Structural Characterization of the Active Site of the PduO-Type ATP:Co(I)rrinoid Adenosyltransferase from Lactobacillus reuteri*

The three-dimensional crystal structure of the PduO-type corrinoid adenosyltransferase from Lactobacillus reuteri (LrPduO) has been solved to 1.68-Å resolution. The functional assignment of LrPduO as a corrinoid adenosyltransferase was confirmed by in vivo and in vitro evidence. The enzyme has an apparent \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{m}^{\mathrm{ATP}}\) \end{document} of 2.2 μm and \batchmode \documentclass[fleqn,10pt,legalpaper]{article} \usepackage{amssymb} \usepackage{amsfonts} \usepackage{amsmath} \pagestyle{empty} \begin{document} \(K_{m}^{\mathrm{Cobalamin}}\) \end{document} of 0.13 μm and a kcat of 0.025 s-1. Co-crystallization of the enzyme with Mg-ATP resulted in well-defined electron density for an N-terminal loop that had been disordered in other PduO-type enzyme structures. This newly defined N-terminal loop makes up the lower portion of the enzyme active site with the other half being contributed from an adjacent subunit. These results provide the first detailed description of the enzyme active site for a PduO-type adenosyltransferase and identify a unique ATP binding motif at the protein N terminus. The molecular architecture at the active site offers valuable new insight into the role of various residues responsible for the human disease methylmalonic aciduria.

B 12 (cobalamin, Cbl) 7 is an essential nutrient for animals, lower eukaryotes, and prokaryotes, but is synthesized exclusively by prokaryotes (1). Adenosylation of the corrinoid ring of Cbl generates coenzyme B 12 (adenosylcobalamin, AdoCbl), an essential cofactor used by enzymes that catalyze intramolecular rearrangements (2)(3)(4), deaminations (5), dehydrations (6), reductions (7,8), and reductive dehalogenations (9). Corrinoid adenosyltransferases play a key role in the biosynthesis of AdoCbl by covalently attaching the 5Ј-deoxyadenosyl moiety from ATP to the Co(I) ion of the corrin ring of Cbl (10,11). These enzymes generate a biologically unique, labile cobaltcarbon bond that is the source of the unusual chemistry associated with the B 12 cofactor. As such, understanding the mechanistic strategies of those enzymes involved in the formation of this carbon-metal bond is of considerable interest. In Salmonella enterica, three separate cob(I)alamin adenosyltransferases have been identified: CobA, PduO, and EutT. All cob(I)alamin adenosyltransferases identified to date belong to one of these three distinct families. CobA is the housekeeping enzyme of S. enterica and is involved in the anaerobic de novo synthesis of AdoCbl. The structure and function of this enzyme has been well characterized (12)(13)(14)(15)(16). The PduO and EutT enzymes, meanwhile, assimilate existing Cbl into AdoCbl. Both PduO and EutT are encoded within large, discreet operons of S. enterica where they play specialized roles in the catabolism of 1,2-propanediol or ethanolamine (17)(18)(19). Despite its specialized role in S. enterica, the PduO-type enzyme is the most widely distributed of these adenosyltransferases with homologues identified in species of archaeotes and prokaryotes, as well as in many eukaryotes, ranging from yeast to humans (19). In animals, PduO is the only cob(I)alamin adenosyltransferase enzyme available for the assimilation of dietary cobalamins into coenzyme B 12 .
Humans, who lack the ability to synthesize Cbl de novo, produce AdoCbl from reduced Cbl through a PduO-type adenosyltransferase (19 -21). Patients with malfunctions in this enzyme suffer from methylmalonic aciduria and metabolic ketoacidosis (22). Initial biochemical, functional, and structural characterization of PduO-type corrinoid adenosyltransferases has recently been reported, including a high-resolution threedimensional crystal structure of the PduO-type enzyme from the archaeon Thermoplasma acidophilum (23), and two additional homologous structures deposited in the RCSB protein data bank (Bacillus subtilis YvqK, 1RTY; and a putative PduO from Mycobacterium tuberculosis, 2G2D). These structures reveal that the enzyme is a trimer with each subunit composed of a five helix-bundle. Unfortunately, the absence of bound substrates in these crystal structures has precluded identification of a definitive location for the active site. Whereas both the CobAand PduO-type adenosyltransferases catalyze the elimination of tripolyphosphate from ATP (a relatively rare mechanism among the vast families of ATP-utilizing enzymes), the CobA-type adenosyltransferases utilize an inverted P-loop motif for ATP binding and elimination of tripolyphosphate while the PduO-type enzymes apparently lack such a motif (12,24,25). Currently, there is no clear structural or biochemical characterization of the ATP binding site in a PduO-type enzyme.
Recently, Lactobacillus reuteri strain CRL1098 was discovered to make B 12 de novo (26), and a gene putatively encoding a homolog of the SePduO enzyme was identified among the cluster of B 12 biosynthetic genes of this bacterium (GenBank TM accession number AY780645). In this study, we report the three-dimensional crystal structure of L. reuteri PduO-type corrinoid adenosyltransferase (LrPduO) complexed with MgATP at 1.68-Å resolution and provide a kinetic characterization of the enzyme. Despite limited sequence identity (31%), the LrPduO protein shares a common fold with the apo PduO protein from T. acidophilum (TaPduO). Our structure model reveals the active site in a deep cleft at the subunit interface made up, in part, of the immediate N terminus of the protein. A description of the active site unveils a previously unseen ATP binding motif and provides a framework for better understanding the various mechanistic strategies employed by cob(I)alamin adenosyltransferase enzymes. Further, this description of ATP binding allows fresh insight into the molecular basis for protein malfunction in mutations of the human corrinoid adenosyltransferase.

EXPERIMENTAL PROCEDURES
Construction of Expression Vectors-To generate an Nterminal (His) 6 -tagged recombinant construct of PduO, the L. reuteri pduO gene (GenBank TM accession number AY780645) was PCR-amplified from L. reuteri strain CRL1098 genomic DNA isolated as described (27). The primers used for amplification were 5Ј-CGGGATCCGTGAAGATTTATA-CAAAAAATGG-3Ј (forward), and 5Ј-GGAATTCTTAGCG-GAAAACGTCTTTACTGTT-3Ј (reverse). The PCR fragment was restriction-digested with BamHI and EcoRI and ligated into the pET-28b vector (EMD Biosciences, Inc.) generating the expression plasmid pPDU19 (Table 1), which expresses the recombinant PduO protein with a 35-amino acid N-terminal tag. To generate a recombinant protein with a rTEV proteasecleavable N-terminal (His) 6 tag, the L. reuteri pduO gene was PCR-amplified from plasmid pPDU19. The primers used for the amplification were 5Ј-AAAAAAACCATGGTGAAGATT-TATACAAAAAATGGTGATAAAGGGC-3Ј (forward) and 5Ј-TTTTTTGCGGCCGCTTAGCGGAAAAC-3Ј (reverse). The PCR fragment was cut with NcoI and NotI and was ligated into vector pTEV3 (Table 1) yielding plasmid pPDU22. The pPDU22 vector directs the synthesis of a recombinant N-terminally tagged protein with the Tev protease site four residues upstream of the first amino acid of the LrPduO protein. The N-terminal amino acid sequence of this construct is MSYYHHHHHHDYDIPTSEN-LYFQGASAPM 1 V 2 . . . where the location of the TEV protease cleavage site is underlined.
Strains, Media, and Chemicals-Bacterial strains and plasmids used in this study are listed in Table 1. Chemicals were purchased from Sigma. Lysogenic broth (LB) (28,29) was used as rich medium to propagate bacteria nonselectively. Vogel and Bonner's no-carbon E (NCE) minimal medium (30,31) was used to assess cobalamin biosynthesis. NCE medium was supplemented with glycerol (22 mM), MgSO 4 (1 mM), dicyanocobinamide ((CN) 2 Cbi, 0.2 M)), kanamycin (15 g/ml), or ampicillin (25 g/ml). Growth behavior was monitored at 650 nm using a Bio-Tek EL808 96-well plate reader. Each well contained 198 l of fresh NCE medium that was inoculated with 2 l of an overnight culture of the strain of interest grown in LB medium; each strain was analyzed in triplicate. Plates (Becton Dickinson) were incubated under aerobic conditions at 37°C over a 24-h period.
LrPduO Protein Production and Purification-Plasmids were transformed into Escherichia coli strain BL21(DE3) for overexpression. Strains were grown at 37°C with shaking in 1 liter of LB medium supplemented with kanamycin (25 g/ml). After the culture reached an optical density (OD 650 ) of 0.6 -0.7, synthesis of the phage T7 RNA polymerase In Vitro Adenosyltransferase Activity Assay-Activity assays were performed as described (37) with the following modifications. The final volume of each reaction was 1 ml. Empty, sealed quartz cuvettes were flushed with oxygen-free N 2 for 5 min. Under a stream of O 2 -free N 2 , Tris-HCl (0.2 M; pH 8.0 at 37°C),    (15 mM), and glycerol (20% v/v) and were flash cooled in a nitrogen stream at 100 K (Oxford Cryosystems, Oxford, UK). The crystals belonged to the space group I23, with one subunit in the asymmetric unit and the unit cell parameters a ϭ b ϭ c ϭ 110.6 Å. A dataset was collected at the SBC 19BM beamline of Advanced Photon Source in Argonne, IL. Diffraction data were integrated and scaled with the program HKL2000 (38). Data collection are summarized in Table 2.
Structure Determination and Refinement-The structure was determined by molecular replacement with the program MOLREP (39) starting from the model for TA1434, a PduO-type cob(I)alamin adenosyltransferase from T. acidophilum (PDB accession identifier 1NOG, Ref. 23), for which there is 31% identity with the protein sequence of LrPduO. Residues 3-181 for the L. reuteri enzyme were built automatically into the electron density by the program ARP/WARP (40), and the structure was refined with the program REF-MAC (41). The initial structure was subject to manual verification. Multiple conformations were added with the program COOT (42). Water molecules were added to the dataset by ARP/WARP with subsequent manual verification. The final model, refined to 1.68-Å resolution, includes residues Lys 2 -Asn 182 of LrPduO and the heteroatom Mg-ATP with the following 15 residues displaying multiple conformations: Gln 19  were disordered and could not be built into the electron density map. Despite the mild red coloration of crystals grown and soaked in the presence of HOCbl, no electron density was observed in the map that could be attributed to fully occupied HOCbl. All figures of molecular structures were generated with the program PyMOL. A Ramachandran plot shows that 96.9% of the residues are in the most favored region, with one outlier that lies in a flexible loop. Refinement statistics are summarized in Table 2.

RESULTS AND DISCUSSION
The LrPduO Protein Has ATP:Co(I)rrinoid Adenosyltransferase Enzyme Activity-Results from complementation studies involving LrPduO protein are shown in Fig. 1. The growth defect of a strain of S. enterica lacking the housekeeping ATP: co(I)rrinoid adenosyltransferase (CobA) enzyme was corrected by the presence of a plasmid encoding the LrPduO protein (Fig.  1, closed versus open triangles). Under the growth conditions used in this study, S. enterica does not express endogenous PduO, nor does it express eutT (44,45). Therefore, the conditions of the above-mentioned experiments demanded that the LrPduO protein adenosylate cobinamide prior to its conversion to AdoCbl (46). Hence, on the basis of these data, it is concluded that LrPduO is involved in corrinoid adenosylation. Given the homology of LrPduO to the N-terminal domain of S. enterica PduO (SePduO; Fig. 2), it was hypothesized that the LrPduO protein catalyzes the formation of the unique Co-C bond between the 5Ј-deoxyadenosyl moiety of ATP and the cobalt ion of the corrin ring.
Kinetics of the Reaction Catalyzed by LrPduO-Direct evidence that LrPduO catalyzes the last step of the corrinoid adenosylation pathway (13) was obtained in vitro. Initial velocity kinetic determinations were performed using chemically reduced cob(I)alamin and ATP as substrates for purified LrP-duO protein that did or did not have an N-terminal tag. Kinetic constants were measured by holding one substrate at saturation while the other was varied (Table  3). These data confirmed that the LrPduO protein has ATP:cob(I)alamin adenosyltransferase enzyme activity. The apparent K m for ATP (K m , 2 M) at saturating cob(I)alamin is similar to the one reported for the human PduOtype adenosyltransferase (K m , 6 M) (20,47), but is lower than that reported for the PduO enzymes from T. acidophilum (K m , 110 M) (23) and S. enterica (K m , 20 M) (48). Notably, the apparent K m for cob(I)alamin at saturating ATP is at least 10-fold lower than any previously reported K m value for any type of ATP:adenosyltransferase (K m , 1-5 M). The apparent k cat for the L. reuteri enzyme (0.03 s Ϫ1 ) is also lower than what has been reported for other adenosyltransferases (k cat ϭ 0.1-0.3 s Ϫ1 ), bringing the k cat /K m values for the L. reuteri enzyme within the range of what has been reported for the other enzymes (k cat /K m ϭ 10 4 -10 5 M Ϫ1 s Ϫ1 ). Our results indicate  that the presence of the 35-residue N-terminal tag has only a small effect on the reaction kinetics. Similar kinetic results have been reported for the N-terminal GST-tagged construct of hATR (47) and the mitochondrial targeting sequence-tagged bovine PduO (19). Based on the location of the active site at the immediate N terminus of LrPduO (see the structural description below), the slightly elevated K m values for the tagged enzyme may reflect a minor impedance of access to the active site with a concomitant reduction in the overall binding affinity.
Overall Structure and Description of the Active Site-The structure of LrPduO was determined for the N-terminally tagged protein using molecular replacement, with initial phases from the crystal structure of TA1434 (PDB accession identifier 1NOG), a PduO-type ATP:cob(I)alamin adenosyltransferase from T. acidophilum crystallized in the absence of ligands (23), with which LrPduO shares 31% sequence identity. The structure was solved for the tagged version of the enzyme, as this particular construct crystallized more readily than did the construct with the proteolytically removed tag. The final structure was refined to 1.68-Å resolution, with an R work of 17.3 and an R free of 20.4. As with the enzyme from T. acidophilum, LrPduO is a trimer consisting of three independent five-helix bundles with an overall topology of 12354 (Fig. 3). The interactions between the individual helices and subunits are similar to those described for the T. acidophilum enzyme (23). Notably, while the immediate N terminus of all previously reported structures of PduO-type adenosyltransferases has been disordered, the structure of LrPduO is unique in that it includes an additional ordered 23 amino acids at the immediate N terminus of the protein (Fig. 4). The newly defined structure of the N terminus spans residues 2-24 of the native polypeptide chain and does not include any contributions from the 35-residue N-terminal tag. Aside from this ordered N-terminal loop, no significant conformational changes accompany ATP binding, with low r.m.s. deviations (0.6 -1.3 Å) for the superposition of the LrPduO structure with three structures of the apoenzyme. The N terminus is wellordered (average B-value, 23.4) and consists of a region of random coil with a helical twist followed by two antiparallel ␤-strands interspaced by a loop of five residues. Electron density is clearly observable for Mg-ATP within this N-terminal cleft with the positions and orientations of the phosphates and the nucleotide ring very well defined (Fig. 5a). The N terminus accounts for the lower half of the enzyme active site with additional residues being contributed from helix four and five of the neighboring subunit. Unlike the inverted P-loop observed for CobA (12), the ATP binding site of LrPduO does not conform to any of the classic nucleotide binding structural motifs as seen in the ATP grasp or protein kinase families. The N-terminal ATP-binding loop is made up of several highly conserved residues and is clamped into position through a salt bridge between two conserved residues: Asp 9 of the N-terminal loop and Arg 157 from helix five of the neighboring subunit. The overall backbone structure of this loop is kept rigid through a pair of backbone hydrogen bonds (Lys 6 -Asp 9 ; Gly 8 -Gly 11 ) that preserve a tight helical twist leading into the first ␤-strand. Both Gly 8 and Gly 11 are conserved among homologs of SePduO, further suggesting a key role in scaffolding the structure of the ATP binding loop. It appears that MgATP binding is responsible for ordering the N-terminal loop. Unfortunately, despite co-crystallization with HOCbl, only small lobes of difference electron density were observed in the vicinity of the active site, indicating low occupancy for this ligand. The binding region for HOCbl, therefore, could not be modeled with confidence. However, identification of the binding site for MgATP does place restraints on the binding site for HOCbl.
ATP Binding at the Active Site of LrPduO-Mg-ATP is bound through direct interactions with 9 residues contributed from adjacent subunits of the trimer (Fig. 5, b and c). A total of 6 residues whose side chains are directly involved in binding are conserved among PduO-type adenosyltransferase enzymes (Fig. 2). This includes the hydroxyl of Thr 13 , which appears to be structurally conserved as either a serine or threonine. The majority of enzyme-substrate interactions are centered on the tripolyphosphate of ATP. As such, these interactions are expected to contribute significantly to ordering the N-terminal loop upon ATP binding. Several of these interactions originate from main chain amide hydrogens and carbonyls on the N-terminal loop, lending importance to the conservation of Gly 8 and Gly 11 in maintaining the overall fold. The phosphate oxygen atoms of the enzyme-bound ATP are positioned to provide 3 coordination bonds to Mg 2ϩ . In addition, two highly ordered water molecules along with the carbonyl oxygen of Asn 156 provide the remaining coordination to the bipyramidal Mg 2ϩ . Whereas many interactions with the triphosphate of ATP originate from a conserved sequence motif at the N terminus, this sequence motif appears to be unique to this class of enzymes, despite its high sequence conservation among the PduO-type adenosyltransferases (Fig. 2). A search of the protein data base using the conserved sequence Thr-(Lys/Arg)-X-Gly-Asp-X-Gly-X-(Thr/Ser), corresponding to residues 5-13 of the LrPduO phosphate binding domain, revealed no significant similarities with existing protein classes. Whereas the conserved amino acid sequence of PduO loosely shares features of the consensus tripolyphosphate binding sequence of the mechanistically similar adenosylmethionine synthetase and inorganic pyrophosphatase (49), a structural comparison with the active site of adenosylmethionine synthetase (50) reveals no striking similarities in the overall fold responsible for ATP binding. PduO catalyzes the elimination of tripolyphosphate from ATP (14,48), a relatively uncommon mechanism among ATP-utilizing enzymes.
The conserved residue Arg 132 is poised to play a critical role in stabilizing developing negative charge in the transition state and also in properly orienting the C5Ј-carbon of ATP for nucleophilic attack through two hydrogen bonding contacts with the bridging oxygen of ribose (2.7 Å) and the bridging oxygen to the ␣-phosphate (2.4 Å). Whereas the precise nature of its catalytic role awaits detailed mutagenic analysis, it is noteworthy that an Arg to His mutation of the structurally equivalent residue in hATR results in a complete loss of enzyme activity (47). Interestingly, CobA has no comparable protein-ligand contacts centered around the C-5Ј carbon, reinforcing the notion that the PduO-and CobA-type adenosyltransferase enzymes are mechanistically distinct despite their common overall reactions and similar efficiencies. Arg 132 , along with the neighboring residue, Arg 128 , displays alternate conformations in the crystal structure. In the absence of HOCbl it is difficult to assess which set of conformations is mechanistically important.
In addition to the contact between the Arg 132 side chain and the bridging ribose, Glu 135 and the main chain carbonyl oxygens of Arg 132 and Arg 14 are the only apparent bonding interactions between the protein and the adenosine base. Of these, only the side chain of Glu 135 contacts the C-6 amine that is specific to adenine. This small number of contacts may facilitate the transfer of adenosine to cobalamin and the subsequent leaving of the first product, AdoCbl. This lack of a stringent set of binding contacts combined with a lack of tight packing around the base moiety also suggests an active site with sufficient conformational freedom to accommodate nucleotides other than ATP. While CobA type adenosyltransferases have long been known to accommodate alternate nucleotides (10,14,25,51), the TaPduO and SePduO accept only ATP (23,48). However, the human PduO enzyme does display moderate catalytic activity with several alternate nucleotides (19).
A survey of LrPduO relative specific activities for a series of alternate nucleotides revealed specificity for a broad set of nucleotides (Table 4) with relative efficiencies similar to those described for hATR. While a detailed structural interpretation of the relative observed rates requires further kinetic analysis, it is interesting to note the low specific activity of ITP compared with ATP. The difference between these two nucleotides is limited to the identity of the functional group at the C-6 position of the purine base. Substitution at C-6 of the hydrogen bond-donating amine of ATP for the hydrogen bond-accepting carbonyl of ITP results in a 100fold reduction in specific activity. Provided that this reduction in activity results primarily from a loss of binding affinity, the reduced specific activity with ITP may be attributed to the loss of a critical hydrogen bonding interaction with the backbone carbonyl of Arg 132 . GTP, which also has a carbonyl at C-6, may recover a portion of this lost binding energy through a favorable potential hydrogen bonding interaction between its C-2 amine and the backbone carbonyl of Val 28 . It is, however, more difficult to structurally interpret the reduced specific activities observed for CTP and UTP on account of the smaller relative size of the purine bases.
A subset of PduO-type enzymes is now emerging that can be defined by the ability of the enzyme to accept alternate nucleotides in place of ATP. In addition, the K m of LrPduO for ATP (3 M) is closer to that reported for the human enzyme (7 M) than it is to those values reported for the SePduO (18 M) or TaPduO (110 M) enzymes. Interestingly, neither humans nor lactobacilli have an identified CobA homolog, leaving the PduO-type adenosyltransferase as the only enzyme available for the conversion of Cbl to coenzyme B 12 . In contrast, the genome of T. acidophilum does contain a gene encoding a putative CobA homolog (30% identical to SeCobA). Whether the presence of a CobA-type enzyme has an effect on the specificity and efficiency of PduO enzymes remains an open question.
The Putative Corrinoid Binding Site-The crystal structure of LrPduO, like that of CobA (12), reveals ATP bound in a deep cleft at the active site. In CobA, the C-5Ј carbon of ribose is positioned such that it is poised for nucleophilic attack by the reduced cobalt atom bound above. A similar scenario would be expected for the binding of cob(I)alamin in PduO. However, despite the high apparent binding affinity of cob(I)alamin for the enzyme at saturating concentrations of ATP, attempts to co-crystallize the enzyme in the presence of 2 mM HOCbl and to soak the crystals in 15 mM HOCbl did not result in any interpretable electron density for this substrate in the final structure. The crystals gained a mild red coloration when co-crystallized with HOCbl and subsequently gained a deeper red coloration upon soaking, indicating some degree of binding. Whereas HOCbl is expected to have a lower affinity than reduced cob-(I)alamin, its dissociation constant for the human enzyme has been measured fluorimetrically as ϳ9 M (52). It is likely that the high concentration of ammonium sulfate in the crystallization solutions significantly reduces the overall binding affinity of HOCbl for the enzyme. Attempts to co-crystallize the TaPduO enzyme with cobalamin have also been unsuccessful (23). Nevertheless, it is likely that the corrinoid substrate is bound immediately adjacent to ATP in a large surface-exposed hole (Fig. 6). PduO catalyzes the direct transfer of the adenosyl moiety from ATP to reduced cobalamin, without proceeding through an enzyme-bound adenosylated intermediate (48). For such a direct transfer to take place, the corrinoid must be bound in close proximity to ATP. The hole immediately adjacent to ATP is large enough to accommodate the corrinoid but the orientation of the tetrapyrrole ring cannot be predicted because side chains lining the pocket will adopt different conformations on binding. Indeed, some lobes of electron density were interspersed throughout this hole at a signal level significantly exceeding background noise ( Ͼ 3) and several residues lining this cavity (Asn 19 , Phe 67 , Ser 83 , Ala 108 , Arg 128 , Arg 132 ) display alternate conformations. Furthermore, this cavity encompasses a localized area of positive electrostatic potential which may serve either to assist in binding the corrinoid substrate negatively charged phosphate moiety and/or to promote docking of  (20). Mechanistic Insights into Methylmalonic Aciduria-The kinetic and structural characterization of the PduO-type enzyme from L. reuteri provides insights into the recent description of mutations in hATR responsible for methylmalonic aciduria (23,47,53,54). Even though the sequence similarity between the human enzyme and LrPduO enzyme is only 39%, the similar kinetic constants and nucleotide specificity provide a structural framework for understanding the biochemical consequences of the mutations. Importantly, the majority of disease-related mutations are clustered around res-idues contributed from helix 5 to the active site (Fig. 7). Interestingly no mutations have been identified within the phosphate binding region on the N-terminal loop.
The R186W mutation of hATR is particularly common in patients, accounting for ϳ30% of sequenced alleles and is associated with early onset of methylmalonic aciduria (54). Purification and characterization of the mutant enzymes R186W and R186A in vitro has been shown to result in a complete loss of activity (23,47). The structurally equivalent residue of the LrPduO enzyme is Arg 128 . 8 Whereas this residue does not directly interact with ATP, it is expected to play a critical role in catalysis as evidenced by its absolute conservation among PduO-type adenosyltransferases and the complete loss of enzyme activity that results from mutations at this position. There are several possible roles for Arg 128 . 1) Its location in the corrinoid binding pocket results in an important specific interaction with the corrin ring. Efforts to co-crystallize LrPduO with various corrinoid substrates are continuing to better define the interactions between the enzyme and the corrinoid substrate. 2) It interacts directly to properly position or stabilize Arg 132 for catalysis, or 3) it forms a critical subunitsubunit contact through a salt bridge with the absolutely conserved Asp 35 on the adjacent subunit. On this note, it is interesting that in hATR the R186W mutation results in apparently unstable protein in vivo (47), suggesting that a stabilizing subunit-subunit contact is lost in these mutant enzymes.
Both R190C and R190H have been identified in patients with methylmalonic aciduria (54). This residue is structurally equivalent to Arg 132 in the LrPduO enzyme and is expected to be essential to catalysis as discussed above. It is also now clear that the E193K mutation found in patients with methylmalonic aciduria (equivalent to Glu 135 in LrPduO) will result in the loss of a specific interaction with ATP in the active site and in the loss of a critical secondary residue responsible for positioning one of two Mg 2ϩ -coordinating water molecules in the active site.
The structure of LrPduO presented here answers many of the questions of how nucleotides bind to this class of adenosyltransferase, but other questions remain. In particular, knowledge of how the corrinoid binds in the active site is required to establish a molecular mechanism for adenosyltransfer and the 8 Arg 128 has two conformations in the LrPduO structure. Each conformation has independently been observed in other deposited structures of PduO. In one conformation, Arg 128 forms an ideal salt bridge with Asp 35 (this single conformation has also been observed in the coordinates 2G2D and 1RTY, for putative PduOs from M. tuberculosis and B. subtilis, respectively) whereas in the alternate conformation, the guanidinium functional group is shifted to position a nitrogen directly above the guanidinium group of Arg 132 . (This equivalent position was observed in the coordinates 1NOG for the T. acidophilum enzyme.) FIGURE 6. Electrostatic surface potential stereoview for LrPduO centered on the putative cobalamin binding pocket. The electrostatic potential ranges from Ϫ20 kT (red) to ϩ10 kT (blue). A space-filling representation of Mg-ATP is colored in green. The electrostatic potential was generated with the program APBS (57) using a protein dielectric constant of 2 and a solvent dielectric constant of 80. role of conserved residues. The current study establishes the foundation for these future investigations.