The crystal structures of the ferric and ferrous forms of the heme complex of HmuO, a heme oxygenase of Corynebacterium diphtheriae.

Crystal structures of the ferric and ferrous heme complexes of HmuO, a 24-kDa heme oxygenase of Corynebacterium diphtheriae, have been refined to 1.4 and 1.5 A resolution, respectively. The HmuO structures show that the heme group is closely sandwiched between the proximal and distal helices. The imidazole group of His-20 is the proximal heme ligand, which closely eclipses the beta- and delta-meso axis of the porphyrin ring. A long range hydrogen bonding network is present, connecting the iron-bound water ligand to the solvent water molecule. This enables proton transfer from the solvent to the catalytic site, where the oxygen activation occurs. In comparison to the ferric complex, the proximal and distal helices move closer to the heme plane in the ferrous complex. Together with the kinked distal helix, this movement leaves only the alpha-meso carbon atom accessible to the iron-bound dioxygen. The heme pocket architecture is responsible for stabilization of the ferric hydroperoxo-active intermediate by preventing premature heterolytic O-O bond cleavage. This allows the enzyme to oxygenate selectively at the alpha-meso carbon in HmuO catalysis.

Biological heme catabolism is conducted by a family of enzymes termed as heme oxygenase (HO), 1 which catalyzes oxi-dative degradation of iron protoporphyrin IX (heme hereafter) to biliverdin IX, iron, and CO in the presence of reducing equivalents (1). In mammalian systems where electrons are supplied by NADPH through NADPH-cytochrome P450 reductase (2), HO is the enzyme responsible for excess heme excretion and iron recycling (3). The product CO has been implicated as a messenger molecule in various physiological functions (4 -6). In pathogenic bacteria, HO is essential for heme-based iron acquisition from a host lacking in free extracellular iron (7)(8)(9).
Major advances have been made in understanding the structure and function of HO by using catalytically active, truncated, water-soluble forms of recombinant HO-1, an inducible isoform of mammalian HO (10 -13). HO is not a hemeprotein by itself but utilizes heme as both a prosthetic group and a substrate. In its catalytic cycle, HO first binds 1 eq of heme to form a ferric heme-HO complex (Fig. 1). The first electron donated from the reducing equivalent reduces the heme iron to the ferrous state. Then O 2 binds to it to form a meta-stable oxy complex. One-electron reduction of the oxy form generates ferric hydroperoxo, which self-hydroxylates the ␣-meso carbon of the porphyrin ring to form the ferric ␣-meso-hydroxyheme intermediate (12,13). This is different from P450 enzymes, in which the O-O bond of the hydroperoxo is heterolytically cleaved to generate a ferryl (Fe 4ϩ ϭO) hydroxylating active intermediate (14). Ferric ␣-meso-hydroxyheme in HO exits as a ferric oxopholin resonance structure that includes a ferrous porphyrin neutral radical (11). Upon reaction with O 2 and one electron, ferric ␣-meso-hydroxyheme is converted to ferrous verdoheme. This conversion has been proposed to be initiated by the dioxygen reaction with the ferrous porphyrin neutral radical rather than that with the heme iron (11). The mechanism of the last oxygenation step, the conversion of verdoheme to biliverdin, is least understood, although the verdoheme iron appears to participate in the oxygen activation process (15) in a manner similar to that described in the first oxygenation step.
Three HO proteins from pathogenic bacteria have been purified: HmuO from Gram-positive Corynebacterium diphtheriae, HemO from Gram-negative Neisseria meningitides, and PigA from Gram-negative Pseudomonas aeruginosa (9,16,17). In comparison to the mammalian HO, none of them is membrane-bound, rather they are soluble and have smaller molec-ular masses: HmuO, 24 kDa; HemO, 26 kDa; and PigA, 23 kDa. HmuO has the highest sequence identity with mammalian HO-1, 33% sequence identity to the first 221 amino acids of human HO-1. Because of the lack of detailed spectroscopic and mutagenesis characterization, the structure and function of HemO and PigA, in comparison to those of HO-1 and HmuO, are not well characterized.
The heme-HmuO complex has an iron coordination structure and electronic state analogous to those found in the mammalian HO complexes. Ferric hydroperoxoheme, 2 ␣-meso-hydroxyheme, and verdoheme are intermediates of HmuO-catalyzed heme degradation, similar to those established for HO-1 (18). Although HmuO and HO-1 share the same molecular mechanism for heme degradation, HmuO is different from HO-1. HmuO has a higher pK a value for the heme-bound water, a less stable oxy form, a slower heme degradation rate, and a higher biliverdin affinity by HmuO (18 -20). NMR studies on the cyanide-bound heme-HmuO complex show that the heme orientation disorder in HmuO is quite different from that in human HO-1 in solution (21).
Structural information is important to understand the molecular mechanisms of mammalian and bacterial HO catalysis. Crystal structures of the heme complexes of human and rat HO-1 and those of HemO have been determined for their ferric and ferrous derivatives (22)(23)(24)(25)(26)(27)(28). Although they have provided insights into HO catalytic mechanisms, the following three issues remain to be elucidated. First is the ambiguity of how protons are delivered from the solvent water to the heme site so that the ferric hydroperoxo species and meso-hydroxyheme are formed. The crystal structures of the mammalian HO-1 heme complex indicate the presence of a water network in the distal heme pocket (22,23). The protons required for ferric hydroperoxo and hydroxyheme formation come from a distal water anchored to the carboxyl group of Asp-140 in HO-1, as revealed by mutagenesis and EPR/ENDOR studies (29,30) as well as the crystal structure of human HO-1 D140A (24). However, the exact route by which the protons are delivered from the water network to the dioxygen-bound heme site prior to the participation from Asp-140 has not been fully identified in the HO-1 crystal structures, although the presence of an extensive hydrogen bonding network has been pointed out in recent NMR studies on the ferric heme complex of human HO-1 (31). Whether a similar residue tethers a distal water molecule or a water network exists in the heme-HmuO system as to deliver protons for the first oxygenation step in HmuO catalysis is unknown.
A second issue to be addressed is the role of the proximal His in HO catalysis. Spectroscopic studies show that both rat HO-1 and HmuO have a neutral imidazole of the proximal His as the heme axial ligand (10,19). Initial crystal structures of the human ferric heme-HO-1 complex have reported the presence of hydrogen bonding interaction between the proximal His imidazole group and nearby residues (22). A strong proximal His hydrogen bonding would promote ionization of the imidazole group which would facilitate cleavage of the hydroperoxo O-O bond (14), resulting in formation of the inactive ferryl species in the first oxygenation step of HO catalysis (12). The proximal His hydrogen bonding interaction needs to be reassessed. The crystal structures of human HO-1 have also shown that the proximal His imidazole plane closely eclipses the ␤and ␦-meso axis of the porphyrin ring (22). Although the implication for this proximal His orientation has never been discussed in the previous crystallographic studies, it would reduce the "push effect" of the proximal His imidazole (14) and assist to stabilize the O-O bond of the iron-bound hydroperoxo (32).
Whether the proximal His of HmuO has such similar orientation remains to be determined. Crystal structure of the heme-HmuO complex would help to clarify these points.
Third is how the proximal and distal heme pockets affect the binding of HO to heme and the release of biliverdin, the final product of the HO catalytic pathway. The human HO-1 crystal structures reveal the presence of two molecules in an asymmetric unit with 2 alternate conformations. In one case, the heme group is tightly sandwiched between the proximal and distal helices; in the other, the distal helix is away from the heme, resulting in less restrained heme-helix contact (22). The latter suggests a more flexible distal helix and has been postulated to promote HO binding to heme and facilitate release of biliverdin (22,23). Because the rat HO-1 and bacterial HemO crystals used for high resolution structural determination have only one molecule per asymmetric unit (26,27), this postulate remains to be proven. An HO crystal containing multiple molecules per asymmetric unit would not only elucidate the significance of a flexible distal heme pocket to HO substrate binding and product release but also provide structural factors responsible for the distal helix flexibility. Thus, crystal structure determination of HmuO, the spectroscopic and enzymatic properties of which have been thoroughly characterized (18 -21, 33), will help to resolve the aforementioned issues. In addition, because of the strong sequence homology between the active sites of HO-1 and HmuO, crystal structures of HmuO would reveal structural features essential for HO catalysis.
To this end, we have determined the crystal structures of the ferric and ferrous forms of the heme-HmuO complex to 1.4 and 1.5 Å resolution, respectively. Although the heme-HmuO complex has an overall protein fold similar to that of the HO-1 complexes (22), there are a number of significant structural differences. The crystal structures of the heme-HmuO complex reveal hydrogen bond network extending from the heme catalytic site to the surface of the molecule through which protons required for oxygen activation could be delivered. Compared with the ferric form, both the proximal and distal helices of the ferrous form are closer to the heme, sterically restricting the orientation of the heme ligand toward the ␣-meso carbon atom of the porphyrin. This finding reinforces our original proposal that the HO regioselective oxygenation is primarily under steric control (34).

EXPERIMENTAL PROCEDURES
Protein Crystallization and Data Collection-Expression and purification of the recombinant HmuO and its reconstitution with heme were carried out as described previously (18). For crystallization, the purified heme-protein complex was concentrated to ϳ20 mg/ml in 20 mM MES buffer, pH 7.0. Crystals suitable for high resolution x-ray structural analysis were obtained by a sitting drop vapor diffusion method from a solution mixed with an equal volume of the protein stock solution and a reservoir solution containing 50 mM MES, pH 5.8, 2.2 M ammonium sulfate, and 230 mM sodium iodide. These crystals belonged to the space group P2 1 with unit cell parameters of a ϭ 53.8 Å, b ϭ 61.8 Å, c ϭ 108.8 Å, and ␤ ϭ 100.9°and contained three HmuO molecules per 1 asymmetric unit. The crystals of the ferrous form of the heme-HmuO complex were prepared by transferring the native ferric crystals into the reservoir solution containing 10 mM sodium dithionite. The unit cell parameters of the ferrous crystals are a ϭ 53.9 Å, b ϭ 62.9 Å, c ϭ 107.0 Å, and ␤ ϭ 101.2°. For cryogenic data collection, crystals were soaked into the reservoir solution containing 25% sucrose and flash-frozen by cold nitrogen stream. The temperature around the crystals was maintained at 98 K throughout the data collection. The high resolution diffraction data sets were collected with a Rigaku JUPITER CCD (210 mm) for the ferric crystals and a Rigaku R-Axis V imaging plate for the ferrous form using 0.9-Å synchrotron radiation at BL45XU (RIKEN beam line) of SPring-8. Data were integrated, merged, and processed with HKL-2000 (35), MOSFLM (36), and SCALA (36).
For phase determination (see below), crystals belonging to the space group C2 (37) were obtained by the sitting drop vapor diffusion method from a solution mixed with equal volumes of a 40 mg/ml protein stock solution and a reservoir solution containing 100 mM MES, pH 6.5, 21% polyethylene glycol 3350, and 200 mM calcium acetate, a condition similar to the one described previously (37). The C2 crystals had unit cell parameters of a ϭ 122.2 Å, b ϭ 43.6 Å, c ϭ 99.3 Å, and ␤ ϭ 131.6°a nd contained two molecules per 1 asymmetric unit. The crystal was mounted on a capillary with its original mother liquid and flash-frozen for the cryogenic data collection at 98 K with a cold nitrogen stream. Only this method, otherwise known as "capillary-cryo" (38), provided diffraction data useful for phase determination. Full data set was collected to 2.7 Å resolution with a Rigaku R-AXIS IV 2ϩ imaging plate using a CuK ␣ radiation generated by a Rigaku Ultra-X rotating anode generator. Data processing and reduction were carried out with MOS-FLM and SCALA (36).
Phase Determination, Refinement, and Structure Analysis-Initial attempt of phase determination with the 1.7 Å resolution diffraction data from the P2 1 crystals did not yield useful results. However, phases for the electron density of the C2 crystals were successfully calculated for the 2.7 Å resolution data by molecular replacement method with AMoRe from the CCP4 suites (36). A search model was constructed from the structure of human HO-1 (23; Protein Data Bank entry 1N45) by mutating residues not conserved between HmuO and human HO-1 to Ala. The rotation and translation searches were performed with the data in 10 to 3.5 Å resolution range. This led to a clear solution for two molecules in the asymmetric unit. The phases were calculated from the molecular replacement solutions up to 3.5 Å and then extended to 2.7 Å by density modification including solvent flattening, averaging between two molecules, and histogram matching with program DM from the CCP4 suites (36). In the end, an electron density map of sufficient quality for model fitting was obtained. An HmuO model was built into the electron density map by using the program O (39) and then refined to fit to the observed structure factors by using rigid body refinement and simulated annealing with a torsion angle by CNS (40) to an R factor of 0.324 and R free factor of 0.380. Throughout the refinement, noncrystallographic symmetry restraints were applied. This roughly refined model was then used as a search model for molecular replacement with the diffraction data from the P2 1 crystals. Molecular replacement solution was successfully obtained with the data in a range of 15 to 4 Å resolution using the program EPMR (41), a program implementing six-dimensional searches with evolutionary algorithm. The crystallographic refinements with simulated annealing and individual B factor refinement with CNS was performed to calculate unbiased model by using the data from 12 to 1.4 Å resolution range, which yielded an R factor of 0.220 and R free factor of 0.235. In the course of the refinement, water molecules were added to the model by manual inspection of their positions in both 2F o Ϫ F c and F o Ϫ F c maps. In addition, one sucrose molecule and six sulfate anions were assigned to the electron density. The model was further refined using the maximum likelihood target with the program REFMAC5 (42). After including additional water molecules, introducing alternative conformations for several residues, and modeling individual anisotropic thermal motions, the R and R free factors dropped to 0.165 and 0.192, respectively. The F o Ϫ F c omit electron density map calculated without the model around the heme pockets and the 2F o Ϫ F c map calculated after several cycles of unrestrained refinement were used to confirm the reliability of the structures around the heme pockets.
The structure of the ferrous heme-HmuO complex was also refined by the same method using the data from 12 to 1.5 Å resolution range. R and R free factors for the ferrous complex are 0.177 of 0.202, respectively. Final refinement statistics for the ferric and ferrous heme-HmuO complex structures are summarized in Table I. Throughout the model building and refinement process, 10% of the reflections were excluded to monitor the R free value. Several residues located at the N-and Cterminal regions were not visible in the electron density map probably due to their disorder. Coordinates and structural factors have been deposited in the Protein Data Bank under accession numbers of 1IW0 and 1IW1 for the ferric and ferrous forms, respectively.

RESULTS
Overall Structure-The crystal structures of the heme-HmuO complex, both in the ferric and ferrous states, have been solved by the molecular replacement method using human HO-1 as a search model. In the heme-HmuO complex crystal, three molecules, A-C, exist in the asymmetric unit and can be superimposed with an overall r.m.s. deviation of 0.41 Å. Similar to its mammalian HO-1 counterparts (22,25), the heme-HmuO complex crystal structure is mostly ␣-helical with the dimensions of ϳ45 ϫ ϳ30 ϫ ϳ25 Å, and its heme substrate is sandwiched between two helices termed as "proximal" and "distal" helices ( Fig. 2).
In contrast, the main chain fold of HmuO is less similar to that of Gram-negative N. meningitidis HemO. Structural comparison shows that the r.m.s. differences for equivalent 161 C␣ carbon atoms of HmuO (ferric form) versus human HO-1 (closed form) is 0.93 Å, whereas the r.m.s. difference for equivalent 133 C␣ carbon atoms of HmuO versus HemO is 1.55 Å (27). This difference is most likely a result of the low degree of sequence identity between HmuO and HemO.
Flexibility of the Proximal and Distal Helices-The flexibilities of the distal helix are quite different between the HmuO and human HO-1 crystal structures. Fig. 3A shows a superimposition of the proximal and distal helices of three molecules in the crystal asymmetric unit of HmuO. The main chain folds of the proximal and distal helices of the three molecules are very similar, indicating that they adopt generally only one conformation. This is different from the human HO-1 crystal structure, in which two conspicuously different conformations of the distal helix are observed (Fig. 3B). One of them is the "open conformation," in which the proximal and distal helices are loosely packed against the heme. The other is the "closed conformation" in which the distal helix is closer to the heme and its water ligand, resulting in a more tightly packed heme pocket (22,23). The single conformation assumed by HmuO is similar to the closed conformation of human HO-1. This single conformation is also observed in the structures of two HmuO molecules in the C2 crystals that we used for the phase determination. Analysis of the crystal packing does not show obvious strong packing interactions in the HmuO crystals that might cause low flexibility of the distal helix. The observed structural difference between HmuO and human HO-1 appears to be sequence-dependent and not due to the crystal packing effect.
Non-conserved residues in the C-terminal regions of the proximal and distal helices account for helix flexibility differences between HmuO and human HO-1. In the case of HmuO, the motion of the proximal and distal helices must be suppressed by tight hydrophobic interactions between two aromatic residues in the distal helix, His-150 and Tyr-151, and the backbone of Leu-33 and Gly-35 in the proximal helix (Fig. 3A). However, these hydrophobic interactions are absent in human HO-1, because the two distal aromatic residues, His-150 and Tyr-151, found in HmuO have been replaced by other relatively small hydrophobic residues, Ala-154 and Leu-155, in human HO-1 (Fig. 3B). The lack of the tight hydrophobic interaction makes the two helices in human HO-1 highly flexible, resulting in two distinct conformations in the crystalline state. The less flexible proximal and distal helices in HmuO are consistent not only with the NMR results (21) but also with the results that the final product biliverdin binds more tightly with HmuO than HO-1 (18,48).
Heme Pocket of the Ferric State-Structure around the heme pocket in the ferric heme-HmuO complex is shown in Fig. 4A. The heme iron is six-coordinate with the imizadole group of His-20 and a water molecule as the axial ligands, consistent with the previously reported spectroscopic and mutagenesis results (19,20,33). The iron-proximal His and Fe-water oxygen distances are comparable with those found in heme proteins with the same coordination structure (Table II), and there is very little displacement of the iron from the mean porphyrin plane (ϳ0.05 Å toward the proximal His).
The electron density for the vinyl and methyl side chains of the heme group is well defined, and a single heme orientation is determined unambiguously as illustrated in Fig. 4B. NMR studies on the cyanide-bound HmuO (21) show that the heme group is orientationally disordered about its ␣-␥-meso axis, as observed in many heme proteins including HO-1 (49). The heme orientation found by NMR is a 3:1 mixture of two rotational isomers at equilibrium with the dominant isomer assuming the same orientation found in the crystal structure (21). Thus, the orientation found in the crystal structure appears to be an energetically stable state of HmuO.
The heme orientation of HmuO is flipped 180°about the ␣-␥ axis with respect to that observed in the crystal structures of human HO-1 and HemO (22,27). This exposes the heme ␤-edge to the solvent and changes the positions of the heme methyl and vinyl side chains from those observed in the structures of HO-1. It is difficult to explain why the favorable orientation of the heme in HmuO is different from those in other HO, because obvious structural differences are not observed in the heme contact regions, especially in the region involving the methyl and vinyl groups of the heme. However, the difference in the flexibility of the proximal and distal helices between HmuO and its mammalian counterparts might be one possible reason for the different heme-binding properties.
The imidazole plane of the proximal ligand, His-20, closely eclipses the heme meso ␤-␦ axis (Fig. 4B), and this is observed in HO-1 and HemO as well. However, it is different from the "eclipsed conformation" found in myoglobin, hemoglobin, and some synthetic iron porphyrin complexes, all of which have an axial imidazole plane that closely eclipses the equatorial Fe-N(pyrrole) bonds (50). Under the "staggered conformation," which describes the imidazole plane orientation found in HmuO, steric repulsion between the imidazole C 2 and C 4 carbon atoms and porphyrin pyrrole nitrogen atoms is minimized, and the orbital-orbital overlap of iron and imidazole bonding interaction is negligible (50). The proximal His imidazole orientation found in the crystal structure substantiates our NMR results reported previously (19) in which the ferric heme-HmuO complex exhibited a smaller spread of the heme methyl hyperfine shifted proton resonance than met myoglobin. The HmuO distal helix is kinked (Fig. 4A) directly over the heme, due to the flexibility introduced by three Gly residues in the conserved sequence of 135 GDLSGG 140 , as seen in HO-1 (22). Because of the kinked helix, a portion of the distal helix from Arg-132 to Val-142 covers the Southern half of the heme plane. The nitrogen of Gly-139 is located about 4 Å from the water ligand, and the carbonyl oxygen of Gly-135 is the only amino acid located within a feasible hydrogen bonding distance to the heme ligand water (2.91-3.03 Å) (Fig. 4C).
Calculation of the F o Ϫ F c and 2F o Ϫ F c maps during structural refinement reveals the presence of an extra electron density on the distal helix just above the heme plane for molecule B. Two alternative conformations of both side and main chains of Ser-138 and Gly-139 (Fig. 4C) might account for this additional electron density. When the occupancy of each conformation is set to 0.5, crystallographic B factors for corresponding atoms in distinct conformations are refined to almost the same values, implying that these two conformations have essentially equal occupancy. Observation of two clearly defined alternative conformations indicates this region, including the main chain of the protein, is flexible.
Two hydrophobic residues, Met-29 and Phe-208, are located next to the heme ␣-meso edge, and together they create a hydrophobic wall. An internal space just above the ␣-meso edge and pyrrole A is filled with a number of water molecules, of which the structures are stabilized by hydrogen-bonding interactions with surrounding hydrophilic residues, Gln-46, Arg-132, Asp-136, Tyr-161, and Asn-204 (Fig. 4D). Among them, Gln-46 and Tyr-161 are not conserved in human HO-1 but converted to Met-51 and Phe-167, respectively, and therefore the regions adjoining the heme pocket in HmuO is more hydrophilic than that in human HO-1.
The imidazole group of the proximal His in HmuO forms hydrogen bonding interaction with a Glu residue nearby (Fig.  5). The O⑀ atom of Glu-24 is located at a hydrogen bonding distance of 2.83 Å to the N␦ atom of His-20 in molecule A. In molecules B and C, they are ϳ4 Å apart, and the direct hydrogen bonding between the proximal His and Glu-24 is rendered unlikely. Instead, in molecule B, they are interacting through a bridging water molecule that is located at 2.84 Å from N␦ of His-20 and at 2.60 Å from O⑀ of Glu-24. In molecule C, a water molecule is located at 2.93 Å from N␦ of His-20, but O⑀ of Glu-21 is closer to the water oxygen atom (2.75 Å) than O⑀ of Glu-24 (3.90 Å).
Three basic residues, Lys-13, Lys-173, and Arg-177, are located in proximity to the heme propionates (Fig. 5). These residues are conserved in mammalian HO-1 as Lys-18, Lys-179, and Arg-183, and their charge interactions with heme propionates have been reported to be significant for proper heme positioning in HO (22,23). Close examination reveals considerable difference in the propionate interaction between HmuO and HO-1. In the HmuO structure, Lys-173 is located 5-7 Å away from the heme propionates, thus unlikely to form hydrogen bonding or charge interaction with the heme propionate. Lys-13 and the heme propionate-6 are 2.89 to 4.25 Å apart. Propionate-6 is solvent-exposed and takes multiple orientations in molecule A. Thus, the interaction between Lys-13 and the heme propionate-6 does not seem to play the dominant role in heme positioning either. The guanidyl nitrogen of Arg-177 is located sufficiently close (ϳ2.7 Å) to form charge/hydrogen bonding interactions with the heme propionate-7. The N⑀ of Arg-177 also forms hydrogen bonding interaction with the  propionate-7 through a bridging water molecule. In HmuO, the interaction involved by Arg-177 appears to be the dominant interaction between the heme propionate and the amino acid residues. The size of the heme pocket of HmuO calculated by the probe-accessible method is 20 -30 Å 3 , which is much smaller than that of HO-1 in the closed conformation (ϳ50 Å 3 ). This is due primarily to the lack of a small hydrophobic inner chamber, an extension of the heme pocket, originally noted in the human HO-1 crystal structure (22). The hydrophobic inner chamber is absent in HemO (27).
Long Range Hydrogen Bonding Network from the Surface to the Active Site-The water molecules found in the distal heme pocket extend their hydrogen bonding network all the way to the surface of the HmuO protein. This hydrogen bonding network is stable, as indicated by the low crystallographic B factors, and connects the heme-ligated water to Asp-86 through Gln-46 (Fig. 6A). Asp-136 O␦ anchors water molecule W3 by hydrogen bonding. Asp-86 is located at the surface of the protein molecule and is accessible to the solvent (Fig. 6B). This long range hydrogen bonding network is absent in the crystal structure of human HO-1. This is because two water molecules, W5 and W6, are not present in the crystal structure of human HO-1 due to the replacement of Tyr-161 and Gln-46 by hydrophobic Phe-167 and Met-51 in human HO-1 (Fig. 4D).
Structure of the Ferrous Form-Reduction of the ferric heme iron into the ferrous one induces local structural change in the regions of the heme-binding site, as depicted in Fig. 7A. The heme iron is five-coordinate without the water ligand in the ferrous state (18). It is displaced from the mean porphyrin plane toward the proximal His by ϳ0.25 Å. The iron-proximal His bond distance changes from 2.0 Å in the ferric state to ϳ2.2 Å in the ferrous form (Table II).
In the distal helix, a significant movement of the backbone structure is noticed, especially in the region from Leu-137 to Val-142, where the C␣ atoms moves a maximum of ϳ1.9 Å. The structural changes occur through removal of the heme-bound water molecule, followed by the conversion of the six-coordinated ferric heme iron to a five-coordinated ferrous one. The resulting empty space allows the distal helix to move closer to the heme plane. This conformational change in the distal helix is stabilized by two newly formed hydrogen bonds, one of which is a hydrogen bond between N⑀ of Gln-141 and the carbonyl oxygen of Tyr-161. As shown in Fig. 7, B and C, the side chain of Gln-141 of the ferrous heme-HmuO complex swings over to place its N⑀ atom within a hydrogen bonding distance (ϳ2.9 Å) of the main chain oxygen atom of Tyr-161. The other hydrogen bond is between the main chain oxygen atom of Ser-138 and main chain nitrogen atom of Gln-141. The distances between these two main chain atoms are 3.0 Å with geometries suitable for hydrogen bonding formation. These hydrogen bonds make the distal helix more rigid in the ferrous form. The multiple conformation of Ser-138 and Gly-139 seen in the ferric form shown in Fig. 4C is absent in the ferrous form. Although the averaged B factors of the C␣ atoms on the whole protein molecules are similar between the ferric and ferrous forms (ϳ16 Å 2 ), that of the C␣ atoms on the distal helix (Gly-122 to His-150) in the ferrous state is ϳ11 Å 2 , lower than the 15 Å 2 of the ferric state. This indicates that the distal helix in the ferrous state is less mobile than in the ferric state.
The proximal helix also moves upon conversion of the ferric heme-HmuO complex to the ferrous state, as shown in Fig. 7A. The aforementioned displacement of the iron atom from the heme plane and the elongated Fe-His bond in the ferrous form pushes the proximal His away from the heme by 0.4 Å with a slight tilt of the imidazole ring. In the ferrous structure, the Glu-21 side chain is rotated toward the imidazole group of His-25 to form hydrogen bonding interaction between them. This interaction shortens the distance between these two residues, resulting in the movement of this portion of the proximal helix toward the heme plane. The distance between the C␣ atom of Glu-24, located in the proximal helix, and that of Glu-139, located in the distal helix, is shortened from 11.1 Å in the ferric form to 8.4 Å in the ferrous form. Because the proximal and distal helices are more closely packed against the heme group, this results in a highly tight heme pocket in the ferrous heme-HmuO complex.
Despite these structural changes, the proximal His still remains in the "staggered" conformation in the ferrous form with a slight tilt due probably to the movement of the proximal helix. In the ferrous structures, the Glu-24 and Glu-21 side chains adopt conformations different from those in the ferric form. The N␦ atom of the proximal His is no longer able to form hydrogen bonding interaction with these Glu residues either directly or through a bridging water molecule in the ferrous form. The proximal axial ligand of the heme-HmuO complex in the ferrous state has a neutral imidazole group, consistent with our resonance Raman results (19).
The interactions between the porphyrin propionates and the three basic residues observed in the ferric structure (Fig. 5) are conserved in the ferrous form. Therefore, the dominant propionates interaction is between the heme propionate-7 and Arg-177 in the ferrous heme-HmuO complex as well. and the distal helices. The former provides the axial His heme ligand, and the latter is kinked over the heme. The distal heme pocket lacks an ionizable side chain that could serve as an acid base catalyst; instead a Gly residue is found immediately next to the bound water ligand. These features are well preserved in the HmuO crystal structures and thus appear to be critical for HO catalysis in general.
The presence of ordered water molecules in the distal heme pocket of mammalian HO-1 and Gram-negative HemO is evident in their respective crystal structures (21,27). However, a long range hydrogen bonding network has not been identified in either of these HO crystal structures. Our HmuO crystal structure results provide the first detection of a hydrogen bonding network that connects from the solvent water to the oxygen-activating heme site. This finding addresses how protons are delivered to enable the first oxygenation step in HO catalysis.
EPR/ENDOR studies on HmuO 2 and HO-1 (30) have revealed that efficient proton transfer to the active site is essential for formation of the ferric hydroperoxo species, for activation of the O-O bond for cleavage, and for heme hydroxylation by direct attack of the remote hydroperoxo oxygen on the ␣-meso carbon. Without this proton transfer, inactive ferryl species is generated instead of ␣-meso-hydroxyheme, the first intermediate of HO catalysis. Resonance Raman studies of the HmuO ferrous oxy form, 3 the immediate precursor of the active ferric hydroperoxo species, show a highly bent Fe-O-O bond similar to that in rat HO-1 (ϳ110°) (34). EPR studies on the oxy cobalt porphyrin-HmuO complex show that the bound oxygen forms a hydrogen bond with the distal pocket water (19). Modeling of the oxy complex by placing O 2 on the HmuO structure with terminal oxygen pointing toward the ␣-meso carbon atom places the terminal oxygen within hydrogen bonding distance with water W3 (Fig. 6A). The hydrogen bonding proton becomes the hydroperoxo proton upon one electron reduction of the oxy form, as we have proposed for HO-1 (30,51). W3 is anchored by the hydrogen bonding with Asp-136 with geometry and distance suitable for efficient proton transfer. Preliminary mutagenesis studies show that disruption of the hydrogen bonding network by replacing Asp-136 by Phe results in the formation of the inactive ferryl species during HmuO catalysis. 4 The hydrogen bonding network with ordered water molecules in the distal pocket probably functions as a ''channel'' to deliver protons required for the formation of the ferric hydroperoxo active species and the hydroxylation of the ␣-meso carbon in HO catalysis. In HO-1, the ordered distal pocket water molecules are present with W3 anchored to the catalytically important Asp-140, which is equivalent to Asp-136 of HmuO (9). However, the hydrogen bonding network identified in HmuO is absent beyond W3 due to the presence of hydrophobic Met-51 and Phe-167 in lieu of Gln-46 and Tyr-161 in HmuO (see Fig. 4D). Recent NMR studies on human HO-1 report the possible presence of an extensive hydrogen bonding network from a noncoordinate distal pocket water to several surface amino acid residues (31), including Asp-92 which corresponds to HmuO Asp-86, the surface amino acid in the HmuO proton delivery system mentioned above. Differences in the organization of the hydrogen bonding network between human HO-1 and HmuO indicate that HO-1 and HmuO use different routes to channel protons required for their catalytic function.
The ferric heme complex of HmuO shows an acid-base transition with a pK a of ϳ9 (19). Acid-base transitions observed in ferric heme proteins with a water ligand are considered to be linked to the ionization of a distal amino acid residue that forms a hydrogen bond with the bound water ligand. The deprotonation of the distal residue, His in most cases, causes ionization of the iron-bound water and results in a predominantly low spin hydroxide form. In the ferric form of the HmuO, the sixth ligand water is not hydrogen-bonded with a side chain of a dissociable residue, such as a distal His found in globins. The water ligand forms a hydrogen bonding interaction with a distal pocket water molecule (Fig. 6A, W2), which is linked to the long range hydrogen bonding network to Asp-86. Deprotonation of the water ligand would take place through the hydrogen bond network. The pK a of the acid-base transition of HmuO (ϳ9.0) is higher than that of HO-1 (ϳ7.8) (10). This difference might be due primarily to the aforementioned difference in the hydrogen bonding network between HO-1 and HmuO.
The HmuO crystal structures show that interaction between Arg-177 and the propionate-7 is the dominant interaction between the heme side chains and the HmuO protein. This is different from HO-1 where three basic residues (Lys-18, Lys-179, and Arg-183) are reported as the key residues for correctly orienting the heme group in the heme pocket through their interaction with the heme propionate groups (22,23). 5 Lys-18 of HO-1 was also implicated to be a key residue in incorporating heme into rat HO-1 by interacting with the heme propionate (53). However, Lys-13 in HmuO, which is equivalent to HO-1 Lys-18, is less important in HmuO incorporation of heme, because its interaction with the heme side chain is either missing or weak in the HmuO crystal structures. Different from HO-1, HmuO might be able to incorporate heme group without the interaction involved with the Lys-13, because the heme-binding site is already formed in the HmuO protein prior to the heme binding due to the more rigid distal helix. The crystal or NMR structure of the apoHmuO protein would provide further insight into the heme incorporation mechanism.
The ferric HmuO crystal structure shows that the steric bulk of the kinked distal helix, which is located within 4 Å above the heme plane, restricts the access of the heme ligand to the ␤-, ␥-, and ␦-meso carbons, leaving only ␣-meso carbon accessible by the bound ligand in HmuO. Indeed, NMR studies on the cyanide-bound HmuO show that the Fe-CN bond is tilted toward the ␣-meso position (21). The 1.7 Å resolution crystal structure of the azide-bound ferric HmuO 6 also shows that azide binds to the iron in the bent/end-on geometry pointing toward the ␣-meso position of the porphyrin ring. Because the distal helix sterically dictates the geometry of the ligand bound to the heme iron in HmuO, the hydroperoxo is forced to point toward the ␣-meso position. An acute Fe-O-O bond angle has been predicted based on resonance Raman studies of oxy HmuO, 3 the immediate precursor of the ferric hydroperoxo species. The highly bent Fe-O-O geometry brings the terminal oxygen of the hydroperoxo ligand in close contact with the ␣-meso carbon atom, thereby facilitating the site-selective hydroxylation. This configuration is consistent with the 1.6 Å resolution crystal structure of the NO-bound ferrous HmuO, 6 which shows that NO binds to the heme iron with an Fe-N-O angle of about 125°w ith the terminal NO oxygen atom located as close as 3.4 Å from the porphyrin ␣-meso carbon. Similar results were also reported recently (24) for the NO bound human ferrous HO-1. These further reinforce the notion that the ␣-regioselectivity in HmuO and HO-1 catalysis is primarily under steric control, as we had suggested previously (30).
Ferrous HmuO exhibits an iron-proximal His stretching frequency at 221 cm Ϫ1 , indicating that the proximal ligand is a neutral imidazole in the ferrous form (19). This is consistent with the ferrous crystal structures that show a lack of proximal His hydrogen bonding interaction. In the ferric form of HmuO, N␦ of the proximal His forms hydrogen bonding interaction with O⑀ of a Glu residue either directly or through bridging water. The proximal His in peroxidase enzymes, such as horseradish peroxidase and cytochrome c peroxidase, forms a strong hydrogen bonding interaction between proximal His N␦ and O␦ of a nearby Asp residue, resulting in an axial ligand with strong imidazolate character (14). The imidazolate character has been reported to produce a strong iron-proximal His interaction, extruding from the porphyrin plane to favor a fivecoordinate ferric high spin species in the peroxidase enzymes (54,55). The heme iron in ferric HmuO is six-coordinate with water ligand, and the heme iron shows very little displacement from the mean porphyrin plane. The HmuO proximal His hydrogen bonding interaction must be much weaker than that in the peroxidase enzymes. The weak hydrogen bonding interaction generates a primarily neutral imidazole axial ligand that is also supported by the presence of a hyperfine-shifted proton resonance of the proximal imidazole NH at 99 ppm for the ferric heme-HmuO complex (19). The imidazolate axial ligand in peroxidase has been postulated to stabilize the high valent oxo active intermediate by facilitating the O-O bond cleavage (14). An imidazolate proximal His would promote formation of a ferryl species, which is inactive in HO catalysis (12). The neutral imidazole axial ligand in HmuO helps prevent hydroperoxo O-O bond cleavage prior to hydroxylation of the ␣-meso carbon of the porphyrin ring. The proximal His in ferric human HO-1 also forms hydrogen bonding interaction similar to that in ferric HmuO (22). In the light of similar spectroscopic and catalytic properties between HmuO and HO-1 (17,18), the HO-1 proximal His hydrogen bonding must also be weak, and a neutral rather than anionic imidazole is the proximal heme ligand.
Weak proximal His hydrogen bonding interactions are present in myoglobin where the proximal His is a neutral imidazole, and the N␦H of which forms hydrogen bonds with the main chain carbonyl oxygen of Leu-89 and O␥ of Ser-92. This hydrogen bonding interaction stabilizes the proximal His orientation (56). The proximal His hydrogen bonding in the ferric heme-HmuO complex might help to stabilize the proximal His orientation into the staggered conformation, which has functional relevance in the ferric form. Due primarily to the steric constraint imposed by the distal helix, the hydroperoxo ligand binds to the iron in the bent/end-on geometry pointing toward the ␣-meso position of the porphyrin ring, as discussed above. In this configuration, where the projection of the hydroperoxo ligand onto the heme plane is nearly perpendicular to the imidazole plane of the proximal His, the -orbitals of the proximal imidazole and the iron-bound hydroperoxo species are orthogonal to the iron d xz and d yz orbitals. This coordination structure decreases the push effect of the proximal His imidazole (14) and stabilizes the O-O bond of the iron bound hydroperoxo, thereby preventing the formation of an inactive ferryl intermediate (32).
The crystal structures of HmuO show that the catalytically critical hydroperoxo intermediate is realized due to the well designed protein architecture. The staggered proximal His orientation and the extended distal pocket hydrogen bonding network with the proton delivery system work in a concerted manner to stabilize the active hydroperoxo species by preventing the formation of the inactive ferryl species. These features together with the kinked distal helix located very close to the heme are significant for regioselective oxidation of the ␣-meso carbon of the porphyrin ring.