Essential Role of Proximal Histidine-Asparagine Interaction in Mammalian Peroxidases*

In heme enzymes belonging to the peroxidase-cyclooxygenase superfamily the proximal histidine is in close interaction with a fully conserved asparagine. The crystal structure of a mixture of glycoforms of myeloperoxidase (MPO) purified from granules of human leukocytes prompted us to revise the orientation of this asparagine and the protonation status of the proximal histidine. The data we present contrast with previous MPO structures, but are strongly supported by molecular dynamics simulations. Moreover, comprehensive analysis of published lactoperoxidase structures suggest that the described proximal heme architecture is a general structural feature of animal heme peroxidases. Its importance is underlined by the fact that the MPO variant N421D, recombinantly expressed in mammalian cell lines, exhibited modified spectral properties and diminished catalytic activity compared with wild-type recombinant MPO. It completely lost its ability to oxidize chloride to hypochlorous acid, which is a characteristic feature of MPO and essential for its role in host defense. The presented crystal structure of MPO revealed further important differences compared with the published structures including the extent of glycosylation, interaction between light and heavy polypeptides, as well as heme to protein covalent bonds. These data are discussed with respect to biosynthesis and post-translational maturation of MPO as well as to its peculiar biochemical and biophysical properties.

In heme enzymes belonging to the peroxidase-cyclooxygenase superfamily the proximal histidine is in close interaction with a fully conserved asparagine. The crystal structure of a mixture of glycoforms of myeloperoxidase (MPO) purified from granules of human leukocytes prompted us to revise the orientation of this asparagine and the protonation status of the proximal histidine. The data we present contrast with previous MPO structures, but are strongly supported by molecular dynamics simulations. Moreover, comprehensive analysis of published lactoperoxidase structures suggest that the described proximal heme architecture is a general structural feature of animal heme peroxidases. Its importance is underlined by the fact that the MPO variant N421D, recombinantly expressed in mammalian cell lines, exhibited modified spectral properties and diminished catalytic activity compared with wild-type recombinant MPO. It completely lost its ability to oxidize chloride to hypochlorous acid, which is a characteristic feature of MPO and essential for its role in host defense. The presented crystal structure of MPO revealed further important differences compared with the published structures including the extent of glycosylation, interaction between light and heavy polypeptides, as well as heme to protein covalent bonds. These data are discussed with respect to biosynthesis and post-translational maturation of MPO as well as to its peculiar biochemical and biophysical properties.
The majority of currently known heme peroxidases, ubiquitous in all kingdoms of life, are members of two superfamilies that arose independently. The superfamily of (archae)bacterial, fungal, and plant heme peroxidases (sometimes called "nonanimal" peroxidase superfamily) is represented by catalase-per-oxidases, ascorbate peroxidases, cytochrome c peroxidases, manganese and lignin peroxidases, and plant secretoric peroxidases (1,2). The second superfamily (named the peroxidasecyclooxygenase superfamily) was defined recently based on the reconstruction of the main evolutionary lines of mammalian heme peroxidases (3). This peroxidase-cyclooxygenase superfamily includes the mammalian peroxidases represented by myeloperoxidase (MPO), 2 eosinophil peroxidase (EPO), lactoperoxidase (LPO), and thyroid peroxidase (3).
Both superfamilies differ greatly in their primary and tertiary structures and also, most strikingly, in the nature of the heme prosthetic group. Mature mammalian peroxidases are posttranslationally modified with the heme covalently linked to the protein via autocatalytic formation of two ester bonds with highly conserved aspartate and glutamate residues (4 -7). MPO is singular in having additionally a sulfonium ion linkage between the heme 2-vinyl group and a conserved methionine. The existence of these three covalent hemes to protein bonds has been correlated with the peculiar spectroscopic, redox, and catalytic properties of MPO (8,9). Other, although more subtle, structural differences between the two heme peroxidase superfamilies concern the H-bonding partners of the essential proximal and distal histidines (4,10).
Closely related with these structural peculiarities is the physiological role and the nature of the substrates of these oxidoreductases. MPO, EPO, and LPO are functionally homologous enzymes involved in host defense. Myeloperoxidase is secreted at inflammatory sites from stimulated polymorphonuclear leukocytes and also monocytes (11), whereas EPO is released from activated eosinophils (12). Lactoperoxidase is found in mucosal surfaces and exocrine secretions such as milk, tears, and saliva (13). In contrast, thyroid peroxidase is involved in the biosynthesis of the thyroid hormones thyroxine and triiodothyronine (14). These four metalloproteins prefer small anionic molecules as electron donors, such as halides (chloride, bromide, and iodide), thiocyanate, and nitrite (8). The corresponding oxidation products (e.g. hypohalous acids or nitrogen dioxide) are (strong) halogenating and nitrating oxidants that play an important role in the innate immune defense system but also contribute to tissue injury in certain inflammatory diseases (15). Due to their important role in (patho)physiology, these enzymes are of strong interest for the pharmaceutical industry (15).
X-ray structures of mammalian peroxidases have been published for one glycoform of canine (16) and human (4,17) myeloperoxidase, as well as for caprine lactoperoxidase (5) (supplemental Table S1). Here, we present the crystal structure of a mixture of MPO glycoforms as obtained directly by purification from human leukocytes. The determined structure exhibits several significant differences compared with those reported in the literature. These variations include heme to protein linkages, sites, and extent of glycosylation as well as the interaction of the C terminus of the light with the N terminus of the heavy polypeptide. X-ray modeling, cross-checked by molecular dynamics simulations, revealed new insights in the interaction between proximal His 336 and Asn 421 . Specifically, the imidazole of His 336 could be present as imidazolate. Its importance in maintaining the physical and catalytic properties of MPO is underlined by the fact that variant N421D exhibited modified spectral properties and completely lost its chlorination activity.

EXPERIMENTAL PROCEDURES
Materials-Buffy coat was kindly provided by the Department of Serology and Transfusion Medicine at the LKH, Graz, Austria. All chromatography columns and media were from Amersham Biosciences. Cetyltrimethylammonium bromide, phenylmethylsulfonyl fluoride, methyl-␣-D-glucopyranoside, EDTA, sucrose, and all other chemicals were from Sigma and at the highest grade available.
Protein Purification-Human MPO was obtained from pooled buffy coat of about 200 different healthy donors. This MPO sample was a mixture of glycoforms and will be referred as hMPOm henceforth. In detail, buffy coat, the layer between plasma and red cells containing the leukocytes, was collected by centrifugation and washed twice with a solution of 155 mM NH 4 Cl, 10 mM Na 2 CO 3 , and 0.1 mM EDTA to lyse the remaining red blood cells. After centrifugation (800 ϫ g, 10 min) leukocytes were resuspended in homogenization solution I (340 mM sucrose, 0.1 mM EDTA, 10 g/ml phenylmethylsulfonyl fluoride) and homogenized in a Potter-Elvehjem homogenizer with a motor-driven Teflon pestle. Centrifugation of the homogenate (500 ϫ g, 10 min), pellet collection, and resuspension in homogenization solution I was repeated until the green colored pellet disappeared.
Finally, the supernatant that contained the released leukocyte granules was pooled and centrifuged at 27,000 ϫ g for 40 min. The pellets, i.e. granules, were collected and resuspended in homogenization solution II (100 mM phosphate buffer, pH 7.8, 100 mM NaCl, 5 mg/ml cetyltrimethylammonium bromide, 10 g/ml phenylmethylsulfonyl fluoride) and homogenized in a Potter-Elvehjem homogenizer. After centrifugation (27,000 ϫ g, 40 min) the clear green supernatant was collected and the remaining pellets were again homogenized as described above.
About 700 ml of crude extract were loaded onto a concanavalin A-Sepharose column (2.6 ϫ 20 cm) equilibrated with 50 mM phosphate buffer, pH 7.0. After washing with 300 ml of equilibration buffer, hMPOm was eluted with 500 ml of elution buffer containing 800 mM methyl-␣-D-glucopyranoside in 100 mM NaCl in 100 mM phosphate buffer, pH 7. The eluate was applied directly onto a cation-exchange CM Sepharose Fast Flow column (2.6 ϫ 20 cm) equilibrated with 25 mM sodium acetate buffer, pH 4.7, containing 200 mM sodium chloride. After washing with 300 ml of equilibration buffer, hMPOm was eluted with 700 mM sodium chloride in 25 mM sodium acetate buffer, pH 4.7, and applied onto a phenyl-Sepharose 6 Fast Flow column (2.6 ϫ 20 cm) equilibrated with 1 M ammonium sulfate in 50 mM phosphate buffer, pH 7.0. Proteins were eluted by decreasing ammonium sulfate concentration linearly from 1 to 0.1 M in 50 mM sodium phosphate buffer, pH 7.0. Fractions containing hMPOm with a Soret maximum at 428 nm and a purity index (A 428 nm /A 280 nm ) higher than 0.85 were pooled and desalted using Centriprep-30 concentrators (cut-off membrane, 30 kDa). Myeloperoxidase in 5 mM phosphate buffer, pH 7.0, was vacuum dried and stored at Ϫ80°C.
Crystallization and Structure Determination-Vacuum dried hMPOm was solubilized with 50 mM Tris-HCl, pH 7.0, 100 mM NaCl up to a protein concentration of 10 mg/ml. Brownish crystals showing a tetragonal bipyramidal morphology were obtained at 20°C by the vapor diffusion hanging drop method over a reservoir containing 10% polyethylene glycol 8000, 8% ethylene glycol, and 0.1 M HEPES, pH 7.5. These crystals belonged to the tetragonal space group P4 3 2 1 2 with unit cell parameters a ϭ b ϭ 110.7 Å and c ϭ 252.3 Å and contained a dimer per asymmetric unit. A native data set was collected using synchrotron beam line ID14.4 at the European Synchrotron Radiation Facility from a crystal flash cooled using a cryoprotectant solution identical to its mother liquor but with increased ethylene glycol concentration (25%). Diffraction data were processed and scaled using the DENZO/SCALEPACK programs (18), respectively (see Table 1). Structure determination was achieved by molecular replacement with MOLREP (19), using as the search model the human MPO structure isoform C with PDB code 1CXP (4). Refinement, performed using REFMAC (20) and manual modeling with the graphic program O (21), gave final R work and R free agreement factors of 23.6 and 25.7%, respectively ( Table 1). The reliability of the electron density maps allowed location of the main chain of all residues even if not present in PDB code 1CXP. Structure factors and coordinates are deposited in the Protein Data Bank (entry code 3F9P) with an imidazolate as the proximal iron ligand.
Molecular Dynamics Simulations-The starting MPO coordinates used in the molecular dynamics simulations were also derived from PDB 1CXP, as this is the human MPO x-ray structure at the highest resolution (1.8 Å) available in the PDB. Although, mature MPO is a dimer, the protomers (each containing a light and a heavy chain) work independently and exhibit spectral and catalytic properties indistinguishable from those of the dimer (10). Thus, for computational convenience, models were constructed with only one protomer, half of the MPO molecule (corresponding to chains A and C in PDB 1CXP). Sugars and acetate and sulfate ions were removed, whereas crystallographic water molecules, chloride, and calcium ions interacting with chains A and C were retained. Hydrogen atoms were added using standard bond lengths and bond angles. The N and C termini of the A and C chains were considered in their charged forms. All Arg, Lys, Asp, and Glu residues were considered in their ionized form except Asp 94 and Glu 242 , which are covalently linked to the heme, and Asp 98 . The proximity between the carboxylate groups of Asp 98 and a propionate (O-O 2.68 Å) suggested that Asp 98 is protonated and this was supported by a preliminary MD analysis. Based on analysis of the local environment around each of the six histidines, all of them (except His 336 , see below) were considered as N ␦ protonated. The protein was solvated with ϳ22,000 water molecules and Cl Ϫ counterions were also included to neutralize the simulation cell. Differing in the protonation state of the proximal heme ligand His 336 , which also interacts with the amide group of Asn 421 , two models were considered, namely a neutral N ␦ -protonated His 336 (His-H), or an anionic His 336 (His Ϫ ).
The AMBER (version 99) (22,23) and TIP3P force fields (24) were used for simulation of the protein and water, respectively. Bonding and van der Waals parameters for the negatively charged His and for Met, Asp, and Glu covalently linked to the heme were provided by the GAFF extension of the AMBER force field (25). Point charges were derived according to the RESP procedure (26). Similarly, point charges were developed for heme-Fe(III). In the fitting, point charges on the porphyrin atoms were fixed at the AMBER heme-Fe(II) values and only the charges on the iron and the four pyrrole nitrogens were optimized.
Molecular dynamics simulations were performed with the NAMD program (27). Periodic boundary conditions were applied. The size of the simulation cell was ϳ102 ϫ 87 ϫ 88 Å. Long range electrostatic interactions were treated by the PME method. A 12-Å cutoff for the real part of the electrostatic and for van der Waals interactions was used. The integration time step was 1 fs. Bonds involving H atoms were kept rigid. After 1 ns of solvent equilibration and energy minimization of the whole system, each model was heated as follows: (i) 100 ps of NVT dynamics at 100 K, fixing the protein backbone; (ii) 100 ps of NPT dynamics (1 atmosphere, 100 K), fixing the protein backbone; (iii) 100 ps of NPT (1 atmosphere, 100 K); (iv) 100 ps NPT (1 atmosphere, 200 K); then, 2.5 ns of the MD simulations were performed at constant temperature (300 K) and pressure (1 atmosphere) (27).
Cloning, Expression, and Purification of Recombinant Wildtype MPO and the Variant N421D-Proteins were produced in two different mammalian cell lines, namely Chinese hamster ovary and HEK293. Transfection into hamster ovary cells, selection and culture procedures for transfected cells, as well as protein purification protocols were described in detail previously (28). Alternatively both wild-type and N421D were expressed in HEK293 cell lines. Briefly, full-length hMPO cDNA was introduced into pcDNA3.1(ϩ) vector as the EcoRI-NotI cassette. For mutation of Asn 421 a 1534 Kpn2I-NotI fragment was replaced by its mutated counterpart. The oligonucleotide primers used were: sense, 5Ј-CCTGC(T/ C)TAGACATGCAGCG-3Ј and antisense, 5Ј-GCATG(T/ C)TAGAGCAGGCAGG-3Ј. After ligation and amplification in Escherichia coli JM107 and control sequencing, for transfection of HEK293 cells, standard selection and cultivation protocols were followed with minor modifications (29). Single stable producing colonies were selected for production in triple flasks (Invitrogen) for 4 weeks. Supernatant was collected every 2-3 days, passed over a 0.22-m filter, and stored at 4°C. Purification included cation exchange chromatography (Sepharose CM fast flow; equilibration, 20 mM phosphate buffer, pH 7.5; gradient elution, 20 mM phosphate buffer, pH 7.5, 100 -500 mM NaCl) followed by fraction pooling, sample concentration by ultrafiltration, and finally, size exclusion chromatography (Sephadex 200).
Circular Dichroism (CD) Spectrometry-Circular dichroism spectrometry was performed on a PiStar-180 (Applied Photophysics, Leatherhead, UK) spectrometer equipped with a thermostatic cell holder. The instrument was flushed with nitrogen with a flow rate of 5 liters min Ϫ1 . For recording far UV spectra (180 -260 nm) the quartz cuvette had a path length of 1 mm. Experimental conditions were: spectral band width, 5 nm; step size, 1 nm; scan time, 15 min; buffer, 5 mM phosphate buffer, pH 7.0, 25°C. For comparison of secondary structure content both recombinant wild-type and N421D were brought to identical concentrations ascertained by identical absorption at 280 nm. Each spectrum was automatically corrected with the baseline to remove birefringence of the cell.
Enzyme Activity Assays-Enzymatic activity was determined polarographically and spectrophotometrically. Due to the fact that in the mutant protein heme occupancy was incomplete (see below), protein concentration was always calculated via heme absorbance to compare wild-type and mutant specific activities. Heme concentration was calculated via absorbance at the corresponding Soret maxima using an extinction coefficient of 91,000 M Ϫ1 cm Ϫ1 (9).
Consumption of hydrogen peroxide in the presence of halides was continuously monitored using a H 2 O 2 electrode.
The platinum electrode was covered with a hydrophilic membrane and fitted to the Amperometric Biosensor Detector 3001 (Universal Sensors, Inc.). The applied electrode potential was 500 mV at pH 5.0 and the electrode filling solution was prepared fresh half-daily. The electrode was calibrated against known concentrations of H 2 O 2 . In detail, the consumption of 100 M H 2 O 2 in the presence or absence of 100 M bromide or 100 mM chloride in 100 mM phosphate buffer, pH 5.0, was followed at 25°C. Reactions were started by addition of either 250 nM wild-type MPO or N412D. One unit of activity is defined as consumption of 1 mol of H 2 O 2 per min at 25°C.
Alternatively, halogenation and peroxidase activity were measured spectrophotometrically or fluorimetrically (Hitachi U-3000 and F-4500). The monochlorodimedone (MCD) assay was used to elucidate the chlorination and bromination activi-ties (30). Experimental conditions used were: 100 M MCD, 100 M bromide, or 100 mM chloride, 50 nM wild-type MPO or N421D, 100 mM phosphate buffer, pH 5.0 (25°C). The reaction was started by addition of 100 M H 2 O 2 and followed as absorbance decrease at 290 nm. Rates of halogenation were determined from the initial linear part of the time traces using an extinction coefficient for MCD of ⑀ 290 ϭ 19.9 mM Ϫ1 cm Ϫ1 (30). One unit of activity is defined as halogenation of 1 mol of MCD per min at 25°C.

RESULTS
Structure of hMPOm-A mixture of MPO glycoforms (hMPOm) isolated from human leukocytes was crystallized as bipyramidal tetragonal crystals reaching about 0.5 mm in length. With a solvent content volume of about 55%, these crystals contained, in their asymmetric unit, a whole MPO molecule and diffracted up to 2.8-Å resolution. The structure was solved by molecular replacement using as the search model the coordinates from human MPO isoform C (PDB code 1CXP) (4). As expected the overall structure of hMPOm was very similar to the previously reported MPO structures (4,16,17) (Fig. 1a). Each half of the molecule contains a light and a heavy chain resulting from the removal of a hexapeptide from pro-MPO during proteolytic processing (33,34). In the hMPOm structure, electron density seems almost continuous between the terminal carboxylate (Arg 106 ) from the light chain and the N terminus (Val 113 ) from the heavy chain, suggesting a direct interaction between the two polypeptides (Fig. 1b). In contrast, in the reported MPO structures, the N-and C-ends are distant or some of the corresponding resi- dues missing (4,16,17). The averaged root mean square deviation for the C␣ atoms with respect to the 1CXP structure is only 0.45 Å. Halide and calcium binding sites in hMPOm are mostly unchanged compared with the published MPO structures. However, in hMPOm, extra electron density was found close to Arg 161 that could correspond to a chloride ion. Another chloride might be present in close proximity to main chain nitrogens around Cys 152 , enlarging the cluster of negative ions that bind along the molecular 2-fold axis. Additional electron density at Cys 150 cannot be accounted for by a simple oxidation as proposed for the 1CXP structure (Fig. 1c).
The hMPOm structure appears to be extensively glycosylated (Fig.  1a). The MPO heavy chain sequence contains five potential asparagine glycosylation sites (Asn-X-Ser and Asn-X-Thr) and four (Asn 157 , Asn 189 , Asn 225 , and Asn 317 ) had been biochemically confirmed (10,34). However, in the 1CXP structure (4) only three N-glycosylation sites, with a total of eight monosaccharide moieties, were reported: Asn 189 and Asn 225 with a 1-4-linked N-acetylglucosamine and Asn 317 with a branched (GlcNac) 2 -(Man) 3 structure that has a fucose 1-6linked to the first GlcNac. In the hMPOm structure the electron density indicates the presence of glycosylation also at Asn 157 (Fig.  1d). At least two additional sugar monomers are visible at Asn 189 and Asn 317 , which increases considerably the amount of sugars at the interface between the two halves of the dimer.
In animal heme peroxidases the prosthetic group is covalently linked with the protein. Published structures of MPO (4, 16, 17) demonstrated the existence of three covalent bonds. In hMPOm the ester bond between Asp 94 and a hydroxymethyl group on pyrrole ring C is present, as is the sulfonium ion linkage between the ␤-carbon of the vinyl group on pyrrole ring A and the sulfur atom of Met 243 (Fig. 2a). However, surprisingly, the side chain of Glu 242 had a low electron density indicating high mobility and suggesting that the ester bond is only present in a fraction of the population of protein molecules (Fig. 2b). The electron density at the methyl-substituent of pyrrole ring A suggests its  Table S1). e, after adding hydrogen atoms, these contacts would result in H-H distances ϳ1 Å. hydroxylation, and the heme, although still considerably distorted, appears to be flatter compared with the published structures.
Special care had to be taken while modeling some of the distinct features of mammalian peroxidases in the heme proximal side (Fig. 2c). All previous MPO structures show the carbonyl group of Asn 421 oriented toward the proximal ligand His 336 (hereafter referred as orientation OHis) (supplemental Table S1). However, this places the amine group of Asn 421 very close to the guanidinium group of Arg 333 , with a H-H distance Ͻ1 Å, leading (see below) to a strong repulsive interaction (Fig.  2d). Arginine 333 also interacts with the heme propionic group, which is expected to raise even further the pK a of the guanidinium group, making neutral (unprotonated) arginine very unlikely. Therefore, in hMPOm we considered flipping the amide group of Asn 421 pointing its carbonyl group toward Arg 333 (hereafter referred as orientation OArg). In fact, a few of the LPO structures recently deposited in the PDB already have the proximal asparagine in the OArg orientation (supplemental Table S1). Nevertheless, if His 336 is neutral (His-H) the OArg orientation would lead again to a very unfavorable H-H contact now between the amine group of Asn 421 and the N ␦ hydrogen atom of His 336 (Fig. 2e). Consequently, the OArg orientation appears to require an anionic His 336 , which might be facilitated by lowering the pK a in His 336 as a result of its coordination with the heme iron.
Molecular Dynamics Simulations-Molecular dynamics simulations were performed to get more information about the orientation of Asn 421 and protonation status of the proximal histidine in MPO. The structure of human MPO, at 1.8-Å resolution (4), was used for the analysis.
The first MD simulation was performed with the OHis orientation and a neutral His 336 (His-H) residue (see Fig. 2d), as reported for PDB entry 1CXP. However, during this simulation the amide group of Asn 421 rotated ϳ180°adopting consistently the OArg orientation (Fig.  3a), although transiently exploring an alternative conformation (Fig.  3c). In the final structure, the Asn 421 side chain is slightly displaced with respect to the experimental crystallographic data, so that the carbonyl oxygen makes a bifurcated hydrogen bond with Arg 333 and His 336 ( Fig. 3a and supplemental Fig. S1), whereas the amino group moves away from His 336 . Such a bifurcated hydrogen bond is not present in the x-ray structure (4).
As the assignment of the Asn 421 orientation from x-ray experiments is uncertain, we compared the distances from the amide terminal heavy atoms, irrespective of their nature (O or N), to Arg 333 and His 336 in the high resolution x-ray structure (4) with the resulting conformations from MD simulations (Fig. 3). The closest distances among non-hydrogen atoms of the pairs Arg 333 -Asn 421 and His 336 -Asn 421 (3.00 Ϯ 0.19 and 3.76 Ϯ 0.33, respectively) differ substantially from the experimental structure (2.89 and 2.86 Å, respectively). The amino group of Asn 421 does not form hydrogen bonds and is surrounded by hydrophobic residues.
A second simulation was performed starting again with the OHis orientation for Asn 421 (Fig. 3b), but assuming now a negatively charged His 336 side chain (His Ϫ ). In this simulation the ϳ180°rotation of Asn 421 took place very rapidly reaching the OArg orientation that was then continuously sustained (Fig. 3,  b and c, and supplemental Fig. S2). In the resulting structure both the carbonyl and amino groups of the Asn 421 side chain are forming hydrogen bonds with Arg 333 and His 336 , respectively (supplemental Fig. S1). The closest distances between the pairs Arg 333 -Asn 421 and His 336 -Asn 421 (2.80 Ϯ 0.10 and 3.05 Ϯ 0.14, respectively) are in better agreement with the experimental structure (2.89 and 2.86 Å, respectively) (see also supplemental Table S1) than the values obtained with a neutral His 336 . Therefore, the MD simulations clearly demonstrate (i) that the com- monly assigned OHis orientation is not stable and (ii) that the OArg orientation only reproduces the experimental crystallographic data well when His 336 is anionic.
Biochemical Analysis of the MPO Variant N421D-To probe the role of Asn 421 in MPO catalysis recombinant wild-type MPO and N421D were heterologously expressed in and purified from two different mammalian cell lines (Chinese hamster ovary and HEK293) without obvious differences regarding the spectral and kinetic properties of the corresponding recombinant wild-type and mutant proteins. Fig. 4 depicts the electronic absorption spectra of the recombinant Fe(III) proteins expressed in HEK293 cells. Wild-type MPO had its Soret band at 428 nm and additional bands at 498, 570, 622, and 690 nm, respectively, characteristic of a six-coordinated high-spin heme (9). It must be mentioned that MPO purified from leukocytes (and used for x-ray crystallography in this work) had identical spectral features (not shown). This is important because the recombinant proteins are partially unprocessed monomers (28). Nevertheless, they exhibit identical spectral, redox, and enzymatic features (28,(35)(36)(37)(38) thus being valuable model proteins for studying structure-function relationships.
Interestingly, N421D showed a blue-shifted Soret band at 412 nm with a slight shoulder at 428 nm. The spectral intensity of this shoulder changed slightly depending on the point in time of protein harvest from both mammalian cell cultures. The purity number (A Soret /A 280 ) of the respective proteins were 0.5-0.62 (wild-type) and 0.18 -0.29 (N421D), indicating incomplete heme occupancy in the mutant protein. Thus, to guarantee comparability of enzymatic properties the concentration of active enzyme had to be calculated via Soret absorb-ance as has been demonstrated recently with other mutants (36,38,39) and described under "Experimental Procedures".
Far UV CD spectra (inset to Fig. 4) indicated that the overall structures of both the wild-type and mutant proteins are very similar suggesting that exchange of Asn 421 did not induce major structural rearrangements. If conformational changes did occur, they must have been very localized and minimal and thus were undetected by CD.
Together with modified heme absorbance the enzymatic activity of N421D was decreased compared with the wild-type protein.

DISCUSSION
One of the important structural differences between the two major heme peroxidase superfamilies concern the interaction of the proximal histidine. In non-animal (plant-type) heme peroxidases (1, 2) the proximal histidine is hydrogen bonded with a fully conserved aspartate that acts as a hydrogen bond acceptor deprotonating the N ␦ position of the imidazole. The resulting imidazolate character is mainly responsible for the negative standard reduction potentials (EЈ 0 ) of the corresponding Fe(III)/Fe(II) couples, which are typically within the range of Ϫ200 to Ϫ300 mV. These EЈ 0 values guarantee the stability of the ferric form that is competent in binding and, finally, reduction of hydrogen peroxide. The imidazolate character of the proximal histidine strengthens the iron-imidazole bond and helps to stabilize the oxidoreductase in its higher oxidation (i.e. compounds I and II) states.
In peroxidases belonging to the peroxidase-cyclooxygenase superfamily (3) the proximal histidine interacts with a fully conserved asparagine and, based on published x-ray structures, it was generally assumed that the N ␦ proton is H-bonded to the neutral amide oxygen of Asn 421 (similar to globins, where the proximal His is H-bonded to a neutral backbone carbonyl group). Skepticism about this proximal architecture in mammalian peroxidases was first raised by resonance Raman studies in the low-frequency region (9) that give valuable information about the iron-imidazole bond strength. Resonance Raman data demonstrated that in both MPO and LPO the iron-imidazole stretch is at fairly high frequencies (9) suggesting a substantial imidazolate character of the proximal histidine also in these heme enzymes. Now, with the help of the presented crystal structure of hMPOm, comprehensive and critical analysis of all published structures of mammalian peroxidases (supplemental Table S1) as well as performed molecular dynamics simulations, this apparent contradiction can be explained. The presented data suggest that in mammalian peroxidases the NH 2 group of Asn 421 (MPO numbering) is oriented toward the unprotonated N ␦ of the proximal His, whereas its carbonyl group interacts with the (positively charged) guanidinium group of Arg 333 . A neutral proximal histidyl imidazole would manifest unfavorable contacts. Instead, deprotonated His 336 will strengthen the iron-imidazole bond thereby supporting the resonance Raman investigations (9). In addition, anionic His will stabilize higher oxidation states, an effect that is very important in MPO because it is known to form (transiently) catalytically active redox intermediates with very positive reduction potentials. Myeloperoxidase exhibits anomalously high EЈ 0 values of the redox couples Fe(III)/Fe(II) (ϩ5 mV (9)), compound I/compound II and compound I/Fe(III), respectively (9,37). These peculiar redox properties, induced by the strong electron withdrawing property of the MPO, typical sulfonium ion linkage (39), are a precondition for its physiological role that involves hypochlorous acid production (8).
In LPO, which lacks the sulfonium ion bond, the midpoint potential (EЈ 0 ) of the Fe(III)/Fe(II) couple was determined to be Ϫ190 mV (40) similar to plant-type peroxidases, thus demonstrating a comparable (de)protonation status of the proximal histidine in LPO and plant-type peroxidases. This suggests that the orientation of the proximal Asn in LPO might also be OArg and this is supported by few of the LPO structures recently deposited in the PDB (supplemental Table S1).
The importance of the close His-Asn-Arg interaction in maintenance of both the proximal and distal heme cavity architecture is supported by the observed effects upon exchange of Asn by Asp. The MPO mutant N421D showed a lowered heme occupancy as well as altered modes of heme to protein linkages as suggested by the calculated A Soret /A 280 ratio and the blue (10 nm)-shifted Soret band. In addition, both its halogenation and peroxidase activity were significantly decreased. The chlorination activity was even lost completely (as in LPO or MPO with exchanged Met 243 (9)). This suggests that the oxidation capacity of the relevant redox intermediate compound I of N421D was significantly lower than in the wild-type protein, possibly due to absence of the sulfonium ion linkage as is supported by the blue-shifted (LPO-like) Soret band (9). In addition, absence of the sulfonium bond might lower the rigidity of the distal H-bonding network that has been demonstrated to support substrate (particularly inorganic anions) binding and oxidation.
The structure of hMPOm provided further important differences to published MPO structures. Translated MPO is known to undergo a series of proteolytic processing steps during synthesis and intracellular trafficking (33,34). These include cotranslational cleavage of the signal peptide and glycosylation of the primary translation product. The structure of hMPOm demonstrates glycosylation at four asparagines. This is in contrast to the published structures, but fits with biochemical studies (34). All glycosylated residues are in the heavy subunit and interactions of these sugar chains make up much of the dimer interface.
Transient association with chaperones assists in proper folding and heme insertion thereby forming pro-MPO that is competent to exit the endoplasmic reticulum and is able to reach the trans-Golgi network, late endosomes, and azurophilic granules for final processing and storage. On this route the N-terminal propeptide as well as a hexapeptide (ASFVTG) from the protein surface of pro-MPO is cleaved off forming a light and a heavy polypeptide. Finally, two light-heavy protomer units then interact to produce the mature form of MPO (Fig. 1).
In contrast to PDB entry 1CXP in the structure of hMPOm the light and heavy chain ends remain close in the absence of the hexapeptide. Both polypeptide chains are connected via a salt bridge between C-terminal Arg 106 (light chain) and N-terminal Val 113 (heavy chain) as is evident by a continuous electron density (Fig. 1b).
Another difference with published MPO structures concerns the heme to protein bonds in hMPOm. Electron density maps clearly suggest two alternative conformations of the side chain of Glu 242 with only about 20% having an intact ester bond. As a consequence the heme in hMPOm is flatter compared with the published structures although being still considerably distorted. Thus, in hMPOm the majority of the ester bond with Glu 242 is broken, whereas both the sulfonium ion linkage and the second ester bond are fully present. This allows the following important conclusions. (i) The heme to Glu 242 linkage is the most labile bond and contributes to heme distortion, (ii) its absence has no impact on the stability of the ester bond with Asp 94 and the sulfonium ion linkage, and (iii) its absence has no impact on the spectral features of hMPOm that had its Soret maximum at 428 nm as well as peaks at 570 and 622 nm (compare with Fig. 4).
It is unlikely that the ester bond between Glu 242 and the 1-methyl substituent is only partially formed due to incomplete hydroxylation. Due to the low resolution of hMPOm the hydroxylation at the methyl substituent cannot be absolutely proven. Autocatalytic bond formation needs H 2 O 2 that mediates formation of compound I and, finally, of a carbocation by transient oxidation of the carboxylate groups of Asp 94 and Glu 242 and subsequent hydrogen abstraction from the heme methyl groups (9). A comparable mechanism has been proposed for the autocatalytic formation of the sulfonium ion linkage. These radical mechanisms suggest a more or less concomitant formation of all three bonds in MPO. Apparently, the formation of two covalent bonds at pyrrole ring A by neighboring Glu 242 and Met 243 destabilizes the ester but not the sulfonium ion bond.
It remains unclear why in the published structure (PDB 1CXP) of an MPO isoform the ester bond with Glu 242 was fully present in contrast to hMPOm. Regarding the chemistry of ester bonds a role for pH of crystallization (pH 5.5 for 1CXP versus pH 7.5 for hMPOm) is unlikely as is an impact of glycosylation on the active site structure. It is important to mention that MPO glycoforms purified from neutrophilic granules most probably do not have distinct physiological roles. MPO is coreleased from these granules during phagocytosis together with many hydrolytic enzymes including glycosidases (41). Thus, it is reasonable to assume that MPO glycoforms are an artifact of the isolation procedure in vitro or are produced during phagocytosis in vivo.