Structural Studies of Medicago truncatula Histidinol Phosphate Phosphatase from Inositol Monophosphatase Superfamily Reveal Details of Penultimate Step of Histidine Biosynthesis in Plants*

The penultimate enzyme in the histidine biosynthetic pathway catalyzes dephosphorylation of l-histidinol 1-phosphate (HOLP) into l-histidinol. The recently discovered in Arabidopsis thaliana plant-type histidinol phosphate phosphatase (HPP) shares no homology with the two other HPP superfamilies known previously in prokaryotes and resembles myo-inositol monophosphatases (IMPases). In this work, identification of an HPP enzyme from a model legume, Medicago truncatula (MtHPP) was based on the highest sequence identity to A. thaliana enzyme. Biochemical assays confirmed that MtHPP was able to cleave inorganic phosphate from HOLP but not from d-myo-inositol-1-phosphate, the main substrate of IMPases. Dimers of MtHPP, determined by size exclusion chromatography, in the presence of CO2 or formaldehyde form mutual, methylene-bridged cross-links between Lys158 and Cys245 residues. Four high resolution crystal structures, namely complexes with HOLP (substrate), l-histidinol (product), and PO43− (by-product) as well as the structure showing the cross-linking between two MtHPP molecules, provide detailed structural information on the enzyme. Based on the crystal structures, the enzymatic reaction mechanism of IMPases is accustomed to fit the data for MtHPP. The enzymatic reaction, which requires Mg2+ cations, is catalyzed mainly by amino acid residues from the N-terminal domain. The C-terminal domain, sharing little identity with IMPases, is responsible for the substrate specificity (i.e. allows the enzyme to distinguish between HOLP and d-myo-inositol-1-phosphate). Structural features, mainly the presence of a conserved Asp246, allow MtHPP to bind HOLP specifically.

deprotonated His exists in proteins. The role of His in proteins is related to its ability to switch protonation states within the range of physiological conditions. His is present in active sites of many enzymes, where it coordinates zinc, nickel, or other metal cations. Also, free, unincorporated into proteins His has been reported to chelate nickel ions in plant species that hyperaccumulate this metal (2,3). In bacteria, fungi, and plants, His phosphorylation is a ubiquitous mechanism, used in two-component signal transduction (4). To humans, His is one of nine essential amino acids; thus, unraveling the details of its biosynthesis in plants is of particular interest (5).
His is also one of the least abundant amino acids in proteins, probably because its biosynthesis is metabolically expensive (6), with 31-41 ATP molecules required to produce a single His molecule (7)(8)(9). Plants biosynthesize His similarly to prokaryotes (10). The 11-step pathway, which originates from 5Ј-phosphoribosyl 1-pyrophosphate, is catalyzed by eight enzymes, HISN1-8, three of which (HISN2, -4, and -8) are bifunctional. His biosynthesis is connected to tryptophan (11) and nucleotide (7,11) metabolic pathways. Also, during the course of His production, two other amino acids, glutamic acid and glutamine, serve as nitrogen donors.
Unlike for many other amino acids, most of the enzymes from the His biosynthetic pathway are encoded by single genes (12), and, because the His biosynthetic pathway is absent in animals, plant enzymes involved in His production could potentially serve as targets in the search for herbicides. Of the eight enzymes required for His biosynthesis in plants, seven were identified in the last decades of the twentieth century: HISN1 (13), HISN2 (14), HISN3 (15), HISN4 (16), HISN5 (13), HISN6 (17), and HISN8 (18 -20). Although HISN7 activity in planta was reported in 1971 (13), it was unknown which particular protein was responsible for this action. HISN7 catalyzes dephosphorylation of L-histidinol 1-phosphate (HOLP) 2 to L-histidinol (HOL) (Fig. 1); therefore, a HISN7 enzyme is a histidinol phosphate phosphatase (HPP) (EC 3.1.3.15).
For a long time, it was thought that bacterial HPP enzymes belonged to either of the two superfamilies: (i) containing conserved Asp residues (DDDD) (21,22) or (ii) polymerase and histidinol phosphatases (PHP) (23). It was therefore confusing that plant genomes did not contain sequences coding for HPP proteins from either DDDD or PHP superfamilies. When a third superfamily of HPP enzymes was discovered in Corynebacterium glutamicum (24) and Streptomyces coelicolor (25), it shed new light on the missing link of His biosynthesis. Sequences of the enzymes from the novel HPP superfamily showed significant similarity to myo-inositol monophosphatases (IMPases). In Arabidopsis thaliana, three IMPase or IMPase-like proteins were identified and reported: VTC4, IMPL1 (IMPase-like 1), and IMPL2 (IMPase-like 2) (26). Of those three candidates, only AtIMPL2 was shown to possess HPP activity (27). Moreover, although AtIMPL2 was still able to dephosphorylate, for example, D-myo-inositol-1-phosphate (IMP) (26), HPP activity was proposed to be the only reaction catalyzed in vivo by this enzyme (28). It is also worth mentioning that the growth of Atimpl2 mutants was severely compromised but was rescued by external supplementation with His (28). Similarly to other proteins involved in His biosynthesis, AtIMPL2 contains chloroplast-targeting signal peptide and has been actually localized in chloroplasts (27).
To date, AtIMPL2 has been the only identified plant HPP enzyme. Here we report the identification of a second member of the plant-type superfamily of HPP enzymes and its thorough structural characterization (i.e. the first structure of IMPaselike HPP enzyme from any domain of life). These studies were performed on Medicago truncatula, a model legume plant, and the target enzyme is referred to as MtHPP. Four high resolution crystal structures (complexes with PO 4 3Ϫ , HOLP, and HOL and a structure illustrating a unique cross-linking interaction between two MtHPP molecules) are described within the scope of this research. Presented results illustrate the exact mode of HOLP binding (i.e. which residues interact with the substrate) as well as clarify the requirements for Mg 2ϩ ions. The structures also indicate important differences in the reaction mechanism between IMPases and MtHPP. Moreover, the structural features, which allow MtHPP to bind HOLP, are compared with the IMP binding site of a classic IMPase to explain why IMPases are unable to bind HOLP. MtHPP, as a member of the novel, third superfamily of HPP enzymes, shares little to no similarity with the previously described HPPs from the DDDD and PHP superfamilies. Therefore, the presented work provides new, important details of His biosynthesis in plants. The results can also be extrapolated to those bacterial species that rely on IMPase-like HPPs.

Experimental Procedures
Cloning, Overexpression, and Purification of MtHPP-The total RNA was isolated from M. truncatula roots using the RNeasy plant minikit (Qiagen). SuperScript II reverse tran-scriptase (Life Technologies) with oligo(dT) (15 and 18) primers was utilized to obtain the coding DNA (cDNA). The cDNA served as a template for acquiring the sequence coding for the MtHPP open reading frame by polymerase chain reaction (PCR). The used primers (forward, TACTTCCAATCCAATG-CCATGTCCTCCTCATCTTCGCCTC; reverse, TTATCCA-CTTCCAATGTTACTACCATTGTAATGAATCTAGAGC-CTGTT) allowed the use of a ligase-independent cloning method (29) for incorporating the insert into pMCSG68 vector (Midwest Center for Structural Genomics). The pMCSG68 vector introduces a His 6 tag followed by the tobacco etch virus protease cleavage site at the N terminus of the expressed protein. The final polypeptide carries an SNA N-terminal fragment before the genuine protein sequence. The correctness of the insert was confirmed by DNA sequencing.
Overexpression was carried out in BL21 Gold Escherichia coli cells (Agilent Technologies). The bacteria were cultured with shaking at 210 rpm in LB medium supplemented with 150 g/ml ampicillin at 37°C until the A 600 reached 1.0. The cultures were cooled down to 18°C, and MtHPP production was induced by the addition of isopropyl-D-thiogalactopyranoside at a final concentration of 0.5 mM. The culture was grown for 18 h and then centrifuged at 3500 ϫ g for 20 min at 4°C. Cell pellet from 1 liter of culture was resuspended in 35 ml of binding buffer (50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 20 mM imidazole, 1 mM tris(2-carboxyethyl)phosphine (TCEP)) and stored at Ϫ80°C. The samples were thawed, and the cells were disrupted by sonication using bursts of total duration of 4 min, with appropriate intervals for cooling. Cell debris was pelleted by centrifugation at 25,000 ϫ g for 30 min at 4°C. The supernatant was applied to a column packed with 6 ml of HisTrap HP resin (GE Healthcare), connected to a VacMan manifold (Promega), and the chromatographic process was accelerated with a vacuum pump. After binding, the column was washed five times with 40 ml of the binding buffer, and His 6 -tagged MtHPP was eluted with 20 ml of elution buffer (50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 300 mM imidazole, 1 mM TCEP). The His 6 tag was cleaved with tobacco etch virus protease (final concentration 0.1 mg/ml), and the excess of imidazole was removed by dialysis (overnight at 4°C) simultaneously. The solution was mixed with HisTrap HP resin to eliminate the cleaved His 6 tag and the His 6 -tagged tobacco etch virus protease. The flowthrough was collected, concentrated to 4 ml, and applied on a HiLoad Superdex 200 16/60 column (GE Healthcare) equilibrated with a buffer composed of 25 mM Tris-HCl, pH 8.0, 200 mM NaCl, and 1 mM TCEP.
Generation of Site-directed Mutants-Site-directed mutagenesis of MtHPP was performed using a polymerase incomplete primer extension technique (30). The plasmid isolated from bacterial strain used for the overexpression of the wild-type MtHPP served as a template. A T151A mutation (underlined) was introduced using two primers: GGAGCTAA-GAGCTTTATCACTGGGAAAC (coding strand) and TCTT-AGCTCCATCAATAGGATCTAATACCC (complementary strand). The 245 CDC 247 to 245 AAA 247 mutant (MtHPP-AAA) was generated using primers GGTGCTGCCGCATATGCAT-ATGCTC and ATATGCGGCAGCACCATATAATGGAAT-TTTAAC. The sequences of resulting plasmids were verified by  MAY 6, 2016 • VOLUME 291 • NUMBER 19 DNA sequencing. Both mutated proteins were overexpressed and purified according to the protocol developed for the wildtype enzyme.

Structures of M. truncatula Histidinol Phosphate Phosphatase
Crystallization and Data Collection-MtHPP was concentrated using Amicon concentrators (Millipore) to 19 mg/ml as determined by measuring the absorbance at 280 nm with an extinction coefficient of 46,400. Crystallization screening was performed using a robotic sitting drop vapor diffusion setup (Mosquito). The most promising hits were optimized manually in a hanging drop configuration. Crystals of MtHPP complex with free phosphate grew in 15% PEG 3350, 0.2 M diammonium hydrogen phosphate, pH 8.0. Paraton-N suited as a cryoprotectant. Aiming at the HOLP complex, crystals from the same condition were washed in a crystallization solution supplemented with 2 mM MgCl 2 , 5 mM HOLP, and 20% glycerol and were immediately vitrified in liquid nitrogen. Crystals grown in 30% PEG 3350, 0.1 M BisTris, pH 6.5, 0.2 M ammonium acetate, 10 mM magnesium acetate were incubated (i) for 8 min with 5 mM HOLP and flash-frozen in Paraton-N to obtain structure with HOL or (ii) overnight over the reservoir supplemented with 100 mM formaldehyde to complete cross-linking between Lys 158 and Cys 245 and vitrified in Paraton-N. Diffraction data were collected at 22-ID (complexes with PO 4 3Ϫ or HOL and cross-linked) and 19-ID (HOLP complex) beamlines at the Advanced Photon Source (Argonne, IL). The diffraction images were processed with XDS (31). Statistics of the data collection and processing are summarized in Table 1.
Determination and Refinement of the Crystal Structures-The crystal structure of MtHPP was solved by molecular replacement with PHASER (32), using the closest homolog for which the crystal structure was available, IMPase from Zymomonas mobilis, as the search probe (33) (Protein Data Bank (PDB) entry 4N81, 45% sequence identity). Automatic model building was carried out with the on-line version of ARP/ wARP (34). The atomic coordinates were placed inside the unit cell with ACHESYM server (35). COOT (36) was used for manual fitting in the electron density maps between rounds of model refinement in Refmac (37). Riding hydrogen atoms for the protein chain were included in the refinement with anisotropic atomic displacement parameters. The refinement statistics are listed in Table 1.
Determination of the Oligomeric State in Solution-The MtHPP elution profile from size exclusion chromatography was tested against gel filtration standard (1511901, Bio-Rad). Two independent experiments were performed: (i) injection of 3 ml of MtHPP and standard samples onto a HiLoad Superdex 200 16/60 column (GE Healthcare) at a flow rate of 1 ml/min and (ii) injection of 100-l samples onto a Superdex 200 10/300 GL column (GE Healthcare) at a flow rate of 0.5 ml/min. Cross-linking in the presence of formaldehyde vapors or carbon dioxide atmosphere was tested using MtHPP at a 0.5 mg/ml concentration, dissolved in buffer composed of 25 mM HEPES, pH 8.0, 50 mM NaCl, with or without 1 mM TCEP. An experiment with formaldehyde cross-linking was performed over 24 h by placing an open vial with 300 l of protein solution above a thin layer of formaldehyde (36% solution) in a closed bottle. To test the covalent dimer formation in the presence of CO 2 in a separate experiment, the open vials with protein solu-tions were positioned in a bottle connected (with rubber hosing) to a Styrofoam box with dry ice. The flow of gaseous CO 2 to the chamber containing vials with protein was tested daily by quenching flame from a lighter, and the dry ice was replenished when needed.
Mass Spectrometry-The samples were run through an Agilent 1290 liquid chromatograph with acetonitrile/water/acetic acid solvent on a Poroshell C3 reverse phase column onto an Agilent 6130 quadrupole mass spectrometer. For each sample, 3 l was loaded with a protein concentration of ϳ0.5 mg/ml, dissolved in 25 mM HEPES, pH 8.0, and 50 mM NaCl. Results were analyzed using the Agilent Chemstation software. For details, see the supplemental material.
Enzymatic Tests-The physiological, forward reaction of MtHPP was measured at room temperature. The phosphatase activity was determined by measuring the concentration of inorganic phosphate by an improved malachite green assay (38). Reactions were carried out in 200-l volumes in a buffer containing 20 mM Tris-HCl, pH 8.0, 50 mM NaCl, 2 mM MgCl 2 . Seven concentrations of the substrate (HOLP xLi, catalog no. 41486, Sigma) were used: 50, 100, 200, 400, 600, 800, and 1000 M. Prior tests confirmed that within this range of concentrations, the rate of the reaction is linear for Ͼ4 min. Enzymatic reactions were initiated by the addition of MtHPP to a final concentration of 20 nM and, after 4 min, quenched with 200 l of the color reagent. The samples were diluted with 600 l of H 2 O to a final volume of 1 ml. Standard reactions to evaluate optimal pH, Mg 2ϩ concentration, and Li ϩ inhibition were carried out using 50 M HOLP to minimize the carryover of Li ϩ cations from HOLP ϫ lithium salt. Activity of MtHPP-T151A and MtHPP-AAA mutants versus wild-type MtHPP was evaluated using 200 M HOLP or 200 M IMP (catalog no. 10007777, Cayman Chemicals). To ensure lack of measurable activity of both mutants, the mixture was let react for 10 min in a separate set of experiments. All assays were performed in triplicate. Absorbance was measured at 630 nm using a PharmaSpec UV-1700 spectrophotometer (Shimadzu). The amount of free phosphate was calculated on the assumption of a molar extinction coefficient for malachite green-phosphomolybdate complex of 90,000 M Ϫ1 cm Ϫ1 . Non-linear regression analysis was computed with Prism version 6.07 (GraphPad Software).
Other Software Used-Molecular figures were created with UCSF Chimera (39). Root mean square deviations were calculated in the same program. Ramachandran plot was calculated in Rampage (40). Structural features were recognized with Pro-Motif (41) within the PDBsum server (42). A phylogenetic tree of proteins containing the identical, 11-amino acid sequence motif based on a PSI-BLAST search (43) was constructed in MEGA6 (44). Signal peptides were recognized using the Tar-getP version 1.1 server (45,46).

Results and Discussion
Identification of the Histidinol Phosphate Phosphatase in M. truncatula-To identify the HPP enzyme in M. truncatula, the BLAST server (47) was applied. AtIMPL2 protein sequence (UniProt accession number Q6NPM8) (27,28), without the signal peptide, was used as the search probe. The three homologs in M. truncatula proteome (48), with the highest identity to

Structures of M. truncatula Histidinol Phosphate Phosphatase
AtIMPL2, all annotated as IMPases, were located in the same locus in the genome (Medtr3g117220), which indicated either three splicing forms of the enzyme or artifacts of in silico prediction. The form with highest identity and query cover to the Arabidopsis enzyme, of 75 and 92%, respectively (UniProt accession number G7J7Q5) was therefore chosen as the target protein (Fig. 2). The N-terminal signal peptide, which in vivo targets MtHPP to chloroplast, was determined by the TargetP version 1.1 server (45,46). The cleavage was predicted to occur between Arg 51 and Ala 52 , and the overexpressed construct was designed to yield the final protein with sequence starting from Met 53 .
Enzymatic tests revealed that indeed, in the presence of Mg 2ϩ , MtHPP was able to hydrolyze HOLP to HOL, with the release of inorganic phosphate. The optimal Mg 2ϩ concentration and pH for the enzyme was determined to be 5 mM and 8.4, respectively. However, for enzymatic tests, these values were lower, to reflect conditions in illuminated (daytime) chloroplast stroma, namely 2 mM Mg 2ϩ (49) and pH 8.0 (50, 51).
The obtained Michaelis-Menten constant (K m ) of MtHPP for HOLP hydrolysis of 263 Ϯ 28 M (Fig. 3) is between 180 and 400 M reported for the Arabidopsis (28) and wheat (13) enzymes, respectively. On the other hand, MtHPP catalyzes the reaction at a higher rate than AtIMPL2 because the two catalytic constants (k cat ) are 3.6 Ϯ 0.14 and 1.3 Ϯ 0.2 s Ϫ1 , respectively. Altogether, the catalytic efficiency (k cat /K m ) calculated for MtHPP (13.7 Ϯ 1.9 ϫ 10 3 ) exceeds the value reported for AtIMPL2 (7.9 Ϯ 0.2 ϫ 10 3 (28)). On the other hand, neither MtHPP nor MtHPP-T151A or MtHPP-AAA mutants have shown any detectable activity against IMP, which confirms previous observations that IMPase-like HPPs play a role only in His biosynthesis (28).
Overall Structure of MtHPP-The structures of MtHPP complexes were solved and refined using x-ray diffraction data ranging to high resolution ( Table 1). The obtained electron density maps allowed us to trace without ambiguity almost the entire protein chain. The exceptions are 9 Ϯ 1 (chain-and complex-dependent) N-terminal amino acids and the mobile loop between helices ␣1 and ␣2.
A monomer of MtHPP has an ␣␤␣␤␣ sandwich-like arrangement (Fig. 4). The N-terminal domain, which forms an ␣ ϩ ␤ structure, covers residues from the N terminus to Glu 201 . Two long ␣-helices (␣1 and ␣2) are separated by a mobile loop that is   The linker between the N-and C-terminal domains consists of residues between Val 202 and Asp 209 . An extensive interface between the N-and C-terminal domains results in the rigidity of the entire structure, meaning that there is no hinge between the two domains, and they cannot move independently. The C-terminal domain, residues Leu 210 -Trp 326 , constitutes an ␣/␤/␣ fold, in which the mixed parallel/antiparallel ␤ sheet is sandwiched between helices ␣6, 7 (3 10 helix), and ␣8 from one side (close to the ␤ sheet of the N-terminal domain) and 4, ␣5, and ␣9 from the other. MtHPP Dimeric Assembly Is Stabilized by Intermolecular Lys-CH 2 -Cys Covalent Bonds-Based on size exclusion chromatography, MtHPP as well as mutants MtHPP-T151A and MtHPP-AAA are dimeric in solution, because they elute between the standard molecular mass markers of 158 and 44 kDa, significantly closer to the latter (Fig. 5). The PISA server (55) also suggests stable dimers, based on the monomer-monomer interfaces in the crystal structure. In the tetragonal crystal form (space group P 42 1 2), with one protein molecule in the asymmetric unit, monomers in the dimer are related to each other by a crystallographic 2-fold axis with the symmetry operation y, x, Ϫz. In the monoclinic crystal form (P 2 1 ), where four protein subunits make up the asymmetric unit, two dimers are present (chains A ϩ B and C ϩ D), but the monomer-monomer interfaces are the same as in the tetragonal crystals. The buried surface area upon the dimer formation equals 3470 Å 2 , whereas the surfaces of monomer and dimer equal 11,975 and 20,480 Å 2 , respectively.
In the crystal structures of MtHPP, a fascinating cross-linking interaction is observed within the dimer. Namely, methylene groups are present between N atoms of Lys 158 and S␥ atoms of Cys 245 . This way, the dimer is stabilized by two methylene bridges as they form in a mutual and interchangeable manner (i.e. Lys 158 from one protein subunit is linked to Cys 245 of the other, and the binding also occurs the other way around) (Fig. 6A). In the presented MtHPP complexes, the bond is observed at ϳ60% occupancy in the structures representing crystals that were harvested over 3 weeks after they had been grown (PO 4 3Ϫ and HOL complexes) but absent in crystal less than 2 weeks old (HOLP complex). Interestingly, this covalent dimerization could also be observed in protein solution that was kept in a CO 2 atmosphere where, after 20 days, dimers represent a substantial band on SDS-PAGE (Fig. 6B). It is also apparent that reducing agent prevents formation of covalent dimers, which, in the presence of TCEP, became barely observable after 20 days of the experiment, probably because a significant fraction of TCEP had degraded by that time.
Based on liquid chromatography-mass spectrometry (LC-MS) experiments, the relative abundance of dimers after 20 days in CO 2 atmosphere and in the absence of reducing agent was 17.4% ( Fig. 6C and supplemental material S-3). The molecular mass of 61,358.75 Da is close to the calculated mass of 61,358.2 Da, corresponding to the doubled mass of the monomer, each with the N-terminal SNA linker that resides after cleavage with tobacco etch virus protease (2 ϫ 30,378.3 Da), plus the difference from the two added methylene groups (24 Da) with SNAMSS adduct (577.6 Da) that most probably is an artifact resulting from the MS experiment.
The formation of methylene bridges most probably involves reaction of the primary amine of Lys 158 with CO 2 to form carbamic acid (56). The carbon within the carbamic group possesses a partial positive charge, suitable for a nucleophilic attack by S␥ of Cys 245 (57), but it is unclear how the cross-linking carbonyl is reduced to the methylene group. Theoretically, analogous bonds could form between any exposed cysteine and lysine residues, but they have never been reported. This suggests that the correct geometry (i.e. accurate location of N and S␥ atoms of lysine and cysteine residues, respectively) must be ensured for the bond to form.
Moreover, when MtHPP was subjected to formaldehyde, which is able to cross-link amines and thiols (58), the results were much more radical, although only HCHO vapors were allowed to contact the protein-containing solution in each case. More precisely, when the reservoir solution below the crystals was supplemented with 100 mM formaldehyde, the cross-linking was completed overnight by HCHO vapors diffusing into the crystallization drop (Fig. 6A). Consistently, the protein sample kept in an open vial placed in a closed bottle with formaldehyde at the bottom showed a high fraction of dimers after a 1-day incubation (Fig. 6, B and C). Inhomogeneity of monomers and dimers, observed on SDS-PAGE and MS, can be attributed to the fact that formaldehyde is a very potent agent in terms of reacting with proteins. Hence, MtHPP was modified at multiple sites, which caused relatively wide distribution of molecular masses. Nonetheless, the presence of covalent dimers is clear using both methods for verification.
It is also of note that although neither Lys 158 nor Cys 245 takes part in the enzymatic reaction (see below), both residues are conserved in the Arabidopsis enzyme, which suggests that analogous cross-linking might occur also in AtIMPL2. At this point, we are not able to unambiguously determine how the covalent dimerization influences the enzymatic reaction due to difficulties in obtaining a protein sample containing only covalent dimers. It would be of great interest to investigate the in vivo half-life of MtHPP and correlate it with the time required to the  MAY 6, 2016 • VOLUME 291 • NUMBER 19

JOURNAL OF BIOLOGICAL CHEMISTRY 9965
Lys 158 -CH 2 -Cys 245 bond to form. Unfortunately, whether it is a post-translational modification, a formaldehyde scavenging mechanism, or simply an artifact remains unclear. Nevertheless, both agents, CO 2 and formaldehyde, occur in each living cell, so it is likely that at least a small fraction of MtHPP (or AtIMPL2) could exist as covalent dimers in vivo.

HOLP Dephosphorylation Reaction Occurs at the Interface between N-and C-terminal Domains-Crystal structures of
MtHPP complexes obtained by swift (MtHPP-HOLP complex) or prolonged (MtHPP-HOL) soaking with the substrate (HOLP) or cocrystallization with by-product (PO 4 3Ϫ ) provide very detailed information about the reaction catalyzed by the enzyme. From the MtHPP complex with HOLP, we learn that the substrate is anchored to the protein by an extensive network of hydrogen bonds (Fig. 7A). Many of these bonds are mediated by water molecules that have low atomic displacement param-eters. Namely, the N⑀ atom of HOLP via a water molecule interacts with the amide nitrogen atom of Ser 264 , with a carbonyl oxygen of Leu 222 , and, through another water, with Thr 217 and Ser 219 . N␦ forms a water-bridged hydrogen bond with a carbonyl oxygen of Gly 265 . The N atom of HOLP forms direct hydrogen bonds with carboxyl oxygen atoms of Asp 149 and Asp 246 and a water-bridged interaction with carboxyl oxygen of Glu 263 . The O atom of HOLP, where the hydrolysis occurs, interacts with the Asp 246 by a water molecule, but this bond is weak, because the distance between HOLP O atom and the water molecule is Ͼ3.1 Å. The phosphate moiety interacts with amide nitrogen of Gly 150 , Thr 151 (OP1 atom of phosphate), and O␥ of Thr 151 (OP2). The overall dipole moment of the ␣3 helix, of which Thr 151 is the N-terminal residue, increases MtHPP interaction with the phosphate moiety of HOLP. In addition, two lysines at positions 152 and 267, conserved in AtIMPL2, possibly act as claws that recruit HOLP (by its phosphate group) and guide the substrate at the initial phase of binding. A similar behavior of two Lys residues, surrounding the entrance to the active site, was proposed for NADP(H) binding to M. truncatula ␦1-pyrroline-5-carboxylate reductase (59).
MtHPP orients HOLP in such a way that the phosphate moiety is exposed to Asp 146 and Asp 270 . Those two aspartates coordinate a magnesium cation (at 0.7 occupancy) in the HOLP complex structure. The presence of Mg 2ϩ cation was concluded based on O-Mg 2ϩ distances that refined to 2.11 Ϯ 0.07 Å (eight distances in four protein subunits were taken into account), which is very close to the ideal value of 2.107 (60). Those distances are certainly too short for a hydrogen bond involving a water molecule (Ն2.4 Å), meaning that placement of a water molecule instead of Mg 2ϩ would result in serious clashes. The Mg 2ϩ cation separates Asp 146 and Asp 270 by 3.4 Å in the complex with HOLP. Without Mg 2ϩ in between, as it is in all of the other MtHPP structures presented, the conformation of the Asp 270 side chain is different (Fig. 7), and the carboxyl groups of Asp 146 and Asp 270 directly interact with each other (shortest O-O distance is 2.4 Å). In that case, one of the carboxylic oxygen atoms of the aspartates (146 or 270) must be protonated and act as a hydrogen bond donor.
In the crystal structures of MtHPP complexes, Mg 2ϩ couples Asp 146 and Asp 270 only in the complex with HOLP, and we failed to incorporate the metal into any other complex, regardless of Mg 2ϩ concentration tested (up to 20 mM). Also, in the absence of magnesium, we were unable to obtain MtHPP complex with HOLP neither by cocrystallization nor by soaking. It is therefore likely that Mg 2ϩ enters the active site complexed with HOLP, and both partners only bind together. There are, indeed, many negatively charged residues surrounding the active site that would repulse the negatively charged phosphate moiety of HOLP, and the substrate can only bind when complexed with Mg 2ϩ , with a neutral net charge. Furthermore, it appears that after the hydrolysis has completed, Mg 2ϩ leaves the active center, because in the structure obtained by 8-min soaking in HOLP, the product, HOL, was found in the cavity, but Mg 2ϩ and PO 4 3Ϫ were not. The hypothesis that, in each reaction cycle, Mg 2ϩ enters the active site with the substrate and leaves with free phosphate agrees with the fact that Mg 2ϩ is bound to MtHPP without a typical octahedral coordination sphere. The weakly bound Mg 2ϩ can therefore easily escape the catalytic center. This observation is also consistent with the reported behavior of Mg 2ϩ in human IMPase at a similar site, concluded to bind and dissociate as the catalysis proceeds (53,61,62).
Superposition of MtHPP complexes with HOL and with PO 4 3Ϫ very well visualizes the immediate post-reaction state (Fig. 7B). One must remember, however, that this artificially constructed illustration still lacks Mg 2ϩ . Protein chains from MtHPP-PO 4 3Ϫ and HOL complexes superpose with a very small root mean square deviation of 0.1 Å (258 C␣ pairs within a 3-Å radius). The free phosphate interacts in a manner very similar to that bound to HOLP. There are, however, two additional, water-mediated hydrogen bonds with Asp 146 and Asp 270 that, in this case, also directly interact with each other. HOL is bound nearly identically to HOLP, and the only difference is the hydrogen bond between the hydroxyl oxygen atom of HOL and two peptide amides of Gly 150 and Thr 151 , which is bridged by a water molecule that occupies the same location as OP1 in the HOLP or PO 4 3Ϫ complexes. The fact that HOL is still present in the structure obtained by soaking MtHPP crystals with HOLP, whereas Mg 2ϩ and PO 4 3Ϫ are absent, might suggest a feedback inhibition mechanism. However, not only has it not been shown that HOL occurs in vivo at high concentration, but also HOL probably is rapidly converted to histidine by histidinol dehydrogenase. It is more likely that free His can suppress MtHPP activity, but the His concentration in chloroplast is unknown (10). Nevertheless, it would be disadvantageous for a cell to produce HOLP already with high effort and halt at this point. Consistently, it has been elucidated that the pool of free His is controlled by ATP-phos-  MAY 6, 2016 • VOLUME 291 • NUMBER 19

Structures of M. truncatula Histidinol Phosphate Phosphatase
phoribosyl transferase, the first enzyme in the His biosynthetic pathway (11,63).

MtHPP and IMPases; Comparison of the Reaction Mechanism-
The mechanism of reaction catalyzed by human IMPase was studied extensively in the last decade of the twentieth century due to an increasing amount of evidence indicating that human IMPase is a putative target of lithium therapy. Using numerous biochemical approaches, it was concluded that the hydrolysis of IMP involves a nucleophilic attack of a metal-activated water molecule (52, 53, 61, 62, 64 -66). More precisely, two metal cations were proposed to bind within the active site and increase the nucleophilic character of a water molecule that directly attacks the phosphorus atom of the substrate. However, the structural positions of metal cations in the human IMPase were inferred from the complex with gadolinium cations (Gd 3ϩ ), which are significantly larger and are typically surrounded by more electron-donating ligands and, thus, do not ideally reflect Mg 2ϩ binding sites. Finally, the structure of bovine IMPase, sharing 85% identity with the human enzyme, showed that even three Mg 2ϩ cations may bind within the active site of bovine IMPase (67). To compare IMPases and MtHPP, we superposed fragments of human IMPase (HsIMPA1) in complex with IMP (Protein Data Bank entry 1IMA (65)) with Mg 2ϩ -binding residues of bovine IMPase (Protein Data Bank entry 2BJI (67)) (Fig. 8). The three Mg 2ϩ cations in bovine IMPase bind to the following ( Fig. 8A; watermediated interactions are excluded): site 1, carboxylic oxygen atoms of Glu 70 (corresponding to Glu 128 in MtHPP) and Asp 90 (Asp 146 ) and carbonyl of Ile 92 (Ile 148 ); site 2, carboxylic oxygen atoms of Asp 90 (Asp 146 ) and Asp 220 (Asp 270 ); site 3, carboxyl of Glu 70 (Glu 128 ). However, this structure was obtained in a 20 mM concentration of Mg 2ϩ , which is outside the physiologically relevant range. It is therefore very likely that only Mg 2ϩ binding site 1 with K d Ϸ 300 M (68) and site 2 with K d Ϸ 3 mM (69) are occupied in vivo. More importantly, Mg 2ϩ bound at site 1 was reported to remain tethered throughout the entire catalytic cycle, unlike Mg 2ϩ at site 2 that binds and dissociates over the course of the reaction (53,61,62).
The crystal structures of MtHPP complexes are consistent with the direct displacement of phosphate mechanism, reported for IMPases (70). Although, theoretically, Thr 151 in MtHPP could bear the phosphate and form a phosphoenzyme intermediate, it is very unlikely that this residue alone, without any metal cation, could perform a nucleophilic attack via a mechanism known for serine/threonine phosphatases (71). The residue structurally corresponding to Thr 151 , Thr 95 of human IMPase, was excluded as the nucleophile, when the T95A mutant was shown to sustain partial activity (62). When the MtHPP-T151A mutant, which did not show any detectable HPP activity, was cocrystallized with HOLP, the active site did not contain HOLP nor HOL (not shown). This suggests that in MtHPP, Thr 151 is probably required for efficient binding of the phosphate moiety of HOLP.
MtHPP is an IMPase-like enzyme, although important modifications in the reaction mechanism can be inferred from MtHPP structures. At this point, it is noteworthy that the structures presented herein were derived from enzymatically active crystals. The most significant difference is the lack of Mg 2ϩ cation at binding site 1 of bovine IMPase in any of the MtHPP complexes and binding of Mg 2ϩ at a location similar to site 2 in MtHPP-HOLP complex (Fig. 8). The role of Mg 2ϩ cations at sites 1 and 2 has been puzzling, because initially, it was reported that in bovine IMPase, Mg 2ϩ at site 1 interacted with the phosphate, but Mg 2ϩ at site 2 was proposed to be responsible for activation of the water molecule (72)(73)(74). In that scenario, the nucleophilic attack would start from the site adjacent to the leaving group (HOL) and proceed in a non-inline pathway with pseudorotation, which means that the configuration of the phosphorus atom would be retained. Later, for the same enzyme, using [ 16 O, 17 O, 18 O]thiophosphate substrate analogue, it was shown that it is actually the opposite (i.e. action of bovine IMPase results in the inversion of phosphate configuration, and it is Mg 2ϩ -1 that activates water) (75,76). This contrasts with the case of MtHPP, which does not bind Mg 2ϩ at the position corresponding to site 1 in bovine IMPase. When Mg 2ϩ is bound to MtHPP, the metal separates Asp 146 and Asp 270 , and negative charge of those aspartates is neutralized. However, in bovine IMPase, the Mg 2ϩ -2 cation is bound ahead of the two aspartates (Asp 90 and Asp 220 ), whereas Asp 146 and Asp 270 at the corresponding structural positions in MtHPP coordinate the metal behind the two carboxylic groups (Fig. 8). Obviously, in MtHPP, the Mg 2ϩ -harnessed pair of Asp 146 and Asp 270 may still attract hydrogen of the water molecule, thus providing an active nucleophilic agent, but the water molecule is activated indirectly, by magnesium-bound aspartates, and not directly by aspartate-bound metal, as in bovine IMPase (67). It must be noted that the hydrolysis catalyzed by MtHPP needs to be performed by a water molecule and not by a hydroxyl anion, because the latter would be repulsed by likewise negatively charged phosphate. In the crystal structure of MtHPP-HOLP complex, this "attacking" water molecule is observed at two alternative locations (Figs. 7A and 9), most probably one of which represents the state before activation and the other after activation. Water molecule at the position representing the preactivated state is present in other MtHPP complexes at full occupancy. In MtHPP-HOLP complex, the distance between the activated water and the phosphorus atom of HOLP is 3.1 Å, which is expected for a direct nucleophilic attack. The electron density for the active H 2 O is in fact uninterrupted between the two alternative conformations (preactivated and activated state), separated by ϳ1.7 Å, but then it narrows and continues toward the phosphorus atom of HOLP (Fig. 9), meaning that even more transient states are captured within the crystal (i.e. averaged in the presented electron density maps). The shape of the electron density maps for HOLP and the attacking water molecule (Fig. 9) clearly indicates that the nucleophile attacks from the same face of the phosphorus atom as the leaving group (HOL). In other words, MtHPP catalyzes the reaction that proceeds in a non-inline pathway. The course of the non-inline reaction has been established using non-enzymatic systems (77). It was shown that the intermediate pentacovalent phosphate (78) undergoes a pseudorotation, because the initially apical nucleophile moves to an equatorial position, whereas the leaving group moves from an equatorial to apical orientation. This results in retention of the configuration of phosphorus.
In the last step of the reaction, the oxyanion oxygen atom that remains at HOL after the hydrolysis retrieves the proton from the oxygen of the water molecule that performs the attack and becomes a part of free phosphate or from a different H 2 O/ H 3 O ϩ or from the amino group of HOLP. There are no other potential proton donors, because the carboxyl oxygen atoms of Asp 149 and Asp 246 , although they are within a 4-Å radius, still coordinate amine nitrogen of HOLP and thus must be deprotonated.
Elements of the C-terminal Domain Are Responsible for Specificity toward HOLP-The sequence alignment (Fig. 2) indicates that the N-terminal domains of IMPases and IMPase-like HPPs are more conserved than their C-terminal domains. In fact, identities shared by MtHPP and AtIMPL1 equal 25% for the N-terminal and 16% for the C-terminal domains. As shown above, the residues that interact with the phosphate (Asp 149 , Gly 150 , and Thr 151 ) belong to the N-terminal domain. So does Asp 146 , which coordinates Mg 2ϩ and, according to the mechanism proposed in this paper, activates a water molecule that performs the nucleophilic attack. Therefore, Asp 270 is the only residue directly involved in the enzymatic reaction that resides in the C-terminal domain. Simultaneously, Asp 270 is one of a few C-terminal residues conserved among proteins from the IMPase family.
The crystal structure of MtHPP-HOLP complex shows that binding of the HOL moiety of substrate involves many residues from the C-terminal domain (Thr 217 , Ser 219 , Leu 222 , Asp 246 , Glu 263 , Ser 264 , Gly 265 , and Asp 270 ) and the side chain of Asp 149 (N-terminal domain). The comparison of active sites (Fig. 8) gives a clear vista on many differences between IMP binding by IMPases and HOLP binding by MtHPP, although some substrate-interacting residues are universal (e.g. Gly 265 in MtHPP and Gly 215 in HsIMPA1; Glu 263 /Glu 213 ). The most important modification that enables MtHPP to bind HOLP is the presence of Asp 246 , which, by its carboxyl oxygen atom, forms a hydrogen bond with N atom of HOLP. The O2 atom of IMP, which lies in a similar location, is bound to HsIMPA1 by the backbone amide of Ala 196 . This alanine residue is conserved in other IMPases (e.g. Ala 275 in AtIMPL1; Fig. 2) that do not catalyze hydrolysis of HOLP. It is likely that the amide exposed to the catalytic pocket prevents HOLP from entering to an active site of IMPase because it would result in disfavored amide-amine interaction. Asp 246 of MtHPP is conserved not only in AtIMPL2 but also in prokaryotic IMPase-like HPPs, CghisN and SchisN (24,25), which allows us to postulate that the aspartate corresponding to position 246 in MtHPP sequence is required for binding of HOLP. Consistently, the MtHPP-AAA mutant, in which the genuine 245 CDC 247 fragment was substituted with 245 AAA 247 (as in AtIMPL1), was unable to catalyze HOLP to HOL hydrolysis. On the other hand, the MtHPP-AAA mutant remained inactive against IMP (as wild-type MtHPP), which suggests that the substrate recognition is more complex.
The analysis of amino acid sequences of IMPase-like HPPs and IMPases (Fig. 2) indicates another element that differs between the two protein families. According to PROSITE (79) (Fig. 2). All four so far confirmed IMPase-like HPPs, including MtHPP, lack the PS00630 pattern but contain the N-terminal PS00629 pattern. Both observations, the presence of the Asp residue at a position corresponding to 246 in MtHPP and the lack of the C-terminal consensus pattern, sketch the route for future determination of more IMPase-like HPP enzymes.
MtHPP Bears No Resemblance to HPP Enzymes from DDDD or PHP Families-As mentioned in the Introduction, in nature there exist two other superfamilies of enzymes that catalyze dephosphorylation of HOLP to produce HOL, apart from the IMPase-like HPPs. Members of one of those two other HPP superfamilies are haloacid dehalogenase-like enzymes (80). They contain four invariant aspartate residues; hence the name of the superfamily, DDDD (21). In E. coli and in other protobacteria, DDDD domains responsible for HPP activity are a part of larger and bifunctional proteins, whose N-terminal domains possess HPP activities, whereas C-terminal domains are imidazole glycerol phosphate dehydratases (81). In E. coli HPP enzyme, the N-terminal domain of HisB (HisB-N), with a molecular mass of 19.9 kDa, is an about one-third smaller enzyme than MtHPP. Although HisB-N also uses aspartate residues for anchoring the Mg 2ϩ cation (Asp 10 and Asp 131 ), the arrangement of the active site and the protein fold (Fig. 10) are completely different from that of MtHPP. DDDD enzymes utilize a single metal cation (usually Mg 2ϩ ) to catalyze dephosphorylation, but the reaction proceeds via an aspartyl-phosphate intermediate (82). This covalent enzyme-phosphate adduct has even been captured in a crystal structure of HisB-N (83). A second reaction step, nucleophilic attack of a water molecule, is required to release the phosphate group from Asp and recycle the enzyme. The entire process involves double inversion of configuration of the phosphorus atom (i.e. overall, the configuration is retained).
MtHPP, the IMPase-like HPP, executes a single-step reaction, in which the non-inline nucleophilic attack is performed by an activated water molecule (see above). In MtHPP, the Mg 2ϩ -tethering Asp 146 and Asp 270 are too distant from the phosphorus atom (Ͼ5 Å), which excludes the possibility of Asp-phosphate covalent complex formation. In the case of MtHPP, the retention of the conformation of the phosphorus atom results from pseudorotation and not from double inversion as in HPPs from the DDDD superfamily. The differences in reaction mechanism are reflected by the very diverse turnover rates of IMPase-like and DDDD HPPs. The k cat value for MtHPP is 3.6 s Ϫ1 , whereas HisB-N converts HOLP to HOL nearly 3 orders of magnitude faster, with k cat ϭ 2.14 ϫ 10 3 s Ϫ1 (83). HisB-N also has a ϳ5 times lower K m (54 M) than does MtHPP, making the latter, and all other examples of IMPaselike HPPs, relatively inefficient enzymes.
Some prokaryotes rely on HPP enzymes from the PHP superfamily. One such example is Lactococcus lactis HPP (LlHPP). LlHPP protein is folded into a distorted (␤/␣) 7 barrel and possesses a trinuclear metal center (usually three Zn 2ϩ ) within the active site (84). Two of those metal ions coordinate a hydroxide ion, which functions as the hydrolytic nucleophile. The third Zn 2ϩ coordinates the oxygen atom of HOLP, where the hydrolysis occurs. In LlHPP, the three metal cations in the active center are coordinated by six His side chains, two Asp and one Glu (Fig. 10B). The fourth Zn 2ϩ is bound outside the catalytic center and does not play a role in the enzymatic reaction. LlHPP is slower than E. coli HisB-N from the DDDD superfamily and turns over with k cat ϭ 174 s Ϫ1 , still nearly 50 times faster than MtHPP does. On the other hand, LlHPP is characterized by a relatively high K m ϭ 1.9 mM.
The studies of MtHPP presented here show that it is completely unrelated to either of the two other HPP superfamilies, FIGURE 10. Examples of HPP proteins from DDDD and PHP superfamilies. A, the protein chain of E. coli HisB-N phosphoaspartate complex (Protein Data Bank entry 2FPW) is shown with superposed HOL (gray carbon atoms, ball-and-stick model) and Mg 2ϩ from Protein Data Bank entry 2FPU structure (83). The four conserved Asp residues that give the DDDD superfamily its name are underlined. B, LlHPP, from the PHP superfamily, in complex with L-histidinol-arsenate (gray carbon atoms, ball-and-stick model, 4GK8 (84)). Magenta spheres, locations of Zn 2ϩ cations. DDDD or PHP. MtHPP, as an example of IMPase-like, plant HPPs, is significantly less efficient than examples from the other two superfamilies. This raises the question of what caused this evolutionary curiosity of three unrelated housekeeping enzymes. So far, it appears that there are not many examples of confirmed IMPase-like HPPs; therefore, a thorough evolutionary analysis is difficult. The structural and functional studies of MtHPP presented here give hints to distinguish IMPase-like HPPs from IMPases, which facilitate separation of the two types of enzymes. With more metagenomic data, it should soon be possible to trace the evolutionary origin of plant HPPs.
Author Contributions-M. R. designed and performed the studies, analyzed the results, and wrote the manuscript. Z. D. analyzed the results and supervised the work.