channel into close contact with the acyl-CoA substrate (Fig. 2e). Significantly, the phenolic side chain of the invariant Tyr-110 is shifted Ͼ8 Å from its position in the hMUT apo /hMUT holo structures and, together with the invariant His-265 and Arg-228, positions the substrate carboxyl for catalysis. The hMUT holo and hMUT ter structures also differ in the orientation of the 5Ј-deoxyadenosyl moiety of the AdoCbl cofactor, which is covalently bonded (via the adenosyl methyl carbon) to the cobalt atom in the hMUT holo structure, but in the hMUT ter structure the C-Co bond is broken, thereby generating the cob(II)alamin radical, and displacing the 5Ј-deoxyadenosyl group to adopt at least two alternative, less-ordered conformations in proximity to the substrate ( Fig. 2e and supplemental Fig. S3). Therefore, substrate binding induces structural rearrangement in the cofactor as well as the protein.
Structure of Nucleotide-bound hMMAA-The assembly of the AdoCbl cofactor into hMCM is predicted to involve a GTP-dependent interaction with hMMAA, a member of the G3E family of P-loop GTPases, based on the properties of their bacterial orthologs (20). To provide insight into the role of human MMAA, we have determined the structure of GDP-bound hMMAA at 2.5 Å resolution (Table 1). hMMAA crystallized as a homodimer, in line with solution studies (supplemental Fig. S2). Each hMMAA monomer (Fig. 3a) consists of a central 7-stranded G-domain (aa 143-330) harboring the four GTPase signature motifs (namely Switch I, Switch II, P-loop, and base specificity loop), flanked by an N-terminal extension (aa 75-142) and a C-terminal dimerization arm (aa 331-416). Several loops connecting the Gdomain ␤-strands are disordered, suggesting their intrinsic flexibility. These include aa 182-197 (L1, Switch I motif), aa 212-220 (L2), aa 246 -248 (L3, part of Switch II motif), and aa 268 -279 (L4). The bound GDP was anchored in a surface groove via polar interactions with highly conserved sequence motifs in the base specificity loop (Lys-290, Asp-292) and P-loop (Lys-156, Ser-157, Thr-158) (Fig. 3b). In addition, hMMAA contributed a non-conserved arginine (Arg-330; Leu-243 in MeaB) to line a pocket for the guanine moiety which interacts with the hydroxyl group of the GDP-ribose. No Mg 2ϩ ion was observed in the nucleotide pocket, suggesting that GDP binding is metal-independent.
hMMAA Dimer Has a More Open Configuration than MeaB-The topology of hMMAA is similar to that of the previously reported M. extorquens MeaB structure (15), as monomers of the two proteins can be superimposed with an rmsd of 2.78 Å (Fig. 4a). Like MeaB, hMMAA crystallizes as a homodimer, but the relative placement of the monomers differs between the two proteins. In MeaB, helix ␣12 from the C-terminal dimerization arm forms a ϳ30 o bend at residue Ala-282, midway along the helix, while helix ␣12 in hMMAA GDP is much less distorted (Ͻ5 o bend in the opposite direction at the equivalent residue Asn-363). As a result, the two helices following, ␣13 and ␣14, are displaced by Ͼ10 Å away from the protein core in hMMAA compared with the equivalent helices in MeaB. In both structures, the dimerization arm (␣12-␣14) is closely intertwined with its equivalent arm from the neighboring subunit, and therefore the relative displacement of helices ␣13 and ␣14 impacts directly on the overall shape and dimensions of the resultant dimer. Significantly, superimposing the hMMAA and MeaB dimers using one monomeric subunit reveals dramatically different positional arrangements for the second subunit (supplemental Fig. S4). Consequently, hMMAA GDP dimerizes via the dimerization arms only (Fig. 4b, top) while the central G-domains from both subunits are oriented away from each other (Fig. 4b, bottom), hence adopting an open U-shaped configuration. By contrast, MeaB dimerizes using the dimerization arms and G-domains from both subunits (Fig. 4c), which results in a more closed configuration with a central cavity.
hMMAA Is a P-loop GTPase Modulated by hMUT-The presence of GTPase signature motifs in the central, G-domain of hMMAA suggests that it binds and hydrolyzes GTP. To assess its putative GTPase activity, we incubated recombinant hMMAA with increasing concentrations of GTP and monitored the time-dependent increase in P i produced. Michaelis-Menten kinetics revealed an unexpectedly high K m of 1210 Ϯ 330 M, along with the more expected catalytic constant (k cat ) of 0.030 Ϯ 0.003 min Ϫ1 (Table 2) (Fig. 5a), the latter consistent with the bacterial ortholog MeaB (k cat ϭ 0.039 min Ϫ1 ) (18). Interestingly, when assayed in the presence of hMUT apo , the catalytic efficiency (k cat /K m ) of hMMAA increased more than 80-fold, due to both tighter binding (K m ϭ 74 Ϯ 8 M) and increased activity (k cat ϭ 0.15 Ϯ 0.01 min Ϫ1 ) (Table 2) (Fig. 5a), indicating a functional association between hMMAA and hMUT. Surprisingly, the addition of hMUT holo resulted in a  DECEMBER 3, 2010 • VOLUME 285 • NUMBER 49 less pronounced increase in the catalytic efficiency of hMMAA, by only 8-fold (Table 2) (Fig. 5a), suggesting that the cofactor state of hMUT directly influences its interaction with hMMAA. Additionally, we tested hMMAA G188R , a recombinant hMMAA variant mimicking a homozygous mutation in a cblA patient (29). The GTPase activity of hMMAA G188R alone   was similar to wild-type levels; however, unlike wild-type hMMAA, it was not stimulated by the addition of hMUT apo ( Table 2), suggesting the G188R mutation may interfere with the functional interaction between hMMAA and hMUT. Physical Interaction of hMMAA-hMUT-To test whether hMUT and hMMAA interact directly, analytical size-exclusion chromatography was employed (Fig. 5b). Both hMUT and hMMAA migrated independently with apparent MW of 196 kDa and 78 kDa, respectively, consistent with homodimeric assembly in each case (supplemental Fig. S2). A mixture of hMUT apo and hMMAA preincubated with GMPPNP eluted with early, higher-MW species in addition to the individual protein dimer peaks (Fig. 5b, hMMAA GMPPNP :hMUT apo ). The early species consist of two subpopulations, peaks 1 and 2, corresponding to 1410 kDa and 740 kDa, with a leading shoulder suggesting still higher MW species. SDS-PAGE analysis confirmed co-elution of hMUT and hMMAA from these peaks (data not shown), indicating complex formation. The presence of individual dimer peaks eluting after the high MW species suggests an equilibrium between the complexes and their protein components. Interestingly, the complexes require pre-bound GTP/GMPPNP for their formation, because in the absence of nucleotide (data not shown) or when a mixture of hMUT apo and hMMAA preincubated with GDP was analyzed, no early peaks were observed (Fig. 5b, hMMAA GDP : hMUT apo ). Additionally, no complex was visualized when AdoCbl was pre-bound to hMUT and mixed with hMMAA GMPPNP (Fig. 5b, hMMAA GMPPNP :hMUT holo ), suggesting that hMUT holo interacts less favorably with hMMAA than hMUT apo , a finding consistent with our GTPase data. Finally, while hMMAA G188R migrated independently as a dimer similar to wild-type (supplemental Fig. S2), it did not form a complex with hMUT apo after pre-incubation with GMPPNP (Fig. 5b, hMMAA G188R,GMPPNP :hMUT apo ).

Structures and Interaction of Human MMAA and MUT
Interaction between hMMAA and hMUT in vivo was confirmed by detection of hMUT following immunocapture of Flag-tagged hMMAA (hMMAA-Flag) from hMMAA-Flagtransfected HEPG2 cells (Fig. 5c). Cell lysates were treated with a cross-linking agent, immunoprecipitated with anti-Flag antibody, and then immunoblotted. A band corresponding to hMUT was detected in cells expressing hMMAA-Flag (lane 1), but not in Flag-only cells (lane 2) or cells expressing hMMAA G188R -Flag (lane 3). Together, these data demonstrate that hMMAA and hMUT form a complex in vivo and in vitro, which is abolished with the G188R mutation in hMMAA.

DISCUSSION
Despite having only two Cbl-dependent enzymes, mammals have deployed a complex pathway for the cellular uptake and assembly of Cbl into its destination enzymes. However, the biochemical properties of some individual components in this complex pathway and the molecular basis underlying their enzymatic functions are not well understood. Previous work on bacterial orthologs, including MUT from P. shermanii and MUT and MeaB from M. extorquens, have become the paradigm for predicting the biochemical behavior and interaction of human MUT and MMAA. In this study, we have examined the structures of hMUT and hMMAA, demonstrating a nucleo-tide-dependent interaction of the two proteins and a gating behavior of hMMAA toward cobalamin binding of hMUT, and confirm the occurrence of a protein complex in cells in vivo but not in a GTPase-active mutant protein carrying the G188R substitution. We further demonstrate unique structural and biochemical differences between these two human proteins and their bacterial counterparts, which have mechanistic implications for their functional interaction and role in cofactor assembly.
Human MMAA-MUT Interaction Is Ligand-dependent and Ligand-selective-Consistent with the presence of GTPase signature motifs and enzyme activities of the M. extorquens MeaB (17,18) and E. coli YgfD (19) orthologs, our GTP hydrolysis assay demonstrated moderate GTPase activity for hMMAA. This activity is enhanced by hMUT, as observed for MeaB (18), but unlike MeaB, it is enhanced in a cofactor-dependent manner with 80-fold stimulation by hMUT apo versus 8-fold by hMUT holo . The modulation of hMMAA activity by hMUT confirms a functional association between these two proteins. Further, by size exclusion chromatography, a stable complex of recombinant hMMAA and hMUT could be detected in a nucleotide-selective (GMPPNP-but not GDP-bound) and apoenzyme-dependent manner. Pre-binding AdoCbl to hMUT also interfered with complex formation even with GMPPNP-bound hMMAA. This interaction was confirmed in vivo by the ability to co-immunoprecipitate the two proteins from fibroblast homogenates treated with a cross-linking agent. Complex formation has been reported for the orthologous E. coli pair (YgfD-Sbm) using size-exclusion chromatography and similar co-immunoprecipitation from cell extracts (19) and for the M. extorquens pair (MeaB-MUT) using native-PAGE (17,18). However, the preference for the hMUT apoenzyme over holoenzyme in forming the stable human complex was not observed with the bacterial orthologs and is in keeping with the more pronounced enhancement of GTPase activity by hMMAA in the presence of hMUT apo compared with hMUT holo . The nucleotide-selective nature of hMMAA in the interaction assay, also observed in the E. coli YgfD-Sbm complex (19), indicates a dependence on GTP binding or hydrolysis for complex formation, and may be used to target hMMAA to the hMUT apoenzyme in preference to the already functional AdoCbl-loaded holoenzyme. Our data are in keeping with the recently proposed GTP-dependent "gating" role for MeaB in screening for the active AdoCbl cofactor for incorporation into MUT (20). Therefore, the unique nucleotide-binding prerequisite to the formation of the hMMAA GMPPNP -hMUT apo complex may be relevant physiologically to the acquisition of AdoCbl by the MUT apoenzyme to form the holoenzyme in the human system.
hMMAA and hMUT Dimeric Assembly Has Implications for Their Interaction-As an initial step toward structurally characterizing an hMMAA-hMUT complex, we have crystallized the individual proteins hMMAA and hMUT in various liganded forms. The two proteins show a strong evolutionary conservation between the bacterial and human proteins in all aspects of their structural folds, domain architecture and catalytic machinery (10,15), consistent with their moderate sequence homology (ϳ50%) and, for the MUT enzymes, the preservation of an induced-fit mechanism to communicate the binding of cofactor and substrate with the active site catalytic machinery. Nevertheless, inspection of the hMMAA and hMUT homodimers, which had been expected to be their biological oligomeric state (30,31), reveals significantly different modes of dimeric assembly compared with the corresponding bacterial proteins.
The different hMMAA/MeaB dimeric configurations are the result of relative displacements in their dimerization arms, a polypeptide region less conserved at the sequence level and distant from the G-domain, and hence unlikely to be nucleotide-dependent. Consequently, the more open hMMAA dimer positions its two nucleotide-binding sites more closely to each other (ϳ20 Å) than does the MeaB dimer (50 Å). Also, the hMMAA dimer exposes a much larger surface at the dimer interface than the MeaB dimer with its more constricted central cavity. For MUT, the different dimeric configurations between the human and P. shermanii structures are primarily due to subunit compositions. The psMUT ␣␤ heterodimer (13) harbors only one active site contributed from the catalytic ␣ subunit. This contrasts with the hMUT ␣ 2 homodimer, shown structurally as a square cuboid, harboring two active sites (one from each ␣ subunit) positioned on the same cuboid face. The homodimeric arrangement is also predicted for E. coli (32) and mouse (33).
These structural configurations lead to predictions about the nature of the oligomeric complexes formed between hMMAA and hMUT. Our hMMAA GMPPNP -hMUT apo complex from size exclusion chromatography may represent supramolecular species built from the core structure of 2 hMMAA dimers: 1 hMUT dimer (MW ϳ352 kDa). Hence the 740 kDa and 1410 kDa peaks might represent 2 units (calculated mass 704 kDa) and 4 units (1408 kDa) of the core structure, respectively. The 2:1 stoichiometry is compatible with our structural data, where the nucleotide binding sites and Switch (I and II) motifs from both subunits of the hMMAA dimer are juxtaposed close enough in the central cavity to be contacted simultaneously by one subunit of the hMUT dimer. This differs from the complex formed by M. extorquens MUT and MeaB, which saturated at a ratio of 1 MeaB dimer:1 MUT dimer (18). It is not unexpected that the bacterial orthologs may adopt a different stoichiometric relationship and binding mode, because M. extorquens MUT is a heterodimer and has a smaller accessible surface than hMUT (ϳ45,000 versus 53,000 Å 2 ) and requires only 1 mol of AdoCbl. The nature of the supramolecular complex formation is currently under investigation by co-crystallization and electron microscopy studies.
In the mitochondrial pathway for Cbl assimilation, AdoCbl synthesized by the MMAB adenosyltransferase from cob(I-I)alamin and ATP is transferred to MUT in a manner that involves, but has left enigmatic, a role for MMAA (34,35). In the M. extorquens MeaB system, AdoCbl is transferred directly to MUT without release into the milieu. When GTP-bound MeaB is complexed with MUT, the selectivity for AdoCbl becomes very high, predicting a gating role for MeaB. Our hMUT and hMMAA dimeric structures, combined with the high-MW complexes formed in vitro and demonstrated to occur in vivo, raise similar questions regarding the stoichiome-try between the hMUT dimer and its two functional partners. Human MMAB forms a homotrimer (36) that binds 2 mol of Cbl and 2 mol of ATP (37), and hence one trimer has the capacity to deliver 2 mol of AdoCbl to one hMUT homodimer. It is not known whether the delivery and/or loading of 2 mol of AdoCbl per hMUT homodimer requires one or two hMMAA dimer(s), although the 2 hMMAA: 1 hMUT core structure predicted from size exclusion chromatography suggests that each hMUT subunit is handled independently.
A Patient Mutation in hMMAA Abolishes Its Interaction with hMUT-To establish a physiological relevance for our observed hMMAA-hMUT interaction, we searched for patient mutations in hMMAA that abolish complex formation. To date, eight hMMAA missense mutations have been reported (6,29,38,39), all of which were characterized biochemically in MeaB (15) except the recently identified homozygous G188R mutation (29). Unlike other hMMAA mutations shown to affect protein stability (15), hMMAA G188R can be expressed recombinantly as a soluble homodimer, and exhibits near wild-type GTPase activity, suggesting the mutation has no deleterious effects on its structure and stability. However, the hMMAA G188R GTPase activity is neither enhanced by hMUT nor can the mutant protein complex with hMUT, demonstrating that the mutation affects directly the robustness of its interaction with hMUT. Gly-188 is part of the Switch I motif in the G-domain, located at the dimer interface, and is disordered in our hMMAA structure. Switch I and II are signaling motifs that transduce GTP binding/hydrolysis with downstream effector proteins (40), so that their conformations may depend on whether hMMAA is bound with GTP and/or hMUT. Therefore, our analysis of the G188R mutant, the first characterized to affect hMMAA-hMUT interaction, extends previous suggestions that the MeaB N-terminal extension binds MUT by demonstrating here that the G-domain and the dimer interface of hMMAA are also involved in the interaction, and underscores the importance of the Switch motifs in the functional interplay between hMMAA and hMUT.
Proposed Model for hMMAA-mediated Cofactor Assembly into hMUT-Our studies lead to a model for the role of hMMAA in maintaining the functional integrity of hMUT. As proposed by Banerjee and co-workers (1,20), GTP-bound MeaB distinguishes the transfer of authentic AdoCbl from ATR (bacterial ortholog of human MMAB) to MUT apoenzyme (gate-keeping) and screens for and removes oxidized Cbl from MUT (editing) in a GTP-dependent manner. Our results point to an evolutionary retention of a gate-keeping role for hMMAA, but with some significant differences. We speculate that, in the human system, AdoCbl binding releases hMUT from the hMMAA-hMUT complex in a GTP-dependent manner. Our results also suggest an importance of supramolecular complexes in this process, though at this time, their significance remains unclear. However, the pathological outcome of the G188R patient mutation underscores the essential role of these protein complexes in generating or maintaining functional hMUT. The benefit of the higher order complexes might be to sequester hMUT apoenzyme in proximity of an MMAB/ATR factory that would convert incoming Cbl to AdoCbl for concerted transfer to the hMMAA-hMUT complex. This would minimize loss of AdoCbl to the mitochondrial milieu and would allow hMMAA to perform the gate-keeping function predicted for MeaB. This model, with release of functional hMUT holoenzyme after cofactor assembly, would be less attuned to hMMAA-mediated surveillance on the functional state of cofactor bound to MUT, as proposed for the MeaB-MUT system. It will be interesting to see if this editing function identified for MeaB extends to the human enzyme.