Ligand-induced Conformational Changes in the Capping Subdomain of a Bacterial Old Yellow Enzyme Homologue and Conserved Sequence Fingerprints Provide New Insights into Substrate Binding*

We have recently reported that Shewanella oneidensis, a Gram-negative γ-proteobacterium with a rich arsenal of redox proteins, possesses four old yellow enzyme (OYE) homologues. Here, we report a series of high resolution crystal structures for one of these OYEs, Shewanella yellow enzyme 1 (SYE1), in its oxidized form at 1.4Å resolution, which binds a molecule of PEG 400 in the active site, and in its NADH-reduced and p-hydroxybenzaldehyde- and p-hydroxyacetophenone-bound forms at 1.7Å resolution. Although the overall structure of SYE1 reveals a monomeric enzyme based on the α8β8 barrel scaffold observed for other OYEs, the active site exhibits a unique combination of features: a strongly butterfly-bent FMN cofactor both in the oxidized and NADH-reduced forms, a collapsed and narrow active site tunnel, and a novel combination of conserved residues involved in the binding of phenolic ligands. Furthermore, we identify a second p-hydroxybenzaldehyde-binding site in a hydrophobic cleft next to the entry of the active site tunnel in the capping subdomain, formed by a restructuring of Loop 3 to an “open” conformation. This constitutes the first evidence to date for the entire family of OYEs that Loop 3 may indeed play a dynamic role in ligand binding and thus provides insights into the elusive NADH complex and into substrate binding in general. Structure-based sequence alignments indicate that the novelties we observe in SYE1 are supported by conserved residues in a number of structurally uncharacterized OYEs from the β- and γ-proteobacteria, suggesting that SYE1 represents a new subfamily of bacterial OYEs.

The OYE family has been studied since the 1930s when Warburg and Christian isolated a yellow enzyme from brewers' bottom yeast (Saccharomyces carlsbergensis) (1). Theorell purified the enzyme in 1935 and showed that it was composed of a colorless apoprotein and a yellow dye. The dye turned out to be a vitamin B 2 -derived molecule, FMN, and to be essential for catalysis. When a second yellow enzyme was isolated from yeast, the first enzyme was named old yellow enzyme (OYE), 3 a name that has persisted till this day. Since then, OYE has become synonymous with the study of flavoproteins, but despite decades of research culminating with a dissection of the catalytic mechanism (2)(3)(4), the physiological role of OYE has only recently begun to emerge (5)(6)(7).
The OYE family has grown steadily in recent years and now features the bacterial nitro-ester reductases PETN reductase (8), glycerol trinitrate reductase (9), XenA/B reductase (10), the bacterial morphinone reductase (11), the YqjM from Bacillus subtilis (41), the plant oxophytodienoic acid reductases (12,13), several yeast OYEs (14 -17), and an enzyme involved in prostaglandin synthesis in Trypanosoma cruzi (18). Although these enzymes originate from different organisms and catalyze different reactions, they share several common functional characteristics. They are able to reduce simple and complex unsaturated aldehydes and ketones (11,19), nitro-esters (8 -10, 20), and nitro-aromatic substrates (21)(22)(23), and yet clear preferences are observed. Furthermore, OYEs can form long wavelength charge transfer interactions with phenolic compounds (24), which in turn exhibit a distinct binding mode in the active site involving hydrogen bonding of the phenolate oxygen to a strictly conserved histidine/asparagine or histidine/histidine pair (25). The reaction mechanism of OYE proceeds by a ping-pong mechanism consisting of an oxidative and reductive half-reaction. Consistent with the plethora of functional similarities among members of the OYE family, the first crystal structures reported revealed a common ␣ 8 ␤ 8 barrel fold with significant variations in the capping subdomain (25)(26)(27). Nonetheless, interesting differences have begun to emerge in recent studies. For example, morphinone reductase lacks the general acid (tyrosine) that protonates the substrate (28). In another case, enzymes possessing two histidines (23) in the active site for substrate binding can catalyze the reduction of TNT to the H-Meisenheimer complex, whereas the enzymes with only one histidine cannot. Finally, YqjM has a flavin that is butterfly-bent in both the reduced and the oxidized state (29).
We have recently probed the genome of Shewanella oneidensis, a Gram-negative ␥-proteobacterium with a remarkable respiratory versatility (30) (www.tigr.org), using yeast OYE1. We identified four OYE homologues (NP718044, NP718043, NP719682, and NP718946) that we named SYE1-4 for Shewanella yellow enzyme (31). SYE1 (NP718044) is the Shewanella OYE homologue with the greatest sequence identity to OYE1. Here, we report structural studies of SYE1 at high resolution and show that SYE1 exhibits a number of novel features that can be mapped to strongly conserved sequence fingerprints in several structurally uncharacterized OYE homologues from the ␤and ␥-proteobacteria. This suggests that SYE1 may be the prototype of a new subfamily of bacterial OYEs.

EXPERIMENTAL PROCEDURES
Materials-The Q-Sepharose FF and Source 30Q matrices and the MonoQ and Superdex 200 columns were purchased from Amersham Biosciences, and the pACYC-Duet-1 vector was from Novagen. Structure screens 1 and 2 were from Molecular Dimensions Ltd., and the additive screen was from Hampton Research. Trizma base, ammonium sulfate, FMN, MES, Na 2 HPO 4 , and NaH 2 PO 4 were purchased from Sigma; HEPES and PEG 400 were from Fluka; NaCl and NADH were from Merck; t-hexenal, t-decenal, and t-dodecenal were from Acros; and isopropyl ␤-D-thiogalactopyranoside was from Duchefa.
Cloning, Expression, and Purification-The sye1 gene was excised from a pGEX-SYE1 construct using EcoRI and NotI and cloned in the first multiple cloning site of pACYC-Duet-1, generating pACYC-SYE1. This cloning strategy puts the sye1 gene out of frame with the His tag. The pACYC-SYE1 vector was introduced into Escherichia coli BL21(DE3) cells. The cultures were grown at 28°C under constant shaking, and SYE1 expression was induced at 0.5 A 600 nm using 0.5 mM isopropyl ␤-Dthiogalactopyranoside. The cultures were left to grow for another 7 h, and the cells were harvested by centrifugation and resuspended using 30 ml of 50 mM Tris, pH 8.0, per liter of culture grown. The resuspended cells were lysed by sonication, and the soluble fraction was clarified by centrifugation. SYE1 was purified in four steps. Lysate containing SYE1 was manually loaded onto a Q-Sepharose FF column equilibrated with 50 mM Tris, pH 8.0. The column was washed with 50 mM Tris, pH 8.0, 50 mM NaCl, and SYE1 was eluted with 50 mM Tris, pH 8.0, 300 mM NaCl. The eluate was dialyzed to remove the salt and was loaded onto a Source 30Q column equilibrated with 50 mM Tris, pH 8.0. SYE1 eluted at 80 -100 mM NaCl. The pooled fractions were dialyzed against 20 mM MES, pH 6.5, and loaded onto a MonoQ column from which SYE1 eluted at 40 -60 mM NaCl. As a final step the SYE1 fractions were pooled, concentrated, and dialyzed against 50 mM Tris, pH 8.0, 100 mM NaCl and loaded onto a Superdex 200 column. The elution profile of SYE1 was consistent with a 40-kDa protein, indicating that SYE1 was a monomer. Final purity was confirmed by silver staining of an SDS-PAGE gel. The pure fractions were pooled and dialyzed against 35 mM Tris, pH 8.0, 70 mM NaCl. The protein was subsequently concentrated to 12 mg/ml and stored in 100-l aliquots at Ϫ80°C.
Crystallization of SYE1-Crystallization trials were set up by the hanging drop vapor diffusion method, using 200 l of reservoir solutions (Structure Screens 1 and 2; Molecular Dimensions Ltd.) and mixing equal volumes of protein solution and reservoir (1 l ϩ 1 l) to make the droplets. Within 2 days needle-like crystals were observed in condition 30 (2% PEG 400, 2 M ammonium sulfate and 0.1 M HEPES, pH 7.5). Optimization of the crystals by varying the pH and the concentrations of PEG 400 and ammonium sulfate was unsuccessful. We therefore screened a series of additives (Additive screen; Molecular Dimensions Ltd.) against the condition with the best needles (100 mM Tris, pH 8.5, 1.8 M ammonium sulfate, 2% PEG 400). Substantially improved crystals were observed in the presence of 0.5% ␤-octyl glucoside. Upon further optimization, diffraction quality crystals were grown in droplets containing 100 mM Tris, pH 8.0 -8.5, 1.75 M ammonium sulfate, 2% PEG 400, and 0.25% ␤-octyl glucoside. This condition gave rise to two distinct crystal morphologies, rectangular block-shaped (Form A) versus hexagonal-shaped crystals (Form B), which diffracted to 1.3 Å versus 2.4 Å resolution, respectively, using synchrotron radiation. We thus chose to focus on Form A crystals. They belong to space group P2 1 2 1 2 1 with a ϭ 48.45 Å, b ϭ 83.60 Å, and c ϭ 87.53 Å and contain one molecule per asymmetric unit with a V m value of 2.24 Å 3 /Da. Data Collection, Structure Solution, and Refinement-X-ray diffraction data were collected at 100 K at the EMBL beamlines X13 and BW7A in DESY (Hamburg, Germany) using single crystals that were previously cryo-cooled in liquid nitrogen following a brief incubation (ϳ20 s) in a cryoprotectant solution (100 mM Tris, pH 8.0, 2 M ammonium sulfate, 2% PEG 400, 20% (v/v) glycerol). All of the diffraction data were processed using the programs Denzo and Scalepack (32). Data statistics are presented in Table 1. The native SYE1 structure was solved by molecular replacement using maximum-likelihood methods implemented in the program PHASER (33). The search model was generated from the coordinates of morphinone reductase (Protein Data Bank entry 1GWJ) which exhibited a 50% sequence identity with SYE1. Nonconserved residues were replaced by alanine (or glycine), and all of the insertions were deleted, as well as all ligands, including water molecules and the FMN cofactor. The correctness of the structure solution was assessed from the quality of the electron density for missing side chains and other unique structural features. Several rounds of model building were subsequently alternated with further crystallographic refinement employing simulated annealing, conjugate gradient minimization, and individual B-factor refinement using a maximum-likelihood target function, as implemented in CNS (34) (see Table 1). A random 5% set of the possible reduced data were selected and used to flag reflections for the calculation of R free values. The same set of reflections was used for the calculation of the R free for all data sets. The FMN cofactor was initially modeled with a planar isoalloxazine ring, but difference electron density maps after the first round of refinement showed clear evidence for a butterfly-bent ring about the imaginary N-5-N-10 axis. The planarity weights were therefore relaxed during subsequent rounds of refinement to facilitate proper refinement of the FMN ring atoms. Ensuing electron density maps indeed contained no residual positive or negative difference electron density for the FMN substructure. A PEG 400 molecule, a sulfate ion, a ␤-octyl glucoside molecule, and water molecules were modeled in the final stages of refinement. The electron density maps allowed us to model the entire PEG 400 molecule, which is consistent with the refined temperature factors of the PEG 400 molecule (see Table 1). All of the ligands were refined using atomic parameters extracted from the HICUP server. Ligand p-hydroxyacetophenone was composed by combining two molecules (phenol and an acetyl group). For all three soaking experiments difference density maps were initially calculated using coefficients F obs,soak Ϫ F obs,native and calculated phases from the high resolution native structure. All of the residues were found in the most favored, additional, or generously allowed regions of the Ramachandran plot (program Procheck (35)) except for Trp 274 , an aromatic amino acid that lines the active site tunnel. The r.m.s.d. values for the different models were calculated by the program SEQUOIA (36), and the figures were drawn with Pymol (www.pymol.org).
Soaking Experiments-Form A crystals of SYE1 were washed two times with 100 mM Tris, pH 8.0, 2 M ammonium sulfate and were subsequently incubated in 100 mM Tris, pH 8.0, 2 M ammonium sulfate, supplemented with 50 mM of the desired ligand. Complex formation in the crystals was monitored via the rapid change of color from yellow to lime-green, in the case of the phenolic ligands and from yellow to colorless in the case of NADH. Incubation times for the phenolic ligands were typically 2-5 min depending on the size of the crystal used and in the case of NADH 30 min to several hours. After incubation, the crystals were flash frozen in liquid nitrogen.
Kinetic Studies-All of the kinetic parameters were determined under strict anaerobic conditions, and all of the assays were performed in triplicate. The reaction mixtures consisting of 0.5ϫ phosphate-buffered saline buffer, pH 7.3 (5.75 g of Na 2 HPO 4 , 1.49 g of NaH 2 PO 4 , and 2.92 g of NaCl/liter), 60 -120 M NADH, and varying concentrations of substrate were prepared within a glove box (Coy Laboratories, Grass Lake, MI) and transferred to sealable quartz cuvettes (Hellma). The concentration of NADH was determined spectrophotometrically using a molar extinction coefficient of 6220 M Ϫ1 cm Ϫ1 at 340 nm. Enzyme solutions were prepared in sealed flasks and made anaerobic by bubbling with N 2 gas. The reactions were initiated by the addition of enzyme to the reaction mixtures using a Hamilton needle. The enzyme concentration was kept constant at 100 nM. SYE activity was assayed by following the oxidation of NADH using a double beam Uvikon 943 spectrophotometer. Apparent steady state kinetic constants were determined by least squares fitting procedures to the standard Michaelis-Menten equation.
Protein Data Bank Accession Codes-The coordinates and structure factors for the four structures of SYE1 reported here are available via the Protein Data Bank (www.rcsb.org) with accession codes 2GOU, 2GQ8, 2GQ9, and 2GQA.

RESULTS AND DISCUSSION
Overall Structure of SYE1-The structure of oxidized SYE1 was determined by molecular replacement methods using morphinone reductase as the search model and was refined at 1.4 Å resolution (Table 1). SYE1, like all OYE homologues characterized to date (26,27,29,37), folds into an ␣ 8 ␤ 8 barrel containing one noncovalently bound FMN molecule located at the top of is the intensity of the i th measurement of reflection h , and ͗I(h)͘ is the average value over multiple measurements.
the ␤ barrel (Fig. 1, A and B). Like morphinone reductase (37) and PETN reductase (27), SYE1 has four extra barrel elements: (a) the N-terminal ␤-hairpin that closes the bottom of the bar-rel (Fig. 1A, bottom arrow); (b) a capping subdomain formed by secondary structure elements between ␤3 and ␣3; (c) an ␣-helix linking ␤4 and ␣4, which in OPR1 (26) is thought to aid in substrate recognition; and (d) the ␣-helix between ␤8 and ␣8 that contributes to the binding of FMN.
In the course of refinement difference electron density maps showed clear additional density for three nonprotein elements: a PEG 400 molecule, a sulfate ion, and a ␤-octyl glucoside molecule (Fig.  1A). PEG 400 was an essential component of the crystallization condition and could be traced in the active site of SYE1 (Fig. 1C). OPR1 also binds PEG in the active site, but in that case the PEG molecule is doubly bent and resembles the substrate 12-oxophytodienoic acid. In SYE1 the PEG 400 molecule is extended, and only the first six atoms (corresponding to two ethylene glycol units) enter the active site tunnel, whereas the rest of the molecule bends around the hydrophobic active site entrance and extends outwards like the tail of a pig (Fig.  1C). Interactions between the PEG molecule and SYE1 are largely based on van der Waals' interactions with the hydrophobic lining of the active site. The sulfate ion and the ␤-octyl glucoside molecule participate in intermolecular contacts between two adjacent SYE1 molecules in the crystal lattice. The ␤-octyl glucoside molecule lies on one side parallel to His 39 and is coordinated on the other side by Lys 261 from a symmetry-related molecule. In contrast to the sugar ring, the tail of the ␤-octyl glucoside molecule is not well defined in the electron density maps. The sulfate ion is coordinated by His 337 and Arg 336 and by His 344 , Arg 301 , and Arg 323 from a symmetry-related SYE1. The last two amino acids are involved in the coordination of the flavin phosphate oxygens.
The  r.m.s.d. over 293 C␣ atoms), and YqjM (1.7 Å r.m.s.d. over 217 C␣ atoms). The largest differences can be found in the capping subdomain and the loop between ␤6 and ␣6. The ␤-hairpin that connects the two ␤-strands of the capping subdomain is orientated in SYE1 toward the inside of the barrel.
The Active Site Structure-The active site of SYE1 is composed of a narrow and shallow tunnel about 10 -15 Å long and is lined by aromatic and hydrophobic residues. At its bottom, the tunnel opens up toward the flavin. Interestingly, Loop 6 4 folds toward the center of the barrel and contributes to a significant part of the lining of the tunnel. This makes the tunnel markedly narrower than any other OYE characterized thus far, which is surprising given the high overall structural similarity with the bacterial homologues PETN reductase and morphinone reductase. We propose that this feature may be of functional importance.
FMN is tightly bound in the protein and exhibits the lowest B-factors in the structure. The amino acids that contact the riboflavin moiety of FMN are conserved among the different known OYE structures (26,27,29,37): the flavin N-5 interacts with the main chain amide of Thr 26 , O-4 is hydrogen-bonded to the O␥ of Thr 26 and the main chain amide of Gly 58 , and N-3 and O-2 interact with the side chain of Asn 100 . O-2 makes an additional hydrogen bond with the amide of His 181 and the guanidino group of Arg 233 ( Fig. 2A).
Differences can be seen in the amino acids binding the ribityl hydroxyl groups and the phosphate oxygen atoms. The first two hydroxyl groups are each hydrogen-bonded to the guanidino group of Arg 233 , and to the main chain carbonyl oxygen of Pro 24 and a water molecule, respectively. The third OH group and the phosphate oxygens are involved in an extensive hydrogen bonding network involving the side chains of Arg 323 and Arg 301 , the main chain amides of these two arginines and of Gly 300 , and three water molecules. In contrast to the si-face of the flavin, the ribityl side chain, and the dimethylbenzene ring, which are all solvent-accessible, the re-face of the flavin is buried in the structure and is surrounded by residues from Loop 1.
The tricyclic ring of the FMN in SYE1 is strongly butterflybent around the hypothetical N-5-N-10 axis in both the reduced and oxidized states (Fig. 2B). Structural analysis of NADH-reduced SYE1 shows that reduction of the enzyme does not result in a further bending of the flavin (see below). A similar feature is reported for the flavin group in Bacillus YqjM (29), yet only 10% of all flavoproteins are thought to have a bent FMN group (38). Both SYE1 and YqjM have a methionine at position 25 just like many other structurally uncharacterized OYE homologues from the ␤-proteobacteria and the ␥-proteobacteria we have identified via a BLAST search using SYE1 4 Loops are numbered after the ␤-strand they follow. Thus Loop 6 lies between ␤-strand 6 and ␣-helix 6. as a probe (Fig. 2C). Interestingly, all of the other structurally characterized OYEs to date have a leucine instead of methionine at position 25, and they all exhibit planar FMNs in the oxidized state. It has been proposed that a bent flavin would destabilize the formation of semiquinones and thus one-electron reactions of the flavin (39). It is interesting to note that semiquinone species were observed during the photoreduction of OPR1 (13) and OYE (40), but not of YqjM (41) and SYE1 (31). In this respect, the behavior of YqjM and SYE1 is similar to other bacterial OYE homologues such as PETN reductase (42), morphinone reductase (43), XenA reductase, and XenB reductase (10), all of which have planar flavin groups. Clearly then, a mechanism other than the bending of the flavin is responsible for the lack of semiquinone formation in these enzymes. Although the importance of a butterfly-bent FMN is not yet known, our analysis suggests that a methionine residue at position 25 may be a general prerequisite for a butterfly-bent flavin group across all OYEs.
Structures of SYE1 in Complex with p-HBA and p-ACE-Our choice of phenolic ligands was limited by the fact that previous experiments had shown that SYE1 is not capable of binding phenolic ligands with high pK a values (pK a Ͼ 9), such as p-cresol (31). We therefore decided to pursue complexes with p-hydroxybenzaldehyde (p-HBA) and p-hydroxyacetophenone (p-ACE). Upon incubation with the phenolic compounds, the crystals turned from yellow to lime-green. Structures of the complexes were determined by difference Fourier analysis and were refined with full occupancy of the ligands in the active site to 1.7 Å resolution ( Table 1).
The two phenolic compounds were well defined in the difference electron density maps, and the orientation of the molecules could be readily determined in the active site. p-HBA and p-ACE are orientated in the same fashion (Fig. 3A). The phenolic oxygen is placed above the C-2 atom of the flavin, forming a hydrogen bond with His 181 and Asn 184 . Although the aldehyde carbonyl of p-hydroxybenzaldehyde in OYE (25) is hydrogen-bonded to Tyr 375 , no bond of this kind is possible because this residue is replaced by Phe 350 in SYE1 (Fig. 3B). Instead, the aldehyde oxygen is hydrogenbonded to Trp 274 from Loop 6, which is forced to adopt an otherwise unfavorable conformation to serve this role. This novel interaction compensates for the loss of Tyr 375 (OYE numbering) because of a mutation to phenylalanine in SYE1. Morphinone reductase has a Trp 279 corresponding to Trp 274 , but it is located toward the outside of the barrel. Two evolutionary scenarios may explain the role of Loop 6; the restructuring of Loop 6 in SYE1 arose when the enzyme had to substitute for the lost Tyr 375 interaction, or Loop 6 appeared first in the active site. Because Trp 274 may substitute effectively for the Tyr 375 interaction, a Tyr to Trp mutation was not harmful for catalysis.
It has been shown for morphinone reductase that a Y356F mutation (equivalent of Tyr 375 ) increases the K m for NADH by a factor of ϳ7. For SYE1 the K m for NADH binding is 9.3 M, which is consistent with the values for other OYE homologues ( Table 2). Thus Trp 274 appears to compensate well for the absence of Tyr 375 both in terms of phenolic ligand binding and NADH binding. Interestingly, several OYE homologues exhibit the Trp 274 and Phe 350 substitutions. We identified the consensus sequence (A/S)E(A/V)DWDDA flanking the Trp 274 residue, whereas the Phe 350 is always preceded by a leucine residue   (Fig. 3B). We propose that these sequence fingerprints, combined with the structural features in the active sites of SYE1, indicate that OYE homologues possessing these features may constitute a new subfamily of OYEs in ␤and ␥-proteobacteria. NADH-reduced Structure-Upon incubation of SYE1 crystals with NADH, the yellow color of the crystals fades away, indicating a reduction of the flavin. However, the occupancy of NADH in the active site was too low to allow its modeling in the active site. This appears to be a recurring theme in the OYE superfamily, because attempts to obtain such complexes in morphinone reductase (37), OPR1 (26), and OYE (25) also failed. It has been noted for other homologues that reduced flavin harbors ordered water molecules above the ring that replace bound ions (Cl Ϫ ) (25,27). In native SYE1, there are no such water molecules because of the presence of PEG 400 in the active site, but upon reduction with NADH, PEG 400 no longer occupies the active site, and instead three ordered water molecules move into its place.
The "Open" Conformation of Loop3 Provides Insights into NADH Binding-The NADH-reduced and the p-ACE bound structures differ little from the PEG bound SYE1 structure, the most notable feature being the disappearance of the ␤-octyl glucoside molecule in the NADH-reduced structure, which brings about a 90°f lipping of His 39 around its hinge. The structure of the p-HBA complex, on the other hand, was not straightforward to interpret. Incubation with p-HBA increased the internal disorder of the crystals, resulting in the disappearance of the electron density of many side chains. As in the NADH-reduced structure, the ␤-octyl glucoside molecule was washed away in the soak, whereas the sulfate ion became highly disordered.
At first glance, a p-HBA molecule could be unambiguously modeled in the active site of SYE1 stacked against the flavin ring in an orientation similar to what we observed for p-ACE. Quite unexpectedly, a second p-HBA molecule (p-HBA-B) was found sandwiched between two Phe residues (Phe 132 and Phe 350 ) at the entrance to the active site at the capping subdomain (Fig. 4, A and  B). The occupancy of this second site has been estimated to be 0.6 based on a comparison of the refined B-factors of p-HBA-B at that occupancy, with the surrounding residues of SYE1. To accommodate for this molecule, Phe 132 was flipped by 90°a round its 1, and the hairpin loop of the capping subdomain was twisted by about 20°around Val 133 and 80°around Gly 141 toward the outside of the barrel. This open conformation of the capping subdomain creates a hydrophobic cleft next to the active site entrance defined by Tyr 68 , Trp 70 , Phe 132 , Phe 142 , and Phe 350 (Fig. 4C). Modeling of a p-ACE molecule at the site of the second p-HBA explains why a similar interaction was not observed in the p-ACE complex. In that case, the methyl group of p-ACE would approach only 2.4 Å from Trp 70 and would cause a steric clash. Trp 70 is part of Loop 2, which is structurally conserved in SYE1, morphinone reductase, and PETN reductase.
The fortuitous discovery of a second binding site for p-HBA at the capping subdomain provides the first direct evidence to date for any OYE homologue that Loop 3 of the capping subdomain can indeed participate in ligand binding. This provides important new insights into the elusive NADH complex and into ligand binding in general. It is tempting to propose that the two binding platforms for p-HBA reflect the binding sites for the nicotinamide and adenine moieties of NADH, respectively. However, careful modeling of NADH, with the nicotinamide ring occupying the site of p-HBA-A at the flavin ring, shows that the aromatic platform presented by Phe 132 and Phe 350 can only serve as the binding site for the ribose group preceding the adenine moiety of NADH. Such a stacking interaction is frequently observed in protein carbohydrate interactions. The adenine moiety of NADH would therefore have to extend toward the exit of the active site channel, but in the absence of an obvious nearby binding cleft, it is difficult to speculate whether the adenine will bind to SYE1 at all. Interestingly, we find that all structurally uncharacterized SYE1 homologues we have identified in our sequence alignments exhibiting a tryptophan residue and a phenylalanine or tyrosine at positions 274 and 350, respectively (Fig. 3B), also have a phenylalanine or a tryptophan at position 132 (Fig. 4D). The strong conservation of such sequence features suggests that this group of OYE homologues likely binds NADH and other ligands in a similar fashion and that SYE1 can serve as a structural prototype for this subfamily of OYEs.
Our observation of a second binding site for p-HBA is consistent with the recent proposal that the distal part of the NADPH moiety interacts with Loop 3 in PETN (27). It has also been proposed that the preference for NADPH in PETN reductase is brought about by the presence of two arginine residues (Arg 142 and Arg 130 ) in Loop 3 (27,44). In morphinone reductase (which prefers NADH), these arginines are replaced by a glutamate residue (Glu 135 ), whereas in SYE1 we observe a lysine (Lys 130 ) and a

Kinetic constants for the binding of t-2-ketanals to SYE1
Apparent kinetic constants were derived from steady state kinetic analyses using the software Graphpad Prism4. Assays were performed under strict anaerobic conditions in 1 ml of 0.5ϫ phosphate-buffered saline buffer, pH 7.3, and 60 -100 M NADH. Enzyme concentration was 100 nM. Also given, is the structure of each t-2-ketanal and of PEG 400 for comparison purposes. glutamate (Glu 139 ). Because SYE1 exhibits a preference for NADH, this theory cannot be extended to all OYE homologues. When we compare the utilization of reducing power among OYE homologues, we find that the ones preferring NADPH accepted both NADPH and NADH, but with a preference for the former (18,31,37,41,45). On the other hand, OYEs that use NADH cannot accept NADPH. We therefore conclude that one should not search for a mechanism that discriminates between NADH and NADPH, but for one that excludes NADPH.
Further Insights into Substrate Selection and Binding-Thus far, activity of SYE1 has only been shown with relatively small molecules: N-ethylmaleimide, acrolein, and 2-cyclohexenone (31), whereas bulkier molecules, such as nitroglycerine and androstadiene/progesterone, exhibit no activity with SYE1. The shallow and narrow landscape of the active site of SYE1 inevitably imposes size restraints on candidate substrates. The identification of a PEG 400 molecule in the active site tunnel of SYE1 prompted us to explore whether SYE1 accepted long chain unbranched ␣,␤-unsaturated aldehydes as a substrate. Indeed, a steady state kinetic analysis showed that SYE1 showed reactivity toward trans-2-hexenal, trans-2-decenal, and trans-2-dodecenal, three naturally occurring components of plant oils with bactericidal properties against Salmonella (46,47) (Table 3). The k cat /K m values of SYE1 for these substrates are rather similar, differing only by a factor of two, in contrast to tomato OPR1, for which the k cat /K m values for t-2-hexenal and t-2-dodecenal differ by a factor of ϳ20 (13). This is probably due to the much narrower active site of SYE1, which likely can only accommodate substrates with short chains. OPR1, on the other hand, binds a complete PEG 400 molecule in its active site (26) and can thus accommodate larger substrates. Furthermore, SYE1 only binds the first six atoms of PEG 400 in the active site tunnel with the rest of the molecule extending outwards. A preference can be seen for the shorter chain of t-2-hexenal. A similar observation was made for OPR1 (13), which has the highest catalytic efficiency with the small molecule t-2-hexenal.
Considering the fact that PEG 400 was absolutely necessary for protein crystallization and that it binds in the active site, one can argue for a physiological substrate similar to t-2-alkenal. trans-2-Alkenals are breakdown products in lipid peroxidation reactions (46), produced from polyunsaturated fatty acids. Substances such as t-2-hexenal and t-2-dodecenal are volatile compounds of plant oils, produced upon bacterial injury to the plant. S. oneidensis, however, is not a plant pathogen, and polyunsaturated fatty acids are only present in psychrophilic bacteria (48,49). An involvement of SYE1 in detoxification of these compounds is thus unlikely. Nonetheless, the identification of a second ligand-binding site at a hydrophobic cleft near the mouth of the active site entrance, created by a restructuring of Loop 3 in the capping subdomain, indicates that the physiological substrate of SYE1 may be a molecule consisting of a bulky (possibly aromatic) head connected to a thin tail. The "tail" of the substrate could then intercalate to the active site, whereas the bulky head could be stabilized at the entrance of the active site.