The genome of P. putida MET94 has not been sequenced to date. Sequence alignments against a nonredundant protein database with blastp  have shown the highest homology with azoreductases from two strains of P. putida: P. putida GB-1 azoreductase R1 (98% identity) and P. putida KT2440 azoreductase R1 (99% identity). To our knowledge, the numbering of orthologue types follows the convention described by the Pseudomonas aeruginosa Community Annotation Project (from the Pseudomonas Genome Project ) and are indexed as ordered in the genome. This means that caution is necessary when comparisons are made: P. aeruginosa, for instance, was reported to have three genes encoding proteins with azoreductase activity (paAzoR1, paAzoR2, and paAzoR3) . Sequence alignment showed that the highest homology with P. putida azoreductase was with P. aeruginosa azoreductase R2 (PaAzoR2) and not with P. aeruginosa azoreductase R1 (PaAzoR1). Because we have no knowledge regarding the genome of P. putida MET94, we prefer not to attribute a number classification to the enzyme studied herein, and refer to it simply as PpAzoR.
Overall structure of the native enzyme
The production and purification of recombinant PpAzoR was performed as previously described . The purified PpAzoR was crystallized under aerobic conditions. The crystal structure obtained belongs to space group F222, containing one monomer (21.4 kDa) in the asymmetric unit. The homodimer indicated by the size-exclusion chromatography experiments reported previously  is generated by a twofold crystallographic symmetry axis. At the dimer's interface, a riboflavin 5′-phosphate (FMN) molecule was modelled into well-defined electron density as revealed by the Fourier difference map. The crystal structure accounts for the majority of the protein sequence, with the exception of the final three residues (Ala201, Ala202, and Ala203), for which no electron density was visible. The overall structure of PpAzoR and a diagram showing the secondary structure elements aligned with the amino acid sequence are shown in Fig. 1A,B. The monomer adopts a flavodoxin-like fold, with a central twisted β-sheet formed by five parallel β-strands (by order: β2, β1, β3, β6, and β7) connected by α-helices. Helices α1 and α6 are located on one side of the central sheet, and on the opposite face are α2, α3, α4, and α5. Sheets β4 and β5 belong to a protrusion that completes the active site of the cognate monomer. Superimposition over all Cβ atoms, with the SSM Superpose algorithm (http://www.ebi.ac.uk/msd-srv/ssm/cgi-bin/ssmserver), of the PpAzoR structure over the structure of the azoreductase from Escherichia coli (EcAzoR) revealed that the main chain carbons of 188 of the 200 residues overlapped with an rmsd of 1.36 Å [Protein Data Bank (PDB) code 2Z98] . The secondary structural arrangement is equally conserved with other azoreductase structures reported: P. aeruginosa (rmsd of 1.48 Å; PDB code 2V9C ), Enterococcus faecalis (rmsd of 1.52 Å; PDB code 2HPV ), and Salmonella typhimurium (rmsd of 1.43 Å; PDB code 1T5B). In the native PpAzoR structure, loop 4, connecting helices 2 and 3, is fully modelled, in contrast to the unstructured equivalent loop 4 in the E. coli and S. typhimurium structures.
Figure 1. Overall structure of PpAzoR1 and amino acid sequence. (A) Amino acid sequence of native PpAzoR with secondary structure elements displayed above the sequence. Residues involved in the binding of the FMN prosthetic group are in bold. (B) Overall structure of the PpAzoR monomer as a ribbon diagram; secondary structure assignments are labelled on the model; the FMN molecule and the poly(ethylene glycol) 400 molecule are in stick representation. (C) Representation of the PpAzoR dimer oriented along the crystallographic twofold axis. The two subunits are in blue and orange. The prosthetic FMN molecules are represented by a stick model, with carbon atoms in yellow, oxygen atoms in red, nitrogen atoms in blue, and phosphorus atoms in orange. The poly(ethylene glycol) 400 molecules (one in each active site, stacked over the FMN) are in stick representation, with carbon in green and oxygen in red. (D) A side view of the dimer rotated by 90º relative to the twofold axis.
Download figure to PowerPoint
The dimer contains two separate active sites that are located at the interfaces between the two monomers (Fig. 1C,D). The interface involves antiparallel side-to-side packing between helices α4 and α5 of each monomer, corresponding to residues 97–109 and 148–164, respectively. It also includes residues 43–51, composing loop 3 (preceding α2), that interact antiparallel with the same stretch of residues on the other monomer. The contact area corresponding to the dimer interface is 1193 Å2, strongly suggesting that this configuration is the biologically relevant dimer. Both monomers contribute as a scaffold for the catalytic centre, supporting the idea that dimerization is essential for biological function. Inspection of the dimer interface revealed a complex net of hydrogen bond interactions (Table 1). Helix α4 is stabilized by a hydrogen bond between residues Asn97 from monomer A and Asp109 from monomer A′, and the reciprocal pair at the other extreme of the helix. The dimer interface also benefits from intermolecular hydrogen bonds from numerous water molecules. In fact, the solvation free energy gain upon assembly (ΔGint), calculated by the program pisa (http://www.ebi.ac.uk/msd-srv/prot_int/pistart.html) to be – 39.6 kcal·m−1, and the calculated Gibbs free energy of dissociation (ΔGdiss), with a value of 12.3 kcal·m−1, are both indicative of a true interface upon dimer formation.
Table 1. Intermonomer interactions of PpAzoR.
|Source residues/atoms||Target residues/atoms||Distance (Å)|
|Ile47/A O||His49/A′ Nε2||2.78|
|His49/A Nε2||Ile47/A′ O||2.78|
|His49/A Nδ1||Thr102/A′ (Oγ1)||2.67|
|Phe50/A (O)||Gln103/A′ (Nε2)||2.98|
|Asn97/A (Nδ2)||Asp109/A′ (Oδ1)||2.85|
|Thr102/A (Oγ1)||His49/A′ (Nδ1)||2.67|
|Gln103/A (Nε2)||Phe50/A′ (O)||2.98|
|Asp109/A (Oδ1)||Asn97/A′ (Oδ1)||2.85|
Each active site contains two redox-active nitrogen atoms of the prosthetic FMN molecule in the oxidized state, assumed from the yellow colour of the protein crystals. Approximately 75% of the FMN molecule is buried in the protein, nested between loops 7 and 11 on the C-terminal end of the central β-sheet, and the isoalloxazine ring is near to planar, adopting a slightly twisted conformation, similarly to what has been described for EcAzoR [15, 17]. The majority of residues coordinating the FMN are very conserved as compared with those involved in the prosthetic group binding of EcAzoR. In brief, the phosphate moiety is well anchored into a pocket comprising residues between β1 and α1, and Tyr96 (Fig. 2). This loop, between β1 and α1, is a typical invariant feature in the flavodoxin fold, bearing a key fingerprint motif [(T/S)XTGXT] and being responsible for interaction with the FMN phosphate moiety  (Fig. 1A). Although this motif has greatly diverged among the known and characterized azoreductases, the functional homology remains. As compared with EcAzoR, Ala16 is the only nonconserved residue in this loop, where the binding occurs through a hydrogen bond between FMN/O1P and the N of the peptide backbone. Otherwise, the remaining features of FMN binding reproduce what is seen for EcAzoR: Ser17 hydrogen bonds to O3P through the backbone N and its side chain Oγ; O3P also interacts with the Oγ of Ser15. O2P interacts with Oγ of Ser9, Cδ1 of Leu11 (and the OH group of Tyr96); and, finally, Cα of Ser15 interacts weakly with O1P. The ribityl moiety forms hydrogen bonds with Thr139 (Oγ with O5′), with Met95 (O with O2′), and with Gly141 (N with O2′). The C7M and C8M region of the FMN is surrounded by a hydrophobic patch composed of Ile10, Leu11 and Tyr96 and Phe50′ and Leu55′ from the complementary monomer. It is curious that, in all three crystal structures of PpAzoR presented here, the Tyr96 is slightly rotated, not presenting parallel packing with the isoalloxazine moiety of FMN, a common detail in many reported azoreductases. Regarding the isoalloxazine region, O2 promotes electrostatic interactions with Gly142 (N); O4 hydrogen binds to Phe98 (N); N1 hydrogen binds to Gly141 (N); and N5 hydrogen binds to Asn97 (N) (Fig. 2). Gly141 and Gly142 are believed to have the additional role of stabilizing the FMNH− intermediate during the enzymatic reaction . It is also worth mentioning the importance of a very conserved lysine, Lys105, which is involved in stabilizing a very strained loop (Pro94–P101) located between β3 and α4. This structurally conserved loop among azoreductases contains a type I′ reverse turn against which the FMN isoalloxazine group is packed, and has been termed the ‘FMN cradle’ [20, 21]. Lys105 stabilizes the loop through hydrogen-bonding interactions between the backbone carbonyl group of the symmetry-related Tyr96′. In turn, the side chain of Lys105 is in a stable conformation (residue average B-factor, 13.03 Å2) owing to a salt bridge with a conserved Asp109. In the native structure of PpAzoR, His144 is turned away from the isoalloxazine ring and does not establish a hydrogen bond between its Nε2 and O4 on the FMN, as seen in EcAzoR (PDB code 2Z98) and S. typhimurium azoreductase (PDB code 1T5B).
Figure 2. Interactions between FMN and amino acids in the active site: schematic diagram showing contacts of the FMN cofactor with surrounding amino acids. The FMN and protein side chains are shown in ball-and-stick representation, with the FMN bonds in purple. Hydrogen bonds are shown as broken green lines; the spoked arcs represent amino acids making nonbonded contacts with the ligand. Each atom element is represented by a sphere of different colour, with a chemical symbol.
Download figure to PowerPoint
The active site: comparison and conformational differences
The native form of PpAzoR was refined to a final value of Rwork = 15.3% (Rfree = 17.9%). In contrast to what was expected, the active site is occupied by a poly(ethylene glycol) 400 molecule. This molecule is positioned above the si-face of the FMN molecule in what has been identified as being the position for substrates. The overall structural similarity found with EcAzoR (PDB code 2Z98) includes many details of the active site. As in EcAzoR, the surroundings of the FMN si-face exposed to the solvent are quite hydrophobic in nature. These hydrophobic residues, provided by both monomers of the functional dimer, are very conserved, both sequentially and structurally (Phe98 and Tyr179; and, from the cognate monomer, Phe50′, Leu55′, Phe118′, Tyr120′, Phe160′, and Phe163′), with the exception of Leu55′. This residue, although sequentially aligned with Leu54 in EcAzoR (not shown here), is structurally aligned with a valine (Val55) in the E. coli structure, owing to an insertion in PpAzoR in this region.
In spite of these striking similarities, there are a few differences in the vicinity of the active site that may influence substrate specificity. The most noteworthy of these concerns the loop between helix 2 and helix 3, spanning residues Gly59 and Ala67 (corresponding in EcAzoR to Arg59–Pro67) (Fig. 1A). In PpAzoR, this loop is bent away from the active site, in contrast to the equivalent loop in E. coli, where some residues offer the possibility of interacting with a substrate . The three residues involved in the turns of their respective loops are ~ 11 Å from each other (Fig. 3). This loop was initially thought to be mobile in EcAzoR, as it was not entirely traceable in the structures that were first determined. Later, in a different space group (PDB code 2Z9D), the full traceability of the loop revealed it to be static, remaining in an identical conformation to that observed for the first E. coli native structure (average B-factor of 31.01 Å2). The equivalent loop in PpAzoR has a similar B-factor value (average value of the backbone Cα, 24.38 Å2), and is not expected to be mobile either, as is confirmed in the structures complexed with AQS and RB5. Nevertheless, this feature is enough to confer a different topology to the active site of PpAzoR, regarding both size and access, and possibly affecting product release. A further clue supporting the importance of the loop comes from another reported structure of an azoreductase, PaAzoR1 (PDB code 2V9C ). This structure shows the equivalent loop (Phe60–Ser68) more proximal to the FMN prosthetic group, resulting in a smaller active site, which is unable to accommodate bulky substrates (Fig. 3).
Figure 3. Topology of the active site: zoom-in view of the active site of native PpAzoR (in dark blue with no transparency). The FMN molecule is in stick representation: C, yellow; N, blue; O, red. The symmetry-related molecule composing the dimer is in faded dark blue, and superimposed with EcAzoR (light yellow, PDB code 2Z9D) and PaAzoR1 (light green, PDB code 2V9C), both equally faded. The divergent loops are indicated by arrows.
Download figure to PowerPoint
Another pivotal feature of the active site topology is the flap comprising β-sheets 4 and 5. This structural element (from the second monomer) places itself over the (noncovalently bound) FMN of the first monomer, and helps to assemble the pocket of the active site. Superimposition of the PpAzoR structure with that of EcAzoR shows these elements to be entirely superposable; however, with PaAzoR1, the overlay shows a shorter flap in PpAzoR, producing a more open and solvent-accessible active site (Fig. 3). This structural element contains an important residue, Tyr120, that aligns structurally with Tyr120 from EcAzoR, and corresponds functionally to Tyr131 in PaAzoR1, and was suggested to play a role in substrate specificity and the architecture of the catalytic pocket [22, 23].
The structure of PpAzoR with AQS bound was solved at 1.50-Å resolution by molecular replacement with the native structure, void of all ligands, as a search model. The structure was refined to final Rwork and Rfree values of 16.9% and 19.3%, respectively (Table 2). A clear positive difference density was visible for the AQS molecule bound within the active site after a single round of refinement with buster  (chemical structure shown in Fig. 4A). AQS was easily built in this density, revealing interactions with the FMN prosthetic group and residues from both monomers (Fig. 4B,C). To investigate the conformational changes of the macromolecules, we superimposed the structure obtained from the cocrystallization of AQS with the native form of PpAzoR. This superimposition demonstrated that the fold of the peptide chains and the FMN environment are strongly maintained throughout the PpAzoR_AQS structure, and no significant conformational changes can be assigned upon comparison with the native form (rmsd on 200 Cα positions, 0.18 Å). A focus on the active site shows, however, that the side chains of some residues have different conformations to accommodate substrate binding. The plane of AQS stacks parallel with the isoalloxazine ring of FMN, a result of the π-stacking interactions between both molecules, with an average distance of 3.6 Å (Fig. 4B). This distance, in particular the distance between N5 of the FMN and the oxygen from the carbonyl (O13) of the anthraquinone ring (3.6 Å), has been shown to be sufficient for hydride transfer [25-27]. Other possible acceptors of the hydride in the quinone substrate are C13 (3.6 Å), C12 (3.9 Å), and C14 (4.0 Å). Our results suggest that the AQS is correctly positioned for the reductive catalysis. This particular orientation of AQS involves interactions with side chains of residues that slightly differ from those previously identified in substrate binding in other azoreductases and quinone oxidoreductases. The conserved Gly142 establishes hydrogen bonds (3.2 Å) with the sulfonate group of AQS, and not Gly141, which has been shown to be the key residue for both azo substrate binding in PaAzoR1  and for NADH binding in EmoB from Mesorhizobium sp. BNC1 . Another conserved hydrophobic residue that is usually involved in contacts with the AQS substrate is Phe163′, which, along with Phe118′ and Phe98, form a very hydrophobic patch in the active site, and have been identified in NAD(P) binding [25, 26]. Ala178, located in a turn of the loop joining β7 and α6, is also involved in hydrogen bonding with the substrate. This residue falls within the turn of the loop, and in its vicinity is Tyr179, which may establish contacts with bulkier substrates.
Table 2. Crystal parameters, X-ray data collection and refinement statistics of native PpAzoR, and of PpAzoR in complex with RB5 or AQS. (Numbers in parentheses are for the highest-resolution shell.).
| ||Native form||RB5||AQS|
|Beamline and detector||ID14-eh4/ADSC Q315R||ID23-eh1/ADSC Q315R||ID14-eh1/ADSC Q210|
|Space group|| F222 || F222 || F222 |
|Cell dimensions (Å)||a = 72.30, b = 95.74, c = 146.54||a = 72.67, b = 94.97, c = 146.95||a = 72.74, b = 95.95, c = 146.52|
|Data collection statistics|
|Resolution (Å)||53.7–1.60 (1.69–1.60)||73.5–1.90 (2.00–1.90)||47.5–1.50 (1.58–1.50)|
|No. of unique reflections||33 249 (4846)||20 109 (2905)||40 377 (5894)|
|Rsym (%)||5.1 (32.1)||8.1 (32.9)||6.3 (28.9)|
|I/σ (I)||8.8 (2.4)||5.3 (2.3)||8.5 (2.7)|
|Completeness (%)||99.1 (100.0)||99.6 (99.6)||99.4 (99.9)|
|Redundancy||4.3 (4.4)||7.1 (4.3)||7.7 (3.7)|
|Refinement and model stastistics|
|No. of reflections used (working set/free set)||33 239/1683||20 092/1026||40 373/2021|
|No. of molecules/average B-factor (Å2)|
|Amino acid residues||200/16.7||200/30.5||200/25.4|
|Poly(ethylene glycol) 400||2/38.7||1/56.1||2/35.5|
|rmsd bond length (Å)||0.010||0.005||0.018|
|rmsd bond angle (º)||1.280||1.687||1.849|
|Ramachandran statistics (%)|
|Additional allowed region||1.9||2.0||1.9|
Figure 4. Active site of PpAzoR occupied by AQS. (A) Chemical structure of AQS (drawn with chemdraw pro13.0, Perkin Elmer, Waltham, MA, USA). (B) Representation of the substrate-binding active site, highlighting the interactions between AQS (purple) and several residue side chains. His144 and Tyr120′ are in double conformation. The fragments of refined poly(ethylene glycol) 400 built into the model are not represented, for the sake of clarity. The blue mesh represents the initial 2Fo − Fc map contoured at 1σ. The FMN molecule is in yellow. (C) Representation of the PpAzoR active site in complex with AQS and poly(ethylene glycol) 400 fragments (in stick representation: C, orange; O, red); each conformation of the Tyr120 side chain is labelled in accordance with the poly(ethylene glycol) 400 fragment to which it is related.
Download figure to PowerPoint
In PpAzoR, His144 shows a mobility that has not been found in other structures with the same conserved residue (E. coli and S. typhimurium) (Fig. 4B). Whereas, in the native form of PpAzoR, the side chain is positioned such as to undergo no interactions with the FMN, with the AQS substrate in place the side chain rotamer is bound to both the FMN and the sulfonate group of AQS. The His144 rotamer (‘away’ from the FMN) found in the native form undergoes an interaction with a water molecule (W1 in Fig. 4B) bridged between itself and Tyr179 (3.0 and 2.1 Å, respectively), and seems to help stabilize the mobile side chain. This rotamer of His144 is still able to interact with the sulfonate group of the AQS substrate. When the active site is occupied, Nε2 of His144 (now pointing towards the FMN) possibly stabilizes the electrophilic character of the semiquinone formed during catalysis, either by donating a proton to O2 of the isoalloxazine or by stabilizing the delocalized negative charge around the N1-C2=O2 region by the positive charge of the imidazole ring . Both His144 rotamers are further stabilized by the neighbouring Thr149, owing to the different rotamer conformations adopted. Another element of interaction suggested to be involved in semiquinone stabilization is the electrostatic interaction between the FMN N1 and the backbone amide N of G141, as has been suggested before [15, 27]. For small quinone substrates, the latter would be the most relevant.
Asn97 has one of the very few polar side chains in the active site. In PpAzoR, the backbone N of this residue undergoes a van der Waals interaction with FMN N5. This has been observed for the four other reported azoreductases from E. coli, P. aeruginosa, S. typhimurium, and En. faecalis. In the structure of the flavin oxidoreductase EmoB, the equivalent residue is Lys81, whose backbone N is also coordinated to the FMN, while the long side chain is available for substrate [or NAD(P)H] hydrogen interactions through Nζ . In PpAzoR, Asn97 is too distant to interact with the AQS substrate; however, it is bound to a well-defined water molecule (W2) (2.9 Å, B-factor of 37 Å2), which, in turn, is bound to O5 of AQS (2.9 Å). This water, one of the very few in a highly hydrophobic catalytic environment, may provide a proton to O5 during substrate reduction by the semiquinone. In the PpAzoR native form, this water is not present, and Asn97 interacts directly with the poly(ethylene glycol) 400 molecule, suggesting the same possibility for bulky substrates. Furthermore, the oxidized structure of EcAzoR also had a water molecule in the same position, bound to Asn97, and this has been suggested to hydrogen bond to N5 of the isoalloxazine upon reduction, helping to stabilize this region .
An unexpected feature in the PpAzoR_AQS structure is the simultaneous presence of poly(ethylene glycol) 400 and AQS in the active site (Fig. 4C). This is undoubtedly a crystallization artefact related to the crystallization condition used. The poly(ethylene glycol) 400 fragment built in the residual positive density is displaced by the substrate, AQS, to the top of the catalytic pocket, further away from the FMN cofactor, as compared with the native structure. The fragment of the poly(ethylene glycol) 400 molecule stacks over the AQS molecule, and pushes Tyr120′ into an alternative rotamer, unseen in all reported azoreductase structures to date (Fig. 4B,C). Alternatively, the poly(ethylene glycol) 400 molecule is further displaced to the side, so as to allow the side chain of Tyr120′ to resume its common orientation relative to the substrate. This residue has been stated to be relevant for the architecture of the active site , and here, indeed, we prove its plasticity, supporting arguments in favour of its role in defining the active site configuration and selection for planar substrates able to π-stack with the FMN cofactor.
The crystal structure of PpAzoR in complex with RB5 was determined to 1.90-Å resolution, and was refined to an RWORK of 18.5% and Rfree of 21.4%, with good stereochemistry (Table 2). Commercially known as Reactive Black 5 or Remazol Black (IUPAC Name: tetrasodium;(6E)-4-amino-5-oxo-3-[[4-(2-sulfonatooxyethylsulfonyl) phenyl]diazenyl]-6-[[4-(2-sulfonatooxyethylsulfonyl)phenyl] hydrazinylidene]naphthalene-2,7-disulfonate), this compound is a double azo dye with four phenyl groups (one substituted naphthalene and two phenyl groups) (Fig. 5A). A clear positive difference density was visible within the active site after a single round of refinement with buster , and RB5 was built into this density, in two different orientations, in order to fully satisfy the positive electron density (Fig. 5B,C). Entirely unexpected was the covalent modification of the cofactor FMN by RB5. Inspection of the maps unequivocally showed that the oxygen (OA1, highlighted in Fig. 5A) from the sulfonate group bound to the alkyl group had reacted with N3 from FMN, forming a covalent bond. RB5 is used in the textile industry to colour cellulosic fibres, and, therefore, its reactive groups are able to form covalent bonds with hydroxyl groups on the fibre . In order to understand what may have promoted this modification, MALDI-TOF/TOF MS experiments were carried out on protein crystals cocrystallized with RB5 (Fig. 6A,B). Analysis of the crystallized complex showed m/z peaks at 21 337, 986 and 906 Da. The first value is close to the theoretical molecular mass of PpAzoR, 21 408 Da. The other two values should correspond to the free dye (expected molecular mass: 991.8 Da) and, possibly, a degradation product of the dye resulting from the cleavage and loss of two SO32− groups (approximate molecular mass: 911.8 Da). We assume that the covalent modification occurred after irradiation of the crystal sample during data collection, as MS analysis of preirradiated protein–RB5 crystals did not show any peaks corresponding to the mass of modified FMN (calculated to be ~ 1450 Da; Fig. 6B).
Figure 5. Active site of PpAzoR occupied by RB5. (A) Chemical structure of RB5; the arrow shows the nucleophilic oxygen that binds to the secondary amine, N3, on the FMN prosthetic group; the broken lines suggest the loss of the sulfonate groups by chemical degradation as observed by MS analysis (drawn with chemdraw pro13.0). (B) Representation of the substrate-binding active site, highlighting the interactions between the FMN chemically modified by RB5 (purple) in conformation A: for clarity of view, the part of the molecule corresponding to the FMN is in yellow, and the part corresponding to RB5 is in purple. The indigo mesh represents the initial 2Fo − Fc map contoured at 1σ. (C) Same representation as before, but with RB5, conformation B, in a different orientation. The blue mesh represents the initial 2Fo − Fc map contoured at 1σ.
Download figure to PowerPoint
Figure 6. MS of PpAzoR cocrystallized with RB5. (A) Spectrum at high mass (15 000–70 000 Da) showing the peak corresponding to protein (free of cofactor). (B) Spectrum at low mass (800–3000 Da) showing the two peaks corresponding to intact and degraded RB5.
Download figure to PowerPoint
Superimposition of PpAzoR_RB5 with the native form of PpAzoR produced an rmsd (over all 200 Cα positions) of 0.25 Å, showing that, in spite of the bulkiness of the substrate, the enzyme backbone is quite rigid, including the loops from both monomers that compose the active site. The surprising affinity of PpAzoR for bulky substrates allows the identification of novel structural elements involved in substrate recognition. Both conformers (A and B) are covalently bound to the FMN by the same oxygen; one conformer is rotated by 180° relative to the other around the first azo double bond following the covalent bond to the FMN, thus maintaining the same planarity relative to the isoalloxazine. In both conformers, the substituted naphthalene group does not stack with the FMN, which is similar to what was observed for the anthroquinone in AQS. Rather, the phenyl group bound to the azo link stacks with the FMN with an average distance of 3.5 Å. Note that, in the second sulfonatooxyethylsulfonyl moiety, lack of electron density compromised the full representation of this group in the model, resulting in a truncation after the first carbon of the ethyl group.
Most of the residues previously identified in ligand binding also participate in RB5 binding in both conformers. Focusing first on the interactions common to both conformers, where they are superimposable, near the covalent bond to the FMN: Tyr120′ is oriented in an almost perpendicular fashion towards the phenyl group, and hydrogen bonds with a carbon of the aromatic ring; the ethyl group (bound to the FMN) undergoes hydrophobic interactions with the side chains of Phe98 and Phe163′; and finally, the reactive oxygen that covalently binds to FMN/N3 interacts with the sulfur from Met95 (Fig. 5B). Leu11 and Asn14 interact with one of the sulfonate groups (conformers A and B). The other sulfonate group (in conformer A) interacts with Gly141 (Cα), Leu177 (C=O), Ala178 (Cα and Cβ) and, unexpectedly, through a weaker electrostatic interaction (3.6 Å), with Arg184. In conformer B, this same sulfonate group is exposed to a solvent area between helix 2/loop 4 and loop 8 (preceding the β4/β5-flap) and is too distant to establish interactions with neighbouring residues (Fig. 5B,C). The most striking differences arise from the contacts between the ‘free’ sulfonatooxymethyl group and the surrounding loops. Different contacts occur, depending on the orientation of the conformer, but both reveal novel interactions with structural elements that have never been previously identified to be involved in substrate binding. In RB5 conformer A, the sulfonatooxymethyl branch interacts with Leu55′ and Val56′ from α2′ and Gly59′ from loop 4′. In the PaAzoR1 structure, the equivalent elements are closer to the active site, limiting the substrate size (Fig. 3); superimposition of PpAzoR_RB5 with EcAzoR (PDB code 2Z9C) shows that clashes with RB5 conformer A would arise in this region (not shown). The same group of RB5 in conformer B is directed to another region, nearer α6 (in the vicinity of Arg184) and loop 13, and interacts with Asn14 (Fig. 5C). Close observation of Arg184 has drawn attention to it as possibly having a role in bulky substrate binding. In particular, inspection of the superimposition of the EmoB structure in complex with NADH (PDB code 2VZJ) with PpAzoR_RB5 shows that loop 1, α-helix 2 (from the second monomer) and loop 13/α-helix 6 are potentially involved in NAD(P)H recognition. In particular, Arg184, which is unique to PpAzoR, may be involved in binding O3 of the ribose moiety, similarly to what is seen for EmoB Gln144. Overall, our results confirm that the topology surrounding the active site contains novel features for substrate recognition and binding. This, in turn, explains the substrate specificity found among different bacterial azoreductases, particularly the difference proposed between PaAzoR1 and PaAzoR2 . As mentioned earlier, PpAzoR aligns best with PaAzoR2, and the structures presented here offer experimental evidence for why PpAzoR (and consequently PaAzoR2) are more efficient catalysts with bulky substrates.
Catalytic properties of PpAzoR
PpAzoR was previously shown to reduce several structurally different azo dye substrates in the absence of oxygen . Two synthetic quinone substrates, 1,4-benzoquinone (1,4-BQ) and AQS, were tested as PpAzoR substrates in order to evaluate the quinone oxidase activity of this enzyme (Table 3). PpAzoR has the interesting feature of using both NADH and NADPH as electron donors, with catalytic efficiencies of approximately the same order of magnitude (Table 3). 1,4-BQ shows higher specificity (kcat/Km) than AQS (a larger substrate), with values similar to those determined for other quinone reductases such as Lot6p  and ChrR . Intriguingly, the recombinant azoreductase from Rhodobacter sphaeroides overexpressed in E. coli did not show any detectable activity with 1,4-BQ, whereas the kinetic parameters for AQS are within a similar order of magnitude as those observed for PpAzoR . Also consistent with our values are those determined for EcAzoR with NADH as electron donor, with a kcat of 1.1 ± 0.3·s−1 and kcat/Km of 1.36 × 105 m−1·s−1 . It is not clear to us why the R. sphaeroides azoreductase is not active on 1,4-BQ; no three-dimensional structure is yet available. However, from the results based on the PpAzoR_AQS structure, and Lot6p from Saccharomyces cerevisiae (PDB code 1TOI) , binding of a substrate as small as 1,4-BQ would involve hydrogen bonding with Ala178 (Val157 for Lot6p), the water bridged between Asn97 (Asn96 for Lot6p) and O5 of the quinone ring, and Tyr120′ (this feature is lacking in Lot6p).
Table 3. PpAzoR kinetic constants for azo substrates (with NADPH) and quinones with NADPH or NADH as electron donor.
|Substrates||Vmax (U/mg)||Km app (mm)||kcat app (s−1)||kcat/Km (m−1·s−1)||Reference|
|Methyl red||1.0 ± 0.04||0.4 ± 0.1||1.0 ± 0.04||3 × 103||  |
|RB5||2.2 ± 0.10||1.4 ± 0.2||2.2 ± 0.06||1.4 × 103||  |
|1,4-BQ/NADPH||117 ± 7||0.08 ± 0.01||114||14 × 105||This study|
|AQS/NADPH||49 ± 3||0.07 ± 0.01||48||7 × 105||This study|
|1,4-BQ/NADH||73 ± 4||0.09 ± 0.02||71||8 × 105||This study|
|AQS/NADH||29 ± 3||0.08 ± 0.02||28||4 × 105||This study|
The steady-state kinetic analysis of PpAzoR in the presence of NADPH and 1,4-BQ or AQS resulted in a family of parallel lines in a double reciprocal plot (Fig. 7), as previously observed for RB5 . This is indicative of ping-pong bi-bi (double displacement) kinetics, where a unique catalytic site is used for both substrates, as has been described for many other azoreductases.
Figure 7. Double reciprocal plots of 1/V0 against 1/[NADPH] and 1/[1,4-BQ]. (A) 1/[NADPH] (white triangle): 0.005 mm (black square), 0.01 mm (white square), 0.015 mm (black circle), 0.03 mm (white circle) and 0.06 mm 1,4-BQ. (B) 1/[1,4-BQ] (black circle): 0.05 mm (white circle), 0.1 mm (black square), 0.125 mm (white square), 0.15 mm (black triangle) and 0.2 mm NADPH.
Download figure to PowerPoint
We propose that the PpAzoR described here is most structurally homologous to PaAzoR2. Azo substrate kinetics also support this assumption in terms of function. Steady-state apparent catalytic constants determined for methyl red reported for PpAzoR (Table 3)  are very similar to those reported by Ryan et al.  for PaAzoR2 and the same substrate. PaAzoR2 shows the highest activities with bulky substrates such as Ponceau BS and amaranth dye, with specificity constants of the same order of magnitude (12.7 × 103 and 2.1 × 103 m−1·s−1, respectively) as have been determined for RB5 (1.4 × 103 m−1·s−1). PpAzoR is also active on Mordant Black 3, 9 and 17, which have ortho-substituted hydroxyl groups, a feature that is claimed to be unique to PaAzoR2 and its increased specificity for bulky substrates . However, PpAzoR is also active on other bulky substrates, such as RB5 and Reactive Yellow 81, that do not have the same chemical features, suggesting that, for PpAzoR, tolerance to substrate size is independent of the position of the hydroxyl/amine group. Another discrepancy observed between PpAzoR and PaAzoR2 is the preference for NAD(P)H: PaAzoR2 shows a strong preference for NADH, whereas our studies show that PpAzoR has the ability to use both NADPH and NADH equally. The lack of a crystal structure of PpAzoR with NADH or NADPH in the active site somewhat limits any final conclusions, but our studies so far admit the possibility of either coenzyme being accomodated within the catalytic site.
The NAD(P)H:flavin reductases, or more specifically, NAD(P)H quinone oxidoreductases, a class to which PpAzoR generically belongs, catalyse a wide range of electron transfer reactions. The exact biological function and electron transfer mechanism of these enzymes are still under discussion. It is unequivocal, from both our studies and the reported literature that, by comparison of kcat/Km values, quinones are much better substrates of PpAzoR than are azo compounds. A linear relationship between the increase in azoreductase mRNA levels and the presence of quinones has provided strong evidence that quinones are indeed the primary physiological substrates for azoreductases . We have studied the structure of PpAzoR in order to shed light on its specificity for non-natural, bulky azo compounds. Previously reported structures clearly show how azoreductases can promote obligatory two-electron reductions: both halves of the reaction involve hydride transfers – first from NAD(P)H to FMN, and then from FMNH2 to the substrate. The three structures here presented provide all of the structural elements that are necessary to promote an azo reduction reaction as described for other azoreductases and quinone reductases in the literature. The charge relay formed by Asn97 and His144 (and/or Gly141) would allow the reaction to take place without unfavourable charge separations. Although the overall structures of short flavodoxin-like fold azoreductases are very similar, small changes in the active site have been shown to significantly affect the substrate specificity. In this respect, we have identified structural elements comprising helix 2 and the residues of the following loop that allow accommodation of bulky substrates. We have also identified two residues, Asn14 and Arg184, that participate in binding/stabilizing of bulky dyes and eventually may be involved in NAD(P)H binding.