Oxyanion Binding Alters Conformation and Quaternary Structure of the C-terminal Domain of the Transcriptional Regulator ModE

The molybdate-dependent transcriptional regulator ModE of Escherichia colifunctions as a sensor of intracellular molybdate concentration and a regulator for the transcription of several operons that control the uptake and utilization of molybdenum. We present two high-resolution crystal structures of the C-terminal oxyanion-binding domain in complex with molybdate and tungstate. The ligands bind between subunits at the dimerization interface, and analysis reveals that oxyanion selectivity is determined primarily by size. The relevance of the structures is indicated by fluorescence measurements, which show that the oxyanion binding properties of the C-terminal domain of ModE are similar to those of the full-length protein. Comparisons with the apoprotein structure have identified structural rearrangements that occur on binding oxyanion. This molybdate-dependent conformational switch promotes a change in shape and alterations to the surface of the protein and may provide the signal for recruitment of other proteins to construct the machinery for transcription. Sequence and structure-based comparisons lead to a classification of molybdate-binding proteins.

Molybdenum is an essential trace element required for the catalytic activity of several enzymes in animals, plants, and bacteria. In some cases the transition metal is complexed with a unique pterin forming the molybdopterin cofactor, and in others it forms part of an iron-molybdenum cluster cofactor (1,2). Escherichia coli acquires molybdenum in the form of MoO 4 2Ϫ by using a high-affinity ABC-type molybdate transporter system encoded by the modABCD operon (3). ModA, the structure of which has been determined (4,5), is similar to the sulfatebinding protein of Salmonella typhimurium (SBP) 1 (6) and a member of the periplasmic-binding protein family. ModA binds and transfers molybdate to ModB at the outer surface of the cytoplasmic membrane. ModB is an integral membrane protein and ModC a membrane associated protein; together they transport the oxyanion across the membrane using an ATP-dependent mechanism. The role of ModD is at present unknown. Exposure of E. coli to high levels of molybdate leads to repression of modABCD by a mechanism that involves the molybdatedependent transcriptional regulator known as ModE (7)(8)(9). The modE gene, situated immediately upstream from the mod-ABCD operon, codes for a transcriptional regulator able to control the uptake and utilization of this transition metal. ModE binds molybdate, and the complex can function as a repressor by binding at a site that overlaps the transcription start of the modABCD operon (10 -12). In addition, the ModEmolybdate complex acts as a positive regulator of several genes including those coding for the molybdoenzymes dimethyl sulfoxide reductase (13) and nitrate reductase A (14), as well as for hyc, the hydrogenase 3 structural operon. It also mediates expression of the moaABCDE operon, which encodes the first enzymes required for the biosynthesis of molybdopterin (15).
E. coli ModE functions as a homodimer of subunits of 262 amino acids. The protein binds MoO 4 2Ϫ and WO 4 2Ϫ with a high affinity (K d ϳ0.8 M in each case) and stoichiometry of 2 oxyanions/dimer. Oxyanion binding produces a large quench (50%) of the intrinsic protein fluorescence accompanied by a blueshift of the emission spectrum maximum, which indicates that a conformational change is induced by these ligands (11). The crystal structure of apo-ModE revealed that the protein forms distinct N-and C-terminal domains (Fig. 1, Ref. 16). The Nterminal domain, comprising 121 residues, is mostly ␣-helical (60% ␣-helix, 20% ␤-strand) with a winged helix-turn-helix DNA binding motif. The C-terminal molybdate-binding domains, which display a pronounced asymmetry, comprise residues 122-262 and contain 60% ␤-strands, which form two ␤-barrels. The first barrel is constructed from a combination of strands ␤6 -␤9 with ␤15 and the second from strands ␤10 -␤14. Each barrel constitutes a subdomain ( Fig. 1) of similar folds, although the order of secondary structure elements with respect to the sequence varies. The subdomain fold is related to the oligomer binding fold (OB-fold (17)), a five-stranded Greekkey ␤-barrel capped by an ␣-helix. The subdomains share significant sequence and structural homology with each other and also with the 7-kDa molybdenum-containing Mop proteins from Clostridium pasteurianum (18,19) and Sporomusa ovata (20). These proteins are implicated in the molybdenum metabolism of a variety of microorganisms (3,11,21). We have previously shown that each subdomain corresponds to the structure of a single Mop unit and termed the C-terminal domain of ModE the DiMop domain (16). The term molbindin has been introduced (19) to identify Mop-like proteins, and we made the further distinction that this refers to proteins composed of only Mop domains.
Although the structure of ModE has been determined, there are a paucity of data concerning how the oxyanion ligands are bound, since we have been unable to determine the structure of the complete ModE-molybdate complex. Recent analysis of the Mop protein from S. ovata provided the first structural details of oxyanion binding to a molbindin, in this case tungstate (20), the results of which will be described later. To study the protein-oxyanion interactions relevant to the function of ModE we cloned and characterized the C-terminal (DiMop) domain. We show that this domain retains the ability to bind oxyanions selectively with properties similar to the intact protein, having determined high-resolution crystal structures of both the molybdate and tungstate complexes. These structures provide high-resolution information on the protein-ligand interactions and allow us to compare the apo-DiMop component of ModE with the oxyanion-loaded domain, a comparison that identifies a conformational change induced by oxyanion binding. Sequence and structural comparisons with other Mop proteins led us to propose a classification of the proteins that utilize the Mop domain.

MATERIALS AND METHODS
Expression of the C-terminal DiMop Domain of ModE-A fragment of the modE gene coding for the C-terminal domain was amplified by polymerase chain reaction using pHW121 as the template plasmid, which carries a fragment covering the modE region in the vector pUC18 (8). The oligonucleotide primer used for the 5Ј-end of the C-terminal domain-encoding DNA was 5Ј-G CGC CAT ATG CAA ACC AGC-3Ј, which incorporated an NdeI recognition site (underlined) and covered the first 3 codons of the C-terminal domain. The primer used for the 3Ј-end of the C-terminal domain was 5Ј-GCG CGG ATC CCA CGC TTA GCG CAG-3Ј, which covered the last 6 codons of modE and contained a BamHI recognition site (underlined). The polymerase chain reactionamplified fragment was cloned as an 0.45-kilobase NdeI-BamHI insert in the N-terminal hexa-His tag expression vector pET15b (Novagen) to create plasmid Plaa7. The expression of protein is under the control of the T7-Lac promoter, and the recombinant DiMop domain carries a thrombin cleavage site between the protein and the His tag.
Plasmid Plaa7 was heat shock-transformed into E. coli BL21(DE3), and transformants were selected on Luria Bertani (LB) agar plates containing 100 g ml Ϫ1 ampicillin. A transformant was grown at 37°C in LB broth with ampicillin to mid-log phase at which point 0.4 mM (final concentration) isopropyl-␤-D-thiogalactopyranoside was added and growth continued with vigorous aeration for a further 4 h. Cells were harvested by centrifugation (2500 ϫ g) at 4°C and were then resuspended in 50 mM Tris-HCl, pH 7.6, 250 mM NaCl, 5 mM benzamidine and stored at Ϫ20°C. Bacterial cells were broken by passage through a French press, the insoluble cell debris pelleted by centrifugation at 4°C (18000 ϫ g) for 15 min, and the cell extract passed through a 0.2-m filter and applied to a 5-ml metal chelate affinity column (Hi-Trap, Amersham Pharmacia Biotech) charged using nickel chloride. The column was washed with four column volumes of 50 mM Tris-HCl, pH 7.6, 250 mM NaCl, and 5 mM benzamidine. The His-tagged protein was eluted with a 0 -500 mM imidazole gradient in the same buffer and then incubated with thrombin (Amersham Pharmacia Biotech) for 12 h at 20°C to remove the histidine tag. The DiMop domain was separated from the thrombin, uncleaved fusion protein, and Nterminal peptide by strong anion exchange chromatography using a RESOURCE TM Q (Amersham Pharmacia Biotech) column on a BioCAD 700E work station. Pooled fractions were concentrated to 20 mg ml Ϫ1 in 50 mM Tris-HCl, pH 7.6, and sample purity was assessed by SDSpolyacrylamide gel electrophoresis and MALDI-TOF (matrix-assisted laser desorption time-of-flight) mass spectrometry (Voyager DE STR; PerSeptive Biosystems). Protein concentration was estimated spectrophotometrically at 280 nm using the calculated molar extinction coefficient of 12660 M Ϫ1 cm Ϫ1 (22). The yield of purified protein was ϳ40 mg liter Ϫ1 of bacterial culture. Following thrombin cleavage the DiMop domain consisted of residues 124 -262 of ModE with a Gly-Ser-His extension at the N terminus. For reasons of consistency we maintained the amino acid numbering of the intact ModE.
Fluorescence Measurements-Intrinsic protein fluorescence spectra were recorded using a PerkinElmer LS50B spectrofluorimeter maintained at 25°C. Excitation was at 295 nm (5 nm bandpass for both excitation and emission) at a scan rate of 100 nm min Ϫ1 with a protein concentration of 29 M of monomer in 50 mM Tris-HCl, pH 7.6. Binding parameters were determined by changes in fluorescence at 347 nm on successive additions of sodium molybdate to a solution of protein (38 M monomer). Corrections were made for dilution effects but not for the inner filter effect because molybdate does not absorb significantly at the wavelengths used.
Crystal Growth and Data Collection-Crystals were grown at 20°C using the hanging drop method by mixing 3 l of protein solution (16 mg ml Ϫ1 protein, 50 mM Tris-HCl, pH 7.6, 10 mM Na 2 MoO 4 or Na 2 WO 4 ) with 3 l of a reservoir solution. Similar crystals were obtained with reservoirs of 400 l of 1.0 M trisodium citrate, 100 mM HEPES, pH 7.6, and 10% mM ethylene glycol or 1.4 M ammonium sulfate in 100 mM Tris-HCl, pH 8.4. The crystals, which attained a size of 0.3 mm in all dimensions over a period of 2 days, were cryoprotected with ethylene glycol and cooled to Ϫ173°C, and the data were then measured using an RAXIS IV image plate with a RU-200 rotating anode (Cu K ␣ '3d 1.54Å). Two data sets were measured, one from a crystal grown in the presence of WO 4 2Ϫ using tri-sodium citrate as the precipitant and the other from a crystal obtained in the presence of MoO 4 2Ϫ using ammonium sulfate as precipitant. The crystals are tetragonal with space group P4 1 with two DiMop domains in the asymmetric unit and about 42% solvent content. All diffraction data were processed, reduced, and scaled using the HKL suite (23) and were then analyzed using CCP4 software (24) with 5% of the reflections used for calculation of the R-free. Further details are given in Table I.
The structure of the DiMop-WO 4 2Ϫ complex was solved by molecular replacement with the program AMORE (25) using a polyalanine search model constructed from a monomer of the DiMop domain from apo-ModE (16). An optimal position for one molecule was found by using an integration radius of 20 Å and all data between 15 and 3 Å resolution (correlation coefficient 0.57). Positioning the second molecule in the asymmetric unit gave a correlation coefficient of 0.56. The resulting electron density maps indicated significant structural changes between the search model and the oxyanion complex. The tungstate positions were identified and incorporated into the model, which was then used to generate phases for use in the warpNtrace mode of ARP/wARP (26). Data to 1.9 Å were used, and a new model containing 268 of 280 residues with an R-factor of 25.1% and an R-free of 27.7% was obtained.
Interactive computer graphics model building was carried out using the program O (27). A Ni 2ϩ ion was located in the same position of the tungstate complex as seen in one of the subunits of the apo-ModE structure and is considered an artifact of the purification procedure (16). Refinement was completed using REFMAC (28) combined with ARPP (29) to place the water molecules. The molybdate complex was solved using the structure of the tungstate-bound form with the water molecules removed and the tungstate replaced with molybdate as a starting model for calculating phases for use in the auto model building procedure in ARP/wARP. The model was refined in a procedure identical to that used for the tungstate complex; relevant statistics are presented in Table I. The molybdate and tungstate complexes are essentially identical, with an r.m.s. deviation of 0.3 Å for superposition of all C␣ atoms. We concentrated on the molybdate-bound structure, although for completeness some details of the tungstate complex are provided. The figures were produced using ALSCRIPT (30), MOL-SCRIPT (31), and RASTER-3D (32).
Coordinates-Atomic coordinates and structure factors have been deposited with the Research Collaboratory for Structural Bioinformatics Protein Data Bank (accession codes 1H9R/1H9S for the tungstate and molybdate complexes, respectively).

Properties of the DiMop ModE Fragment-The C-terminal
DiMop fragment of ModE was expressed and purified in high yield, and analytical gel filtration analysis (data not shown) revealed it to be a dimer in solution. The DiMop fragment retains two of the three tryptophan residues of ModE (Trp 131 , Trp 186 ), and the intrinsic protein fluorescence emission spectrum is similar to that reported for intact ModE with max at 347 nm. Upon addition of 0.5 mM sodium molybdate to the DiMop fragment, the fluorescence emission spectrum was reduced by more than 60% of its value in the absence of molybdate (data not shown). This molybdate-induced quench is slightly greater than that reported for intact ModE (50% (11)), clearly demonstrating that the DiMop fragment retains the ability to bind molybdate. Moreover, it suggests that the third tryptophan (Trp 49 ) of ModE, located in the DNA-binding domain, makes a less important contribution to the molybdateinduced fluorescence quench of ModE.
Titration of the DiMop fluorescence quench provided an estimate of the stoichiometry of molybdate bound per monomer fragment of 1 with a K d of ϳ0.5 M (Fig. 2). These values are in excellent agreement with those reported for molybdate binding to intact ModE; stoichiometry of 1 molybdate per monomer with a K d of 0.8 M. The DiMop fragment effectively retains the molybdate binding capacity and properties of intact ModE; this observation validated our approach to the exploration of ModE molybdate interactions by crystallographic analysis of the ligand-bound DiMop fragment.
The Molybdate Binding Site-The DiMop dimer binds two oxyanions, in sites designated I and II, at the dimerization interface ( Fig. 3) using residues from both polypeptide chains. Each ligand binding site is created by the turn between ␤5 and ␤6 from one DiMop and helix ␣6 and the C-terminal region of strand ␤15 from the other (Fig. 3B). The oxyanion is held in position by accepting nine hydrogens bonds with each oxygen participating in at least two such interactions. Five of the hydrogen bonds are donated from one polypeptide, three from the other, and one from a water molecule (Table II, Fig. 4). The MoO 4 2Ϫ interacts with a single basic residue, Lys 183 . Other interactions from the same subunit involve hydrogen bonds donated from the main chain amides of Arg 128 , Ala 184 , and the side chain OH of Ser 126 . The MoO 4 2Ϫ is bound to the other subunit by hydrogen bonds donated from Ser 166 OH and both the OH and the main chain amide of Thr 163 . Sequence comparisons showed that Ser 126 , Arg 128 , Thr 163 , and Lys 183 are conserved in all Mop-like sequences (Fig. 5), and it was proposed that the residues were involved in oxyanion binding (19). Arg 128 binds the oxyanion using the main chain amide. The side chain forms an intramolecular salt bridge to Glu 218 (not shown) and contributes to the stabilization of the tertiary structure and formation of the ligand-binding pocket.
The charge compensation for the oxyanion is achieved by interactions with a single basic side chain (Lys 183 ) and partial charges contributed from the three main chain NH groups in the binding site. A strong peak in the electron density near the molybdate was assigned as a water molecule, rather than a monovalent cation such as sodium, on the basis of a water-like hydrogen bonding pattern. This water forms hydrogen bonds with the side chain OD1 of Gln 144 from one subunit and the carbonyl oxygen of Thr 232 of the partner subunit. Of note is a particularly short contact, 3.0 and 3.1 Å, between the metals in each site and the carbonyl oxygens of Ser 126 (Fig. 4) (4) and Azotobacter vinelandii ModA (5). The volume of these molybdate-binding pockets is larger than that calculated for the sulfate-binding pocket in SBP, 59 Å 3 . The size of the oxyanion must therefore be important for selectivity.
ModA also exhibits a similar selectivity of tungstate and molybdate over other oxyanions even though it has a fold and binding site similar to SBP (4). ModA binds the oxyanion in a deep cleft formed at the junction of two globular domains. There are seven direct hydrogen bonds, all donated from main chain amides and side chain hydroxyls of neutral protein ligands (4,5), and a contribution to binding from N-terminal ␣-helix dipoles. Although the DiMop domain is a different fold compared with ModA, both proteins use three OH groups from the Ser or Thr residues in ligand binding and have residues that bind ligands via a chelate-type interaction using both main chain nitrogen and side chain OH groups.
Molybdate-specific proteins, irrespective of fold, have larger binding sites and use size in conjunction with charge compensation to determine oxyanion selectivity. The mode of dimerization imposes size restrictions because the binding of a smaller oxyanion would require readjustment of the protein backbone and dimerization interface to optimize the length and angles of the hydrogen bonds involved in ligand binding.
The Overall Structure of the Ligand-bound DiMop Domain and Comparisons with the Apo Form-Like intact ModE, the ligand-bound DiMop forms a homodimer with a secondary structure similar to the apoprotein (Figs. 1 and 3). An obvious difference between the two structures is a decrease in the asymmetry between the partner subunits that was observed in the apoprotein structure (16). A least-squares superposition of the C␣ atoms of chain A onto chain B gives an r.m.s. deviation of 1.0 Å for the apoprotein, whereas for the DiMop-molybdate complex the r.m.s. deviation is lowered to 0.6 Å. Residues 138 -144 were excluded from this superposition because they are disordered in chain B of apo-ModE.
The binding of molybdate induces a change in the quaternary structure that, as a first approximation, can be described as the rigid body movement of the four Mop domains against each other; when aligned on Mop1A (as described in the legend for Fig. 6), the center of mass of Mop1B moves about 3.0 Å between the apo and MoO 4 2Ϫ -bound state. As this movement is almost perpendicular to the line connecting the centers of mass of Mop1A and -1B, the distance between the Mop1s remains practically constant. In the same alignment the center of mass of Mop2A moves about 1.9 Å, mostly toward Mop1B on the other chain. Overall these movements lead to the molybdatebound protein being more compact than the corresponding part of the apo structure, with the buried surface area increasing from 2 ϫ 750.5 Å 2 for the apo-dimer to 2 ϫ 1442.5 Å 2 for the ligand-bound form. For consistency residues B138 -B144 were removed from the complex structure prior to the calculation. The reduction of exposed surface is caused only in part by the rigid body movements; a significant contribution is also made by a number of loops that move relative to their surroundings. The largest change (with C␣ movements up to 12.6 Å) is observed for residues 138 -144 (loop I, Fig. 6). As mentioned above, this loop is partially disordered in the apoprotein with no density observed for one chain. However, it becomes ordered upon molybdate binding, moving to cover another flexible loop

FIG. 3. The dimeric assembly and architecture of the DiMop domain of ModE complexed with molybdate (shown as van der Waals spheres with molybdenum in cyan and oxygen in red)
occupying binding sites I and II. The same color scheme is used as in Fig. 1. A, view with Mop1B and Mop2B in a similar orientation to the corresponding subunit shown in Fig. 1. B, view orthogonal to A with selected elements of secondary structure labeled. formed by residues 163-166 (loop II). This smaller loop is important because its movement is crucial for the formation of the molybdate binding site, one-half of which is constituted by residues 163 and 166 (see above) after the loop has moved Ϸ2 Å toward the anion. This movement closes the molybdate binding site and leads to the formation of new hydrogen bonds across the intersubunit interface, e.g. between the backbone oxygen of Ser 126 and the hydroxyl of Ser 166Ј , as well as the NH2 of Arg 169Ј or between the guanidino moiety of Arg 128 and the Thr 163Ј hydroxyl (prime is used to denote a residue on the partner subunit). The latter two residues are also involved in molybdate binding and are conserved in virtually all known Mop domains (Fig. 5). Residues 152-155 form a third mobile loop (loop III). In the apoprotein it contacts the symmetryrelated loop of the Mop2 of the same chain (residues 224 -226), and its movement upon molybdate binding follows, to a limited extent, that of the Mop2. We noted that all mobile loops are located on the Mop1 domain and that Mop2 actually behaves like a rigid body.
We can now describe the probable order in which molybdate binding and the subsequent structural changes take place. We propose that molybdate initially binds to the half-site formed by residues 126, 128, 183, and 184 of one protein chain. The bound molybdate then effects the movement of loop II of the partner chain and with that consequently the rigid body movement of the partner subunit. This seems a reasonable assumption, as the binding site components of residues 126, 128, 183, and 184 are practically unchanged in the two structures and can be considered a preformed binding site, whereas loop II moves and is clearly influenced by molybdate binding. Further support for this half-site being responsible for initial ligand binding comes from the unequal distribution of protein-ligand interactions; as described above this is where three of the four oxygens of the molybdate are bound, whereas the loop II component provided by the other subunit binds only one.
The closing of the binding site by the movement of loop II then could allow the hitherto solvent-exposed and flexible loop I to rearrange itself and cover loop II, which leads to the formation of a number of new hydrogen bonds, e.g. Gln 145Ј backbone oxygen and nitrogen with Ile 162Ј backbone nitrogen and oxygen, respectively, or Val 143Ј backbone oxygen with Ala 164Ј backbone N, and also improves the hydrophobic packing. The now ordered loop I might then attract the loop formed by residues 210 -216 of the other chain, thereby effecting the observed movement of that Mop2 domain and leading to the formation of further hydrophobic contacts. Finally the loop III of the other chain moves to "follow" the Mop2 domain movement.
The fluorescence perturbation experiments indicate that the environments of one or both tryptophans in the DiMop domain change on binding MoO 4 2Ϫ , and the crystal structures provide an explanation. When the apo and ligand-bound forms of the DiMop domains are compared, there is no significant change in the environment of Trp 131 (not shown), which is located on strand ␤6 and positioned in a cavity formed by the residues linking ␤7 and ␤8. Trp 186 , however, is at the dimerization interface on a 3 10 helix. In apo-ModE the Trp 186 of subunit A is aligned parallel to Trp 186 from the B subunit and separated by about 6 Å (Fig. 1). The adjustment of the DiMop domains on binding MoO 4 2Ϫ alters the side chain conformation, renders the tryptophans more buried, and reduces their distance to 3.3 Å (Fig. 3B). In apo-ModE, Trp 186 contributes 1% of the DiMop dimerization interface, but in the ligand-bound form this now increases significantly to 5% of a greater surface area. We therefore attribute the spectral changes observed on binding oxyanion to alterations associated with the environment of Trp 186 . It follows then that changes in the environment of Trp 186 on the addition of molybdate lead to the changes observed in the near UV CD spectrum noted for ModE (11).
Implications for Other Mop Proteins-Mop domains occur in a variety of proteins other than ModE; the molbindins consist of either one (e.g. Mop from Hemeophilus influenzae or Mop I-III from C. pasteurianum) or two (e.g. ModG from A. vinelandii) Mop domains. The function of these small proteins is unclear, but it has been suggested that they store molybdate/ tungstate and are thus involved in maintaining homeostasis of these anions. As molybdate binding occurs at the interface between two Mop units, it is obvious that functional molbindins must be at least dimeric. The structure of a single Mop protein from S. ovata (20) revealed a hexameric arrangement (Fig. 7A) that provides three binding interfaces, similar to those described for DiMop, at which six WO 4 2Ϫ are bound. The highly symmetric assembly creates two additional binding sites located on a 3-fold axis and formed by residues from three neighboring Mop domains (A/C/E and B/D/F) and therefore do not occur in ModE (Fig. 7).
A Mop-like sequence is also observed as a C-terminal extension in the ATP-binding subunit of the mod-encoded molybdate transporter (e.g. ModC from E. coli or ModD from A. vinelandii). Again the function of the Mop domain is unclear, but by analogy with similar extensions, e.g. of the maltose transporter ATP-binding protein MalK (34), we suggest that it could allow regulation of transporter activity depending on intracellular molybdate levels. Such rapid feedback regulation would complement the transcription level control exercised by ModE. Fig. 5 shows a sequence alignment of Mop domains from a representative selection of Mop-carrying proteins. With the exception of some of the Mop2s from ModE proteins (see below), all sequences show conservation of the residues involved in molybdate binding. In a number of cases Ser 166 (numbering for the Mop1 domain of E. coli ModE) is replaced with an alanine, but from the structural data it seems perfectly conceivable that in those cases a water could take the place of the serine hydroxyl group.
Based on the known structures, we propose that Mop-containing proteins can be divided into at least four groups depending on the quaternary arrangement of their Mop moieties. The first group comprises the single-Mop molbindins. These Mops dimerize, and the dimers then trimerize to form a hexameric assembly as shown in Fig. 7A. Each pair of dimers provides a molybdate binding interface like that observed in ModE, altogether supplying six binding sites. Two additional MoO 4 2Ϫ binding sites are located on the 3-fold axis. Thus, the hexameric Mop proteins bind eight molecules of molybdate, all of them at subunit interfaces. DiMop molbindins with two Mop domains on a single polypeptide constitute the second structural group. Even though no structural data are currently available, it can be expected that DiMop proteins will trimerize to form an assembly similar to the Mop hexamer, the main difference being that in this case the participating dimers are covalently linked (Fig. 7B). Likewise the DiMop molbindins should also posses eight molybdate binding sites.
E. coli ModE is a member of the third group. Here only two DiMop chains dimerize to form an assembly that can be superimposed almost perfectly onto two "'dimers" from the S. ovata Mop structure (Fig. 7C). As the "missing" dimer participated in the formation of not only the two "3-fold" sites but also two of the binding interfaces, it is clear that ModE-like proteins can FIG. 6. Alignment of the ModE molybdate-binding domain in its apo form (blue) and oxyanion-bound form (green). Two "intermediate" states that were generated by Cartesian morphing (using the program LSQMAN (35)) are shown to illustrate the transition. The two proteins were aligned on the C␣ atom of the rigid core of Mop1A and -B comprising residues 124 -137, 145-151, 156 -162, and 167-184. The color gradient from blue-green to red reflects increasing C␣ distances based on the separate alignments of the rigid core for every Mop domain. Even though Mop2s are essentially rigid bodies, for comparability they were aligned on the C␣ atoms of residues 196 -209, 217-223, 227-233, and 238 -253. Unless stated otherwise all alignments and center-of-mass calculations use these sets of residues. The molybdate binding sites are labeled as in Fig. 3. bind only 2 mol of molybdate/mol of (dimeric) protein. It is furthermore clear that two of the four Mop domains (one from each dimer, dark green and purple in Fig. 7C) are no longer involved in oxyanion binding. Accordingly it is not surprising that in some ModE-like proteins (e.g. the ones from E. coli or H. influenzae, Fig. 5) the Mop2 domain has undergone amino acid changes that render molybdate binding impossible. Consider the Mop2 domain of E. coli ModE; the residue corresponding to Mop1 Ser 126 is Asn 198 , which, in the observed conformation, would for steric reasons be unable to bind molybdate. Arg 128 is substituted by Asp 200 , its side chain facing toward the hypothetical binding site, preventing the approach of an oxyanion. Asn 252 from the Mop2 domain is the equivalent residue to Lys 183 of Mop1, and again a residue important for molybdate binding is lost. The other two binding site residues, Thr 163 and Ser 166 , are replaced by Pro 234 and Glu 237 , respectively, which abolishes all hydrogen bonding and introduces another negative charge incompatible with oxyanion binding.
We note that Mop proteins with a proposed storage function maximize the number of molybdates bound per Mop unit (ϭ1.33), whereas the oligomerization state of other Mop proteins with a function in signaling/feedback provide only two binding sites, thus ensuring that no molybdate is "wasted" (e.g. E. coli ModE binds 0.5 molybdates/Mop).
The fourth group of Mop proteins are the ATP-binding components of the molybdate-transporting ABC transporter system, for example E. coli ModC and A. vinelandii ModD. On the basis of sequence alignments (Fig. 5), these proteins carry a Mop domain that should be able to bind molybdate. To do so it must at least dimerize with a Mop domain-containing protein, and it is possible that this involves another subunit of ModC itself.
Conclusions-Our analysis shows that the molybdate binding properties of the C-terminal DiMop fragment of E. coli ModE are the same as for the intact protein; it has provided details of oxyanion binding together with the conformational and quaternary changes that result. Structural data have revealed that E. coli ModE discriminates between oxyanions based on size and charge. This molecular recognition is vital for the biological role of ModE in sensing and then controlling the internal molybdate concentration in E. coli by regulation of the transcription of the modABCD operon. The DiMop domain undergoes a ligand-induced conformational change that produces an alteration to the surface of the DiMop dimer and by implication to the surface of ModE. We previously suggested a role for the C-terminal domain of ModE in the recruitment of partner proteins to form the complexes necessary for the regulation of transcription (16). The conformational alteration that the C-terminal domain undergoes when oxyanion binds may represent a molecular switch that regulates the recruitment of such partner proteins necessary for the positive regulation of transcription. The possibility exists that this confor-mational switch in the C-terminal domain may extend to the N-terminal domain and influence the interactions with DNA, although further work will be required to investigate this hypothesis.