Structural Snapshots of Escherichia coli Histidinol Phosphate Phosphatase along the Reaction Pathway*

HisB from Escherichia coli is a bifunctional enzyme catalyzing the sixth and eighth steps of l-histidine biosynthesis. The N-terminal domain (HisB-N) possesses histidinol phosphate phosphatase activity, and its crystal structure shows a single domain with fold similarity to the haloacid dehalogenase (HAD) enzyme family. HisB-N forms dimers in the crystal and in solution. The structure shows the presence of a structural Zn2+ ion stabilizing the conformation of an extended loop. Two metal binding sites were also identified in the active site. Their presence was further confirmed by isothermal titration calorimetry. HisB-N is active in the presence of Mg2+, Mn2+, Co2+, or Zn2+, but Ca2+ has an inhibitory effect. We have determined structures of several intermediate states corresponding to snapshots along the reaction pathway, including that of the phosphoaspartate intermediate. A catalytic mechanism, different from that described for other HAD enzymes, is proposed requiring the presence of the second metal ion not found in the active sites of previously characterized HAD enzymes, to complete the second half-reaction. The proposed mechanism is reminiscent of two-Mg2+ ion catalysis utilized by DNA and RNA polymerases and many nucleases. The structure also provides an explanation for the inhibitory effect of Ca2+.

The histidine biosynthetic pathway serves as a model system for better understanding of the fundamental metabolic, physiological, and genetic processes in bacteria (1). This pathway is identical in both Escherichia coli and Salmonella typhimurium and has been thoroughly characterized (1,2). The sixth and eighth steps of histidine biosynthesis are catalyzed by imidazole glycerol phosphate dehydratase (IGPD, 3 EC 4.2.1.19) and histidinol phosphate phosphatase (HPase, EC.3.1.3.15) respectively (1) (Scheme 1). In protobacteria, including E. coli and S. typhimurium, the IGPD and HPase activities are encoded by a single gene (3)(4)(5), whereas in archaea, eukarya, and most bacteria they are encoded by two separate genes (3). The bifunctional HisB enzyme has been proposed to be the result of a fusion of two independent cistrons that occurred recently in evolution (3).
Biochemical and genetic studies of the HisB enzyme together suggest that both of its enzymatic activities are independent of one another and reside in separate domains (6 -8). The HPase activity is found within the N-terminal domain (residues 1-167, HisB-N), whereas the C-terminal domain (residues 168 -356) exhibits the IGPD activity. The phosphatase activity requires the presence of metal ions such as Mg 2ϩ , Mn 2ϩ , Co 2ϩ , or Zn 2ϩ but is inhibited by calcium (9). Based on the presence of four invariant aspartic acid residues, HisB has been classified as a member of the haloacid dehalogenase-like hydrolase (HAD) family within the "DDDD" superfamily of aspartyl-phosphate utilizing phosphohydrolases/phosphotransferases (10,11). The phosphoryl transfer catalyzed by phosphotransferases and phosphatases from the HAD family occurs via a phosphoaspartate intermediate and involves one metal ion (12). In the first half-reaction the catalytic aspartate (first in the conserved DXDX(T/V) motif I) aided by the metal ion attacks the phosphorus leading to cleavage of the substrate phosphoester bond, formation of a phosphoaspartate intermediate, and the release of product. The second aspartate plays the role of general acid to protonate the leaving group (13). A conserved Ser/Thr present in motif II is involved in binding the phosphoryl group. The consensus sequence (G/S)DXX(N/T)D comprises motif III in which the first aspartate is involved in metal coordination while the second aspartate in involved in orienting a lysine residue that in turn interacts with the phosphoryl group. In the second half-reaction, a water molecule activated by a general base (second aspartate of motif I) attacks the phosphoaspartate intermediate regenerating the active site aspartate and releasing inorganic phosphate. The other members of the DDDD superfamily include nonspecific phosphatases, phosphoglycolate phosphatases, phosphoserine phosphatases, and trehalose-6-phosphatases (3). Several enzymes belonging to the HAD family have been structurally and enzymatically characterized, including phosphoserine phosphatase from Methanococcus jannaschii (14), class B acid phosphatase from E. coli (15), human mitochondrial deoxyribonucleotidase (16), ␤-phosphoglucomutase from Lactococcus lactis (17,18), and DNA 3Ј-phosphatase from Saccharomyces cerevisiae (19).
Interestingly, some bacteria utilize an unrelated enzyme to catalyze the same histidinol phosphate phosphatase reaction. Unlike E. coli HisB, this enzyme is monofunctional and belongs to the polymerase and histidinol phosphatase superfamily characterized by the presence of four conserved motifs, each containing a histidine residue (20). Members of the polymerase and histidinol phosphatase superfamily (e.g. HisB from Thermus thermophilus HB8 (21)) are structurally unrelated to the HAD family enzymes and employ a different catalytic mechanism.
The enzymes involved in the histidine biosynthesis pathway have recently been the subjects of intense structural investigation with crystal structures of all the E. coli enzymes (or their close orthologs) having been determined, with the exception of HPase (HisB-N). Here we report the crystal structures of E. coli HisB-N and several enzyme-ligand complexes. Based on these structures and the accompanying biochemical data we propose a novel mechanism for the dephosphorylation of histidinol phosphate mediated by two Mg 2ϩ ions. This proposed mechanism is distinct from that previously reported for other HAD family enzymes, where only one Mg 2ϩ ion is required for catalysis. We also show that in the presence of Ca 2ϩ the reaction is arrested following the formation of the phosphoaspartate intermediate.
A 1-liter culture was grown in LB medium containing 100 g ml Ϫ1 ampicillin for 2 h at room temperature, protein expression was induced with 100 M isopropyl 1-thio-␤-D-galactopy-ranoside, and the culture was incubated for an additional 6 h. Cell pellets were resuspended in 40 ml of lysis buffer (50 mM Tris-HCl (pH 7.5), 10 mM imidazole, 400 mM NaCl, 5% (v/v) glycerol, 10 mM ␤-mercaptoethanol, Complete TM protease inhibitor mixture (Roche Diagnostics, Laval, Canada) and 1% (v/v) Triton X-100) and sonicated on ice for a total of 2 min with cycles of 15 s on and 15 s off. The lysate was centrifuged at 100,000 ϫ g for 40 min at 4°C, and the supernatant was incubated with DEAE-Sepharose (Amersham Biosciences) pre-equilibrated with the lysis buffer. The flow-through was loaded on a 1-ml nickel-nitrilotriacetic acid column (Qiagen, Valencia, CA), washed with buffer (50 mM Tris-HCl (pH 7.5), 400 mM NaCl, 5% (v/v) glycerol, 10 mM ␤-mercaptoethanol, 40 mM imidazole) and eluted with buffer containing 200 mM imidazole. The protein buffer was exchanged to 20 mM Tris-HCl (pH 7.5), 200 mM NaCl, 5% (v/v) glycerol, and 5 mM dithiothreitol and concentrated by ultrafiltration to 20 mg ml Ϫ1 . The molecular mass measured by ESI-MS showed that the N-terminal methionine had been removed.
Dynamic light scattering measurements were performed using a DynaPro plate reader (Protein Solutions, Charlottesville, VA) at room temperature. Molecular weight was also estimated by size-exclusion chromatography on a Sephadex G-75 HR10/30 column. One-dimensional 1 H NMR spectra for mutant and wild-type enzymes were recorded on a Bruker DXR 500-MHz spectrometer and compared with wild-type enzyme to ensure proper folding of the mutant protein.
For structure determination a native crystal grown in the presence of MgCl 2 was soaked for 1 min in 1 M NaBr and flashcooled in the N 2 cold stream (Oxford Cryosystem, Oxford, UK). We refer to these crystals as HisB-N⅐Mg. The crystals obtained from condition b are referred to as HisB-N⅐Ca. We next soaked crystals obtained from condition a in 50 mM histidinol phosphate for 4 h or 50 mM histidinol (product) for 16 h. In both cases only histidinol was observed in the electron density map (HisB-N⅐Mg/histidinol). Soaking a crystal grown in condition b in 50 mM histidinol phosphate for 20 min resulted in the phosphoaspartate complex (HisB-N⅐Ca/pAsp), and the crystals from a were soaked in 0.2 M MgSO 4 for 16 h (HisB-N⅐Mg/sulfate).
Data Collection and Refinement-Data collection for the Br Ϫ SAD experiment was carried out at a wavelength of 0.9202 Å, and the remaining data were collected at a wavelength of 1.1 Å using a Quantum-4 charge-coupled device detector (Area Detector Systems Corp., San Diego, CA), at the X8C beamline, National Synchrotron Light Source, Brookhaven National Laboratory. Data processing and scaling were performed with HKL2000 (24). The seven-site Br Ϫ substructure was determined with SHELXD (25), and phases were derived with SHELXE (26). Solvent flattening with RESOLVE (27) led to phases with a figure-of-merit of 0.81. ARP/wARP (28) automatically built 95% of the expected residues. The model was completed manually using XtalView (29) and refined at 1.7-Å resolution with REFMAC5 (30) using all reflections. Difference electron density maps showed the presence of ions in various datasets, and they were modeled as Na ϩ , Ca 2ϩ , Mg 2ϩ , Br Ϫ , or Zn 2ϩ according to the observed geometry, coordinating atoms, composition of the mother liquor, and refined B-factors. There are two molecules in the asymmetric unit. The final model of HisB-N⅐Mg contains residues 3-19, 22-116, and 120 -164 for subunit A, 4 -116 and 120 -166 for subunit B, and 440 water molecules. The final R work was 0.187, and R free was 0.215 ( Table  1). The model has good stereochemistry as analyzed with PRO-CHECK (31).
Other structures were isomorphous to the native crystals and were refined, starting from the HisB-N⅐Mg model, at resolutions ranging from 1.75 to 2.2 Å. The metal ions were modeled based on the content of the mother liquor. The R work was ϳ0. 18 and R free was ϳ0.22 (Table 1). Residues 19 -22 and 117-120 were less well defined in the electron density maps and were modeled only in some molecules. The N-terminal His tag was disordered in all crystals.
HPase Assay and Steady-state Kinetics-The HPase assay was carried out as described (32) with release of inorganic phosphate detected as a malachite green⅐phosphomolybdate complex (33). Each 100-l reaction mixture contained 25 mM Tris-HCl (pH 7.5), 70 nM enzyme, 25 M Mg 2ϩ , and 200 M histidinol phosphate. After 15-min incubation at room temperature the reaction was stopped by the addition of 20 l of molybdate reagent (1.75% (w/v) ammonium heptamolybdate in 6.3 N sulfuric acid). A volume of 20 l of 0.035% (w/v) malachite green in 0.35% (w/v) polyvinyl alcohol (M r 14000) in water was added after a 10-min incubation for color development. The absorbance of the malachite green⅐phosphate complex was recorded at 610 nm after 30 min. A molar absorption coefficient of 110,620 M Ϫ1 was used to determine the concentration of phosphate released. Values for K m and k cat were derived through non-linear regression analysis of the Michaelis-Menten equation using duplicate measurements. Eleven different substrate concentrations (10 -110 M) were used for each set of experiments. A reaction time course was also monitored by incubating 10 M enzyme (native and mutants) with 2-4 mM histidinol phosphate and 5 mM MgCl 2 /CaCl 2 . 3-l aliquots were withdrawn at various time points for phosphate estimation.
Mass Spectrometric Analysis-Formation of the phosphoaspartate intermediate of HisB-N was followed by ESI-MS using an Agilent 1100 Series LC/MSD (Agilent Technologies, Palo Alto, CA). A 50-l reaction mixture containing 1 mM histidinol phosphate, 5 mM MgCl 2 or CaCl 2 , and 50 M HisB-N was incubated at 21°C for 30 min. A 2-l aliquot of the reaction mixture was injected directly at 5-min intervals, and the appearance of the phosphoaspartate enzyme adduct was monitored in the positive mode using isocratic conditions of 30% (v/v) acetonitrile containing 0.1% (v/v) formic acid. Extended X-ray Absorption Fine Structure Measurements-The extended x-ray absorption fine structure spectra for HisB-N were collected at beamline X9B at the National Synchrotron Light Source using a double crystal silicon (111) monochroma-tor. High order harmonics were suppressed (Ͻ0.1%) by a nickelcoated reflecting mirror. Data were collected in fluorescent mode using a 13-element Ge array detector (Canberra). All measurements were made with the HisB-N samples in 20 mM Tris-HCl (pH 7.5), 0.2 M NaCl, 5% (v/v) glycerol, and 5 mM dithiothreitol at a protein concentration of 20 mg ml Ϫ1 in an He-displex cryostat operating at 20 K. Several scans of 40 -60 min each were collected and summed together.
Isothermal Titration Calorimetry-ITC measurements were carried out on a MicroCal VP-ITC titration calorimeter (MicroCal, Inc., Northampton, MA) using the VPViewer software for data acquisition and instrument control. Chelex-treated buffer solution (25 mM Tris-HCl, pH 7.0) was used for the metal binding studies, and stocks of metal solutions were prepared using the same buffer. Protein samples treated with 5 mM EDTA for 30 min were chromatographed on a Superose-12 16/60 column with Chelex-treated buffer to remove any bound metal ions. All solutions were degassed prior to the experiment. In a typical experiment HisB-N at a concentration of 260 M was added to the reaction cell, maintained at 20°C, and stirred at a constant rate of 260 rpm to ensure rapid mixing. A solution of 5 mM MgCl 2 or CaCl 2 was loaded into a 300-l syringe, and 45 injections of 6 l each were added to the protein sample in 5-min intervals. Data were analyzed using the MicroCal ORIGIN (version 7.0) software provided by the manufacturer. The thermodynamic values K d , ⌬S o , and ⌬H o for a two-site model were obtained by non-linear curve fitting.

RESULTS AND DISCUSSION
Molecular Structure-The two HisB-N monomers in the asymmetric unit can be superposed with a root mean square deviation of ϳ0.35 Å for 155 C␣ atoms, indicating that no significant conformational differences exist between them. Each monomer has approximate dimensions of 47 ϫ 34 ϫ 26 Å 3 and constitutes a single domain containing a Rossmann fold (Fig. 1a). The central part of the monomer is formed by a five-stranded parallel ␤-sheet (␤5-␤4-␤1-␤2-␤3) flanked on one side by three ␣-helices nearly parallel to the ␤-strands and on the other side by one ␣-helix and a random coil region resembling a distorted helix (between ␤3 and ␤4) (Fig. 1a). A long ⍀ loop (Asp 12 -Glu 34 ) spans strand ␤1 and helix ␣1 and narrows at the ends due to hydrogen bonds between two short ␤-strands (␤1a: Leu 15 -Ser 17 and ␤1b: Ala 32 -Glu 34 ).
The crystal structure together with dynamic light scattering and size-exclusion chromatography indicate a dimeric organization of HisB-N (Fig. 1b). The monomers pack back-to-back via helix ␣1 (residues Phe 28 -Leu 42 ) and helix ␣2 (Leu 76 -Gln 84 ) to form the dimer with dimensions of ϳ47 ϫ 68 ϫ 26 Å 3 . Dimerization mainly involves hydrophobic interactions, with only eight direct hydrogen bonds. Water-mediated hydrogen bonds are observed at the periphery of the dimer interface. The total surface area buried upon dimer formation is ϳ1250 Å 2 corresponding to ϳ8% of the monomer surface. The substrate-binding site is located on the protein side opposite to the dimerization interface.
Metal Binding Sites-Difference electron density maps calculated from datasets collected for various HisB-N crystals showed the presence of two or three strong peaks (Ͼ8) near each molecule. Based on the presence of Ca 2ϩ , Mg 2ϩ , or both Mg 2ϩ and Na ϩ as well as the observed ligand coordination, these sites were modeled as metal ions (Fig. 1).
Zinc Binding Site-The Zn 2ϩ ion is present in all HisB-N molecules. Its presence was confirmed by extended x-ray absorption fine structure analysis, which showed strong absorption at the K edge of zinc (9668 eV, data not shown). This site is situated within the loop following strand ␤5. The side chains of Cys 94 , His 96 , Cys 102 , and Cys 104 tetrahedrally coordinate the Zn 2ϩ . The average Zn 2ϩ distance to the cysteine thiol groups is 2.3 Å and to ND1 His-96 is 2.1 Å (Fig. 1a). The temperature factor for this Zn 2ϩ , (ϳ16 Å 2 ), is similar to the average B-factor of the surrounding protein atoms. Since Zn 2ϩ was not part of the crystallization conditions, its presence in the crystal structure indicates that HisB-N acquired Zn 2ϩ during expression in E. coli and that this metal is required for maintaining the structural integrity of the protein. The Zn 2ϩ site is ϳ14 Å away from the catalytic Asp 10 .
Metal Site 1-HisB-N⅐Mg crystals show a metal-binding site near the active site Asp 10 coordinated by six O atoms with octa-hedral geometry (Fig. 2a). The average metal-O distance of 2.1 Å confirms the presence of Mg 2ϩ . This site is common to both monomers and is situated above strands ␤1 and ␤4. The Mg 2ϩ is coordinated by the side-chain oxygens O␦2 Asp-10 and O␦1 Asp-131 , the main-chain oxygen O Asp-12 , and three water molecules. In crystals of HisB-N⅐Ca grown in the presence of 50 mM CaCl 2 , this site is occupied by a Ca 2ϩ and is coordinated by the same residues as described for the Mg 2ϩ site, except that the average Ca 2ϩ -O distance is 2.35 Å.
Metal Site 2-In the HisB-N⅐Ca/pAsp complex there is a second metal ion with octahedral coordination (in molecule B only) ϳ6 Å away from site 1. This site is located between strands ␤2 and ␤4. The metal ion is coordinated by O␦2 Asp-12 , O Phe-23 , and four water molecules. The average ion-O distance is 2.35 Å, indicative of Ca 2ϩ . FIGURE 2. a, the omit F o Ϫ F c electron density map for HisB-N⅐Mg contoured at 3. The residues depicted were excluded from the phase calculation. Mg 2ϩ occupies site 1 and Na ϩ occupies site 2; b, the omit F o Ϫ F c electron density map for HisB-N⅐Mg/sulfate contoured at 3. The depicted residues were excluded from phase calculation. Mg 2ϩ occupies site 1, water W2 occupies site 2, a hexa-hydrated Mg 2ϩ ion is near the protein surface; c, hydrogen bonding interactions within the active site of HisB-N. The HisB-N forms a complex with histidinol. Site 1 is occupied by Mg 2ϩ , and site 2 by the amino group of histidinol with its hydroxyl group occupying the position of water W1; d, the HisB-N⅐phosphoaspartate intermediate.
This site is also occupied in the crystal of HisB-N⅐Mg soaked briefly in 1 M NaBr. In this case, the OE2 Glu-18 atom replaces one of the water molecules in the coordination sphere ( Fig. 2a) with an average ion-oxygen distance of 2.45 Å, substantially longer than that expected for Mg 2ϩ -O but in accordance with the distance expected for Na ϩ -O coordination.
In crystals grown from CaCl 2 (HisB-N⅐Ca) or soaked with MgSO 4 (HisB-N⅐Mg/sulfate) a tetrahedrally coordinated water molecule occupies this site (Fig. 2b). In the latter struc-ture we observe a fully hydrated Mg 2ϩ ion (octahedral coordination by six water molecules, average distance 2.1 Å) ϳ5 Å from site 2 and near the protein surface. This cluster is hydrogen-bonded to the side chains of Arg 132 , Arg 105 , and Asp 58 and to the sulfate ion.
ITC Measurements of Metal Binding Affinities-We measured binding of Mg 2ϩ and Ca 2ϩ to HisB-N by ITC. Titration of HisB-N in the absence of substrate with either Mg 2ϩ or Ca 2ϩ confirmed the presence of two metal-binding sites with binding constants differing ϳ50-fold (Table 2). Based on the crystal structures, site 1 is likely the primary site with ϳ100 nM affinity for the first Mg 2ϩ and site 2 is the secondary site with an affinity of ϳ3 M. The affinity for the first Ca 2ϩ ion is 77 nM and ϳ6 M for the second. The relatively low affinity of site 2 concurs with our finding that in apo-HisB-N at low Ca 2ϩ concentration a water molecule occupies this site. Furthermore, the presence of Ca 2ϩ in site 2 in HisB-N⅐Ca/pAsp but not in HisB-N⅐Ca at an identical Ca 2ϩ concentration indicates that the affinity of site 2 is enhanced upon formation of the phosphoaspartate intermediate.
HisB-N⅐Histidinol Complex-The difference Fourier map for crystals of HisB-N soaked with either histidinol phosphate or histidinol in the presence of Mg 2ϩ showed identical electron density in the substrate-binding site of molecule B consistent with the histidinol moiety, thus indicating retention of enzymatic activity in the crystal. No density for histidinol was observed in molecule A, because this site is blocked by the Arg 165 side chain from a neighboring molecule.
The histidinol is positioned within the cleft formed at the top of the central ␤-sheet over strand ␤1 (Fig. 1). Its imidazole ring stacks against the flat face of the Asp 58 side chain on one side and against the backbone and C␤ atoms of Phe 23    Phosphoaspartate Intermediate-The dephosphorylation of substrates catalyzed by enzymes from the HAD superfamily takes place through the formation of a covalent phosphoaspartate intermediate (13)(14)(15)(16)18). This intermediate was trapped in HisB-N⅐Ca/pAsp when a crystal was soaked in histidinol phosphate in the presence of Ca 2ϩ . The difference electron density clearly showed the covalent attachment of a phosphoryl group to the side chain of Asp 10 . This phosphoryl group participates in an extensive hydrogen-bonding network to the main chain of Arg 11 , Asp 12 , and Asn 56 and to the side chains of Lys 106 and Thr 55 . The terminal OP1 oxygen atom participates in the coordination sphere of Ca 2ϩ in site 1 leading to some rearrangement of waters and a change in Ca 2ϩ coordination from hexadentate to heptadentate (average Ca 2ϩ -O distance of 2.40 Å, Fig. 2d). Simultaneously, the side chain of Arg 132 rotates outward creating a wider opening of the substrate binding site.
Sulfate Binding Site-In the crystal structure of HisB-N⅐Mg/ sulfate, obtained by soaking a native crystal in 200 mM MgSO 4 , we observed one sulfate ion in the substrate-binding cleft. This ion was found in the upper part of the cleft, ϳ5 Å above the catalytic Asp 10 , close to the protein surface. The sulfate is hydrogen-bonded to the side chains of Gln 56 , Lys 106 , and Thr 134 and to the hydrated Mg 2ϩ cluster near the protein surface (Fig. 2b). The side chain of Lys 106 adopts a slightly different orientation from that observed in the HisB-N⅐Ca/pAsp complex. We propose that the position of this sulfate ion delineates the exit route of the phosphate following release from Asp 10 . The tight binding of the sulfate to the hexahydrated Mg 2ϩ via hydrogen bonds suggests that the Mg 2ϩ plays a role in phosphate release. Sequence and Structural Comparisons-The E. coli HisB-N domain has counterparts in many bacteria, with sequences for over 220 HPases presently known. This protein family belongs to the very large haloacid dehalogenase-like (HAD) superfamily, which includes more than 6000 proteins (PFAM, PF00702). Specifically, HisB-N belongs to the HAD IIIA subfamily of aspartate-nucleophile hydrolases that are characterized by the lack of a "cap domain" over the substrate-binding site (34). The HAD superfamily is distinguished by the presence of an N-terminal motif containing the aspartate nucleophile, a central motif containing a conserved lysine (or arginine) and a C-terminal motif containing a conserved serine/threonine and aspartate (11) (Fig. 3). In HisB-N these residues are Asp 10 , Lys 106 , and Asp 131 , respectively. The sequences of proteins annotated as HPases or histidinol phosphatase-related show additional absolutely conserved residues, corresponding in HisB to Arg 11 , Asp 12 , Ser/Thr 55 , Asn 56 , Gln 57 , Arg 105 , Pro 107 , and Gly 130 . The Cys 94 -X-His-(X) 5 -Cys-X-Cys 104 motif forming the Zn 2ϩ -binding site (Fig. 1a) is present in the majority of these proteins, suggesting that most of them require stabilization by Zn 2ϩ . In addition, Glu 18 , Gln 24 , Asp 58 , and Arg 132 are present in all bifunctional HPases but not in some histidinol phosphatase-related proteins. Their conservation in only the HPases suggests a role in defining substrate specificity.
The closest structural homologs of HisB-N are the phosphatase domain of mouse 5Ј-polynucleotide kinase-3Ј phosphatase (PDB code 1YJ5 (35)), mouse magnesium-dependent phosphatase-1 (PDB code 1U7P (36)), and acid phosphatase (AphA) from E. coli (PDB code 1N8N (37)). Enzymes from subfamily III often form oligomers where the substrate-binding site of one subunit is partially occluded by another subunit (PDB 1J8D (38)). In other cases, such as E. coli AphA, the substrate-binding site is more open, accounting for the observed lack of specificity of this enzyme. HisB-N employs an ⍀ loop following strand ␤1 (in particular Phe 23 and Asp 58 -Phe 65 ) and the Zn 2ϩ -binding loop (Cys 94 -Lys 106 ) to create a deep and narrow substrate binding site for accommodating histidinol phosphate. The Mg 2ϩ observed in other HAD enzymes is also present in HisB-N at site 1 and is coordinated by the conserved residues Asp 10 , Asp 12 , and Asp 131 . The metal ion observed in site 2, near the catalytic site, is unique to HisB-N and has not been observed in other HAD enzymes.
Mutagenesis of Asp 12 and Glu 18 -Asp 12 and Glu 18 were mutated to alanine or serine to determine their potential as water-activating nucleophiles. The Asp 12 3 Ala, Asp 12 3 Ser, Glu 18 3 Ala, and Glu 18 3 Ser mutants showed similar elution profiles as wild-type protein by gel-filtration chromatography, and the alanine mutants showed 1 H NMR spectra characteristic for folded proteins. Under our standard assay conditions these mutants showed no measurable activity.
To measure residual HPase activity we repeated the assay at 140-fold higher protein concentration (10 M) and longer incubation times. Under these conditions, the Glu 18 3 Ala mutant showed ϳ5% of the wild-type activity (Fig. 4) indicating that Glu 18 , while important, is dispensable. This side chain participates in the coordination of a metal ion at site 2, and its mutation would be expected to weaken metal binding. Increasing the concentration of Mg 2ϩ from 5 M to 1 mM led to a dramatic increase of activity of this mutant (Fig. 4b), suggesting that Mg 2ϩ at site 2 is essential for catalysis.
Conversely, the Asp 12 3 Ala mutant showed only traces of activity, even with overnight incubation in the presence of 1 mM Mg 2ϩ , showing that the presence of the Asp 12 side chain is essential for enzyme activity. When assayed in the presence of Ca 2ϩ neither the Glu 18 3 Ala nor the Asp 12 3 Ala mutant showed any detectable activity.
Mechanism of Action-The various structures we have determined represent snapshots along the HisB-N reaction pathway (Fig. 5a). The HisB-N⅐Mg model (see "Experimental Procedures") shows the initial state of the enzyme. In the resting state, a water molecule (W2) rather than a Na ϩ ion occupies site 2, as observed in the HisB-N⅐Ca structure. Two other key sites within the active site are occupied by water molecules (W1 and W3) in the HisB-N⅐Mg and HisB-N⅐Ca structures. We have modeled histidinol phosphate bound to HisB-N based on the HisB-N⅐histidinol and HisB-N⅐Ca/pAsp structures. Substrate binding displaces these waters, with W1 replaced by the histidinol hydroxyl, W2 by the histidinol amino group, and W3 by the phosphate oxygen (Fig. 5a). This orientation of the substrate presents its phosphoryl group for nucleophilic attack by the side chain of Asp 10 . The reaction proceeds through a penta- valent transition state (14,17) resulting in a covalent phosphoaspartate intermediate with an inversion of configuration of the phosphorus, and release of histidinol. Direct coordination of the metal in site 1 to the phosphoryl group neutralizes its negative charge and stabilizes the covalent intermediate. Additional stabilization of the phosphoryl group is provided by the side chain of Lys 106 , the backbone NH of Asp 12 and Asn 56 , and bridging water molecules (Fig. 5a). The HisB-N⅐Ca/pAsp structure provides direct evidence for the formation of a phosphoaspartate intermediate in the reaction catalyzed by HPase and also shows that the release of histidinol after the formation of phosphoaspartate facilitates binding of the second metal at site

A similar phosphoaspartate intermediate was previously
observed in the structure of ␤-phosphoglucomutase (18). The second step, hydrolysis of the phosphoaspartate, is performed by a water molecule W1, which occupies the position of the histidinol hydroxyl group in the substrate, and is within the coordination sphere of the metal in site 2.
The presently known structures of HAD phosphohydrolases/ phosphotransferases suggest that the aspartate (Asp 12 in HisB) positioned two residues after the catalytic aspartate (Asp 10 in HisB) acts as a general acid donating proton to the leaving group (15). The same aspartate is presumed to play the role of general base in the second half-reaction, activating the water molecule W1. This aspartate is directed toward the position occupied either by the oxygen of the scissile P-O bond in the substrate (enzyme⅐substrate complex) and, hence, is in a position to protonate the leaving group, or by water W1 (phosphoaspartate intermediate) ready for inline attack on the phosphoaspartate (Fig. 5b). In AphA from E. coli (37) and sugar phosphatase from Bacteroides thetaiotaomicron (39) this orientation of the aspartate side chain is stabilized through a salt bridge to an arginine. However, in HisB-N the carboxylate of Asp 12 is rotated by ϳ70 o away from the crucial water W1 and is directed toward site 2 (Fig. 5b). Furthermore, this orientation of Asp 12 is stabilized by a salt bridge to Arg 11 (NE Arg-11 . . . O␦1 Asp-12 . . . NH 2  , conserved in HPases but distinct from other HAD enzymes), by a hydrogen bond to a bridging water W4 (present in all HisB-N structures), and the interactions with the water W2 or metal occupying site 2 (Fig. 5b). These structural restrictions confine the orientation of Asp 12 away from W1 and put into question the ability of Asp 12 to act as a general acid protonating histidinol and a general base activating water W1. Mutation of Asp 12 to an alanine or a serine leads to an inactive enzyme, however, this may reflect a destabilization of the active/substrate binding site, because these mutants express poorly and could not be crystallized even when micro-seeded with crystals of the native enzyme. This behavior suggests that Asp 12 is essential for the integrity of both the active and substrate-binding sites and could play an indirect rather that direct role in the enzymatic mechanism. If this is indeed the case, it is unclear where the proton required for histidinol protonation comes from. We propose that in HisB-N, the metal ion at site 2 (usually Mg 2ϩ ) is essential for the second half-reaction by assisting in the deprotonation of water W1 (its ligand), possibly by aiding Asp 12 , whose O␦1 atom is adjacent to water W1 in the octahedral coordination environment (Fig. 5). Water W1 also forms a hydrogen bond to O Asn-56 . The latter interaction likely helps in proper positioning of water W1 for nucleophilic attack on the phosphoaspartate. The importance of the metal in site 2 is supported by mutations of Gln 18 that participates in coordination of the metal in site 2. Gln 18 3 Ala or Gln 18 3 Ser mutation reduces the affinity of site 2 for the  ion. The release of phosphate and the metal from site 2 restores the enzyme to its initial state. The conservation of Glu 18 and all residues forming metal binding sites in the sequences of all histidinol phosphatases but not other HAD enzymes strongly suggests that the proposed novel mechanistic variant is common to this entire subfamily. This mechanism, in which two-Mg 2ϩ ions are employed in catalysis, and the water is activated directly by the metal ion, is highly reminiscent of the mechanism employed by polymerases (exonuclease domain), ribonucleases, and alkaline phosphatase to break the phosphoryl bond in nucleic acids (reviewed in Ref. 40).
Mode of Inhibition by Calcium-HisB-N shows high catalytic efficiency in the presence of metal ions (Table 3). In the presence of 50 M Mg 2ϩ or Co 2ϩ the enzyme had similar k cat /K m (ϳ3.8 ϫ 10 8 s Ϫ1 M Ϫ1 ), whereas the activities in the presence of Zn 2ϩ or Mn 2ϩ were 30 and 50% lower, respectively. No enzymatic activity was detected in the presence of either Ca 2ϩ or EDTA (100 M). The effect of EDTA could be reversed by the addition of excess metals. A similar activity profile has been observed for HisB from Salmonella typhimurium (9).
The loss of enzymatic activity correlates with an increase in the atomic radius of the metal (calcium being significantly larger than other metals tested), which can be rationalized based on the crystal structure of HisB-N⅐Ca/pAsp. In the presence of Ca 2ϩ , the phosphoaspartate intermediate is trapped in the crystal, indicating that the first half-reaction is unaffected and that inhibition occurs at the second half-reaction. The transfer of the phosphoryl group from histidinol phosphate to Asp 10 occurs as in the presence of Mg 2ϩ , with one oxygen of the phosphoaspartate intermediate coordinating the metal in site 1.
However, the larger radius of Ca 2ϩ versus Mg 2ϩ causes a small rearrangement of water molecules, leading to a hepta-coordinated Ca 2ϩ (Fig. 2b). Consequently, the phosphoryl group is likely shifted by ϳ1 Å, relative to its position in the presence of a smaller Mg 2ϩ ion, and occupies a suboptimal position for nucleophilic attack by water W1. In addition, the Ca 2ϩ ion in site 2 also rearranges the coordination sphere compared with a smaller ion (e.g. Mg 2ϩ ). The side chain of Glu 18 is excluded from the Ca 2ϩ coordination sphere and is replaced by a water molecule to which it is hydrogen-bonded. Ca 2ϩ has previously been shown to inhibit another HAD enzyme, human phosphoserine phosphatase (41). However, based on the crystal structure, a different mechanism of inhibition was proposed, in which the bidentate coordination of Ca 2ϩ to Asp 20 (equivalent to Asp 10 in HisB) prevents the aspartate from acting as a nucleophile.
To investigate formation of the phosphoaspartate during catalysis in vitro we have followed the reaction by mass spectrometry and detected the phosphoaspartate by ESI-MS during a reaction time course. In the presence of Mg 2ϩ and histidinol phosphate only one species with a molecular mass of 19,903 Da was present. This mass corresponds well to the enzyme with the N-terminal methionine removed (theoretical molecular mass is 19,909 Da). In the presence of Ca 2ϩ and histidinol phosphate we observed the accumulation of a species (ϳ30% of the wildtype enzyme) with a mass of 19,983 Da, corresponding to that expected for the phosphoaspartate enzyme adduct (ϩ79 Da), consistent with the crystal structure of HisB-N⅐Ca/pAsp.
Dynamics of the Active Site-Comparison of all the HisB-N molecules provides insight into the dynamics of the substratebinding cavity at various stages of the histidinol phosphate dephosphorylation reaction (Fig. 6). Two residues show substantial movements that affect the size of the entrance to the active site cleft, namely Arg 132 and Glu 18 . In the apo model the active site is relatively open for binding the substrate. Upon binding of the substrate, the side chains of Arg 132 and Glu 18 move inward resulting in narrowing the entrance to the binding site and capturing the substrate (Fig. 6b). After the first halfreaction, i.e. the formation of a phosphoaspartate intermediate, the side chain of Glu 18 rotates away from the active site and Arg 132 swings out, resulting in a widening of the cavity and enabling the release of histidinol (Fig. 6c). The hydrolysis of phosphoaspartate is accompanied by the restoration of the Arg 132 side chain to its original position and the release of the phosphate (possibly together with a hydrated Mg 2ϩ ion) guided by the side chain of Lys 105 .
The opening and closing of the active site is not uncommon to enzymes belonging to the HAD superfamily, especially in enzymes containing the cap (sub)domain, although the rearrangement is largely restricted to the cap domain. Similarly to HisB, bacterial acid phosphatase AphA from E. coli, lacking the cap domain, also exhibits the opening and closing of the active site based on the absence or presence of a phosphorylated substrate, through a loop movement representing the "open" and "closed" states (15). Notable examples of domain shifts observed in members possessing the cap domains include phosphoserine phosphatase from Methanococcus jannaschi (42) and a phosphatase from Thermotoga maritima (43).
Oligomeric State of the Full-length HisB-We asked the question of the organization of the dimeric N-terminal HisB domain in the context of the full-length HisB. The C-terminal IGPD domain of HisB shows ϳ50% sequence identity to monofunctional IGPDs from Filobasidiella neoformans (44) and Arabidopsis thaliana (45) both of which form 24-mers in the presence of Mn 2ϩ . We have found that in the presence of 5 mM Mn 2ϩ HisB also associates into large oligomers that eluted from Sephacryl S-300 gel filtration column near the void volume corresponding to ϳ900 kDa, and dynamic light scattering showed apparent molecular mass of ϳ800 kDa. The phosphatase activity of HisB in the presence of Mn 2ϩ remained unaltered, indicating that HisB oligomerization has no effect on HisB-N activity. Based on the structure of HisB-N and the structure of corresponding IGPD domains from orthologous protein we constructed a model of the bifunctional HisB in which the IGPD domain is a 24-mer. A unique way of matching the N-and C-terminal domains was apparent (Fig. 7). This model suggests that the basic unit of HisB is a hexamer.