Insight into the Role of Escherichia coli MobB in Molybdenum Cofactor Biosynthesis Based on the High Resolution Crystal Structure*

Two proteins, which are co-transcribed in Escherichia coli (MobA and MobB), are involved in the attachment of a nucleotide moiety to the molybdenum cofactor to form active molybdopterin guanine dinucleotide. Although not essential for this process, the dimeric MobB increases the activation of molybdoenzymes, incorporating this cofactor by a mechanism that is not understood. The structure of MobB has been elucidated in two crystal forms, one of which has provided a model at 1.9-Å resolution with Rwork and Rfree values of 21.5 and 28.7%, respectively. The MobB subunit displays an α/β-fold arranged into a major and minor domain, the latter of which is inserted between the major and minor domains of the partner subunit, creating an elongated dimer constructed around a 16-stranded β-sheet. Structural homologues have been identified, and they include a number of nucleotide-binding proteins. Comparisons indicate that although the phosphate-binding regions are highly conserved, MobB lacks the elements of structure required to interact with and efficiently bind a nucleotide base. In the present structure, a sulfate is bound to the Walker A phosphate-binding motif of MobB. The possibility that MobB forms a complex with the nucleotide-binding MobA, the protein with which it is co-transcribed, is explored, and modeling suggests that such a MobA:MobB complex is feasible. This hypothesis is supported by recent biochemical evidence indicating that MobB interacts with several proteins involved in various stages of molybdenum cofactor biosynthesis including MobA. We propose therefore that MobB is an adapter protein that acts in concert with MobA to achieve the efficient biosynthesis and utilization of molybdopterin guanine dinucleotide.

Molybdenum is an essential trace element associated with a diverse range of redox-active enzymes (1). Such molybdoenzymes catalyze basic metabolic reactions in the carbon, nitrogen, and sulfur cycles exploiting the redox chemistry of this second row transition metal. With the exception of the nitrogenases that contain an iron-molybdenum-sulfur cluster, molybdenum is integrated into the enzymes as the molybdenum cofactor (Moco), 1 which contains mononuclear molybdenum coordinated to an organic cofactor termed molybdopterin (MPT). Although Moco is the active form of the cofactor present in eukaryotes, most bacterial enzymes require the attachment of a nucleotide moiety onto the terminal phosphate group of the MPT side chain to be functionally active (2). In Escherichia coli, the active form of Moco is molybdopterin guanine dinucleotide (MGD, Fig. 1), molybdopterin covalently linked to GMP. The biosynthesis of Moco itself has been extensively studied and can be described in four stages (3): (a) conversion of a guanine nucleotide into a derivative termed precursor Z, (b) conversion of precursor Z into MPT, (c) chelation of molybdenum by MPT producing the molybdenum cofactor Moco, and (d) attachment of a nucleotide moiety to Moco, thus forming the active guanine dinucleotide form. Although it is not yet understood how Moco is inserted into molybdoenzymes, due to the intrinsic instability of the chemical species concerned, the intermediates and cofactor remain bound to proteins during the biosynthetic process until the final incorporation into the apomolybdoenzymes.
In E. coli, there are five genetic loci involved in the biosynthesis of active Moco (moa, mob, mod, moe, and mog), and mutations in any of these result in the loss of all molybdoenzyme activities. It has been shown that the moaA and moaC gene products are required for the biosynthesis of precursor Z from the guanine nucleotide (3); that MPT is synthesized from precursor Z by the gene products of moaD, moaE, and moeB (4); that moeA and mogA are necessary for the incorporation of molybdenum (5,6); and that modABC are required for the acquisition and transport of molybdenum into the cell (7). The final step of this biosynthetic pathway requires the mob locus, which encodes functions that catalyze the synthesis of MGD from MPT and a GMP donor, most probably GTP (8). The mob locus in E. coli contains two genes, mobA and mobB, arranged as a single transcription unit (8,9). MobA is a well characterized monomeric enzyme capable of the synthesis of MGD from MPT and GMP, and it has been shown to activate the molybdoenzyme nitrate reductase (10). MobB is not essential for MGD production or activation of nitrate reductase; however, its presence significantly enhances the yield of the active enzyme (11). The amino acid sequence of MobB (see Fig. 2A) reveals a putative nucleotide-binding motif, the Walker A motif ((G/ A)XXXXGK(S/T), see Ref. 12). This motif is frequently referred * This work was supported by awards from Biotechnology and Biological Sciences Research Council (UK) and The Wellcome Trust. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The to as the phosphate binding or P-loop, or alternatively, the G 1 motif of bacterial GTPases (13) and is often predictive for nucleotide triphosphate binding, suggesting that the protein might be involved in the binding of the guanine nucleotide. MobB does display weak binding of GTP (K d of 2.0 M) and low GTPase activity (K m ϭ 7.5 M, turnover rate ϭ 3 ϫ 10 Ϫ3 min Ϫ1 , see Ref. 14).
The crystal structures of MoaC (15), the MoeB-MoaD complex (16), MoeA (17,18), MogA (19), and MobA (20,21) have been solved and, with the exception of MobA, all form oligomers. This, along with biochemical evidence showing that MoeB and MoaD (22) and MobA, MobB, MogA, and MoeA (23) are involved in heterogeneous complex formation, indicates that protein-protein interactions are critical in molybdenum cofactor biosynthesis. The biological function of MobB is, unlike the other proteins discussed above, not clear. To investigate the role that MobB might contribute to Moco biosynthesis, we have determined the high resolution crystal structure of the E. coli protein.

MATERIALS AND METHODS
Cloning and Expression-The mobB gene from E. coli was amplified from genomic DNA by PCR. Two oligonucleotide primers were designed (5Ј-CTA-GCA-CAT-ATG-ATA-CCG-TTA-CTC-GCC-3Ј and 5Ј-GCA-CC-A-CTC-GAG-TTC-GTT-CTG-CTT-TTG-3Ј) to introduce unique restriction sites for NdeI and XhoI, respectively (shown in bold). The PCR product was cloned into the pCR-Blunt II-TOPO vector using the Zero Blunt TOPO PCR cloning kit (Invitrogen). Positive clones were sequenced, and then following excision, the mobB gene was ligated into expression vector pET15b (Novagen), which ultimately produces a protein carrying a hexahistidine tag at the N terminus. The new plasmid was heat-shock-transformed into E. coli BL21 (DE3) and selected on Luria-Bertani agar plates containing 100 g/ml ampicillin (LB/ampicillin). Bacteria were cultured in LB/ampicillin media and grown at 37°C to an optical density of 0.6 at 600 nm. Expression was induced with 0.4 mM isopropyl-␤-D-thiogalactopyranoside, and the temperature was reduced to 20°C. Cell growth continued for 16 h before cells were harvested by centrifugation (2500 ϫ g) for 20 min at 4°C. In addition to the expression of native MobB, selenomethionine (SeMet) MobB was expressed in E. coli B834 (DE3) cells. Cultivation of this variant was done in M9 minimal media supplemented with the usual amino acids except that SeMet (50 g/ml) was substituted for l-methionine. Protein expression was inducted, and cell growth was carried out in the same manner as for native MobB (above).
Purification-Once harvested, the cell pellets were resuspended in 50 mM Tris-HCl buffer, pH 7.5, and lysed by three passages through a French pressure cell at a pressure of 900 p.s.i. The insoluble cell fraction was separated by centrifugation (38,000 ϫ g) for 20 min at 4°C, and the supernatant was filtered before being applied to a Ni 2ϩ -resin HiTrap column (Amersham Biosciences) that had been equilibrated with 10 column volumes of the buffer. Elution of the protein was achieved by running a linear gradient of 0 -500 mM imidazole. Fractions were collected and analyzed by SDS-PAGE, and those containing MobB were pooled and dialyzed overnight against 50 mM Tris-HCl, pH 8.5. The N-terminal His 6 tag was cleaved with thrombin (10 units/mg) in 50 mM phosphate-buffered saline for 5 h at 20°C. The sample was then loaded onto a Poros HQ strong anion exchange column (Applied Biosystems). This column was equilibrated in 50 mM Tris-HCl, pH 8.0, and elution was achieved with a linear gradient of 0 -500 mM NaCl in the same buffer. The fractions containing MobB were again pooled, analyzed, and dialyzed overnight into 50 mM Tris-HCl, pH 8.0, and then concentrated to 10 mg/ml for use in crystallization trials.
Crystallization-Two crystal forms of MobB were obtained using the hanging drop vapor diffusion method. Monoclinic blocks (0.25 mm maximum dimension) were occasionally grown from a 4-l drop that consisted of 2 l of protein solution containing 5 mM MgCl 2 , 5 mM GDP, and 5 mM dithiothreitol as well as 2 l of the reservoir solution comprising 150 mM (NH 4 ) 2 SO 4 , 10.5% (w/v) polyethylene glycol 4000, and 15% (w/v) glycerol. A hexagonal crystal form (rods of 0.15 ϫ 0.15 ϫ 0.5 mm 3 ) proved more reproducible than the monoclinic form and was obtained using a reservoir of 100 mM NaAc and 2 M (NH 4 ) 2 SO 4 at pH 4.6.
Data Collection and Processing-A medium resolution multiwavelength anomalous dispersion data set was collected from the hexagonal crystal form of the SeMet protein at station ID29 at the European Synchrotron Radiation Facility using a CCD Area Detector Systems Corp. Q210 detector. These SeMet crystals belonged to the space group P6 4 22 with unit cell dimensions a ϭ 241.5, c ϭ 49.2 Å. The asymmetric unit contains two monomers (subunits A and B) giving a combined molecular mass of ϳ39 kDa with 77% solvent content and a Matthew's coefficient (24) V M of 5.4 Å 3 /Da. High resolution diffraction data were collected from a native monoclinic crystal on PX 9.6 at the Synchrotron Radiation Source, Daresbury Laboratory using a CCD ADSC Q4 detector. This crystal form has space group P2 1 with a ϭ 52.4, b ϭ 64.8, c ϭ 54.2 Å, and ␤ ϭ 97.7°. Here the solvent content for two molecules in the asymmetric unit was 48% (V M of 2.4 Å 3 /Da). Prior to each data collection, crystals were transferred through a cryoprotectant consisting of 50% glycerol and 50% of the relevant reservoir solution before being flashed-cooled to Ϫ173°C in a stream of nitrogen. The data were indexed, integrated, and scaled using either MOSFLM (25) or DENZO and SCALEPACK (26). Data manipulation was achieved using the CCP4 suite (27), and details are listed in Table I. a remote was used in the refinement of the P6 4 22 data. Numbers in parentheses correspond to the statistics in the highest resolution shell. b R sym ϭ ⌺͉I Ϫ ͗I͉͘/⌺I, where I is the measured intensity and ͗I͘ is the average intensity summed over all symmetry equivalent reflections.
is the reference structure factor from infl. and F PH is the structure factor of the derivative. Structure Solution and Refinement-The initial structure of a MobB subunit was determined using the multiwavelength anomalous dispersion method of phase determination targeting the K-edge of selenium and using the hexagonal crystal form. MobB has five methionine residues per subunit, and a homodimer constitutes the asymmetric unit. The program SOLVE (28,29) identified 10 selenium positions that provided phases to 2.4 Å with an overall figure-of-merit of 0.48. The 10 selenium positions corresponded to the location of 5 SeMet residues in one subunit and 4 in the other subunit with one SeMet side chain displaying a dual conformation. Density modification (without the use of non-crystallographic symmetry averaging) using RESOLVE (30) raised the figure-of-merit to 0.58. An initial round of model building, using O (31), provided only 165 out of a possible 348 residues, 118 of which were associated with a single subunit. One subunit was reasonably well defined in the experimental density, but the other was of poor quality. This model was used for molecular replacement with the P2 1 data set to utilize the higher resolution data. The program AMoRe (32) provided a clear molecular replacement solution for two subunits using data within the resolution limits of 15-3 Å, which, after rigid body refinement, gave an R-factor of 47.7% and correlation coefficient of 44.1%. Prior to refinement using REFMAC (33), 5% of the data were set aside for the calculation of the R free (34). Refinement was carried out with a bulk solvent correction but without the use of non-crystallographic symmetry restraints. Rigid body refinement of the dimer was used to extend the resolution to 1.9 Å, and several cycles of refinement, inter-dispersed with rounds of model building, solvent, and sulfate identification, produced a model with an R work of 21.4% and R free of 28.7%. For completeness, this model was used to initiate refinement with the hexagonal data by superimposing the dimer onto the well defined subunit described above. Rigid body refinement and several rounds of graphics inspection, rebuilding, and refinement gave an R work of 31.1% and an R free of 33.8%. PROCHECK (35) was used to assess the geometry of models during the analysis, and refinement statistics are presented for the monoclinic crystal form in Table II.

RESULTS AND DISCUSSION
The Structure of MobB-Initial phases to 2.4 Å were obtained using the multiwavelength anomalous dispersion method for phase determination on the hexagonal crystal form. These crystals diffracted only to medium resolution and were poorly ordered but nevertheless enabled us to construct the first model of MobB. This model proved suitable for molecular replacement into and refinement with the high resolution native monoclinic data. Although the hexagonal crystal structure was subsequently refined, the data and resulting model are inferior to those derived from the monoclinic form, and consequently, our discussion concentrates on the 1.9-Å resolution structure.
MobB consists of 175 residues with an approximate molecular mass of 19.4 kDa. The refined model comprises residues 6 -45 and 58 -174 for subunit A and residues 6 -175 for subunit B; this accounts for 327 out of a possible 350 residues for this homodimer. Secondary structure elements were assigned by a combination of automated methods, using PROMOTIF (36), and visual inspection. They are mapped onto the amino acid sequence in Fig. 2A.
The MobB subunit exhibits a mixed ␣/␤-fold with an overall topology of eight ␤-strands and seven ␣-helices (Figs. 2 and 3) and is divided into two domains. The major domain is constructed with a core six-stranded twisted parallel ␤-sheet with strand order 8-7-6-1-5-2, two helices on one side of the sheet (␣1 and ␣7), three shorter helical sections on the other side (␣3, ␣4, ␣6) with a single turn of helix (␣5) at the C-terminal edge of the sheet. The minor domain consists of a distorted ␣2, then antiparallel strands ␤3 and ␤4. We note that ␣2 is only ordered in one of the two subunits that comprise the asymmetric unit (Fig. 4).
The symmetric dimer of MobB is formed by the insertion of a minor domain from one subunit between the major and minor domains of the other. Strands ␤3 and ␤4 from one subunit are positioned between ␤2 and ␤4 of the partner and arranged such that although ␤2 is parallel to ␤3, the ␤4 strands are antiparallel to each other (Figs. 2b and 3). This arrangement produces a dimer, with overall dimensions of ϳ90 ϫ 38 ϫ 26 Å 3 , dominated by an elongated ␤-sheet composed of 16 strands. The sheet is flat in the central section of the dimer and twisted at either end (Fig. 3). The non-crystallographic symmetry 2-fold axis is perpendicular to the central section of the extended ␤-sheet.
Although GDP was present in the crystallization conditions, we note that there was no electron density corresponding to the putative ligand in either crystal form. However, sulfate ions were identified binding at the Walker A motif that lies between ␤1 and ␣1. The oxyanion binding is likely influenced by the helix macrodipole of ␣1 (Fig. 3) (37). Further details of the sulfate-binding site will follow.
Structural Homologues-Sequence homology among GTPbinding proteins (G-proteins) is well characterized (38), and the GDP-binding site is typically described by three highly conserved sequence motifs (39). The first, the Walker A motif (13), occurs at residues 14 -21 ( Fig. 2A). This phosphate-binding region adopts a well conserved structure in G-proteins, consisting of a ␤-strand and ␣-helix with a loop between them (39). The helix dipole interacts favorably with the nucleotide phosphates. The conformation of the region surrounding the Walker A motif and the position of the sulfate ion in MobB follow the general trends displayed for phosphate binding by G-proteins. The second motif, DxxG, is linked to a conformational switch in G-proteins dependent on whether GDP or GTP is present. Although E. coli MobB carries the sequence DxxG close to the phosphate-binding loop at residues 51-54, we note that such a motif is absent in MobB homologues. The third motif, which is predicted to determine the specificity of an enzyme for guanine (NXXD), is not present in MobB.
In addition to the conservation of three motifs, G-proteins share a common core structure of five ␣-helices and a sixstranded ␤-sheet. The ␤-sheet has five parallel strands and one antiparallel strand that, unusually for ␣-␤-proteins, is placed at the edge of this sheet. Since MobB displays some of the basic structural elements of G-proteins, a search for structural homologues was considered necessary to investigate the possible function of the protein.
An architectural comparison, using DALI (40) and DEJAVU (41), indicted that the MobB subunit is structurally similar to  43), and also the GTPase domain of the signal recognition particle receptor, FtsY (Protein Data Bank accession code 1FTS; Z score ϭ 7.6; see Ref. 44). The Z-score is a measure of the statistical significance of the best domain-domain alignment and was determined by DALI. Typically, two dissimilar proteins will have a Z-score less than 2.0, and the MobB subunit matched with itself gives a Z-score of 27.3.
The three structural homologues that were identified display similarities to the Ras-related GTPases. They were aligned on the most complete subunit of MobB using LSQKAB (45), and secondary structure matching was carried out using the secondary structure matching server at the European Bioinformatics Institute (www.ebi.ac.uk/msd-srv/ssm/cgi-bin/ ssmserver). The root mean squared fit was calculated from the C␣ atoms of matched residues at the best superposition of the structures. The fractions of the C␣ atoms in the MobB subunit that overlay on NIP, MinD, and FtsY are 71, 64, and 54% with root mean squared deviations of 3.2, 2.2, and 2.0 Å, respectively. The secondary structure elements of MobB found to align with all three structural homologues were ␤1, ␣1, the P-loop, ␤2, ␣4, ␤5, ␤6, ␤7, ␤8, and ␣7, which constitutes almost the entire major domain of MobB. Since the searches for structural homologues revealed homology only with the major domain, a further search was carried out using only the minor domains as organized in the dimer, but no structural homologues were identified.
A superposition reveals that the phosphate-binding region (␤1, the P-loop, and ␣1) assumes an almost identical conformation in all four structures (Fig. 5). ADP is present in the crystal structures of both NIP and MinD, and the nucleotides superpose well (not shown). The sulfate in the MobB structure occupies the same position as the ␤-phosphate group of the bound ADP molecules. This sulfate accepts hydrogen bonds from the main chain amides of Gly 16 , Gly 18 , Lys 19 , and Thr 20 , and we note that both NIP and MinD have identical residues at equivalent positions that interact with ␣or ␤-phosphate oxygen atoms (Fig. 6). In both NIP and MinD, a threonine (Thr 18 , MinD numbering) binds the ␣-phosphate oxygen using the side chain hydroxyl and main chain amide group (43), and in MobB, the equivalent residue (Thr 21 ) interacts with a water molecule occupying the same position as an ␣-phosphate oxygen in the MinD complex (Fig. 6). It is likely that MobB would bind diphosphate in similar fashion to NIP and MinD.
Although their interactions with oxyanions appear similar, the structural comparison of the four homologues reveals significant differences between MobB and the others in the nucleotide-binding pocket. MobB lacks a section of ϳ20 residues between ␤8 and ␣6 (Fig. 5). In MinD and NIP, these residues extend the ␤8 and ␣6 equivalents of MobB and form a helixloop combination that creates the nucleotide-binding pocket. In MinD, the helix is ␣9, and the extended helix is ␣10 (42,43). In particular two residues (with MinD numbering) Pro 198 and Asp 200 , which are positioned on these extensions, are conserved in MinD and NIP and provide hydrogen bonding interactions between the main chain and the nucleotide. MobB is also different in the ␤7-loop-␤8 section. In MinD, NIP, and FtsY, the strand corresponding to MobB ␤7 is extended toward the adenine position, creating an integral part of the nucleotide-binding pocket. This loop consists of about 10 amino acids (leading into ␣8 of MinD) with a structurally conserved asparagine in all three homologues (not shown). In MinD, this Asn 171 directly contacts the base. In FtsY, further hydrogen bonds to the adenine are provided by an additional asparagine present in this loop (not shown) (44). From the structural overlay, it is obvious that MobB does not have the necessary structural features that are utilized by MinD, NIP, and FtsY to efficiently bind the nucleotide base. We conclude that MobB, on its own, would not be an efficient nucleotide-binding protein.
Protein-Protein Complexes-MobB has the correct sequence and structural elements for the efficient binding of the phosphate tail of a nucleotide but not of the base. In contrast, a crystal structure of MobA, the protein that is co-transcribed with MobB, reveals that MobA has little interaction with the phosphate tail of the nucleotide but strong interactions with the base (21). Also, near the putative phosphate-binding region, there is a disordered loop that is highly conserved in MobA sequences and that is termed the consensus loop (20). This loop is glycine-rich but does not contain a phosphatebinding (Walker A) motif. Moreover, the N-and C-terminal ends of the consensus loop are directed away from the phosphate tail of the bound GTP and could not contribute to its binding (21). Consequently, the possibility of MobB binding the phosphate moiety of GTP in the MobA:GTP complex was explored.
The sulfate bound to MobB occupies the equivalent position of the ␤-phosphate of ADP in the MinD complex (see above). Therefore, the ADP coordinates were taken from the aligned MinD structure and (without further manipulation) placed into the MobB structure. This ADP molecule was superimposed onto the bound GTP in the MobA complex. Since MobB is a dimer, another molecule of MobA was placed onto the second subunit of MobB in a similar fashion to generate a model for a putative MobA:MobB complex (Fig. 7).
In this model, the GTP fits neatly into a pocket formed at the protein-protein interface with no steric hindrance. There is only one area where steric clash of C␣ atoms might occur, around residues 121-125 in MobB and 154 -157 in MobA (Fig.  7). These regions along with several other sections of the proteins at the putative interface are conserved in homologous proteins. In MobA, this includes residues 75-79, 128 -134, 150 -158, and 182-192 (20), and in MobB, this includes resi-  dues 41-50 and 106 -111 ( Fig. 2A). Residues 121-127, 142-145, and 155-159 of MobB are also located at the putative interface (Fig. 7). In addition, a number of disordered and presumably flexible regions of both MobA and MobB are positioned at this interface. In MobB, the N-and C-terminal ends of the disordered residues A46-A57 sit near this interface, and although the corresponding residues of the partner subunit (B46-B57) are well ordered, this section of the structure is stabilized by crystal lattice contacts. In the P6 4 22 crystal form of MobB, this region is disordered in both subunits, which suggests that this section of the protein is conformationally labile. The same conclusion applies to the consensus loop of MobA, positioned at this putative MobA:MobB interface, which is poorly defined in the crystal structures solved with or without nucleotide bound. The high sequence conservation of sections of MobA and MobB that display conformational flexibility and that are placed at the putative MobA: MobB interface suggests that our model may have some merit and hints that such a protein-protein association may be beneficial for these proteins to act in concert during MGD biosynthesis. The function of MobB might therefore be that of an adapter protein during MGD biosynthesis.
Specific protein-protein associations have already been shown to play a central role in the early stages of molybdenumcofactor biosynthesis (46), and our modeling has suggested that MobB might form a protein-protein complex with MobA. It is particularly gratifying that during the course of this analysis, Magalon et al. (23), with the use of a bacterial two-hybrid approach, were able to demonstrate that MobB interacts with MogA, MoeA, and MobA in vivo. These observations are compatible with our postulated function of MobA as an adapter protein.
Concluding Remarks-The crystal structure of MobB and comparisons with structural homologues provide clear evidence that MobB by itself is not designed for binding and processing nucleotides with high efficiency, although it does have a phosphate-binding loop characteristic of some nucleotide-binding proteins. Access to three-dimensional coordinates for MobA and MobB allowed modeling studies, which suggest that a MobA:MobB protein complex is feasible and that such a complex might enhance the efficiency of the conversion of Moco to MGD. Our analysis illustrates the value of three-dimensional knowledge in investigating a functional role for MobB and points the way forward for further biochemical studies required to clarify details of the role of this protein role in MGD biosynthesis.