Comparison of Apo- and Heme-bound Crystal Structures of a Truncated Human Heme Oxygenase-2*

Heme oxygenase (HO) catalyzes the first step in the heme degradation pathway. The crystal structures of apo- and heme-bound truncated human HO-2 reveal a primarily α-helical architecture similar to that of human HO-1 and other known HOs. Proper orientation of heme in HO-2 is required for the regioselective oxidation of the α-mesocarbon. This is accomplished by interactions within the heme binding pocket, which is made up of two helices. The iron coordinating residue, His45, resides on the proximal helix. The distal helix contains highly conserved glycine residues that allow the helix to flex and interact with the bound heme. Tyr154, Lys199, and Arg203 orient the heme through direct interactions with the heme propionates. The rearrangements of side chains in heme-bound HO-2 compared with apoHO-2 further elucidate HO-2 heme interactions.

a critical role in heme and iron homeostasis, but the products of heme degradation have important physiological functions (1,2). Iron is the essential catalytic center for heme and many nonheme metalloenzymes and has crucial effects on cytoprotection (3,4). Although CO is toxic at levels above ϳ500 ppm, it also is a signaling molecule (5,6). CO has anti-apoptotic, antiinflammation, and anti-proliferation properties (7)(8)(9). Bilirubin is an antioxidant that is involved in resistance to oxidative stress (10,11).
Two HO isoforms have been reported in mammalian cells, heme oxygenase-1 (HO-1) and heme oxygenase-2 (HO-2) (12). The mammalian HO-1 and HO-2 exhibit similar catalytic activities and share 55% identity and 76% similarity in humans (13,14). Another isoform, HO-3, was recently reported in rat; however, no activity was detected for this isoform, and it is considered a pseudogene (15). Plants contain several HOs, including one named HO-2, but these lack the C-terminal membranespanning region seen in mammalian HOs and are more closely related to the bacterial and insect HO-1 than to the mammalian HO-2 (13,16).
Although they are both membrane-bound proteins and their catalytic activities are similar, HO-1 and HO-2 are expressed in discrete tissues and exhibit very different patterns of expression (17). Although human HO-1 is an inducible protein whose expression is up-regulated by heat shock and oxidative stress conditions, HO-2 is constitutively expressed (17)(18)(19)(20)(21)(22). HO-1 has been found in most tissues, with particularly high expression levels in liver and spleen; however, HO-2 is mainly found in the brain, testes, and carotid body (23)(24)(25)(26).
HO-1 and HO-2 exhibit two regions of sequence divergence: one is around residue 127 and the other is near the C terminus between residues 240 and 295 (HO-2 numbering throughout unless otherwise stated). In both HO-1 and HO-2, the C termini appear to play regulatory roles, albeit of quite different natures. Cleavage of the 52 amino acids at the C terminus of HO-1 affects nuclear import, leading to the activation of genes associated with protection against oxidative stress (27). In HO-2, the C-terminal region between residues 255 and 287 contains two heme regulatory motifs (HRMs), which form a thiol/disulfide redox switch that regulates heme binding affinity (28).
Because full-length HO-1 and HO-2 are relatively unstable, most structure-function studies have used truncated proteins in which the membrane region has been removed. The truncated mammalian HO-2 can be expressed at high levels in Escherichia coli and, as was first shown for HO-1, has a specific activity that is identical to that of the full-length protein (29).
The first three-dimensional structure of human HO-1 was determined for the truncated protein in 1999 (30). Subsequently, many HO-1 crystal structures have been reported, including rat HO-1, biliverdin-associated HO-1, ferrous/ferrous-NO-bound HO-1, and some HO-1 variants (31)(32)(33). These structures led to significant advances in the study of the HO-1 catalytic mechanism (34). The HO-2 used in this study is the soluble portion of HO-2 with a mutation of Cys 127 to an alanine and is missing residues 264 -316 and will be referred to as HO-2 in this report. Here we report the first apo-and hemebound HO-2 crystal structures at 2.4-and 2.6-Å resolution, respectively.

Cloning, Overexpression, and Purification of Human HO-2-
The full-length human HO-2 cDNA in a pGEX-4T-2 (Amersham Biosciences GE Healthcare) GST gene fusion vector was kindly provided by Dr. Mahin D. Maines (University of Rochester, School of Medicine). The truncated HO-2 lacking the three HRMs and the membrane binding region was generated by replacing Cys-127 with Ala-127 and putting a stop codon "TAA" right before Cys-265, which will be referred to as HO-2 in this report. The QuikChange site-directed mutagenesis protocol from Stratagene (La Jolla, CA) was used for variant generation. A glutathione affinity column was used to help purify HO-2, because the human HO-2 fragment was fused to GST in this construct.
HO-2 purification was performed as previously described with slight modification (28). The newly generated construct was transformed into E. coli strain BL21(DE3) for protein overexpression. Cells were grown in LB medium containing 100 g⅐ml Ϫ1 ampicillin at 37°C and induced with 0.3 mM isopropylbeta-D-thiogalactopyranoside (Gold Biotechnology, Inc., St. Louis, MO). 5-Aminolevulinic acid hydrochloride (Sigma-Aldrich) was added to help the induction of HO-2. After cell harvest, the pellet was resuspended in 5 volumes of lysis buffer (50 mM KCl, 50 mM Tris-HCl, pH 7.5), containing 0.04% (v/v) Triton X-100 (Sigma-Aldrich), protease inhibitor (1 tablet/50 ml of extraction solution, Roche Applied Science, Indianapolis, IN), lysozyme (0.5 mg⅐ml Ϫ1 , Sigma-Aldrich), and DNase I (5 units⅐ml Ϫ1 , Sigma-Aldrich) at 4°C. Cells were lysed by sonication, and the supernatant was loaded onto a glutathione-Sepharose 4B column (Amersham Biosciences GE Healthcare). After washing the resin with a buffer containing 50 mM KCl, 50 mM Tris-HCl, pH 7.5, the HO-2-GST fusion protein was eluted with the same buffer containing 10 mM reduced glutathione. Then the glutathione in the fractions was removed through dialysis, and the GST domain was cleaved by limited proteolysis using thrombin. The HO-2 was separated from the GST tag with another glutathione-Sepharose 4B column. The protein was then concentrated to 5 mg⅐ml Ϫ1 for crystallization. To make the heme-bound HO-2, aliquots of a freshly prepared heme solution (0.1 M NaOH solution containing 10% Me 2 SO) were slowly diluted into a purified apoHO-2 protein solution to give a 2:1 molar ratio of heme to protein. Less than 5% (volume ratio) of the heme solution was added to prevent any noticeable change in pH of the final heme-protein solution. All purification steps were carried out at 4°C.
X-ray Data Collection and Structure Solution-X-ray diffraction data for the apoHO-2 structure was collected at a wavelength of 0.97933 Å at the General Medicine and Cancer Institutes Collaborative Access Team (GM/CA-CAT) beamline 23-ID-D located at the Advanced Photon Source at Argonne National Laboratory. The diffraction data for the heme HO-2 structure was also collected at the GM/CA-CAT beamline 23-ID-B utilizing a wavelength of 0.97946 Å. Both data sets were indexed and scaled using HKL2000 (35). Molecular replacement was carried out with MOLREP using human heme oxygenase-1 (Protein Data Bank code 1N45) as the initial model (36). An initial apoHO-2 model was partially built with ARP/ wARP resulting in a structure with 54.7% of the possible residues positioned (37). The apoHO-2 structure was completed with several rounds of manual building with Coot and restrained refinement in REFMAC (38,39). The refined apoHO-2 structure was used as the molecular replacement model for the heme-bound HO-2 structure. The heme-bound HO-2 model was then completed as described for apoHO-2. The refinement process for apoHO-2 was monitored using an R free value based on 5.1% of all the independent reflections. The same set of reflections was used to calculate R free in the hemebound HO-2 structure.
The quality of the final structures were assessed using MOLPROBITY and PROCHECK (40,41). The final model of apoHO-2 was refined to a resolution of 2.40 Å with an R-factor

Crystal Structure of Apo-and Heme-bound HO-2
of 0.205 and an R free of 0.253. Heme-bound HO-2 was refined to a resolution of 2.61 Å with an R-factor of 0.195 and an R free of 0.256. The 16-conformer heme-bound HO-2 structure had an R-factor of 0.156 and an R free of 0.244. All pertinent information on data collection, refinement, and model statistics of apo-and heme-bound HO-2 are summarized in Table 1. All figures were generated using PyMOL. 3

HO-2 Overall
Structure-The HO-2 that was used in this study lacked the two C-terminal HRM domains (265-316), and had Cys-127 mutated to Ala. These alterations to HO-2 were done to increase the likelihood of crystallization. Two molecules of HO-2 were observed in the asymmetric unit of both apoHO-2 and heme-bound HO-2. Out of the 264 residues in the truncated apoprotein, 212 residues were observed in molecule A, and 217 residues were observed in molecule B. Molecule A of apoHO-2 was lacking electron density in both the N-terminal (Met 1 -Met 28 ) and the C-terminal (Ser 242 -Lys 264 ) regions. The B molecule of apoHO-2 was lacking several N terminus (Met 1 -Met 30 ) and C terminus (Leu 249 -Lys 264 ) residues. There were two partially resolved molecules of Triton X-100 observed near the heme-binding site in both apoHO-2 molecules. Similarly, in the heme-bound HO-2 structure, 213 residues were observed in molecule A and 218 residues were observed in molecule B. The A molecule of heme-bound HO-2 was lacking electron density for Met 1 -Met 28 and Thr 243 -Lys 264 . The B molecule of heme-bound HO-2 was missing density for Met 1 -Arg 29 and Leu 249 -Lys 264 . In all previous HO-1 structures, the C-terminal residues also have not been observed.
In the B molecule of heme-bound HO-2, the heme was modeled at 75% occupancy. Elongated positive electron density was observed near the heme in both molecules of the heme-bound HO-2 structure. Compared with the A molecule the heme pocket of the B molecule was more disordered. The four pyrrole rings of the heme in the B molecule were restrained during refinement to form a plane. There was no need to restrain the heme in molecule A beyond the default REFMAC parameters. In addition to REFMAC restrained refinement the hemebound HO-2 was also modeled using ensemble refinement using CNS (Protein Data Bank code 2RGZ) (43,44). The disorder of the heme in the heme pocket of the B molecule can be seen in the 16-conformer ensemble refinement (supplemental Fig. 1).
Distal Helix-A flexible, highly conserved sequence (Gly 159 -Asp-Leu-Ser-Gly-Gly 164 ), which is an integral part of the heme binding pocket, allows the distal helix to bend and come into close proximity to the heme (30,32,45,46,(51)(52)(53)(54). This glycine-rich flexible region in heme-bound HO-2 allows the backbone atoms of Gly 159 to interact with the heme iron through a

Crystal Structure of Apo-and Heme-bound HO-2
water molecule. Gly 159 in apoHO-2 is in the same position as observed in heme-bound HO-2. The catalytically important Asp 160 that resides on the distal helix is relatively unchanged in both apo-and hemebound HO-2 (55,56). The most striking differences between the distal helix in apoHO-2 and hemebound HO-2 are found in the amino acids near the bound heme. In heme-bound HO-2, Tyr 154 repositions to form a hydrogen bond with the propionate located in the heme pocket. The side chains of Thr 155 and Arg 156 in heme-bound HO-2 move away from the heme allowing more room in the heme pocket (Fig.  3). The expansion of the heme pocket caused by the movement of Thr 155 and Arg 156 is not observed in apoHO-2.
In the previously solved hemebound human HO-1 structure, the two HO-1 monomers in the asymmetric unit were observed in slightly different conformations (30). Molecule A of the heme-bound HO-1 was considered to be in a closed conformation while molecule B was in an open conformation, which was proposed to be important in binding heme and/or releasing products (30). The difference between the open and closed conformation is the position of the distal helix with respect to the heme pocket. In the open conformation, the distal helix has moved away from the heme pocket causing a slightly larger heme pocket compared with the closed conformation. In the HO-2 structures the movement of the distal helix occurs at Gly 163 . Gly 163 moves away form the heme causing the heme pocket to adopt the open conformation. By comparing the overall r.m.s.d. of the heme-bound HO-1 and HO-2 ␣-carbons it is apparent that both apoHO-2 and heme-bound HO-2 are in the open conformation ( Table 2). The movement of the distal helix at Gly 163 is seen when the apo-and heme-bound HO-2 structure is compared with the closed heme-bound HO-1 structure (supplemental Fig. 2). All other significant backbone atom deviations between HO-2 and HO-1 are located in the loops between helices.
Proximal Helix (His Ligand Side)-Unlike the distal helix, the proximal helix lacks any flexible region that would allow the helix to translocate during heme binding. The heme iron-coordinating histidine, His 45 , resides on the proximal helix (Fig. 4). His 45 of heme-bound HO-2 coordinates the heme iron through an imidazole nitrogen at a distance of 2.1 Å in the A molecule. The distance between the imidazole nitrogen of His 45 and the heme iron was restrained during refinement to a distance of 2.1 Å. A hydrogen bonding interaction with Glu 49 serves to stabilize the iron-coordinating histidine. A carboxyl oxygen from Glu 49 forms a hydrogen bond with the imidazole nitrogen of

Crystal Structure of Apo-and Heme-bound HO-2
His 45 that is not coordinating the heme iron at a distance of 3.5 Å in molecule A and 3.8 Å in molecule B. The interaction between His 45 and Glu 49 is not present in the apoHO-2 structure. Without a bound heme, the imidazole ring of His 45 of apoHO-2 is no longer restrained and adopts a different conformation than the one that is observed in heme-bound HO-2 (Fig.  4). There is little change in the proximal helix with the exception of the movements of His 45 and Glu 49 .
Hydrophobic Region-The residues that make up the hydrophobic pocket, Val 54 , Phe 53 , Phe 57 , and Phe 234 remain unchanged in both apo-and heme-bound HO-2. In the hemebound HO-2, the heme ␣-meso edge is positioned so that it is directed toward the residues that make up the hydrophobic region of the heme pocket. Unexplained positive electron density was observed between the hydrophobic region and the ␣-meso edge of the heme in the heme-bound HO-2 structure. We were unable to identify the compound giving rise to the density, and the density was left un-modeled (Fig. 5A). In the same region that the unexplained density was found in heme-bound HO-2, a molecule of Triton X-100 was observed in apoHO-2. The Triton X-100 molecule occupied a position near the hydrophobic region of the heme pocket where the heme ␣-meso edge would occupy (Fig.  5B). The tetramethylbutyl group of the Triton X-100 was in the same position as the unexplained density in heme-bound HO-2.
Heme Protein Interactions-In HO-2 and other known HOs, the ␣-, ␤-, and ␥-meso edges of the heme group are positioned in the interior of the heme pocket with the ␦-meso edge exposed to the solvent. In order for the hydroperoxo form of HO-2 to specifically catalyze electrophilic attack on the heme ␣-mesocarbon, the heme group needs to be properly oriented in relation to the heme pocket. The way that HO-2 seems to orientate the bound heme is through interaction with the propionate groups. One of the heme propionate groups is buried in the heme pocket, whereas the other is solvent-exposed. In the heme-bound HO-2 structure, there are three residues that play crucial roles in heme coordination (Tyr 154 , Lys 199 , and Arg 203 ). Tyr 154 is located on the distal helix with its OH group 2.7 Å in molecule A and 2.9 Å in molecule B from one of the carboxyl oxygens of the buried propionate. Located on the A6 helix, Arg 203 coordinates the other carboxyl oxygen of the buried propionate at a distance of 2.4 Å in the A molecule and 2.6 Å in the B molecules. The solventexposed propionate is coordinated to Lys 199 and could transiently hydrogen bond with Lys 38 . Like Arg 203 , Lys 199 is located on the A6 helix and interacts with the solvent-exposed propionate at a distance of 3.0 Å in both the A and B molecules. While Lys 38 is too distant to form a hydrogen bond with the solventexposed propionate in the crystal structure, a shift in the position of the propionate could cause the formation of a hydrogen bond. These hydrogen bonds correctly orient the heme and place the ␣-meso edge carbon in position for hydroxylation (Fig. 4).

DISCUSSION
The two structures we present are of the core domain of human HO-2. These truncated human HO-2 structures contain the region responsible for heme recognition, binding, and degradation, elucidating the important structural features of HO-2 and improving our understanding of the catalytic mechanism of HO-2. The structures indicate that the catalytic cores of HO-1 and HO-2 are structurally conserved, and the major differences underlying the diverse physiological functions of HO-1 and HO-2 may lie in the C terminus of the two proteins.
Hydrophobic Cavity-The conformation of the HO-2 heme pocket can be directly affected by the hydrophobic cavity. One possible reason that both HO-2 structures were observed in an open conformation could be due to the Triton X-100 in apoHO-2 and the un-modeled electron density in heme-bound HO-2. The Triton X-100, which occupies a position near the hydrophobic cavity, might hinder the heme pocket from assuming a closed conformation. Although there is no definitive evidence, it seems likely that the extra electron density that was observed adjacent to the heme in the heme-bound HO-2 structure is a polyethylene glycol dimethyl ether 500 molecule from the crystallization solution. The polyethylene glycol dimethyl ether 500 of heme-bound HO-2 could act in the same way as the Triton X-100 in apoHO-2 and force the distal helix into the open conformation. This is not the first time that something has been found in close proximity to the hydrophobic region. A crystal structure of HO-1 in complex with biliverdin shows that the biliverdin binding site is adjacent to the hydrophobic pocket, and, compared with the heme biding site of HO-1, it has moved further into the interior of the heme pocket (52). A portion of the biliverdin molecule in HO-1 occupies the same position as Triton X-100 in apoHO-2 and unaccounted electron density in heme-bound HO-2. It appears that this large cavity in HOs plays a role in hydrophobic substrate binding.
Comparison of HO-1 and HO-2-There is a high level (45%) of sequence identity between fulllength HO-1 and HO-2 (Fig. 6), with 55% identity in the conserved core regions of HO-1 and HO-2 exhibit. The heme pocket regions in HO-1 and HO-2 are highly conserved. With such a high level of identity, it is not surprising that the core domains of HO-1 and HO-2, which are responsible for the heme binding and degradation, share a similar fold. Both HO-1 and HO-2 structures consist mostly of ␣ helices. Overall, the secondary structure of HO-2 closely resembles apoHO-1 and heme-bound HO-1.
It has been shown that mutating Asp 140 to an Ala converts HO-1 into a peroxidase (56). It is proposed that the Asp 140 to Ala mutant cannot stabilize the peroxide bond and will lead to an Fe 4ϩ -O oxyferryl center instead of the Fe 3ϩ -OOH reactive intermediate (57). Asp 140 stabilizes the Fe 3ϩ -OOH intermediate through a hydrogen bonding network with water molecules located near the heme and the Fe 3ϩ -OOH intermediate. In the high resolution heme-bound HO-1 structure, the water molecules thought to be involved in this hydrogen bond network were observed in the closed conformation (51). The open conformation of heme-bound HO-1 has a much less ordered solvent structure with some water thought to be at a partial occupancy. This disorder in the open confirmation may explain why the solvent structure is not seen in the heme-bound HO-2 structure. Even though the entire hydrogen network is not resolved in the heme-bound HO-2 structure, Asp 160 still adopts the same conformation as Asp 140 in the heme-bound HO-1 model and is thought to act in a similar fashion.
Although the C-terminal HRM region shows the most significant sequence divergence between HO-1 and HO-2, most of the key amino acids involved in heme binding, Lys 38 , His 45 , Glu 49 , Tyr 154 , Gly 159 , Asp 160 , Gly 163 , Lys 199 , and Arg 203 are observed in similar positions in the heme-bound HO-2 structure when compared with heme-bound HO-1. The major differences in the crystal structure of apoHO-2 and heme-bound HO-1 are seen in the helices that directly interact with the heme. The proximal helix of apoHO-2 moves slightly down-