The complex machinery of human cobalamin metabolism

Vitamin B12 (cobalamin, Cbl) is required as a cofactor by two human enzymes, 5‐methyltetrahydrofolate‐homocysteine methyltransferase (MTR) and methylmalonyl‐CoA mutase (MMUT). Within the body, a vast array of transporters, enzymes and chaperones are required for the generation and delivery of these cofactor forms. How they perform these functions is dictated by the structure and interactions of the proteins involved, the molecular bases of which are only now being elucidated. In this review, we highlight recent insights into human Cbl metabolism and address open questions in the field by employing a protein structure and interactome based perspective. We discuss how three very similar proteins—haptocorrin, intrinsic factor and transcobalamin—exploit slight structural differences and unique ligand receptor interactions to effect selective Cbl absorption and internalisation. We describe recent advances in the understanding of how endocytosed Cbl is transported across the lysosomal membrane and the implications of the recently solved ABCD4 structure. We detail how MMACHC and MMADHC cooperate to modify and target cytosolic Cbl to the client enzymes MTR and MMUT using ingenious modifications to an ancient nitroreductase fold, and how MTR and MMUT link with their accessory enzymes to sustainably harness the supernucleophilic potential of Cbl. Finally, we provide an outlook on how future studies may combine structural and interactome based approaches and incorporate knowledge of post‐translational modifications to bring further insights.

compartments in which it is modified for use by the two client enzymes. It is no surprise that these proteins form a dynamic, physically connected metabolic network, in which the underlying molecular interactions are beginning to be mapped. This is also accompanied by a series of remarkable structural studies demonstrating the coevolution of unique functional domains to enact Cbl sequestration, chaperoning and modification.
The two Cbl-dependent client enzymes serve key, wide-ranging metabolic functions in the cell. At the mitochondrion, MMUT funnels the products of propionate catabolism into the tricarboxylic acid cycle for energy production. In the cytosol, MTR generates precursors essential for protein translation and post-translational methylation. Just as these cellular pathways are metabolically linked, so are the underlying enzymes physically linked through protein-protein interactions. The concept of a multi-enzyme 'metabolon' has emerged to depict the intricate and intimate structure-function relationship between the constituent proteins. 1 This connectivity has important implications for pathway regulation of metabolic flux, biosynthesis and cellular signalling. It also helps explain how loss of function of individual enzymes, often in the form of inherited genetic disorders, result in metabolic perturbations with cascading effects that are further reaching than might be expected from loss of a single enzymatic reaction.
In this review, we provide an overview of the players and processes involved in intracellular cobalamin metabolism. Although other excellent reviews provide detailed clinical 2,3 or biochemical 4,5 descriptions of these pathways, here we focus on how the three-dimensional structure and physical interactions of these enzymes influence the function, regulation and activity of both pathways. We also try to provide insights into how these features have been uncovered and how these discoveries have shaped our understanding of these metabolic processes.
2 | THE UNIQUE STRUCTURE OF VITAMIN B 1 2 Cobalamin is the largest vitamin required by humans and is also the only known biomolecule with a stable cobalt-carbon bond. It is through the interplay of the oxidation state (+1, +2 or +3) and coordination chemistry (4-, 5-or 6-bonds) of the cobalt atom, that endows cobalamin the versatility and utility as a cofactor. 6,7 Over six decades have elapsed since the first elucidation of the Cbl molecular structure by Dorothy Hodgkin, revealing a macrocyclic structure (corrin ring) surrounding the central cobalt atom. 8 The corrin ring is embellished with propionate side chains, one of which is connected to a 5,6-dimethyl-benzimidazole (DMB) ribonucleotide loop ( Figure 1A). This latter appendage can, quite unusually, coordinate back to the central cobalt atom creating a lower (α-axial) ligand. Formation of an additional upper (β-axial) ligand with any number of 'R-group' molecules completes the potential six-coordinate geometry of the cobalt ( Figure 1A). In humans, the two most important biological R-groups ( Figure 1B) are 5 0 -deoxyadenosyl-(AdoCbl) and methyl-(MeCbl), which constitute the cofactor forms for MMUT and MTR, respectively. Other physiologically relevant R-groups include hydroxo-(OHCbl) and cyano-(CNCbl) (Figure 1B), the latter of which is officially 'vitamin B 12 ' although it is derived from an artefact of industrial purification. 6,7,9 Complete biosynthesis of Cbl requires approximately 30 enzymatic steps. It is therefore exclusively performed by certain bacteria and archaea 10 which use as cofactors Cbl as well as more than a dozen Cbl-like analogues (cobamides) that differ from Cbl at the α-axial ligand. By contrast, Cbl is the only cofactor form used by eukaryotes 11 and therefore, selectivity for Cbl among other Cbllike compounds which are present in natural food sources and gut microbiota represents the first challenge to human Cbl metabolism.

| STRUCTURAL BASIS OF COBALAMIN ABSORPTION
Oral ingestion is the entry point into Cbl metabolism, following which Cbl undergoes an elaborate route of absorption, transport and uptake via three Cbl-binding proteins (and their respective membrane receptors): haptocorrin (HC), intrinsic factor (IF) and transcobalamin (TC). The three proteins act in succession, at specific points of the food-to-tissue delivery route. Inherited mutations of genes encoding IF (CBLIF 12,13 ), the IF receptor (CUBN, AMN 13,14 ) and TC (TCN2 15 ) are causative of disease, while the clinical relevance is uncertain for mutations of genes encoding the TC receptor (CD320 16 ) as well as HC (TCN1 17 ), which has no dedicated receptor.
Secreted by salivary glands, 18,19 HC (previously called R-binder or transcobalamin-1) first binds Cbl and its analogues freed from dietary protein in the stomach, to protects them from gastric acid hydrolysis. [20][21][22] The HC:Cbl complex travels through the upper gastrointestinal tract to the duodenum, where HC is degraded by pancreatic proteases, releasing Cbl. Free Cbl in the duodenum is bound by IF which has specificity for Cbl and selects against Cbl-like analogs. 23 It is noteworthy that glycosylation of HC and IF is crucial to protect these proteins and Cbl (reviewed in Ref. [24]). The IF:Cbl complex continues through the small intestine until it is internalised by enterocytes of the distal ileum through recognition by a receptor consisting of cubilin and amnionless (cubam). 25 Cubilin and amnionless are mutually dependent for correct apical membrane targeting, 26,27 and a recent co-structure demonstrated their interaction via the N-terminal coiled-coil domains of cubilin which form a heterotrimer that is anchored onto the corresponding three-faced β-helix of amnionless. 28 Following enterocyte internalisation, Cbl is transported through the cell and exported into the blood. Approximately 80%-90% of plasma Cbl is bound by HC, whose role may be to facilitate Cbl storage and recycling or the sequestration of Cbl analogs, 5 analogous to its similar role in enterohepatic circulation. 29 The remaining 10%-20% of plasma Cbl is bound by TC. 30,31 Interaction of TC:Cbl with the receptor megalin mediates renal reabsorption, 32 while recognition by the transcobalamin receptor (TCblR) is responsible for uptake in liver and other tissues. 33 Fulfilling a similar cellular role of chaperoning Cbl during transit, HC, IF and TC seem to have a common evolutionary origin, and each protein carries a single Cbl molecule per protomer with sub-picomolar affinity. 34 It is no surprise that they have strongly overlapping 3D structures (RMSD: 2 Å 35 ) (Figure 2A), each composed of an N-terminal alpha-domain and a C-terminal beta-domain connected by a flexible linker. [35][36][37] Cbl binds at the domain interface, with both domains participating approximately equally to binding ( Figure 2B). In all cases, Cbl is found in the 'base-on' state, where the DMB is coordinated as an α-axial ligand 38 ( Figure 2C). A threestep Cbl-binding mechanism has been proposed for all three proteins, 23 whereby an encounter with Cbl brings the otherwise disjointed alpha-and beta-domains of each protein together. Cbl is initially contacted by the betadomain, 35,39 followed by dynamic juxtaposition of the two domains to sandwich the bound Cbl. 38 Despite clear evidence for the capability of IF and TC, but not HC, to differentiate between Cbl and its analogs, 23 the current binding mechanism and structural data could not explain the underlying molecular basis.
More recent structures of TC:CD320 and IF:cubilin complexes have shed light on the recognition of Cbl by receptors ( Figure 2D). The strong (1 nM) affinity of TC:CD320 appears to be driven by the calcium-dependent binding of two LDLR-A domains in CD320 to the alpha-domain of TC. 40 His56, His104 and His154 of TC have been suggested to modulate the pH dependent interaction with CD320. The IF:cubilin interaction is also calcium dependent, involving charged contacts between CUB domains of cubilin and a conserved lysine/arginine finger from IF. Interestingly, CD320 binds only to the alpha-domain of TC, 40 while cubulin interacts with both alpha-and beta-domains through different CUB domains 41 ( Figure 2D). The IF:cubilin binding mechanism makes it obvious why only Cbl bound IF (holo-IF) is able to bind the cubilin containing receptor, as the alpha-and beta-domains are only placed in close enough proximity for cubilin binding if Cbl is present.

| INTRACELLULAR LYSOSOMAL EFFLUX VIA ABCDAND LMBD1
Following endocytosis, internalised Cbl remains bound to TC. Release of Cbl from TC for intracellular trafficking and function appears to occur via proteasomal degradation of TC following progressive acidification from endosomes into lysosomes. The subsequent transport of free Cbl into the cytosol requires two integral membrane proteins: lipocalin-1 interacting membrane receptor domaincontaining protein 1 (LMBD1, encoded by LMBRD1) and adenosine triphosphate (ATP)-binding cassette subfamily D member 4 (ABCD4, encoded by ABCD4). Disturbed function of either protein results in lysosomal accumulation of free Cbl 42,43 and is associated with the inherited defects cblF (LMBD1) 44 and cblJ (ABCD4). 42 It was initially expected that only one integral membrane protein would be required to transport Cbl into the cytosol. However, a series of cellular and biochemical investigations have illuminated the important roles of both proteins. LMBD1 is heavily glycosylated and targets to the lysosome. [44][45][46] ABCD4, by contrast, is unglycosylated and when over-expressed is located in the ER 46 or autophagosomes. 45 However, when co-expressed with wild-type LMBD1, but not LMBD1 whose lysosomal targeting sites have been ablated, ABCD4 is lysosomally targeted. 45,46 This implies physically interaction between these proteins, as shown in vitro 47 and in live cells. 45 These findings suggest a chaperoning role of LMBD1, recruiting ABCD4 to the lysosome and protecting it from proteolytic degradation. They also indicate ABCD4 may be the active lysosome-cytosol Cbl importer. Support for that role was recently provided by demonstration that ABCD4 can conduct the ATP-dependent transport of Cbl out of liposomal membranes in vitro. 48 The cryo-electron microscopy (cryo-EM) structure of ABCD4 49 reveals a homodimer with each subunit containing six transmembrane helices to traverse the lysosomal membrane and a cytosol-facing nucleotide-binding domain (NBD) that binds and hydrolyses ATP ( Figure 3A). The presumed position of the two NBDs in the cytoplasm implies that ABCD4 is a 'trans-acting' exporter (i.e., lysosomal Cbl enters the ABC transporter in trans to cytosolic ATP) contrary to classical 'cis-acting' exporters (i.e., substrate cargo and ATP access the ABC transporter from the cytosol in cis) ( Figure 3). This is consistent with the activity of ABCD4 reconstituted in liposomes, whereby Cbl captured within ABCD4-embedded liposomes was released following provision of ATP to the surrounding medium. 48 However, this is an extremely unusual action for a classical ABC exporter which pumps in the opposite direction. Such example includes human ABCD1 which exports very long-chain fatty acids from the cytosol into peroxisomes, 50 and retains close genetic and structural homology to ABCD4 as they belong to the same type I exporter family ( Figure 3B). On the other hand, the bacterial Cbl transporter BtuCD, genetically and structurally dissimilar to ABCD4 (Figure 3B), is a Cbl importer whose function mirrors that predicted for ABCD4. 51 Such a disconnect between structure and function is currently difficult to reconcile. Further structural and biochemical evidence is required to fully understand how ABCD4 contributes to the release of lysosomal Cbl to the cytosol.

| CYTOSOLIC MODIFICATION AND TRANSPORT BY MMACHC AND MMADHC
The journey of Cbl does not end with its intracellular entry into the cytosol. The enzyme MMACHC has been shown to bind ABCD4 and LMBD1 in vitro, 47 and therefore may bind Cbl at the point of cytosolic entry. Deficiency of MMACHC (encoded by MMACHC) is associated with the inherited defect cblC, 52 the most frequent inborn error of Cbl metabolism. 53 Much akin to a welcoming reception, the enzyme MMACHC greets whatever Cbl forms that it encounters. The structures of MMACHC, determined in the apo-state 54 and bound to different Cbl forms (including MeCbl, 54 AdoCbl, 55 antivitamin B 12 56 ), demonstrate how MMACHC modifies the canonical nitroreductase fold, including three novel protrusions, to bind Cbl in the 'base-off' state, while opening a large cavity above the corrin ring that accommodates Cbl forms with nearly any R-group attached ( Figure 4A).
The base-off binding of Cbl weakens the cobalt-Rgroup bond, enabling MMACHC to function also as a processing enzyme that removes the attached R-group via dealkylation (e.g., MeCbl or AdoCbl 58 ) or decyanation (e.g., CNCbl 59 ). Such activities require an electron donor, typically glutathione, whose coordination depends on three strongly conserved arginines (Arg161, Arg206 and Arg230), 56 of which p.Arg161Gln represents the most prevalent disease causing missense mutation. 53 We 55 and others 60 found MMACHC to dimerize in vitro and in crystallo, although the monomeric state is preferred. Interestingly, in the crystallised MMACHC homodimer, one of the novel protrusions caps off the Cbl R-group in the active site cavity in an inter-monomer fashion, 55 a feature that may facilitate catalysis ( Figure 4D).
Delivery of MMACHC-processed Cbl to MTR and MMUT requires MMADHC (encoded by MMADHC), associated to the cblD defect. Dysfunction of MMADHC, uniquely among Cbl disorders, can present as a deficiency only of MTR, only of MMUT, or combined deficiency of both, dependent on the MMADHC genotype. 61,62 This is clear confirmation that MMADHC situates itself at the traffic junction between the cytosolic (MTR) and mitochondrial (MMUT) destination routes. The structure of MMADHC 57,63 revealed closest structural homology to MMACHC (Figure 4B), despite a lack of obvious sequence conservation. However, unlike MMACHC, MMADHC has not tailored the nitroreductase fold to accommodate Cbl binding, as it is missing the protrusions required to surround the molecule ( Figure 4C). This is consistent with the reported lack of Cbl binding capability of purified recombinant protein. 60 The wealth of evidence for direct physical MMACHC:MMADHC interaction 57,60,64-67 is consistent with the notion that MMADHC acts as a chaperone for MMACHC-bound Cbl. Indeed, the MMACHC: MMADHC interaction appears to be dependent on prior Cbl processing activity of MMACHC, whereby MMADHC binds MMACHC only following R-group removal. 57,65 Structural modelling guided by small angle x-ray scattering data indicates a 1:1 interaction of MMACHC:MMADHC, suggesting MMACHC homodimerization may be a prelude to functional heterodimerization with MMADHC. This model, supported by more recent AlphaFold 68 predictions ( Figure 4E), also implicates the MMADHC loop Thr182-Trp189, whose mutation is associated with combined MTR and MMUT deficiency, to create part of the MMACHC-MMADHC interface and directly interact with MMACHC-bound Cbl. Li and colleagues 66 recently demonstrated that MMADHC can coordinate Cys261 to cob(II)alamin bound to MMACHC, and solved the structure of MMADHC complexed with Cbl in this coordination state. Interestingly, in their structure, the only other interactions between MMADHC and Cbl consist of the Cys261-cobalt bond, and a hydrogen bond between threonine 187 (part of the Thr182-Trp189 loop) and an acetamide side-chain of the corrin ring.
Together, these data provide an indication of how cytosolic Cbl is sequestered, modified and prepared for transport across the cell. Unprocessed Cbl is bound by MMACHC, which uses reductive cleavage to remove the β-axial ligand, and a MMACHC homodimer could facilitate catalysis by stabilising the β-axial ligand through the cross-over caps ( Figure 4D). Following cleavage, this MMACHC:MMACHC homodimer is replaced by an MMACHC:MMADHC heterodimer whereby a helixturn-helix motif of MMADHC containing Cys261 takes over the space left by one MMACHC cap and provides a β-axial cobalt bond to the Cbl-bound MMACHC in the heterodimer ( Figure 4E). Meanwhile, the MMADHC loop containing Thr182-Trp189 secures interaction both with MMACHC and the Cbl molecule. From here, Cbl is ready to be delivered to its two client enzymes. How this is achieved, and which mechanisms control the direction Cbl is transported (towards mitochondrial MMUT or cytosolic MTR), remain to be elucidated.

| METHYL-COFACTOR SYNTHESIS AND UTILISATION
The cytosolic end-user of Cbl, in the methyl form (MeCbl), is the enzyme MTR (encoded by MTR), the deficiency of which is associated with the inherited cblG defect. 69 MTR takes charge of the final step in methionine biosynthesis, with far-reaching contributions to protein translation (as amino acid) and global methylation (as precursor to the methyl donor). MTR catalyses an overall reaction that involves the transfer of a methyl group from 5-methyltetrahydrofolate (MeTHF) to Cbl in the +1 reduced state (cob[I]alamin), forming methylcob(III)alamin that in turn transfers its methyl to homocysteine (Hcy), ultimately producing methionine.
To accomplish the multiple intricate reaction steps, MTR has evolved as a modular enzyme consisting of five domains, four of which individually bind Hcy, MeTHF, Cbl and AdoMet, along with a small cap domain. 70 Most structural data on MTR relies on crystal structures of fragments from the bacterial homologue, MetH, with limited information from human MTR. The Hcy and MeTHF binding domains were structurally characterised from an N-terminal fragment of Thermotoga maritima MetH, 71 and the equivalent from human MTR (PDB ID: 4CCZ; Figure 5A). These two homologous structures revealed the Hcy and MeTHF domains as two tightly interacting (βα) 8 barrels linked by a rigid linker. As the active sites of the two domains are separated by ≈50 Å, and act as one rigid module, the structures imply that significant domain movements will take place for the two active sites to each interact with the Cbl cofactor during the catalytic cycle. 71 The only other known structure from human MTR is of the AdoMet-binding (re-activation) domain revealing a shovel-like fold 72 ( Figure 5B). Substantial structural differences between it and Escherichia coli MetH activation domain, despite sequence homology, may influence dimerization or protein interactions of human MTR, 73 although the AdoMet binding pocket is highly conserved. 72 Insights into how MTR binds Cbl are revealed from multiple fragment structures of E. coli MetH. 70,[74][75][76] The Cbl-binding domain uses a Rossmann like-fold to bind Cbl in both the His-on and His-off states. 75 The bound Cbl is shielded by a four-helix cap domain ( Figure 5C,D), which shifts by 26 Å when MTR undergoes reactivation by AdoMet (and presumably during the normal catalytic cycle). This conformational change allows the activation domain to interact directly with the Cbl domain with the cap domain sequestered away ( Figure 5D).
Overall, this ensemble of structures suggest a highly dynamic enzyme and allow us to piece together the conformational landscape of MTR ( Figure 5E). The normal catalytic cycle involves 'back-and-forth' movements of the Hcy and MeTHF domains as they in turn interact with the Cbl domain. Such movement facilitates the MeTHF domain in the methyl transfer step between cob(I) alamin and 5-methyltetrahydrofolate, and facilitates the Hcy domain in the methyl transfer step between methylcob(III)alamin and homocysteine. 77 The conformational dynamics of MTR also prepare the cap and activation domains for interacting with the Cbl domain during the occasional cycle of AdoMet-dependent Cbl reactivation, as described in the next section.

| MTR REACTIVATION (SUPRA-) COMPLEX
Occasionally (e.g., every 200-1000 turnovers), the labile MTR-bound cob(I)alamin is oxidised to inactive cob(II) alamin, which requires re-activation via a methyl transfer from S-adenosylmethionine (AdoMet) to regenerate methylcob(III)alamin ( Figure 5E). 65 This reactivation cycle is aided by methionine synthase reductase (MTRR; EC 2.1.1.135). Human MTRR (encoded by MTRR) harbours two flavin-binding domains on its N-terminus that accommodate NADPH and FAD, respectively. 78 A crystal structure of its C-terminus revealed a connecting domain and an FMN-binding domain 79 ( Figure 6A). The reactivation cycle begins with sequential electron transfer within the MTRR protein, that follows the direction from NADPH, to FAD, to FMN and then to the MTR protein.
Through a direct interaction between FMN domain of MTRR and activation domain of MTR, 80 the donated electron from MTRR reduces the MTR-bound cob(II)alamin to cob(I)alamin, before it is remethylated using S-adenosylmethionine (AdoMet) to become cob(III)alamin 81 ( Figure 6B).
Like the normal catalytic cycle of MTR, the AdoMetdependent reactivation cycle is hugely facilitated by conformational flexibility of MTR through its cap and activation domains. Here the cap domain flips on top of the oxidised cob(II)alamin until an electron becomes available for reduction to cob(I)alamin ( Figure 5E) The activation domain then displaces the cap and enables its bound AdoMet to act as a methyl donor for cob(I)alamin, producing methylcob(III)alamin that re-enters the catalytic cycle involving once again the Hcy and MeTHF domains 82 ( Figure 5E).
Recent research on the interaction between MTR and MTRR have shown that besides the electron transfer, MTRR has a role of favouring the catalytic activity of MTR through the tandem arginines on its N-terminal. 83 Interaction with the FMN domain of MTRR can also cause the stabilisation of the apo form of MTR and assist in the incorporation of cobalamin. 84 MMACHC and MMADHC may also be key players of a putative reactivation supra-complex, as this 1:1 heterodimer is critical for the cobalamin delivery to MTR and MUT. 57 Co-immunoprecipitation and proximity ligation assays with MTR and MMACHC have shown that they have a close protein-protein interaction that happens at the same time or after MMACHC and MMADHC interaction. 85 More interestingly, the interaction also occurs with the truncated form of MTR which might indicate a regulatory role in the cobalamin conversion to MeCbl or AdoCbl, besides the delivery of cobalamin to MTR. 85 Further co-immunoprecipitation and DuoLink proximity assays revealed interactions of not only known interactions, such as MTR and MTRR, MMACHC and MMADHC, but also novel interactions between MTR, MMACHC and MMADHC. Silencing experiments using siRNA targeting these four genes have confirmed their interaction. 64 Although the absence of MMADHC does not affect the interaction between MMACHC, MTR and MTRR, the silencing of any of the latest three can decrease the interaction with all other four, suggesting that they might form a stable complex composed at least by MTR, MTRR and MMACHC 64 ( Figure 6B).

| MITOCHONDRIAL PROCESSING AND UTILISATION
It is currently not known how and when Cbl digresses from the cytosol and enters the mitochondria. However, once inside, three proteins are responsible for its modification, implementation and utilisation. These successive roles are fulfilled by MMAB (encoded by MMAB), MMAA (encoded by MMAA) and MMUT (encoded by MMUT), respectively, corresponding to the inherited defects cblB, cblA and mut.
The protein MMAB is required for the generation of AdoCbl from Cbl. This reaction is performed through its ATP:cob(I)alamin adenosyltransferase activity using ATP as a co-substrate. Much of our understanding of MMAB function has come from studies using bacterial orthologues. These have demonstrated generation of AdoCbl proceeds in a step-wise fashion. First, ATP is bound to one of three potential active-sites, each of which occur at the trimeric subunit interfaces ( Figure 7A). ATP-binding induces the active-site to form a binding pocket which forces incoming Cbl into an unusual 4-coordinate (i.e., base-off, no upper-ligand) cob(II)alamin. 86 This renders Cbl more amenable to reduction to cob(I)alamin, 87 required for supernucleophilic attack of the 5 0 -carbon of ATP. The resulting nucleophilic displacement reaction generates AdoCbl and inorganic triphosphate, 88 captured crystallographically 89 ( Figure 7A). Following triphosphate release, AdoCbl becomes solvent exposed ( Figure 7A), potentially in preparation for transfer to MMUT. Release of AdoCbl from MMAB is induced following ATP binding to an adjacent active site, beginning the cycle anew. 90 Although ATP prompts ejection of AdoCbl from MMAB, AdoCbl release is facilitated by the presence of MMUT and modulated by the presence of MMAA. 91 Human MMAA is a homodimer (Figure 7B), whereby each subunit contains a G-domain, encompassing a GTPspecific binding site as well as three nucleotide sensitive 'switch' loops, an N-terminal extension of unclear significance and a C-terminal dimerization arm. 92 The conformation of the switch I and switch II loops are dependent on, and influence, GTP binding and cleavage, while the switch III motif appears to mediate interaction between MMAA and MMUT. [93][94][95] The biological significance of MMAA seems to be wholly dependent on its interaction with MMUT. Many studies have demonstrated physical interaction between MMUT and MMAA (or their bacterial counterparts). 91,92,[96][97][98][99] While MMUT induces GTPase activity of MMAA; MMAA in turn regulates transfer of AdoCbl from MMAB to MMUT (as mentioned above), protects MMUT from oxidative inactivation, and reactivates inactivated MMUT by cofactor exchange. 100,101 The importance of these roles are underlined by variants in MMAA which, despite leaving its intrinsic GTPase activity intact, cause disease by disrupting physical or biochemical communication with MMUT. 91 The much-coveted molecular view of the MMAA:MMUT interaction remains elusive, although structure of bacterial IcmF, a bifunctional enzyme with fused MMAA-like and MUT-like domains, has offered a glimpse of possible interfaces between these proteins 102,103 and low resolution cryo-EM structures suggest multiple oligomeric conformations. 104 MMUT is the AdoCbl end user, taking advantage of the radical nucleophile of AdoCbl to catalyse isomerisation of methylmalonyl-CoA to succinyl-CoA, a reaction that funnels the products of propionate metabolismgenerated from the breakdown of branched-chain amino acids (methionine, threonine, valine and isoleucine), odd-chain fatty acids and the side chain of cholesterolinto the tricarboxylic acid (TCA) cycle. Human MMUT is a homodimer, whereby each subunit contains an N-terminal substrate binding site, a C-terminal Cbl binding site and the active-site at their conjunction ( Figure 7C). Structural snapshots of MMUT in the unbound (apo-), AdoCbl-bound (holo-) and AdoCbl plus malonyl-CoA bound (ternary-) states have demonstrated the progressive structural rearrangements following cofactor and substrate binding. 92 These follow an induced fit mechanism, whereby formation of the holo-structure results in re-arrangement of the C-terminal binding site around the incoming AdoCbl molecule, including positioning of His627 to coordinate as an α-axial cobalt ligand, while the N-terminal substrate binding site stays largely static. By contrast, subsequent formation of the ternary-structure promotes constriction of the N-terminal TIM barrel around the incoming substrate, sealing off the channel pore from surrounding solvent and positioning active-site residues. As described in the paragraphs above, MMUT function depends on direct interaction with at least MMAB and MMAA. Analogous to MTR, MMUT is part of a multi-enzyme complex. 105 In addition to MMAB and MMAA, this complex contains methylmalonyl-CoA epimerase (MCEE, encoded by MCEE), which produces the MMUT substrate L-methylmalonyl-CoA and whose deficiency is associated with MMUT dysfunction. 106 Recent evidence also demonstrates proximal anaplerotic and TCA cycle enzymes including alpha-ketoglutarate dehydrogenase, an enzyme regulated by succinyl-CoA and is dysregulated in MMUTdeficiency, 107 and glutamate dehydrogenase to be part of this complex. 105

| CONCLUSIONS, OPEN QUESTIONS AND OUTLOOK
The past 10-15 years have witnessed a structural biology revolution, with subsequent structures of Cbl processing enzymes providing many insights into the transport and processing of this rare cofactor within human metabolism. Recently determined structures of TC, HC and IF have demonstrated their similarity and shared Cbl binding mechanism. In the case of IF, it has also been shown how its binding interaction with the cubilin is dependent on the holo-form. The structure of ABCD4 has given us a clear view of ATP-binding, but a complicated interpretation of lysosomal Cbl efflux. Individual x-ray crystal structures of MMACHC and MMADHC have shown how each protein has repurposed an ancient fold to bind (MMACHC) and interact (MMADHC) with Cbl, potentially through a dimer swapping mechanism. Finally, we have been shown at the molecular level the conformational flexibility of MTR and MMUT, how these enzymes position Cbl to maximise enzymatic benefit, and how they rely on a set of supportive interacting proteins to maintain functionality.
Despite these advances, we still do not yet understand a number of mechanistic specifics, for example, how Cbl is exported from the lysosome intracellularly. Key to such knowledge may be a better understanding of how ABCD4 and LMBD1 assemble, and potentially how MMACHC interacts with this complex. We have yet to uncover the molecular triggers determining intracellular Cbl partitioning towards MeCbl or AdoCbl synthesis nor how these cofactor forms are physically delivered to MTR and MMUT. On the cytosolic side, full-length structures of MTR in conjunction with MTRR may reveal the secrets of how reactivation and MeCbl formation occurs, while inclusion of MMACHC and MMADHC may unravel the process and determinants of cofactor loading. At the mitochondrial end, still to be elucidated is how Cbl enters the mitochondria, the role of MMADHC, and how MMAB, MMAA and MMUT assemble as a complex along with which additional partners (e.g. MCEE) may be included. Additionally, with the notable exception of MMACHC, very little is known about regulation of Cbl metabolic proteins, at the expression, post-translational and degradation levels that could further modulate their functionality and interaction. Overall, we have a good understanding of the individual players in human Cbl transport and processing but have limited understanding in how they all piece together.
This review appeared at a time when machine learning-driven Alphafold 108 and RoseTTAfold 109 servers have begun the revolution of homology modelling, towards structure coverage of all single proteins in the entire proteomes. Peering into the future, experimental structural studies should follow the niche where it creates impact and knowledge for example, in mechanistic discoveries and elucidation of interactomes and the effect of diseasecausing mutations. To do so, these next-generation studies should take advantage of recent technical advances, such as tandem expression systems in eukaryotic cells, improved stabilisation methods for purified recombinant or endogenous protein complexes, and the increased resolution and accessibility of cryo-EM to determine bi-or multi-protein complexes at atomic level resolution. In parallel, advances in mass spectrometry (such as crosslinking MS, native MS) have made systems level identification of protein-protein 110 and protein-metabolite 111 interactions possible, enabling the capture of all complex member proteins without the need for targeted (over-)expression systems. These are further benefitted from CRISPR-editing of endogenous complexes 112 to study the effects of disease-associated mutants and for their purification. Overall, there is strong reason for optimism that the current outstanding questions of Cbl metabolism may soon be resolved. With them, the potential to create targeted therapies that are able to fully compensate for the loss of individual enzymes in these pathways may be realised.