Structural Insights into the MMACHC-MMADHC Protein Complex Involved in Vitamin B12 Trafficking

Background: Two intracellular proteins, MMACHC and MMADHC, functionally interact for cobalamin trafficking. Results: MMADHC crystal structure reveals protein-interacting regions and unexpected homology to MMACHC; mutations on either protein interfere with complex formation via different mechanisms. Conclusion: Complex formation likely depends on prior cobalamin processing and can be broken by disease mutations. Significance: MMACHC-MMADHC heterodimerization forms the essential trafficking chaperone delivering cobalamin to client enzymes.

a combined HC and MMA defect (HCϩMMA) (4). MMACHC and MMADHC are two proteins involved in the early Cbl processing steps that are shared by both mitochondrial and cytosolic targeting routes. MMACHC gene mutations are the most common cause of Cbl metabolic disorders, resulting in HCϩMMA (8). Genetic defects of MMADHC, by contrast, uniquely cause any of the above three phenotypes (9,10).
A possible explanation for the heterogeneous MMADHC phenotypes is the presence of distinct functional domains at the MMADHC protein level, responsible for trafficking to either mitochondrial MUT (of AdoCbl) or cytosolic MS (of MeCbl), or both. This hypothesis is supported by analysis of metabolic phenotypes from known disease mutations (11) and alaninescanning mutagenesis (12). These studies dissect the 296amino acid (aa) human MMADHC polypeptide functionally into the N-terminal 115 aa that is required for the mitochondrial route but dispensable for cytosolic trafficking, the Cterminal 180 aa that contribute to both routes, and two regions therein (aa 197-226 and 246 -259) that have greater influence on the cytosolic than the mitochondrial route. Current knowledge points to a role of MMADHC in regulating Cbl delivery at the branch point between the mitochondrial and cytoplasmic target enzymes, albeit via an incompletely understood mechanism.
Furthermore, despite the MMADHC gene discovery seven years ago (10), few of the biochemical properties and functions of the MMADHC protein are known. Although previously postulated to contain Cbl and ATP binding motifs (10), MMADHC has been shown not to bind Cbl (13) or to hydrolyze ATP (14). Further biochemical exploration has been hampered in part by a lack of structural knowledge or sequence homologues. The inability of MMADHC to bind Cbl has implicated protein-pro-tein interaction as the means of its Cbl-targeting role. This is supported by recent evidence from phage display (13,15), bacterial two-hybrid studies (15), native PAGE (14), and surface plasmon resonance (13,15) that MMADHC interacts directly with MMACHC in vitro (13,15), and the interaction does not require the N-terminal 115 aa of MMADHC (14). Because MMACHC has been shown structurally (16,17) and biochemically (18,19) to bind various Cbl forms, and to process the upper Cbl ligand by reductive decyanation (19) or dealkylation (20) using the flavin cofactors FMN/FAD and GSH, respectively, it seems plausible that MMACHC-bound Cbl can partner with MMADHC to be ferried to the two destination enzymes.
To evaluate the importance and molecular basis of the MMACHC-MMADHC interaction, we have defined a minimal MMACHC interaction module of MMADHC, and demonstrated that experimental and patient missense mutations disrupt this interaction. We have further determined the structure of this interaction module from mouse MMADHC to 2.2 Å resolution, and established the 1:1 stoichiometry of the MMACHC-MMADHC heterodimer using small angle x-ray scattering (SAXS). Finally, based on the combined findings, we propose the first structural model of the MMACHC-MMADHC interaction.

Experimental Procedures
Recombinant Production of MMACHC and MMADHC-DNA fragments encoding human (Homo sapiens, IMAGE clone: 3826071) and mouse (Mus musculus, IMAGE clone: 3493526) MMADHC, harboring different N-and C-terminal boundaries, were amplified and subcloned into pNIC28-Bsa4 vector (GenBank TM accession number EF198106) in-frame with a tobacco etch virus protease cleavable N-terminal His 6 tag. Constructs of human HsMMACHC in pNIC28-Bsa4 vector were prepared previously (16,21). Site-directed mutations were constructed using the QuikChange mutagenesis kit (Stratagene) and confirmed by sequencing. Cloning and site-directed mutagenesis primers are available upon request. Proteins were expressed in Escherichia coli BL21(DE3)R3 and purified by affinity (Ni-Sepharose; GE Healthcare) and size-exclusion (Superdex 200; GE Healthcare) chromatography. For crystallization, MMADHC proteins were further purified by ion exchange chromatography (Resource Q; GE Healthcare). Selenomethionine (SeMet)-derivatized proteins were expressed using SelenoMethionine Medium Complete (Molecular Dimensions) and purified as above.
Crystallization and Structural Determination-Purified SeMet-derivatized and native MmMMADHC ⌬128 were concentrated to 15-20 mg/ml, and crystals were grown by sitting drop vapor diffusion at 4°C. The mother liquor conditions are summarized in Table 1. Crystals were cryo-protected in mother liquor containing ethylene glycol (25% v/v) and flash-cooled in liquid nitrogen. X-ray diffraction data were collected at the Diamond Light Source beamline I04 and processed using XIA2. Selenium atoms were located using SHELXC/D (22) for initial phase calculation in SHELXE, and subsequently for automated building with BUCCANEER (23). The structure was solved by selenium single-wavelength anomalous diffraction phasing and refined using PHENIX (24), followed by manual rebuilding in COOT (25). The SeMet MMADHC model was used to solve the native structure by molecular replacement using PHASER (26). The final refined model consists of aa 132-157, 169 -236, and 246 -296 of MmMMADHC. Atomic coordinates and structure factors for MmMMADHC ⌬128 have been deposited in the Protein Data Bank with the accession code 5A4R. Data collection and refinement statistics are summarized in Table 1.
Small Angle X-ray Scattering-For SAXS, MMACHC and MMADHC were cleaved by tobacco etch virus protease to remove the N-terminal His 6 tag, followed by reverse affinity purification. Purified proteins were then incubated separately or together with the appropriate ligands and further purified by size-exclusion chromatography. Concentrations of proteins used were 24 mg/ml (MMADHC), 18 mg/ml (MMACHC), and 17 mg/ml (complex). Scattering data were collected at the Diamond Light Source B21 beamline. Data collection was performed either in-line with size-exclusion chromatography (KW404 column, Shodex) or in batch mode where the peak fractions were collected and immediately subjected to SAXS by flowing sample through an in-vacuum quartz capillary of 1.6-mm diameter. Data were collected using a Pilatus 2M detector (DECTRIS, Baden, Switzerland) at a sample-detector distance of 3914 mm and a wavelength of ϭ 1 Å. The range of momentum transfer of 0.1 Ͻ s Ͻ 5 nm Ϫ1 was covered (s ϭ 4sin/, where is the scattering angle). For the in-line mode, 1-s exposures were collected, whereas for batch samples, a comparison of eighteen 10-s exposures was performed. Radiation damage was checked for the batch mode by monitoring changes in radius of gyration in each frame, where no significant changes were observed. The data were radially averaged, and the scattering of the buffer was subtracted. The forward scattering I(0), radius of gyration R g , pair distribution of the particle P(r), and maximum dimension D max were analyzed using Scatter (27) and the ATSAS suite of programs (28).
Solution Characterization of MMACHC and MMADHC-Blue native-PAGE was performed with 25 M protein (MMACHC and/or MMADHC) alone or in the presence of 50 M cobalamin (MeCbl, AdoCbl, CNCbl) and/or 8 mM ligand (GSH, FMN, FAD), which was loaded onto the native-PAGE gel system, after preincubation in the dark at room temperature for 1 h, and then run following the manufacturer's instructions (Life Technologies). All blue native-PAGE experiments were performed at least twice independently. Analytical gel filtration (16), FMN/FAD binding by intrinsic fluorescence quenching (17), and differential scanning fluorimetry (21,29), were performed as described previously. For native mass spectrometry, 75 l of sample containing ϳ4 mg/ml MMADHC and MMACHC incubated in the presence of 0.5 mM MeCbl and 2 mM reduced GSH underwent four desalting steps using Micro Bio-Spin columns (Bio-Rad) that were pre-equilibrated with 50 mM ammonium acetate buffer (pH 6.5). Following desalting, samples were loaded into a 1.0-ml gas-tight positive displacement syringe (Hamilton) that was inserted into the syringe pump. Samples were then directly infused with a constant flow rate of 6 l/min through a PEEK capillary tubing (inner diameter, 0.005 inches) into a Q-TOF 6530 mass spectrometer attached to a standard electrospray ionization source (Agilent Technologies). The mass spectrometer was operated in positive ion mode, using the 1-GHz detector mode with a scan range of 100 -20,000 m/z and a fragmentor voltage of 430 V. Following data acquisition, results were evaluated by using the Mass-Hunter Qualitative Analysis software (Agilent Technologies).

MMADHC aa 154 -296 Is an MMACHC Interaction
Module-To better understand the required regions, ligand dependence, and stoichiometry of their in vitro interactions, we generated a series of human (Hs) MMADHC and MMACHC truncation proteins by recombinant expression (Fig. 1A). Using blue native-PAGE (BN-PAGE) (Fig. 1B), full-length MMACHC (HsMMACHC FL ) alone migrated as two bands (lane 1), consistent with its known monomeric and dimeric forms, whereas the dimeric band became more prominent in the presence of MeCbl and GSH (lane 2, white dot). The HsMMADHC proteins with various N-terminal truncations migrated as single bands (lanes 3-8) corresponding to their respective monomers, corroborating previous investigations (13). Complex formation, as judged by the appearance of an additional band in BN-PAGE as compared with single-protein controls, was observed when HsMMACHC FL preincubated with MeCbl and GSH was added to HsMMADHC ⌬61 or HsMMADHC ⌬123 , or to a lesser extent, HsMMADHC ⌬153 (lanes 9 -11, white asterisk). However, no complex band was observed with N-terminal truncation of HsMMADHC beyond aa 154 (HsMMADHC ⌬157 , HsMMADHC ⌬167 , or HsMMADHC ⌬172 ; lanes [12][13][14] under the same conditions. The HsMMADHC-HsMMACHC complex bands migrated at a position intermediate of the estimated monomeric and dimeric HsMMACHC bands, suggesting a possible 1:1 heterodimeric complex. HsMMACHC ⌬C , which lacks the C-terminal Pro-rich region (aa 236 -282; Fig. 1A), was sufficient for complex formation with HsMMADHC ⌬123 (Fig. 1C, lane 2). Size-exclusion chromatography using MmMMADHC ⌬128 and HsMMACHC ⌬C proteins further confirmed the regions required for interaction and the 1:1 stoichiometry of the complex (Fig. 1D). Therefore, the MMADHC region C-terminal to aa 154 and the MMACHC region without the Pro-rich C terminus is sufficient for direct protein-protein interaction. Notably, the MMADHC C-terminal region is the most evolutionarily conserved section of the polypeptide from human to Caenorhabditis elegans (Fig. 1E), suggesting selected retention of this interaction module in evolution.
The MMACHC-MMADHC interaction is further studied by native mass spectrometry, where the complex of HsMMACH-C FL and HsMMADHC ⌬123 , incubated in the presence of MeCbl and GSH, generated two peaks consistent with a 1:1 stoichiometry, one missing the 15 atomic mass units of the methyl group (Fig. 1G, 52,875 atomic mass units) and the other missing 16 atomic mass units, corresponding to the methyl group plus the hydrogen from GSH (Fig. 1G, 53,181 atomic mass units). Thus, R-group removal appears to be a prerequisite for MMACHC-MMADHC interaction.
MMADHC Contains a Nitroreductase-like Fold with Homology to MMACHC-We next characterized MMADHC by x-ray crystallography to elucidate the molecular properties of its MMACHC interaction module. To maximize the probability of crystallization, we adopted a cross-species approach that involved the study of recombinant human, M. musculus, and Xenopus laevis MMADHC proteins, coupled with an extensive survey of construct boundaries. From Ͼ40 purified proteins, we successfully crystallized M. musculus MMADHC aa 129 -296 (MmMMADHC ⌬128 ; Figs. 1A and 2A), a region bearing 93% identity with the human sequence. In our BN-PAGE assay, MmMMADHC(e.g. MmMMADHC ⌬128 ) can substitute the equivalent human construct in stable complex formation with HsMMACHC (not shown). Despite repeated attempts, the equivalent HsMMADHC construct did not crystallize. The structure of MmMMADHC ⌬128 (Fig. 2, A and B), determined to 2.6 Å resolution by single-wavelength anomalous dispersion phasing for the selenomethionine (SeMet) derivative, and to 2.2 Å by molecular replacement for the native protein (Table 1), is a mixed ␣/␤ domain composed of a central four-stranded antiparallel ␤-sheet (␤1, ␤2, ␤6, ␤3) flanked by a 14-turn helix (␣B) and a short ␤-turn (␤4-␤5) (Fig. 2C).
A structural comparison of MmMMADHC ⌬128 with other proteins using DALI (33) revealed that the core structure is derived from the nitroreductase (NTR) fold (Fig. 3A), primarily found in flavoenzymes and oxidoreductases that utilize FMN/ FAD cofactors and form homodimers by ␤-strand exchange, as well as MMACHC (Fig. 3B). Nevertheless, MmMMADHC ⌬128 has extensively modified the NTR fold with three loop protrusions (Pr1-Pr3), rendering it functionally different from classical NTRs. Firstly, MmMMADHC ⌬128 does not bind FMN/FAD in solution (data not shown) because the canonical binding site found in NTRs is disrupted by part of the MMADHC Pr2 loop protrusion (Fig. 4A). Secondly, although in crystallo MmMMADHC ⌬128 reveals a symmetry-related dimer, its dimeric interface does not match that of the conventional NTRs (Fig. 4B). Indeed, MmMMADHC ⌬128 is a monomer in solution by BN-PAGE, size-exclusion chromatography (not shown), and SAXS (see "A Structural Model of the MMACHC-MMADHC Complex"), consistent with HsMMADHC constructs migrating as monomers in BN-PAGE (Fig. 1B).
To our surprise, the closest structural homologue of MMADHC is its interaction partner MMACHC (DALI (33) Z-score, 5.2), despite a lack of obvious sequence conservation. Like MMADHC, MMACHC contains three loop protrusions, at equivalent spatial positions, in the core NTR fold (16, 17) (Fig. 3B). MMACHC also lacks the classical FMN binding site present among NTRs, while retaining the ability to bind flavins in solution (17), likely involving a non-canonical binding site (16). The MMADHC and MMACHC structures superimpose only moderately and within the core NTR secondary structure elements (C ␣ -r.m.s.d. 3.0 Å, 96 aligned aa), beyond which they differ substantially (Fig. 3B). For example, MmMMADHC ⌬128 does not contain the four-helix cap domain and differs in length and conformation of two protrusions as compared with MMACHC, all of which form part of the MMACHC Cbl binding pocket (16,17). These extensive structural differences from MMACHC likely account for the reported lack of Cbl binding capability of MMADHC (13). We additionally found no sequence elements or homology consistent with ATPase activity. (11,12) revealed specific MMADHC  regions responsible for the intracellular targeting of the two cobalamin cofactors. These mutations correlate with the MMA, HC, or HCϩMMA phenotypes based on their amino acid location (Fig. 5A). Missense mutations causing decreased MeCbl production (HC phenotype) or both AdoCbl and MeCbl production (HCϩMMA phenotype) in cells are found within three sequence stretches at the MMADHC C-terminal half (Fig. 5A, stretches 1-3). When mapped onto our MmMMADHC ⌬128 structure, these three stretches cluster in spatial proximity and are surface-accessible (Fig. 5B). We investigated whether these mutations impact on the MMACHC-MMADHC interaction by introducing onto the HsMMADHC ⌬123 construct a selection of HC and HCϩMMA causing mutations from the three stretches (Stretch 1: p.T182N, p.M186A, p.W189A; Stretch 2: p.D226A; Stretch 3: p.L259P). The recombinant mutant proteins behaved similarly to wild type in terms of expression level, protein solubility (not shown), and thermal stability (Fig. 5C), where only p.D226A led to slightly unstable protein. We found interaction of HsMMACHC FL (incubated with MeCbl and GSH) with HsMMADHC ⌬123-T182N and HsMMADHC ⌬123-W189A to be severely decreased (Fig. 5D, lanes 3 and 5), but retained for HsMMADHC ⌬123-M186A , HsMMADHC ⌬123-D226A , and HsMMADHC ⌬123-L259P (Fig. 5D, lanes 4, 6, and 7). These findings suggest that residues within Stretch 1 (Thr-182, Trp-189) occur at the binding interface with MMACHC.

Patient and Experimental MMADHC Mutations Knock Out Interaction with MMACHC-Our previous mutagenesis studies in patient fibroblasts
A Structural Model of the MMACHC-MMADHC Complex-We next utilized SAXS, a low-resolution structural characterization method in solution (34), to construct a model of the MMACHC-MMADHC interaction. To ensure that the SAXS data most closely resembled the ordered aa regions revealed from the MMACHC and MMADHC structures, we reconstituted a complex from untagged MmMMADHC ⌬128 and untagged HsMMACHC ⌬C proteins in the presence of MeCbl and GSH. The Guinier approximation and pair distribution function P(r) calculated from the scattering data (Fig. 6, A and  B) suggest a radius of gyration (R g ) of 24.7 Å and maximal intraparticle dimension (D max ) of 80 Å for the complex, both parameters larger than those from either protein component alone (Fig. 6C). Further, the elution time of the protein complex from the in-line HPLC column was earlier than either protein alone,  giving rise to a calculated molecular weight consistent with a 1:1 stoichiometry (Fig. 6C). The ab initio envelope of the MmMMADHC ⌬128 -alone sample fits well to one monomer (Fig. 6D). MMACHC is known to exist in a monomer:dimer equilibrium ( Fig. 1) (16), and the monomeric species selected for SAXS analysis is consistent with one protomer in the ab initio envelope (Fig. 6E). However, for the MMACHC-MMADHC complex, the ab initio envelope (Fig. 6F) gives an elongated dimension of 80, 40, and 40 Å. One protomer from each of HsMMACHC ⌬C and MmMMADHC ⌬128 would account for all the mass of the envelope, further indicating a 1:1 heterodimeric stoichiometry that agrees with our BN-PAGE, size-exclusion chromatography, and native mass spectrometry data (Fig. 1). Although the relative orientation of both proteins cannot be ascertained by SAXS, we applied our mutagenesis and interaction data (Fig. 5) to direct protein docking of the complex using the ZDOCK server (35), given that residue Thr-182 of MmMMADHC should be involved in the interaction. With this constraint, the top seven models generated from ZDOCK (Fig. 6F) all yield a good fit ( 2 ϳ0.03-0.04) to the measured scattering. These models are consistent with an MMACHC-MMADHC interface that involves the Cbl binding region of MMACHC and aa 182-189 (Fig. 5, stretch 1) of MMADHC (Fig. 6G). A similar docking run was performed with HADDOCK (36), yielding consistent models (data not shown).

Discussion
The role of MMACHC within the early stages of the Cbl processing pathway has previously been clarified from the available structural and biochemical evidence of its Cbl binding and processing activities (16 -20). By contrast, the role of MMADHC, tasked with directing Cbl to the destination enzymes yet unable to bind Cbl itself, is much less clear. This study presents the first structural characterization of MMADHC to provide novel insights into its unknown function, with the key findings that: (i) MMADHC functions by binding Cbl-laden MMACHC, but only post-Cbl processing, i.e. after the upper axial ligand of Cbl has been removed; (ii) the MMACHC interaction module of MMADHC contains an inge- niously modified NTR fold that abolishes homodimerization, favors heterodimerization with another modified NTR fold from MMACHC, and accounts for the reported lack of Cbl binding (13); and (iii) missense mutations of MMADHC that reduced both AdoCbl and MeCbl production in cells can be explained at least in part by an abrogated interaction with MMACHC.
Our data illustrate the application of protein-protein interaction in the early Cbl-targeting steps of the pathway for both the mitochondrial (AdoCbl to MUT) and the cytosolic (MeCbl to MS) targeting routes. Such protein-protein interaction is essential to the targeting process, and is disrupted by a known disease mutation of MMACHC that precludes B 12 processing and by mutations of MMADHC that alter the binding interface. We propose a mechanistic model for the MMACHC-MMADHC complex whereby the adaptation of the NTR fold by both proteins favors heterodimerization to form the "Cbl trafficking chaperone." In this model, MMACHC in the Cbl-free form could exist in a monomer-homodimer equilibrium (Fig. 7A). The self-association of MMACHC into homodimer, likely with a high dissociation rate that precludes its isolation in vitro (13,16,17), is enriched in the presence of Cbl ligand (Fig.  7B) and may serve to close the enzymatic active site for proper Cbl processing (16) (Fig. 7C). Cbl-laden MMACHC (monomer/ homodimer), with the upper axial ligand now removed, preferentially binds monomeric MMADHC as a 1:1 heterodimer rather than binding to itself, resulting in the "trafficking chaperone" that delivers processed Cbl to its target destinations (Fig. 7D).
Interestingly, only the C-terminal 154 aa of MMADHC is required for the interaction with MMACHC. As such, the exact function of the MMADHC N terminus remains to be determined. Predicted to be largely disordered, the Nterminal ϳ115 aa harbors a mitochondrial targeting sequence (aa 1-12) and has been shown to be required for cellular AdoCbl production only (11). Thus, this protein region is likely only required for function within, or direction to, the mitochondria.
Because MMADHC mutations that abolished direct interaction with MMACHC result in the combined HCϩMMA phenotype (11,12), it is conceivable that disruption of the MMACHC-MMADHC complex will result in an inability to deliver Cbl to both destination enzymes MS and MUT. A possible role of MMACHC in the mitochondria, where it has not been detected either on its own or in complex (37), remains undetermined. The dysfunction caused by MMADHC missense mutations, which result in a loss of MS activity only, i.e. the so-called cblD-HC mutations (10), also remains unexplained. Because MMACHC has recently been shown to bind various isoforms of MS (38), the cblD-HC mutations could affect a larger, yet unidentified multi-protein complex, likely including at least MMACHC and MS, but possibly also methionine synthase reductase.
In conclusion, we have structurally characterized the MMACHC-interacting module of MMADHC and demonstrated that specific MMADHC mutations that cause combined loss of AdoCbl and MeCbl production in cells can interfere with MMACHC interaction. Further studies will clarify the affinity of these proteins for each other, and how this proteinprotein interaction results in Cbl delivery in the cellular context.