Adapting to oxygen: 3-Hydroxyanthrinilate 3,4-dioxygenase employs loop dynamics to accommodate two substrates with disparate polarities

3-Hydroxyanthranilate 3,4-dioxygenase (HAO) is an iron-dependent protein that activates O2 and inserts both oxygen atoms into 3-hydroxyanthranilate (3-HAA). An intriguing question is how HAO can rapidly bind O2, even though local O2 concentrations and diffusion rates are relatively low. Here, a close inspection of the HAO structures revealed that substrate- and inhibitor-bound structures exhibit a closed conformation with three hydrophobic loop regions moving toward the catalytic iron center, whereas the ligand-free structure is open. We hypothesized that these loop movements enhance O2 binding to the binary complex of HAO and 3-HAA. We found that the carboxyl end of 3-HAA triggers changes in two loop regions and that the third loop movement appears to be driven by an H-bond interaction between Asn27 and Ile142. Mutational analyses revealed that N27A, I142A, and I142P variants cannot form a closed conformation, and steady-state kinetic assays indicated that these variants have a substantially higher Km for O2 than WT HAO. This observation suggested enhanced hydrophobicity at the iron center resulting from the concerted loop movements after the binding of the primary substrate, which is hydrophilic. Given that O2 is nonpolar, the increased hydrophobicity at the iron center of the binary complex appears to be essential for rapid O2 binding and activation, explaining the reason for the 3-HAA–induced loop movements. Because substrate binding-induced open-to-closed conformational changes are common, the results reported here may help further our understanding of how oxygen is enriched in nonheme iron-dependent dioxygenases.

It is estimated that ϳ30% of the mass of woody plant material on earth is comprised of aromatic compounds that are known for their unusual stability when compared with similar aliphatic hydrocarbons (1). One strategy devised by nature to overcome the stability of aromatic compounds is to utilize iron-bound oxygen directly to cleave the ring, breaking aromaticity. The ring-cleaving dioxygenase enzymes are abundant and versatile. These enzymes utilize a mononuclear, nonheme iron ion to orchestrate electron transfer between an organic substrate and oxygen molecule with catalytic reaction rates rapid enough for their catalytic intermediates to escape spectroscopic characterization (2). A fundamental question is how these enzymes accomplish rapid O 2 binding, a prerequisite for the observed catalytic rates, whereas local O 2 concentrations in aqueous solution are relatively low (200 -300 M at 1 atm pressure and room temperature), and the rate of oxygen diffusion from air to solution is slow relative to the rate of catalysis.
Here, we chose 3-hydroxyanthranilate 3,4-dioxygenase (HAO), 3 a ring-cleaving dioxygenase, as a model system to study this problem. HAO is a nonheme, iron-dependent extradiol dioxygenase belonging to the cupin superfamily (3)(4)(5). This enzyme cleaves the aromatic ring of its substrate adjacent to the substitution groups. It activates and inserts dioxygen between C3 and C4 of its substrate 3-hydroxyanthranilate (3-HAA), ultimately producing quinolinic acid via the intermediary product ␣-amino-␤-carboxymuconate-⑀-semialdehyde (Fig. 1A). HAO is present in the kynurenine pathway, where it is essential for tryptophan degradation and de novo NAD ϩ biosynthesis in eukaryotic organisms and a few bacteria (6 -10). This enzyme is also found in the 2-nitrobenzoic acid catabolic pathway for energy production in some bacteria (11). HAO from Cupriavidus metallidurans has been shown to operate at a catalytic turnover rate of 25 s Ϫ1 (3,12) until near depletion of O 2 . A second iron-binding site at the protein surface has been proposed to be an iron storage site for the protein to ensure activesite iron occupancy when the enzyme is dealing with a substrate that is also a metal chelator (13).
Previous structural studies of HAO have allowed atomic level understanding of the overall protein structure and the active site architecture (3)(4)(5). A substrate analog, 4-Cl-3-HAA (ClHAA) was tested and defined as an irreversible inhibitor to prevent the active site from binding with the native substrate (12). The crystal structures of HAO from C. metallidurans were also reported previously (3,13). In this study, a closer inspection of the substrate-free (PDB entry 4L2N) and substratebound (PDB entry 1YFY) structures of WT HAO (wtHAO) showed that the enzyme exhibits two distinct conformations (Fig. 1B), which are hereafter termed open and closed states. The substrate-bound structure is in the closed conformation, showing significant structural differences relative to the ligandfree, open state of three specific loop regions surrounding the catalytic iron center, which is composed of a ferrous ion coordinated by a His 2 -Glu 1 motif. The crystal structure of HAO in complex with O 2 and ClHAA (PDB entry 1YFW) (3,12) also shows the closed state with three loops moving toward the active site (Fig. 1C). Therefore, the questions raised here are the following: 1) What is the driving force for the loop movement? 2) How are the loop movements on the three loop regions coordinated? 3) What is the potential biological significance of the loop movement in the context of oxygen binding? To investi-gate these questions, we identified and mutated the essential residues involved in the loop movement, determined kinetic parameters for both oxygen and 3-HAA, and performed X-ray crystallographic studies. The results revealed how the nonheme iron enzyme handles its substrates with disparate polarities through a primary substrate-induced change of the active site hydrophobicity.

Insights inferred from sequence and structural alignments
We noted that three largely hydrophobic loop regions in HAO move toward the catalytic iron center, forming a closed conformation upon HAO binding with the primary substrate (PDB entry 1YFY) compared with the ligand-free counterpart (PDB entry 4L2N) (3,13). Similar structural features were obtained in hundreds of repeated data sets of the ligand-free and substrate-bound structures, indicating that the organic substrate-binding induced closed conformation is an intrinsic feature of the enzyme. The three loop regions consist of residues 21-28 for loop region A (Leu 20 -Leu 21 -Lys 22 -Pro 23 - Most of these residues are highly conserved across the protein sequences of HAO from various origins (Fig. 1D).
A salient feature is that all three loops are composed of mostly hydrophobic residues (bolded); 17 of the total 26 residues (65%) are aliphatic (Fig. S1). Each loop also contains one or two proline residues. Proline is a cyclic hydrophobic residue and is commonly utilized as a helix breaker and an aid in formation of turns. Two consecutive proline residues are present in both loop A and loop C (i.e. Pro 23 -Pro 24 and Pro 147 -Pro 148 , respectively). The proline fragment endows the two loops with increased hydrophobic structure. It is interesting to note that the two negatives make a positive, i.e. the Pro 147 -Pro 148 fragment, a double helix breaker, is part of an ␣-helical structure. Loop B is a highly hydrophobic fragment with only one proline residue (Pro 45 ). Upon 3-HAA binding-induced loop movement, loop A undergoes the largest movement among the three loops with a shift toward the active site of ϳ8.6 Å in average when compared with the corresponding loop regions in the substrate-free form.
Another notable feature of the loop regions in the closed form of HAO is that the side chains of hydrophobic residues of Val 25 , Val 41 , Ile 142 , Val 142 , Leu 139 , and Leu 146 are sequestered toward to the active site, where the catalytic iron ion is located (Fig. S1, B and D). Meanwhile, most of the side chains of the hydrophilic residues point away from the active site: Lys 22 , Arg 28 , His 46 , Lys 140 , Thr 144 , and Asp 145 (Fig. S1, B and C). When the three loops are gathered toward the catalytic iron ion, the ligand-bound iron center becomes significantly more compact and hydrophobic, which is distinct with respect to the more open and spacious conformation observed in the ligandfree structure. The movement of loop A is triggered by the interaction of the substrate carboxylate group with the side chains of Asn 27 and Asn 45 and the backbone of Val 25 , which form the potential H-bond or salt-bridge network in both substrate-bound and ClHAA-bound structures (Fig. 2). Loops A and B appear to move synergistically because of the nearly pairwise backbone residue interactions and like a "double-door" to the active site cavity. As shown in Fig. S1C, an H-bonding network forms among the backbones of residues Pro 23 , Pro 24 , and Gly 26 in loop A and Gly 42 , Pro 44 , and His 46 in loop B, gathering loops A and B toward the active site.

Single-point mutation of Asn 27 and Ile 142 diminished the catalytic efficiency by decreasing the oxygen capturing capacity
One question arising from the structural comparison is how loop C is triggered to move toward the catalytic iron center. This loop region is more distant from the active site in the open conformation. However, the entire ␣-helix to which loop C is connected rotates ϳ10° (Fig. S1B). After binding 3-HAA or ClHAA, the side chain of Asn 27 and the backbone of Ile 142 moves closer to the iron center by forming hydrogen-bonding interactions with each other (Fig. 2) (3). Specifically, the distance between the backbone amide of Ile 142 from loop C and the side chain of Asn 27 from loop A decreases substantially, from ϳ8.0 to 2.6 Å, which results in loops A and C directly moving together and toward the active site. Thus, Asn 27 and Ile 142 appear to be the critical residues for a coordinated conformational change during 3-HAA binding to the iron ion.
To determine the connection and especially the effect of the loop movement on the reaction catalyzed by HAO, we mutated Asn 27 and Ile 142 to proline and alanine. The kinetic parameters of N27A, I142P, and I142A were determined for both substrates, 3-HAA and O 2 in separate experiments. With respect to the organic substrate, 3-HAA, two different buffer environments were involved. One was the oxygen-saturated buffer by bubbling oxygen (Ͼ1.3 mM) during measurement, and the other used regular buffer without any treatment and contained 286 Ϯ 8 M molecular oxygen. Under the regular buffer environment, the mutant N27A showed a 6-fold decrease in the k cat and a 7-fold increase in the K m (3-HAA) relative to the WT enzyme (Table 1), exhibiting a substantially reduced catalytic efficiency, 2.5% of wtHAO. However, using O 2 saturated buffer, the k cat of N27A reached the same level of wtHAO, indicating that the primary reason for a substantial negative impact of the N27A mutation in catalysis with the regular buffer is a decreased oxygen-binding affinity.
The I142A and I142P variants showed 53-and 9.5-fold decreased catalytic efficiency, k cat /K m , toward 3-HAA with the regular buffer compared with wtHAO, respectively. The K m (3-HAA) (239.7 M) of I142A increased 10-fold, whereas the k cat decreased by ϳ5-fold compared with these of wtHAO.

A mechanism for O 2 enrichment in an extradiol dioxygenase
The K m (3-HAA) of I142P increased 4.3-fold. With oxygen-saturated buffer, the apparent constants k cat of I142A and I142P for 3-HAA are 13.9 Ϯ 0.9 and 24.8 Ϯ 2.6 s Ϫ1 , respectively, showing a partial or a full rescue of the catalytic efficiency. Interestingly, both mutants exhibited a near comparable catalytic activity with that of wtHAO only with oxygen supplementation.
Because of the apparent yet significant increasing of hydrophobic environment in the active site, we hypothesize that the major benefit of the loop movement is to attract the other nonpolar substrate, molecular oxygen. We also determined the K m value of O 2 . With a saturating amount of 3-HAA, the K m value increased ϳ9-fold, from 124.8 M for wtHAO to 1144 M for N27A and 1120 M for I142P, whereas the k cat remained similar (Table 1). An attempt to determine the K m for O 2 of I142A mutant failed because of an inability to saturate the enzyme (K m Ͼ 2,227 M).

Mutation of Asn 27 evades substrate-binding-induced loop movement
To assess whether or not the closed conformation would still be formed upon substrate binding after the alteration of the primary residue involved in orchestrating the loop movement, the mutant N27A was crystallized under similar conditions reported for wtHAO (3,13). The holo crystal structures of N27A in the ligand-free form was refined to 1.90 Å ( Table 2). Comparison of the structure of N27A ( Fig. 3A) with wtHAO (PDB entry 4L2N) revealed no apparent deviations, including the metal ligands and second sphere residues, as indicated by a small RMSD value of 0.090 Å between the two structures of all C␣ carbons. We then attempted to obtain a ligand-bound structure of this mutant. N27A crystals were soaked with 3-HAA or ClHAA anaerobically. The crystal structure of N27A in complex with ClHAA was solved at 2.30 Å (Fig. 3B). The omit map in the active site clearly illustrates the location of carboxylate and amine groups of ClHAA because of its strong saltbridge interactions with residue Arg 99 (Fig. S2). However, the phenyl ring and chloride electron densities were incomplete. Compared with ClHAA-bound wtHAO structures (PDB entries 1YFW and 1YFX), there is also negative density map around the chloride of ClHAA (3). Unfortunately, the ES complex structure remained elusive, even after extensive attempts. Compared with the single-crystal electronic absorption spectrum of ligand-free N27A, a N27A crystal soaked with ClHAA for 5 min exhibits an apparent absorbance peak at 328 nm (Fig.  3C), which is characteristic of ClHAA, indicating that ClHAA is indeed bound to N27A. The solution state UV-visible spectra of ClHAA-bound wtHAO and N27A showed similar binding affinity for the inhibitor (Fig. S3). The absorbance maximum of the inhibitor is also centered around 328 nm, which agrees well with the single-crystal UV-visible data.
The electron density map of ClHAA-bound N27A reveals that the inhibitor binds to the active-site iron with an ironhydroxyl distance of 2.7 Å, similar to what was observed in the ClHAA-bound structure of wtHAO. However, the superimposition of ClHAA-bound N27A and wtHAO structures reveals several differences. As compared with wtHAO, the phenyl ring of the inhibitor moved up with an angle of 23°in the mutant structure, whereas the carboxylate group was fixed in both structures by a salt bridge with Arg 99 , and the hydroxyl group of

Effect of Ile 142 on loop regions conformation change
Because the mutation of Ile 142 to proline and alanine also significantly decreased the oxygen binding and activation capacity (Table 1), a question raised was whether I142P and I142A would also be incompetent to present a closed conformation after binding with the substrate or ClHAA. Considering the H-bond interaction is observed between the backbone of Ile 142 and the side chain of Asn 27 , the I142A mutant was A mechanism for O 2 enrichment in an extradiol dioxygenase

in which the sum is over all the i measured reflections with equivalent Miller indices hkl, I(hkl)
is the averaged intensity of these i reflections, and the grand sum is over all measured reflections in the data set. c According to Karplus and Diederichs (28). d All positive reflections were used in the refinement. e According to Engh and Huber (29). f Calculated by using MolProbity (30). g NA, not applicable.

A mechanism for O 2 enrichment in an extradiol dioxygenase
expected to have a more subtle effect than N27A or I142P mutants on the oxygen binding and activation efficiency. The potential effects of deletion of the side chain of Ile 142 on loop regions movement were investigated by solving the ligand-free, enzyme-substrate, and enzyme-inhibitor complex structures.
The superimposed ligand-free structures of I142A (resolution 1.90 Å) and I142P (resolution 2.22 Å) and wtHAO showed high similarity with an RMSD of 0.141 and 0.161 Å from all the C␣ atoms aligned in these structures. After soaking with ClHAA for 10 min, an I142A-ClHAA complex was obtained and refined at 2.31 Å (Fig. 4A and Fig. S4). The single-crystal electronic absorption spectrum of the I142A crystal soaked with ClHAA shows absorbance at 328 nm caused by binding of the inhibitor to the HAO variant (Fig. S5). Like the N27A mutant, the overall structure of the I142A-ClHAA complex also exhibited high similarity with wtHAO with an RMSD of 0.144 Å. The binding mode of ClHAA to the active-site iron ion of I142A was identical to the N27A-ClHAA complex except that a water molecule is still bound to iron ion. However, the distance between the side chain of Asn 27 and Ala 142 backbone was 7.7 Å, making it too far to form a hydrogen-bond interaction. Three loop regions of the I142A-ClHAA complex align well with the open form of ligand-free wtHAO structure (PDB entry 4L2N in Fig. S6). Moreover, the crystal structure of I142A in complex with 3-HAA was obtained and refined to 2.60 Å (Fig. 4B). The electron density map reveals that 3-HAA binds to the active site of I142A in a bidentate chelation pattern. The superimposition of these two structures reveals a 45°rotation of the phenol rings of the substrates in the active site. The loop regions can be well aligned with the substrate-free structure (Fig. S6). The distance between Ala 142 and Asn 27 is 7.6 Å, not eligible for formation of an H-bond.
We also crystallized I142P and soaked its crystals with 3-HAA and ClHAA. As a result, the complex structures of I142P bound with 3-HAA or ClHAA were determined and refined to high resolution ( Table 2). Fig. 4 (C and D) clearly show a completed omit density map for both of ClHAA and 3-HAA in a similar binding mode to each other in the active site but distinct from those observed for the wtHAO. The ligands are anchored in the active site by interacting with the side chains of Arg 99 and Glu 110 . However, they are not directly ligated to the catalytic iron ion. Additionally, the two water ligands remain coordinated to the metal ion. Because residue 142 is part of an ␣-helix structure, and it sits at the start point of the loop C region, the helical structure of I142P in loop C is shorter than that of the wtHAO, resulting in a relatively larger change of the loop C than that in I142A. However, the overall structures of ClHAA-and 3-HAA-bound I142P show high similarity to wtHAO with an RMSD of 0.180 and 0.166 Å, respectively. Moreover, the loop regions well align with the substrate-free, open-form of wtHAO despite a shorter helix in loop C (Fig. S7). The distances for the two complexes between Pro 142 and Asn 27 are 8.5 and 8.4, respectively, and hence these residues are not able to interact with each other. Of note, the I142P alteration leads to a new binding mode for 3-HAA and ClHAA, which explains the 330-to-350 nm shift of the absorbance maximum in the ClHAA-bound I142P solution optical spectrum (Fig. S3).
A perusal of the differences between the open and closed forms of HAO revealed that not only does the backbone of Ile 142 form a hydrogen bond with Asn27 in the closed form, but its bulky, hydrophobic side chain also points toward the active site. Compared with the ligand-free structure, deletion of the side chain of Ile 142 led to a larger cavity with seven ordered water molecules between Ala 142 and the iron center in the ligand-free structure (Fig. S8). The I142P structure also has extra solvent between the Pro 142 and the iron center. The ligand-free structures of wtHAO reported were refined at higher resolutions; therefore the water cluster in the mutant structure is less likely because of resolution differences. It appears that mutation of I142 disrupts a water-free, putative oxygen binding channel connecting protein surface to the catalytic iron ion. Therefore, the K m values for O 2 of I142A and I142P increased dramatically.

Protein dynamics during HAO reaction
Molecular oxygen is known to have a central role in complex living systems. Understanding how nature is able to selectively utilize O 2 , a nonpolar reagent, to perform specific chemistry in aqueous solution at a high rate of catalysis is a question of great fundamental importance and deserving of study (1). It is apparent that the substrate of HAO, 3-HAA, with three polar functional groups, has considerable hydrophilic character. After binding with hydrophilic 3-HAA, how the HAO binary complex attracts a second, hydrophobic substrate is a question of fundamental importance. Here, we demonstrated that binding of 3-HAA induces loop dynamics of HAO to move to make for a more hydrophobic environment to efficiently capture molecular oxygen. The protein dynamic conformational change has loop regions moving by 8.6 Å and totally involves more than 15% of the protein residues in HAO. It is historically believed that O 2 would randomly diffuse into the protein scaffold, and enzymes that utilize O 2 have evolved to develop interior hydrophobic regions for O 2 binding. HAO presents a more intelligent way to facilitate O 2 binding to its active site. The O 2 -binding channel is not permanently composed; it is dynamically formed in response to substrate binding and changes along with the reaction process. Presumably, the open conformation is for exchange of substrate and product, and the closed conformation is for rapidly incorporating oxygen.

The putative O 2 binding cavity in the active site of HAO
The loop regions covering the active site of HAO form an apparently dynamic hydrophobic cavity. In the ternary com-plex of HAO-ClHAA-O 2 structure, this is where the O 2 binds with one of the oxygen atoms anchored at the catalytic iron ion. The cavity in the ligand-free, open-form HAO structure (Fig.  S9A) appears to be connected with the solvent from two different directions, potentially serving as a tunnel to facilitate ligand binding to the active site. In contrast, the cavity in the closed form wtHAO structure (Fig. S9B) exhibited significant differences as compared with the open form. The radius of the cavity decreases by approximately half, and the cavity no longer connects to the bulk solvent, presumably to minimize the impact of bulk solvent to the reactive intermediates during the catalytic reaction.
As was mentioned previously, Ile 142 and Asn 27 are located ϳ9 and 11 Å away from the catalytic iron ion, respectively. Elimination of the hydrogen-bonding interaction between loops A and C (mutant N27A) creates a variant that does not show the closed conformation upon substrate binding. The catalytic efficiency (k cat /K m ) of N27A for both molecular oxygen and 3-HAA is dramatically decreased compared with WT ( Table 1). The hydrogen bond formed between Asn 27 and Ile 142 is hypothesized to be the driving force to gather the loop regions together, which is triggered by the chelation of 3-HAA or ClHAA. Therefore, the dynamic action of the loop regions are critical for efficiently performing the dioxygenation. Meanwhile, the decreased hydrophobic character of loop region C (mutant I142A) could also lead to the elimination of loop movement, and determination of kinetic parameters with respect to oxygen is not possible. Without the proper positioning of Asn 27 and Ile 142 , the movement of loop regions A, B, and C would not occur, so the hydrophobic environment made by the "closed" form of the loop regions is broken, which leads to the reduction of oxygen capturing capacity of the mutants. Therefore, the major contribution of the loop regions is providing a sealed hydrophobic environment for binding and activation.

The loop regions in HAO from other biological origins
A comparison of available structures of 3-HAA dioxygenases can be found in Fig. S10. With the exception of an extra domain in HAO from mammalian sources, the overall structures show high similarity from a broad range of species. Compared with HAO from C. metallidurans, the RMSD values of HAO from Homo sapiens, Bos taurus, and Saccharomyces cerevisiae are 0.594, 0.813, and 1.072 Å, respectively. Moreover, the loop regions are among the highly conserved regions. Therefore, the loop regions required for the dioxygenation of HAO are one of the conserved structural characteristics during the evolution of HAO. Interestingly, holo-HAO from bovine crystallized in the closed form with a hydrogen bond between Asn 24 and Leu 137 (5). The isoleucine from loop C in prokaryotes is replaced by another similar residue of leucine in HAO from several eukaryotes (Fig. 1D). Based the alignment results, it is reasonable to believe that the dynamic action of loop regions for attracting hydrophilic or hydrophobic substrates has been conserved in HAOs during evolution.

A mechanism for O 2 enrichment in an extradiol dioxygenase How do the nonheme iron enzymes handle two or more substrates with distinct polar natures?
It is common in dioxygenases that flexible loop region movement is induced by interactions with substrates to be more ordered to cover the active site. Such an enzyme-substrate interaction can be observed and visualized by X-ray crystallographic structure determination. In an extradiol dioxygenase LapB, a loop containing His 248 and Tyr 257 was reported to hydrogen bond with its substrate, causing lower flexibility in the ligand-bound form (14). These two residues have essential roles during catalysis. In quercetin 2,3-dioxygenase, after binding its substrate (flavonol), a flexible loop around the active site became more ordered to stabilize the loaded substrate, and then the activation of C2 atom of flavonol was made to bind oxygen (15).
The nonheme iron and ␣-ketoglutarate (␣KG)-dependent enzyme family catalyzes a variety of chemical reactions coupled to the decarboxylation of ␣KG to succinate. They meet the similar challenge of dealing with both a hydrophilic substrate (␣KG or other substrates) and a hydrophobic substrate (oxygen). The common strategy is that they usually react with oxygen after binding a primary substrate. Taurine and ␣KG-dependent dioxygenase catalyzes taurine to aminoacetaldehyde and sulfite. Structural comparison reveals that the binding of taurine induced an ␣-helix region containing hydrophobic residues to move forward to the active site by forming a hydrogenbonding network (16). The strategy of conformational change around the active site is similar with that of HAO. The ␣-helix region would help to form a more hydrophobic environment in the active site and benefit the capture of oxygen for performing the dioxygenation. The CurA halogenase also belongs to the nonheme Fe(II)/␣KGdependent family (17) and catalyzes a cryptic chlorination on a specific acyl carrier protein. Their crystal structures with various states indicate "open" and "closed" conformations. After binding with ␣KG, the disordered "lid" (open form) containing 27 residues on the top of the active site became a more ordered closed form, which was triggered by the salt bridge between ␣KG and Arg 241 . The open form was proposed to organize the binding sequence in which the cosubstrate ␣KG and Cl Ϫ need to bind to the iron ion prior to the oxygen and the acyl carrier protein. Although the lid region does not contain a high percentage of hydrophobic residues, its dynamic action is critical for the chlorination.
Phenylalanine hydroxylase (PAH) is a nonheme iron-and pterin-dependent enzyme that catalyzed the oxidation of phenylalanine to tyrosine. PAH binds pterin and L-Phe first and oxygen (hydrophobic) last (18). Structural study reveals that binding of pterin in the active site of PAH induces a loop region to move in the direction of the iron center by forming hydrogen-bond networks with the residues on the loop. Therefore, the PAH-pterin complex provides an ideal orientation for binding oxygen in between pterin and the iron ion (19).
Unlike the nonheme iron-dependent enzymes mentioned above, the loop regions in HAO do not have an obvious interaction with 3-HAA in the active site; rather they provide a hydrophobic environment for the dioxygenation caused by the high percentage of nonpolar residues in the loop regions. For other dioxygenases, including catechol dioxygenase, 2,3-dihydroxybiphenyl dioxygenase, biphenyl dioxygenase, and cysteine dioxygenase, substrate binding-induced loop region movement has not been reported or does not exist.

Structural requirement for capturing O 2 in iron-dependent proteins
The most well characterized oxygen binding to an iron protein is in heme systems. Hemoglobin is well known as an oxygen-transport metalloprotein in the blood cells, carrying oxygen from the respiratory organs to the rest of the body. Because of the functional requirement, hemoglobin has a high oxygenbinding capacity (20). The relationship between the oxygenbinding and structural information has been investigated since the hemoglobin crystal structures were determined. The histidine gate hypothesis was proposed by Perutz and Mathews (21) to demonstrate the HisE7 in the distal region controls the ligand migration rate by a conformation change from an open to a closed status. Boechi et al. (22) described that the hydrophobic effect around HisE7 is the major driving force for oxygen uptake other than steric hindrance from HisE7. Moreover, Cohen and Schulten (23) compared the characters of O 2 migration pathway in 12 monomeric globins from a broad range of species. The pathway for O 2 migration shows various configurations from different globins; however, the located residues are conserved as mainly containing hydrophobic residues (Trp, Phe, Leu, Ile, etc.). Therefore, the hydrophobic environment of cavity networks of globins facilitate ligands traveling through the protein matrix.

Concluding remarks
Although nature has more than one strategy to solve the low-oxygen-concentration problem in solution for a rapid oxygen consumption reaction at the enzyme active site, the unique feature of protein dynamics induced by substrate binding reported here may be the first structurally defined example for facilitating O 2 binding to the catalytic iron along with the primary organic substrate in each catalytic cycle. With a rapid turnover rate, the O 2 level could drop significantly after several seconds of oxygenation reaction, and the O 2 diffusion to the solution may not be fast enough and thus may cause a seemingly hypoxic condition. If this occurs, the increased hydrophobicity would become more important to attract O 2 . Given the vast number of metal-dependent oxygen utilization enzymes, the protein dynamic-hydrophobicity strategy may be one of nature's common strategies to direct molecular oxygen toward organic substrate-bound metal centers.

Materials
The substrate 3-HAA and the inhibitor ClHAA were purchased from Sigma-Aldrich and Enamine Ltd. (Monmouth Jct., NJ), respectively.

Site-directed mutagenesis and protein preparation
The HAO gene from C. metallidurans was used as the template (GenBank TM accession no. CP000353) (3) to construct single mutants by the PCR overlap extension mutagenesis method. The forward primers were 5Ј-TTGAAGCCGCC-CGTCGGCGCCCGGCAGGTGTGGCAGGAC for N27A and 5Ј-GAGGTGCAGCTCAAGAGCGCCGTCACCGACCTCC-CGCCT for I142A, and 5Ј-GAGGTGCAGCTCAAGAGCCC-GGTCACCGACCTCCCGCC for I142P. The plasmids of mutants were verified and transformed to the expression host cells, Escherichia coli BL21(DE3). The purification strategy of the two mutants' proteins was the same as wtHAO (3). The wtHAO and mutants were anaerobically reconstituted with 3 eq. Fe(NH 4 ) 2 (SO 4 ) 2 and excess Fe 2ϩ was removed by Superdex 75 column before performing kinetic assay and crystallization.

Kinetic assay of wtHAO and mutants toward both substrates
The kinetic assays for wtHAO and mutants toward the substrate of 3-HAA were measured by monitoring the absorbance increase caused by the product formation (␣-amino-␤-carboxymuconate-⑀-semialdehyde, ⑀ 360 nm ϭ 47,500 M Ϫ1 cm Ϫ1 ) with an Agilent 8453 diode-array spectrophotometer (3,13). The O 2 -saturated buffer was made through bubbling with pure oxygen gas during the analysis. The kinetic assays toward the substrate of oxygen were monitored with a Clark-type oxygen electrode (Oxygraph, Hansatech Instruments) with saturating 3-HAA concentration. All assays were done at room temperature.

Crystallization, data collection, processing, and refinement
The mutants were crystallized by the hanging-drop method in an anaerobic chamber as described previously (3,13). After incubation with 20 mM of the inhibitor ClHAA for ϳ5-20 min, the soaked crystals were flash-cooled in liquid nitrogen. The diffraction data sets were collected, processed, and scaled by HKL-2000 (24). The structures of mutants were solved by molecular replacement by using the wtHAO (PDB entry 4L2N) as the template and refined by employing the Phenix 1.11.1-2575 (25) and Coot 0.8.3 (26). PyMOL (27) was used in drawing structural figures.