2.1. Purification and crystallization

Wild-type 2-aminophenol-1,6-dioxygenase from Comamonas sp. strain CNB-1 (APD_CNB-1) was purified from laboratory-cultured bacterial cells as described previously (Wu et al., 2005). Selenomethionine-substituted protein was produced from the same strain grown in M9 minimal medium supplemented with L-selenomethionine. The methionine-pathway inhibition protocol was used to obtain maximal incorporation (Doublié, 1997).

The purified apoenzyme was concentrated to approximately 10 mg ml^-1 in a solution consisting of 50 mM Tris pH 8.0, 5% glycerol, 2 mM DTT, 0.2 mM EDTA prior to crystallization trials. Crystals of metal-free APD were grown in 25% PEG 3350, 0.2 mM sodium chloride, 0.1 M sodium cacodylate pH 6.5 by the hanging-drop method at room temperature under ambient aerobic conditions. Crystallization drops for the active enzyme were set up under the same conditions in the presence of 5 mM FeSO₄. Complex crystals were prepared by cocrystallization of the holoenzyme with the addition of 5 mM 2-aminophenol (2AP) or 4-nitrocatechol (4NC).

2.2. Data collection

The crystals used for data collection were dipped into cryoprotectant (reservoir solution supplemented with 10% glycerol) for roughly 10 s before being mounted in nylon CryoLoops (Hampton Research) and flash-cooled in a stream of liquid nitrogen at 95 K. X-ray data collections for active APD and the complex with 4NC were carried out on beamline NW12A at KEK, Photon Factory, Japan using a wavelength of 1.0 Å. Multiwavelength anomalous diffraction (MAD) data for the selenomethionyl derivative were also collected on beamline NW12A using three wavelengths: 0.9792, 0.9794 and 0.9642 Å. Native data sets for the apoenzyme and the complex with 2AP were collected using an in-house Rigaku R-AXIS IV⁺⁺ image plate and a Cu K radiation source ( = 1.5418 Å). All diffraction data were evaluated using MOSFLM and SCALA from the CCP4 program suite (Winn et al., 2011; Rossmann & van Beek, 1999). Further details of APD purification, crystallization and data collection are given in Li et al. (2012).

2.3. Structure determination and refinement

The structure of metal-free inactive APD was determined using the MAD technique. Phase calculation and density modification were performed using the SOLVE/RESOLVE program suite (Terwilliger, 2004). A partial model was automatically established and the rest of the model was manually built using the graphics package O (Jones et al., 1991). The structure was refined using CNS (Brünger et al., 1998). The refined model was subsequently used as a search template in structure determinations of the active enzyme and the APD-2AP and APD-4NC complexes using molecular replacement. Refinement of these structures was performed using CNS and REFMAC5 (Murshudov et al., 2011) with manual modelling between refinement cycles. The final models were validated using the CCP4 program PROCHECK (Laskowski et al., 1993) to control stereochemical quality. Two subunits and two subunits forming a moderately compact tetramer are present in the asymmetric unit of all crystals. The statistics of data collection, MAD phasing and structure refinement are given in Tables 1 and 2. All figures representing APD structures were prepared using the PyMOL molecular-graphics program (http://www.pymol.org ).

Table 1 Summary of MAD data collection and phasing for Fe-free APD Values in parentheses are for the highest resolution shell.   Peak Inflection Remote Data collection   Wavelength (Å) 0.9792 0.9794 0.9642   Space group C2 C2 C2   Unit-cell parameters     a (Å) 270.67 269.85 272.05     b (Å) 48.76 48.51 48.73     c (Å) 108.82 108.59 109.68      (°) 90 90 90      (°) 109.86 110.10 109.68      (°) 90 90 90   Resolution (Å) 69.17-2.60 (2.74-2.60) 68.68-2.70 (2.85-2.70) 69.50-2.70 (2.85-2.70)   Completeness (%) 99.2 (93.0) 99.9 (99.1) 99.8 (98.7)   Unique reflections 41350 (5544) 36876 (5264) 37677 (5355)   Multiplicity 7.0 (5.4) 3.6 (3.4) 3.6 (3.3)   Rmerge (%) 8.8 (42.4) 9.3 (45.4) 9.7 (47.2)   Average I/(I) 18.7 (3.0) 12.1 (2.2) 11.8 (2.0) Phasing   Se-atom sites 29   Overall figure of merit 0.63

2.4. Ligand modelling and refinement

In the enzyme-substrate complex structure, the lactone intermediate 3-iminooxepin-2(3H)-one and the product 2-aminomuconic 6-semialdehyde were modelled into the subunits D and B, respectively. To obtain the best fit to the diffraction data, F_obs - F_calc electron density contoured at 3.5 was used for this modelling work. Likewise, the suicide inhibitor 4NC was modelled into both catalytic subunits of the enzyme-inhibitor complex structure. The ligand dictionaries for 3-iminooxepin-2(3H)-one and 2-aminomuconic 6-semialdehyde were generated from imported PDB files using the CCP4 program Monomer Sketcher after running REFMAC5 to idealize the ligand structures and that for 4NC was downloaded from the HIC-Up database (Kleywegt, 2007). The atom indices of all ligands were renumbered to maintain consistency with those shown in Fig. 1. Model bias introduced by ligand modelling was minimized by real-space refinement in the annealed OMIT maps generated in the absence of active-site ligands. The distances between the amino or hydroxyl groups of the ligands and Fe were not restrained during refinement. In the last round of refinement, a ligand-free model was generated by setting the occupancies of all atoms of the ligands to zero and allowing them to be refined for ten cycles. The resultant positive density (>3.5) in an OMIT difference map (F_obs - F_calc) was checked to determine whether it overlays well with ligand 2F_obs - F_calc density after ten cycles of full model refinement.

2.5. Enzyme-kinetics measurements

Catalytic kinetics measurements of APD on 2AP and catechol were performed at room temperature by monitoring the increase in absorbance at 380 nm for 2AP (Wu et al., 2005) and at 375 nm for catechol. The reaction solution buffered at pH 8.0 contained 0.16 mM 2AP or catechol. The K_m value was calculated from Lineweaver-Burke plots using the following molar extinction coefficients: 15 100 M^-1 cm^-1 for 2AP (Lendenmann & Spain, 1996) and 33 000 M^-1 cm^-1 for catechol (Iwagami et al., 2000). The K_m of O₂ on different substrates was obtained polarographically from plots with an oxygen concentration ranging from 0 to 1200 µM. Reaction buffers containing different oxygen concentrations were prepared by vigorously bubbling with a humidified mixture of O₂ and N₂ gas for at least 60 min prior to the experiment, and the dioxygen partial pressure was controlled by adjusting the O₂:N₂ ratio in the gas mixture.

For the inhibition experiment, the purified enzyme was incubated with 0.2 mM catechol before measuring the remaining catalytic activity towards 2AP. Subsequently, the reaction solution was applied onto a desalting column (HiTrap 5 ml, GE Healthcare) to remove catechol and the enzyme was incubated with 2 mM FeSO₄ and 5 mM ascorbate for 2 h at 277 K before measuring the recovered enzymatic activity.

2.6. Site-directed mutagenesis

The ORF encoding full-length APD was inserted into the expression plasmid pET28a, which was subsequently transformed into Escherichia coli strain BL21 (DE3) to produce the recombinant enzyme. A series of constructs for mutations in the subunit were made by the PCR method. The mutations included H13A, H62A, E251A, Y129F and H195Q. The mutant proteins were purified using the same protocol as used for the native enzyme.

3.1. Overall structure and subunit fold

We crystallized the APD protein from Comamonas sp. strain CNB-1 with and without ferrous ion and in complexes with 2AP and with the suicide inhibitor 4-nitrocatechol (4NC). The structures were refined at resolutions from 2.3 to 2.7 Å (Table 2). The overall structure of APD shows an ₂₂ assembly with a crescent shape (Fig. 2a). The tetramer is composed of two tightly bound heterodimers, each of which forms an independent catalytic unit (Fig. 2b). The two subunits contact each other in a back-to-back manner, but there is a lack of direct contacts between the subunits. The and subunits of APD are related in sequence (Fig. 3a) and, as expected, display almost identical folds (Figs. 2c and 2d). Both subunits comprise ten -strands, nine -helices and two 3₁₀-helices, with the major -sheet sandwiched between two helical clusters.

Table 2 Summary of X-ray data collection and refinement Values in parentheses are for the highest resolution shell.   Fe-free APD Active APD APD-2AP APD-4NC Data collection   Wavelength (Å) 1.5418 1.0000 1.5418 1.0000   Space group C2 C2 C2 C2   Unit-cell parameters     a (Å) 269.73 265.39 270.24 272.73     b (Å) 48.26 47.47 48.39 48.26     c (Å) 108.44 107.38 108.55 108.12      (°) 109.44 109.47 109.57 109.61   Resolution (Å) 39.78-2.40 (2.53-2.40) 47.95-2.70 (2.85-2.70) 39.87-2.30 (2.42-2.30) 69.3-2.50 (2.64-2.50)   Completeness (%) 91.8 (82.5) 98.1 (93.7) 97.8 (86.5) 91.0 (81.6)   No. of unique reflections 48049 (6247) 34673 (4756) 58265 (7456) 42370 (5494)   Multiplicity 4.5 (4.4) 3.2 (2.6) 5.2 (3.9) 3.2 (2.9)   Rmerge (%) 7.3 (32.3) 10.3 (33.3) 10.9 (35.6) 4.7 (33.6)   Average I/(I) 20.6 (4.4) 11.5 (2.7) 15.7 (2.7) 18.2 (3.2) Structure refinement   Rwork (%) 18.5 17.5 19.0 20.5   Rfree (%) 24.6 24.7 23.3 27.0   R.m.s.d. bond angles (°) 1.47 1.12 1.03 1.09   R.m.s.d. bond lengths (Å) 0.007 0.007 0.006 0.007   Ramachandran plot     Core region (%) 86.6 89.8 89.8 88.2     Allowed region (%) 12.7 9.6 9.3 11.4     Generously allowed region (%) 0.4 0.6 0.7 0.2     Disallowed region (%) 0.3 0 0.2 0.1

Figure 2 The overall structure and subunit fold of APD. (a) Ribbon model of an APD tetramer with an 22 assembly. (b) Ribbon model of an dimer. (c) Fold of the subunit. (d) Fold of the subunit. The iron centre is represented by a magenta sphere.

Figure 3 Sequence and structural comparison between APD and 4,5-PCD. (a) Structure-based sequence alignment of the and chains of APD (APD-B and APD-A, respectively) and the chain of 4,5-PCD. The secondary-structure elements of APD and 4,5-PCD are labelled on the top and bottom lines, respectively. (b) Structure superimposition on the subunit of both enzymes: APD is in green and 4,5-PCD is in gold. The Fe atom is represented by a magenta sphere. (c) Superimposition of the dimeric structures, with the subunit of APD in red and that of 4,5-PCD in light blue. (d) Overlay of active sites. The model of APD is colour coded by atom (C in grey, N in blue, O in red and Fe in magenta) and that of 4,5-PCD in gold. The labels for APD and 4,5-PCD are in black and gold, respectively.

3.2. Comparison with protocatechuate 4,5-dioxygenase

In the extradiol dioxygenase superfamily, APD and 4,5-PCD are thought to be evolutionarily related, although 4,5-PCD catalyzes the cleavage of vicinal hydroxyl compounds while APD is active towards noncatechol substrates (Vaillancourt et al., 2006). Both enzymes have ₂₂-type subunit compositions. Despite the low sequence identity of below 15% between their catalytic subunits (Fig. 3a), the two enzymes basically share the same fold, with an r.m.s.d. of 1.7 Å on structure superimposition with 229 aligned C atoms (Fig. 3b). The active sites of both enzymes are located in a cleft in the subunit close to a surface extensively covered by the subunit. The iron in APD is coordinated in a distorted pyramidal geometry by residues His13, His62 and Glu251, which correspond to His12, His61 and Glu242, respectively, in 4,5-PCD.

Despite these similarities, some significant differences between APD and 4,5-PCD were observed. Firstly, APD appears to have evolved through a gene-duplication event as its two subunits are homologous in sequence, but the two subunits of 4,5-PCD are unrelated. Secondly, the quaternary structure differs between them, although both exist as stable ₂₂ tetramers. APD displays a crescent arrangement, while 4,5-PCD shows a more compact assembly (Sugimoto et al., 1999). Even so, there is a lack of direct contacts between the two subunits in both structures. Thirdly, the accessibility of substrates toward the active sites varies between the enzymes because of the different fold and orientation of the subunit in the structures. In 4,5-PCD the iron centre is completely covered by a lid composed of two helices from the subunit, while its counterpart in APD is only partially covered by the subunit (Fig. 3c). Consequently, the active site of APD is more open than that of 4,5-PCD.

3.3. The iron centre

Perfect electron density for the iron was observed in the density map of the Fe-containing structure but not in the Fe-free structure. The ferrous centre is present in the subunit only, as the subunit lacks the 2-His-1-carboxylate facial triad for iron coordination in its sequence. The Fe(II) ion is located midway in a groove between the central -sheet and a helical cluster composed of 1, 2 and 4 in the catalytic subunit (Fig. 3b) and is coordinated by His13, His62 and Glu251. Another two residues, Tyr129 and His195, are involved in forming the second shell of the iron-coordination sphere. The geometric arrangement of the Fe ligands is identical in APD and 4,5-PCD (Fig. 3d). The only notable difference is that in the second shell of the iron-coordination sphere Tyr129 in APD is replaced by a histidine residue (His127) in 4,5-PCD. The substitution at this position may at least partially account for the different substrate preferences of the two enzymes.

3.4. Ligand modelling in the APD-2AP complex structure

Although the enzyme-substrate crystal was grown with the addition of 2AP, we thought it unreasonable to model the substrate itself in the complex structure because (i) the crystal was grown under aerobic conditions and 2AP has been reported to be very susceptible to oxygen and (ii) inactivation of APD by an excess of 2AP has never been observed in enzymatic assays by ourselves or other laboratories; this is a feature that is distinct from most extradiol dioxygenases, which can be inhibited by their own substrates, e.g. HPCD and 4,5-PCD. In fact, we have observed that the enzyme remains active in catalyzing the fission of 2AP even in crystallo, albeit with a much slower rate than in solution.

After cycles of refinement of the protein model of the complex structure, ligand density in the F_obs - F_calc map became very clear at contours above 3.5. In one of the catalytic subunits (D), intact circular density was clearly observed opposite His62 and Glu251, and spherical density perfect for a superoxide ion or solvent molecule was present opposite to His13. The circular density showed imperfect planarity and the encircled contour seemed to be more appropriate to accommodate a seven-membered ring than a six-membered ring. It appears that O-O bond cleavage had already occurred since no density to accommodate a dioxygen molecule was observed around Fe(II). Considering all of these experimental observations, a lactone intermediate, 3-iminooxepin-2(3H)-one (Fig. 1f), was modelled into the density opposite His62 and Glu251. Consistent with this, a superoxide ion was placed into the density opposite His13 since one O atom from the dioxygen was already incorporated into the substrate ring (Fig. 4a).

Figure 4 Fe ligandation in the structure of APD in complex with 3-iminooxepin-2(3H)-one and 2-aminomuconic 6-semialdehyde. (a) Lactone-intermediate complex (subunit D) fitting the electron density. (b) Stick model of the lactone intermediate and superoxide binding in the Fe(II) coordination sphere in subunit D. (c) Product complex (subunit B) fitting the electron density. (d) Stick model of product and a solvent molecule binding in the Fe(II) coordination sphere in subunit B. The grey 2Fobs - Fcalc maps are contoured at 1.0 and the green Fobs - Fcalc ligand-omit maps are contoured at 3.5. Element colour indices: carbon, grey (protein) or yellow (ligand); nitrogen, blue; oxygen, red; iron, magenta. Hydrogen bonds and 2AP-Fe2+ bonds are represented by dashed lines. Numbers indicate bond distances in Å.

However, the ligand density in the other catalytic subunit (B) appeared to be different. The circular electron density shown in the F_obs - F_calc map opposite His62 and Glu251 appeared broken in comparison with that in subunit D. This density also remained discontinuous in the 2F_obs - F_calc map even after ten cycles of restrained refinement after modelling different intermediates in this subunit, which indicates that a ligand with an opened ring is more likely to be present here. Therefore, the product of the 2AP fission reaction, 2-aminomuconic 6-semialdehyde, was modelled in this site (Fig. 4c). The fit of this ligand to the density did not seem to be as good as the fit in subunit D, probably because of the higher ligand flexibility resulting from ring opening. The occupancy of this ligand was accordingly adjusted to 0.7 to maintain reasonable B factors parallel to those of the ligand in subunit D. A solvent molecule was added trans to His13 since both O atoms from the dioxygen have been incorporated into the product.

Ligand modelling in both catalytic subunits led to a significant decrease in R_free, good fitting to the electron density and reasonable Fe-ligand distances and other stereochemical parameters, all of which indicate correct modelling.

3.5. Symmetric intermediate/product binding

Presumably, the correct modelling means that the substrate has been oxidized in the crystal such that the hydroxyl substituent of 2AP has become a carbonyl group in the intermediate 3-iminooxepin-2(3H)-one and a carboxyl group in the product 2-aminomuconic 6-semialdehyde, while the amino substituent of 2AP has become an imine in both ligands. The iron is assumed to be in the ferrous state in both subunits since the enzyme apparently remains active towards its substrate. As shown in Figs. 4(b) and 4(d), both ligands bind the iron in such a way that O^2AP of the substrate takes up the position opposite to Glu251 and N^2AP occupies the site trans to His62, leaving an adjacent position trans to His13 for dioxygen binding. Almost equal Fe²⁺-O and Fe²⁺-N distances of 2.4-2.5 Å indicate symmetric binding of both ligands to the iron (Fig. 4 and Table 3), which is reasonable because of the non-ionic form of the ligands subsequent to O-O bond cleavage. In addition, the Fe-ligand distances are obviously longer than the averaged distances in the published crystal structures of extradiol dixoygenases containing deprotonated hydroxyls, yet are in good agreement with those in the case of carbonyl/carboxyl O and imine N.

Table 3 Bond distances (Å) in the Fe-coordination spheres Distance Subunit B (1) Subunit D (2) Average APD-3-iminooxepin-2(3H)-one   Fe-N His13   2.1     Fe-N His62   2.2     Fe-O Glu251   2.0     Fe-proximal Oligand#   2.4     Fe-Nligand   2.5   APD-2-aminomuconic 6-semialdehyde   Fe-N His13 2.2       Fe-N His62 2.2       Fe-O Glu251 2.0       Fe-proximal Oligand# 2.5       Fe-Nligand 2.4     APD-4NC complex   Fe-N His13 2.1 2.1 2.1   Fe-N His62 2.1 2.2 2.1   Fe-O Glu251 1.8 2.0 1.9   Fe-O24NC+ 1.8 2.0 1.9   Fe-O14NC§ 2.4 2.3 2.3 #The proximal Oligand corresponds to the hydroxyl O in the substrate, i.e. O2AP. +O14NC occupies the same binding position in the Fe-coordination sphere as N2AP. §O24NC occupies the same binding position in the Fe-coordination sphere as O2AP.

The vicinal His195 effectively stabilizes the bound ligand in both cases by forming hydrogen bonds to the imine group. However, Tyr129, another key residue involved in the second Fe-coordination shell, does not directly contact the ligands, as is the case observed in other published structures (Shu et al., 1995; Vaillancourt et al., 2006; Kovaleva & Lipscomb, 2007), where Tyr129 invariably forms a tight hydrogen bond to the deprotonated hydroxyl of the substrate. In this case, the side chain of Tyr129 forms a hydrogen bond to His62 instead (Figs. 4b and 4d). The unusual behaviour of this residue provides another piece of evidence for an almost complete enzymatic reaction.

3.6. Asymmetric inhibitor binding

In the enzyme-inhibitor structure, the suicide inhibitor 4NC fits the electron density perfectly (Fig. 5a). It binds the iron with the same orientation as the intermediate, i.e. O1^4NC (corresponding to N^2AP) occupies the Fe-coordination site trans to His62 and O2^4NC is positioned opposite Glu251. In this complex structure, 4NC binds Fe in an asymmetric manner, as the distances O1^4NC-Fe and O2^4NC-Fe are unequal (Fig. 5b and Table 3). The O2^4NC-Fe bond length (1.9 Å) clearly indicates the deprotonated state and the monoanionic form of O2^4NC. A tight hydrogen bond between O2^4NC and OH^Tyr129 is formed to stabilize the monoanion of 4NC (Fig. 5b), which is in good agreement with published structures in which asymmetric binding occurs as a result of deprotonation of one hydroxyl group of the bound ligand (Kovaleva & Lipscomb, 2007, 2008a; Sato et al., 2002). Consequently, a 45° rotation of the Tyr129 side chain between the structures is revealed on the superimposition of the two complex structures (Fig. 5c), indicating a high degree of flexibility and an important role of this amino acid in the process of enzyme turnover.

Figure 5 The complex trapped in the active sites of the APD-4NC structure. (a) 4NC fitting the electron density. (b) 4NC binding the iron oxidized from the ferrous state to the ferric state. Hydrogen bonds and the 4NC-Fe3+ interaction are represented by dashed lines. Numbers indicate bond distances in Å. (c) The Fe-coordination sphere with overlaid ligands: intermediate, product and 4NC. The APD intermediate model is denoted with atom-indexed colours; the product ligand is in cyan and the APD-4NC model is in green.

In contrast to the enzyme-substrate complex, the iron in the APD-4NC structure is presumed to be oxidized to Fe(III) because (i) the Fe-ligand distances are remarkably shorter than those in the other structures (Table 3) and conform to reported Fe³⁺-ligand distances (Sugimoto et al., 1999) and (ii) inactivated APD recovers its enzymatic activity to 100% upon incubation with Fe(II) and a reducing agent (see below).

3.7. Enantiomeric Fe coordination

Although we did not straightforwardly observe substrate binding based on electron density, the binding mode of the substrate could indubitably be deduced from the complex structures determined in this study. Like the substrates of all other extradiol dioxygenases, 2AP would bind Fe in an asymmetric manner upon deprotonation and ionization of its hydroxyl substituent. The substrate binding in APD would be in such a way that -O^- occupies the position trans to Glu251 and the -NH₂ group resides at the site trans to His62, leaving an adjacent site trans to His13 for dioxygen binding (Fig. 6a).

Figure 6 Comparison of Fe-ligand coordination geometry at the active site prior to O2 binding in some 2-His-1-carboxylate enzymes. (a) The APD-4NC complex structure, (b) the 4,5-PCD structure (PDB entry 1b4u ; Sugimoto et al., 1999), (c) HPCD (PDB entry 2iga ; Kovaleva & Lipscomb, 2007) and (d) the -KG-dependent enzyme clavaminate synthase (CAS--KG-PCV; PDB entry 1drt ; Zhang et al., 2000). Schematic representations of the Fe-coordination scheme are shown in the middle and simplified representations showing Fe coordination with two O- anions from the carboxylic side chain of the Glu residue and a hydroxyl substituent of the bound ligand are shown at the bottom.

By comparison with published structures of 2-His-1-carboxylate enzymes, the three-dimensional arrangement of ligand binding to Fe in APD differs from that in all other enzymes except for 4,5-PCD (Fig. 6b). The configuration of Fe coordination shown in these two enzymes are effectively enantiomeric to those in other extradiol dioxygenases, e.g. HPCD (Fig. 6c), and -KG-dependent enzymes, e.g. clavaminate synthase (Fig. 6d). One His and the Glu residue chelating Fe(II) exchange their locations in the APD and 4,5-PCD structures with respect to the other enzymes, forming a striking difference between these structures. Owing to this ligand-position exchange, the deprotonated hydroxyl of bound compounds is positioned uniquely trans to a Glu in APD and 4,5-PCD, rather than a His as in all other 2-His-1-carboxylate enzymes (Fig. 6, middle). As a result, the two monoanions coordinating Fe before O₂ binding, one from Glu and the other from the ligand, are aligned so as to form a linear O^--Fe²⁺-O^- species in the minor group of extradiol dioxygenases represented by APD and 4,5-PCD (Fig. 6, bottom). In addition, the site for dioxygen binding is opposite a His in this group, which differs from HPCD but is identical to -KG-dependent enzymes. All of these variations highlight the versatility of the 2-His-1-carboxylate platform that is shared among mononuclear nonhaem Fe(II) enzymes.

3.8. Kinetics assays

In kinetics measurements, K_m and V_max of 2AP and catechol for APD were calculated from Lineweaver-Burke plots (Fig. 7) to be 17.1 µM and 19.0 µM min^-1 mg^-1, respectively; in comparison, those of catechol were 6.84 µM and 1.12 µM min^-1 mg^-1, respectively. These data indicated that catechol has a threefold higher affinity for APD than 2AP, but that its catalytic efficiency was one order of magnitude lower than that of the latter. The results obtained in our experiment agreed well with previous reports for APDs from other sources (Takenaka et al., 1997; Davis et al., 1999). The K_m values for oxygen with the above compounds were also measured, but no remarkable variance was observed between them. K_{m, O2} under 2AP and catechol incubation was 77.5 ± 5.1 and 58.3 ± 9.6 µM, respectively, which indicated that 2AP binding to the iron centre increases its affinity for O₂ with a comparable efficiency to monoanionic catechol binding.

Figure 7 Lineweaver-Burke plots of APD with 2AP (a) and catechol (b) under aerobic conditions, as well as those with a controlled O2 concentration from 0 to 1200 µM using 2AP (c) and catechol (d) as the substrates. Assays were carried out in phosphate buffer at pH 8.0. The concentration of purified APD was 0.015 mg ml-1.

3.9. Activity inhibition and recovery

It has been reported that the enzymatic activity of APD from Pseudomonas pseudoalcaligenes JS45 (APD_JS45) or Pseudomonas sp. AP-3 (APD_AP-3) catalyzing the ring opening of 2AP was inhibited in the presence of catechol analogues (Davis et al., 1999; Takenaka et al., 1997). In agreement with these published observations, we found that APD_CNB-1 completely lost its activity towards 2AP after incubation with catechol or 4NC for 100 s. The catalytic activity towards 2AP, however, was restored to 100% after removal of the catechol by applying the reaction mixture to a desalting column and subsequent incubation with 2 mM Fe²⁺ and 5 mM ascorbate on ice. These data provided biochemical evidence for the oxidation of Fe(II) to Fe(III) in enzyme inhibition.

3.10. Site-directed mutagenesis

Key amino acids involved in Fe coordination include His13, His62 and Glu251 in the first shell and Tyr129 and His195 in the second shell (Figs. 4 and 5). As expected, any single mutation of these residues led to a complete loss of enzymatic activity towards 2AP. Among the effective mutants, Y129F is particularly noteworthy as it highlights the important role of this residue as a hydrogen donor. The abolition of catalytic activity on its mutation implies a possible role of this residue in the deprotonation of the hydroxyl group of 2AP and the formation of a hydrogen bond between O^2AP and OH^Tyr129 during enzyme turnover, although this was not observed in crystallo.