Discovery, Characterization, and Kinetic Analysis of an Alditol Oxidase from Streptomyces coelicolor*

A gene encoding an alditol oxidase was found in the genome of Streptomyces coelicolor A3(2). This newly identified oxidase, AldO, was expressed at extremely high levels in Escherichia coli when fused to maltose-binding protein. AldO is a soluble monomeric flavoprotein with subunits of 45.1 kDa, each containing a covalently bound FAD cofactor. From sequence alignments with other flavoprotein oxidases, it was found that AldO contains a conserved histidine (His46) that is typically involved in covalent FAD attachment. Covalent FAD binding is not observed in the H46A AldO mutant, confirming its role in covalent attachment of the flavin cofactor. Steady-state kinetic analyses revealed that wild-type AldO is active with several polyols. The alditols xylitol (Km = 0.32 mm, kcat = 13 s–1) and sorbitol (Km = 1.4 mm, kcat = 17 s–1) are the preferred substrates. From pre-steady-state kinetic analyses, using xylitol as substrate, it can be concluded that AldO mainly follows a ternary complex kinetic mechanism. Reduction of the flavin cofactor by xylitol occurs at a relatively high rate (99 s–1), after which a second kinetic event is observed, which is proposed to represent ring closure of the formed aldehyde product, yielding the hemiacetal of d-xylose. Reduced AldO readily reacts with molecular oxygen (1.7 × 105 m–1 s–1), which confirms that the enzyme represents a true flavoprotein oxidase.

Carbohydrate oxidases are highly valuable biocatalysts for analytical and synthetic purposes. Chemical methods cannot compete with the exquisite regio-and/or enantioselectivity by which these enzymes oxidize polyols. Applications in which carbohydrate oxidases are used are, for example, biosensors for blood sugar, synthetic routes toward chiral building blocks, sweeteners, and flavors. Oxidase-mediated catalysis also leads to formation of hydrogen peroxide, a property that is used in a number of applications, such as bleaching processes and wastewater treatment (1,2). At present, only a limited number of carbohydrate oxidases have been identified, which restricts the biocatalytic exploitation of this class of redox enzymes. The best known representative is glucose oxidase, which in fact is the most widely applied redox enzyme.
Besides galactose oxidase, which contains copper as cofactor, all presently known oxidases acting on carbohydrates contain a flavin cofactor. Examples of such flavoprotein oxidases are glucose oxidase, L-gulono-␥-lactone oxidase, xylitol oxidase, hexose oxidase, lactose oxidase, glucooligosaccharide oxidase, and pyranose oxidase. Except for glucose oxidase and pyranose oxidase, all of these oxidases belong to a specific group of flavoproteins, the vanillyl-alcohol oxidase (VAO) 2 family. Members of this family share a similar overall structure consisting of two domains (3). One domain binds the adenine part of the FAD cofactor and is called the FAD-binding domain, whereas the other, called the cap domain, covers the isoalloxazine moiety of the cofactor and forms the major part of the active site around the isoalloxazine ring system. A special feature of this flavoprotein family is the fact that a relatively large number of VAO members bind the FAD cofactor in a covalent manner. This is also the case for all of the above mentioned VAO-type carbohydrate oxidases. In fact, the recent elucidation of the structure of glucooligosaccharide oxidase has revealed the first example where a flavin cofactor is covalently linked to two amino acid residues (4). It has been shown that a covalent FAD-protein linkage can have a significant effect on the redox behavior of the flavin cofactor (e.g. increasing the redox potential) (5). This is in line with the observation that most VAO-type covalent flavoproteins act as an oxidase (6). In these cases, the FAD cofactor is typically tethered to a histidine residue, and this linking histidine can be readily identified by sequence motif recognition. Hence, the ability to identify covalent VAO homologs by sequence analysis can be used as a tool to find novel oxidase genes.
Most of the characterized carbohydrate oxidases have been isolated from fungi, whereas only two from bacterial origin have been described (7,8). Here we describe the discovery and characterization of an alditol oxidase (AldO) from the actinomycete Streptomyces coelicolor A3 (2). We identified the putative aldo gene (SCO6147) in the genome sequence of this bacterium by searching for VAO homologs. The predicted protein sequence of AldO contains a histidine that is expected to form a covalent FAD-histidyl linkage. Therefore, it was anticipated that the protein would have a high probability to exhibit oxidase activity. In this report, we show that the enzyme is indeed a typical * This work was performed within the framework of Integration of Biosynthesis and Organic Synthesis. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains an appendix. 1  Cloning and Expression of the aldo Gene-The aldo genes were cloned into pBADNdeI and pBAD-MBP for expression. The latter is a pBADNdeI-derived vector in which the malE gene, including a Factor Xa protease cleavage site from pMAL-c2x, is inserted (NdeI-HindIII).
To facilitate TOPO cloning, 3ЈA-overhangs were generated by incubating the amplified aldo genes with 1 l of Taq DNA polymerase at 72°C for 15 min. For pBADNdeI cloning, the TOPO-aldo clones were digested with NdeI and HindIII, after which the aldo DNA was isolated from an agarose gel. The fragments were subsequently ligated into an NdeI-and HindIIIdigested pBADNdeI vector, yielding pBAD-aldo(krad) and pBAD-aldo. For pBAD-MBP cloning, the TOPO-aldo clones were digested with EcoRI and PstI, after which the aldo genes were isolated from an agarose gel. The aldo genes were subsequently ligated into an EcoRI-and PstI-digested pBAD-MBP vector, yielding pBAD-MBP-aldo(krad) and pBAD-MBP-aldo.
E. coli TOP10 cells were transformed with the constructed expression vectors and grown at 17°C for 3 days in terrific broth medium supplemented with 50 g/ml ampicillin and 0.02% (w/v) L-arabinose. Induction of pBAD-aldo(krad) and pBAD-aldo clones yielded AldO containing a 25-amino acid C-terminal extension consisting of a c-Myc epitope and a His 6 tag. Induction of pBAD-MBP-aldo(krad) and pBAD-MBP-aldo clones yielded MBP-AldO fusion proteins containing a Factor Xa protease cleavage site.
Purification of MBP-AldO-Cells from three 0.5-liter cultures were collected by centrifugation at 4,000 ϫ g and 4°C for 15 min and resuspended in 30 ml of 50 mM KP i buffer (pH 7.5). The cells were disrupted by sonication and subsequently centrifuged at 23,000 ϫ g and 4°C for 30 min to remove cell debris. The supernatant was divided into two 20-ml aliquots to prevent overloading during the purification procedure. After loading the supernatant onto a Q-Sepharose column, the column was washed until the A 280 was lower than 0.05. MBP-AldO was eluted by applying a 0 -1 M KCl linear gradient. The yellow fractions were pooled and concentrated using an Amicon stirred cell and YM-30 membrane (Millipore). To remove KCl from the concentrated enzyme solution, it was loaded onto a HiPrep 26/10 desalting column (Amersham Biosciences).
Analytical Methods-All experiments were performed at 25°C and in a 50 mM KP i buffer, pH 7.5. The S.D. value in the experiments is 5%, unless stated otherwise. Oxidase activity and steady-state kinetic parameters were determined by coupling the production of H 2 O 2 by (m-)AldO to a horseradish peroxidase-mediated oxidation of 4-aminoantipyrine and 3,5-dichloro-2-hydroxybenzenesulfonic acid. This results in formation of a pink to purple colored product, which can be measured at 515 nm (⑀ 515 ϭ 26 mM Ϫ1 cm Ϫ1 ) (9). The reaction mixture contained 50 mM KP i buffer, pH 7.5, 0.1 mM 4-aminoantipyrine, 1 mM 3,5-dichloro-2-hydroxybenzenesulfonic acid, 3 units of horseradish peroxidase, and 15 nM AldO or m-AldO.
The extinction coefficient of m-AldO was determined by comparing the absorption spectra before and after incubation with 0.1% SDS. For this, it was assumed that the flavin spectrum of the unfolded m-AldO was equal to that of free FAD (⑀ of 11.3 mM Ϫ1 cm Ϫ1 at 450 nm) (10).
For product identification, the conversion of xylitol by m-AldO was followed by HPLC analysis using a Shodex SUGAR SP0810 (8.0-mm inner diameter ϫ 300 mm) column. Optical rotation of the product was determined by polarimetric analysis using a Schmidt and Haensch Polartronic MH8 apparatus.
Redox potentials were measured by using the method described by Massey (11). A cuvette containing m-AldO (5-10 M), xanthine (400 M), KCl (107 M), benzyl or methyl viologen (2.0 M), and redox dye (5-16 M) was made anaerobic by flushing with argon. Under a constant flow of argon, 68 nM xanthine oxidase was added, and spectra were collected every 2.5 min during reduction using a PerkinElmer Life Sciences Lambda Bio 10 spectrophotometer.
Kinetic Measurements-Pre-steady-state studies were performed with an Applied Photophysics stopped-flow apparatus, model SX17MV. Spectral data were collected at time intervals of 2.5 ms using a diode array detector. A photomultiplier detector was applied to follow single wavelength traces in time. Spectral data obtained by diode array measurements were deconvoluted using the program Pro-K of Applied Photophysics. From the deconvoluted spectra, several wavelengths were selected to study the observed phases individually. The first two phases of the reductive half-reaction were monitored by following the absorption at 320 and 405 nm, respectively, after anaerobically mixing the enzyme solution with varying concentrations of xylitol. The first phase was monitored at 320 nm, because here the second and third phase showed no change in absorbance, whereas 405 nm is an isosbestic point in the first phase enabling analysis of the second phase. The third phase was monitored at 452 nm. The oxidative half-reaction was followed at 452 nm by mixing a solution containing reduced enzyme with a solution containing either 0.25 or 1.25 mM oxygen (in the presence or absence of 100 mM D-xylose). For anoxic conditions, all solutions contained 1.0 mM 4-ethylphenol and were made anaerobic by flushing with nitrogen. VAO was added to a final concentration of 100 nM to remove any residual oxygen. Traces obtained by photomultiplier measurements were fitted to an exponential function (Equation 1), where A represents absorption, C is a constant, and k is an observed rate constant. The observed rates for substrate concentration dependent reduction of m-AldO were fitted using the following equation (Equation 2).
Mechanistic Calculations-Mathematica 5.2 software was used to derive rate equations from the kinetic scheme by the determinant method (12) (see supplemental materials). The kinetic parameters K m and k cat as well as the redox state during catalysis can be calculated according to the following equations. For a ping-pong mechanism, K m and k cat values can be calculated using Equations 3 and 4.
For a ternary complex mechanism, K m and k cat values can be calculated using Equations 5 and 6.

RESULTS
Expression, Purification, and Biochemical Characterization of AldO-By a PSI-BLAST search for putative oxidases using VAO-type flavoprotein sequences, a gene (SCO6147) from S. coelicolor A3(2) was identified that encodes such a flavoprotein. The protein consists of 418 amino acids with a calculated mass of 44,347 Da (excluding FAD) and therefore represents one of the smallest members in the VAO flavoprotein family. Among the known VAO-type oxidoreductases, it shows highest sequence similarity with xylitol oxidase from another Streptomyces isolate (78% sequence identity) and pig L-gulono-␥-lactone oxidase (30% sequence identity). Except for sequence conservation typical for a VAO homolog, no other sequence motifs can be identified, suggesting that it is a cytosolic protein. Alignment of the protein sequence with known VAO-type protein sequences revealed the presence of a histidine residue (His 46 ), which is typically involved in covalent attachment of the FAD cofactor ( Fig. 1). Recent papers report on VAO-type oxidases containing a covalently bound FAD attached to two amino acids (4,13,14). Next to an 8␣-N-histidyl link, a second 6-Scysteinyl linkage to the isoalloxazine moiety of FAD is present in these oxidases. Sequence alignments did not reveal such a conserved cysteine that would be involved in covalent FAD attachment.
To verify that the enzyme from S. coelicolor is functional as an oxidase and explore its enzymatic properties, the corresponding gene was amplified from genomic DNA by PCR, cloned, and expressed. For initial expression experiments, the (aldo) gene was subcloned in a pBAD vector. Using the pBADaldo expression plasmid, recombinant active AldO was only expressed at a very low level, as observed by SDS-PAGE analysis. The respective protein band is fluorescent at pH 4, indicative of the presence of a covalently bound flavin cofactor (Fig. 2,  lanes A and B) (15). To increase the level of expression, the protein was expressed fused to MBP, since it is known that N-terminal fusion to MBP often boosts expression of soluble protein (16). The resulting pBAD-MBP-aldo vector, linking MBP to the N terminus of AldO, drastically increased the expression level (Fig. 2, lane D). From 1 liter of culture broth, an impressive amount of 350 mg of MBP-AldO was purified by one anion exchange chromatography step. Purified MBP-AldO runs as a single band upon SDS-PAGE at about 87 kDa, which nicely agrees with a calculated mass of 88,068 Da (including FAD). The protein band was again found to be fluorescent at pH 4, indicative of a histidyl-bound FAD cofactor (Fig. 2, lanes  C and D).
The covalent binding of the flavin cofactor was also confirmed by the observation that upon denaturation of the enzyme, the redox cofactor coprecipitated with the protein, yielding a bright yellow pellet after centrifugation. To test the location of the histidyl-FAD linkage, the H46A mutant was prepared and expressed. The H46A mutant protein was no longer fluorescent after separation by SDS-PAGE. The expression level was comparable with that of the wild type MBP-AldO; however, part of the protein was insoluble. The H46A mutant was purified by anion exchange chromatography in the presence of 100 M FAD. After purification, the H46A mutant was found not to contain FAD, and also no oxidase activity was observed. This confirms that His 46 is crucially involved in covalent FAD binding.
To determine whether the fused MBP-AldO protein behaves similar to the native AldO protein, we also prepared AldO with-out the MBP tag. For this, MBP-AldO was treated with modified trypsin lacking chymotrypsin activity. This yielded the two separate proteins: MBP and AldO. Using amylose resin, MBP could effectively be removed from this mixture. ESI-MS analysis showed that the cleavage had occurred at the expected target site for proteolysis resulting in a homogenous preparation of AldO. The determined mass of 45,665 Ϯ 2 Da corresponds nicely with the expected mass of AldO of 45,663 Da, which includes an N-terminal extension of 4 residues (Ile-Ser-Glu-Phe) that were part of the MBP-AldO linker and one acetate molecule. By gel permeation chromatography, it was confirmed that AldO and MBP-AldO both behave as monomers. The flavin absorbance spectrum (300 -650 nm) of AldO prepared by proteolytic cleavage was identical to that of the MBP-tagged AldO. This indicates that the microenvironment of the flavin cofactor is not influenced by the MBP fusion protein. Furthermore, it was found that isolated AldO was reactive with the same range of polyols as found for MBP-AldO (see below). Hence, we conducted all further analyses with the MBP-AldO fusion protein, referred to here as m-AldO.
The absorption spectrum of m-AldO shows two absorption maxima at 348 and 452 nm, which are typical for an oxidized flavin cofactor. The relatively low wavelength for the 348 nm absorption maximum is a typical feature of histidyl-bound flavin cofactors (10). The extinction coefficient of the covalently linked flavin was determined by comparing the absorption spectra of m-AldO before and after treatment with 0.1% SDS (10). From these spectra, an extinction coefficient for m-AldO was determined: 12.5 mM Ϫ1 cm Ϫ1 at 452 nm. A common characteristic of flavoproteins acting as an oxidase is the flavin reactivity with sulfite (17). To test whether m-AldO also exhibits this reactivity, it was titrated with Na 2 SO 3 (Fig. 3,  inset). This revealed that the enzyme readily forms a sulfite adduct with a K d of 59 Ϯ 5 M. The addition of this nucleophile resulted in the disappearance of the absorbance maxima in the visible area, which is indicative of formation of a flavin N-5 sulfite adduct. It has been suggested that there is also a correlation between oxygen reactivity and the formation of a red anionic semiquinone upon partial reduction of the FAD cofactor (e.g. with a strong white light source) (18). The FAD cofactor of m-AldO also shows the ability to stabilize the red anionic semiquinone when reduced by light. The appearance of the red anionic semiquinone form of FAD, typified by an intense absorbance band at 380 nm, was also observed during redox potential measurements (Fig. 3, spectrum 3). The redox potential of the flavin in m-AldO was determined by using the xanthine oxidase method (11). For the oxidized/semiquinone couple, a redox potential (E 1 ) of Ϫ4 mV was determined by using methylene blue (ϩ11 mV) as a reference dye, whereas a redox potential (E 2 ) of Ϫ213 mV for the semiquinone/reduced couple was determined by using anthraquinone-2-sulfonate (Ϫ225 mV). The midpoint redox potential Substrate Identification and Steady-state Kinetic Analysis-In order to identify substrates for this m-AldO, a large number of potential substrates were tested. The test set of 86 substrates contained a wide range of aromatic and aliphatic amines and alcohols. For activity screening, a generic chromogenic assay was employed in which hydrogen peroxide, inherently formed upon oxidase activity, is used by a peroxidase to form a colored product. The activity screening revealed that only a few polyols are readily accepted as substrate by m-AldO. The alditols sorbitol and xylitol are the best substrates. By HPLC analysis, it was found that m-AldO performs selective oxidation on one of the primary hydroxyl groups of xylitol. Furthermore, polarimetric analysis showed that only D-xylose is formed upon oxidation of xylitol.
m-AldO shows the highest catalytic efficiency with xylitol (Table 1). Although a similar k cat is observed with sorbitol, the K m value with this polyol is almost 5 times higher than with xylitol. Steady-state experiments were also performed with 1.25 mM O 2 , which resulted in a 1.5-fold increase of both k cat and K m . The observed k cat values for the alditols are in the same range as found for other oxidases and their substrates, which suggests that these polyols or related polyols are the physiological substrates for m-AldO. Besides alditols, only a low activity (conversion rate Ͻ1 s Ϫ1 ) was observed with other aliphatic primary alcohols, diols, and carbohydrates. It is clear that aliphatic polyols are preferred and that stereochemistry plays a role in substrate recognition. Next to this, the chain length of the substrate is also important for optimal reactivity. While xylitol is the best substrate with a 5-carbon skeleton, L-threitol (4 carbons), D-sorbitol (6 carbons), and D-mannitol (6 carbons) are all worse substrates because of relatively high K m values. Based on these results, we used xylitol as a model substrate to explore the kinetic behavior of m-AldO in more detail.
Using stopped-flow absorbance spectroscopy, we determined the redox state of m-AldO during steady-state catalysis by monitoring flavin absorbance. Upon mixing 0.25 mM O 2 , 5.0 mM xylitol, and 11.6 M AldO at 25°C, we observed within the first 30 ms a fast decrease of absorbance at 452 nm to reach the steady-state phase. The steady-state phase lasted for about 300 ms, which complies with the determined k cat value, and was followed by a rapid and full reduction of the flavin due to the excess of xylitol. During steady-state, 32% of the flavin was found to be in the oxidized form, indicating that the reductive and oxidative half-reactions are almost balanced.
Pre-steady-state Kinetic Analysis-To study the pre-steadystate kinetics of m-AldO, we followed the spectral changes of the covalently linked flavin by using a stopped-flow instrument equipped with diode array detection for spectral scans and a photomultiplier for single wavelength measurements. By investigating the reductive and oxidative half-reaction separately, information was obtained concerning the kinetic mechanism of m-AldO.
The reductive half-reaction was monitored by anaerobically mixing the enzyme with varying concentrations of xylitol. After mixing, three phases were observed, leading from oxidized to fully reduced flavin (Fig. 4). The inset in Fig. 4 shows the corresponding concentration profiles that were obtained from the deconvolution with Pro-K. To study the rates of the observed phases separately, strategic wavelengths were selected for every phase in such a way that only the desired phase showed changes in absorption. To follow the first phase, the absorption at 320 nm was monitored, and the third phase was studied at 452 nm, because the largest spectral difference occurred at that wavelength. The second phase was followed at 405 nm, which is the isosbestic point of the two spectra for the transition in the first  . Spectra were recorded every 2.5 ms for 1 s. Three phases can be distinguished in the deconvoluted spectra. The first phase is the reduction of FAD, leading to an intermediate spectrum (spectrum 2); the second phase leads to appearance of fully reduced flavin (spectrum 3); and the third phase represents product release, leading to spectrum 4. The inset shows the relative concentrations of the four species during anaerobic reduction.

TABLE 1 Steady-state kinetic parameters for m-AldO
phase, allowing a rate determination without the influence of the first phase.
The first phase showed the largest decrease in flavin absorption at 450 nm, indicating that this kinetic event reflects flavin reduction. A hyperbolic relation was observed between the rate of this first phase and an increasing concentration of xylitol, indicating that xylitol binding is the step preceding this redox reaction. The kinetic data for the first phase could be fitted by a simple hyperbolic function, indicating that reduction is virtually irreversible (Equation 2). Fitting yielded a reduction rate constant (k red ) of 99 s Ϫ1 and a K d of 1.3 mM. The spectrum formed upon flavin reduction (spectrum 2 in Fig. 4) does not resemble a typical fully reduced flavin spectrum. This suggests that the initially formed oxidation product, the open form of xylose, causes some spectral perturbations, possibly by forming a charge transfer complex, as is indicated by the increase in A 520 (19).
Following this first phase, there is a second fast phase leading to a fully reduced flavin spectrum (spectrum 3 in Fig. 4). This second phase also displayed a hyperbolic relation between the observed rate and xylitol concentration, leading to an observed rate of 51 s Ϫ1 at saturating xylitol concentrations. A similar kinetic event has been observed in the reductive half-reaction of another polyol oxidase, amadoriase I from Aspergillus sp. (20). The data were fitted by using a hyperbolic function with a nonzero intercept of 7 s Ϫ1 (k Ϫ3 ). This suggests that this second kinetic event is reversible. By subtracting k Ϫ3 from the maximum observed rate, a k 3 value of 44 s Ϫ1 can be calculated. We propose that the observed second phase corresponds to the ring closure of open D-xylose, which is formed in the first phase. It is energetically favorable to form the ring-closed hemiacetal of D-xylose. It should be noted that the k 3 and k Ϫ3 are observed rates that are in fact a complex function of the other rates of the reductive half-reaction. Extracting the true kinetic values for k 3 and k Ϫ3 is complicated due to the fact that the rate of reduction (k red ) and the observed k 3 are in the same order of magnitude. The identified kinetic events during the reductive half-reaction were simulated using the determined values for the step preceding and following this phase (k 3 /k Ϫ3 ) and the proposed mechanism (Scheme 1). It was shown that the observed rates are merely lower limits of the true k 3 and k Ϫ3 .
The third observed phase is a relatively slow process of 3.5 s Ϫ1 (k 4 ) and involves only marginal spectral changes. This last phase could represent product release. The rate constant for this process is relatively slow when compared with the k cat (13 s Ϫ1 ), implying that product release from this complex is not part of the catalytic cycle. This suggests that m-AldO follows a kinetic route in which the product is released from another binary or ternary complex. The anaerobic reduction of m-AldO was also followed by monitoring tryptophan fluorescence.
However, no significant change in fluorescence was observed in time.
Taken together, for the reductive half-reaction, a model is proposed as depicted in Scheme 1. In this model, E ox represents oxidized m-AldO, S represents the substrate, PЈ represents the open form of D-xylose, P represents the closed form of D-xylose, and E red represents the reduced form of m-AldO.
The oxidative half-reaction was monitored by mixing reduced enzyme with varying concentrations of molecular oxygen. To determine the reoxidation rate constant of reduced m-AldO with molecular oxygen, m-AldO was first anaerobically reduced with 3 eq of xylitol. After full reduction of m-AldO, reoxidation was followed with diode array detection and was found to be a monophasic process (Scheme 2).

DISCUSSION
In this paper, we report the discovery of AldO by genome mining. AldO was identified while searching for VAO homologs in the available genome sequence data bases. Based on sequence homology with other sequence-related flavoprotein oxidases, AldO was predicted to contain a covalent FADhistidyl linkage and therefore had a high chance to represent an oxidase. This study has shown that AldO is indeed a covalent flavoprotein oxidase primarily acting on alditols. The physiological function of this oxidase is yet unknown. Overlap in substrate specificity is found with a xylitol oxidase isolated from another Streptomyces strain (7). Also several homologs have been identified in other actinomycete genomes (sequence identity 81% for Streptomyces avermitilis; 78% for Streptomyces scabies; 52% for Stigmatella aurantiaca; 49% for Arthrobacter sp. FB24, Acidothermus cellulolyticus 11B, and Kineococcus radiotolerans SRS30216). For the gene from K. radiotolerans, we have demonstrated that the respective protein also acts as an alditol oxidase (data not shown). Although the expression of the K. radiotolerans AldO was poor in E. coli and could not be boosted by MBP fusion, it was found that the protein also contains covalent FAD and acts as an oxidase on the alditols xylitol and sorbitol. This indicates that the above mentioned bacteria (mainly actinomycetes) all harbor AldO orthologs. Apparently, a selected number of bacteria employ alditol oxidases for a yet unknown catabolic or anabolic route. They may be involved in modification of a secondary metabolite. Actinomycetes are SCHEME 1. Proposed model for the reductive half-reaction of m-AldO. In this model, E ox represents oxidized m-AldO, S represents the substrate, PЈ represents the open form of D-xylose, P represents the closed form of D-xylose, and E red represents the reduced form of m-AldO. SCHEME 2. Proposed models for the oxidative half-reaction of reduced m-AldO with and without product bound. In this model, E red represents the reduced m-AldO, E ox represents oxidized m-AldO, and P represents the closed form of D-xylose.
known for their ability to produce a range of secondary metabolites, which often contain a polyol moiety (22). The enzyme could also be part of a catabolic route for sorbitol or xylitol degradation. Inspection of the genome of S. coelicolor showed that the aldo gene is not flanked by genes that are related to carbohydrate modifications. However, the ortholog in S. avermitilis is located upstream of several genes related to carbohydrate degradation (i.e. a xylose repressor, a sugar transporter, a ␤-galactosidase, and a ␤-1,4-xylanase). Nevertheless, S. coelicolor A3 (2) is unable to grow efficiently on sorbitol, suggesting that AldO is involved in a more dedicated route (23).
To obtain large amounts of AldO, the gene was subcloned in a pBAD-MBP vector in such a way that AldO was expressed containing an N-terminal maltose-binding protein. In a single purification step, 350 mg of m-AldO per liter of culture was obtained. From spectral, electrospray ionization-mass spectrometry, and gel permeation analysis of m-AldO and AldO obtained by trypsin cleavage, it was clear that MBP has no influence on the properties of AldO. Hence, all experiments were performed with m-AldO.
Based on sequence homology, His 46 was expected to be involved in covalent FAD attachment. The lack of cofactor binding by the H46A mutant showed that His 46 is required for covalent FAD linkage. This is supported by the preliminary x-ray analysis of AldO crystals, which show that AldO contains 8␣-N 1 -histidyl-bound FAD to His 46 (24). 3 Next to this, the mutant protein no longer displayed oxidase activity. From these results, it appears that the covalent FAD linkage serves multiple purposes. In addition, it is known from other flavoprotein oxidases that a covalently linked flavin can increase the redox potential and in that way facilitate usage of molecular oxygen as electron acceptor (5). For m-AldO, a midpoint redox potential of Ϫ109 mV was determined, which is higher than the average midpoint redox potential found for flavin-dependent dehydrogenases. This is in line with the observation that m-AldO shows all characteristics of a true oxidase.
As has been described for multiple other flavoprotein oxidases, m-AldO shows reversible reactivity with sulfite (K d ϭ 59 Ϯ 5 M), which forms a flavin N-5 sulfite adduct. When reduced by light, m-AldO forms and stabilizes the red anionic semiquinone form of FAD. Both observations are typical for oxidases and imply the presence of a positive charge near the flavin N(1) locus. A high reactivity of the reduced enzyme with molecular oxygen also is an indication for a protein to be a true oxidase (21). This is clearly the case, since bimolecular reoxidation rate constants of 1.7 ϫ 10 5 M Ϫ1 s Ϫ1 and 1.4 ϫ 10 5 M Ϫ1 s Ϫ1 were measured for reduced and product-bound reduced m-AldO, respectively. Steady-state analysis has shown that m-AldO is active with a narrow set of alditols. The highest catalytic efficiency was found with xylitol (K m ϭ 0.32 mM, k cat ϭ 13 s Ϫ1 ). Upon oxidation of the alditol by m-AldO, a primary alcohol is converted into the corresponding aldehyde. By HPLC and polarimetric analysis, it was shown that m-AldO acts as a primary alcohol oxidase, yielding solely D-xylose as product in the case of xylitol as substrate. This indicates that m-AldO has a high regioselectivity. It also appears that the chain length of the substrate is important for optimal substrate recognition and reactivity as m-AldO has a preference for C5 and C6 alditols.
Pre-steady-state kinetic experiments were conducted by studying the reductive and oxidative half-reactions separately. The reductive half-reaction showed three phases, leading from oxidized to fully reduced flavin. The first phase corresponds to the irreversible reduction of the flavin by the substrate with a reduction rate constant of 99 s Ϫ1 . A plausible explanation for the observed second phase is the following. During the first phase, flavin-mediated oxidation of xylitol takes place, and the initially formed D-xylose product will be in the open form. However, it is energetically favorable for D-xylose to adopt the closed hemiacetal conformation (Scheme 3). Therefore, it is tempting to assume to that the second kinetic phase, which is observed after flavin reduction, reflects such a rearrangement (the cyclization) of the product in the active site. Consequently, the second deconvoluted spectrum (spectrum 2 in Fig. 4) represents the binary complex of reduced m-AldO and the open form of the D-xylose product. The fact that this second spectrum is different from the fully reduced flavin spectrum can be explained by the close proximity of the formed aldehyde product, the open form of D-xylose, to the FAD cofactor. Interactions between the reduced cofactor and the bound product lead to perturbations of the flavin spectrum. A slight but significant increase in absorption at 520 nm in the second spectrum suggests that the interaction involves a charge transfer complex (19). Increasing xylitol concentrations have an effect on the observed rate of the rearrangement of the aforementioned binary complex. This suggests that xylitol can bind to the enzyme-product complex in such a way that it enhances the proposed ring closure. The third phase most likely represents product release from the reduced enzyme and has a rate constant that is too low to be catalytically relevant.
Reoxidation of reduced m-AldO yielded a bimolecular rate constant of 1.7 ϫ 10 5 M Ϫ1 s Ϫ1 , whereas a bimolecular rate constant of 1.4 ϫ 10 5 M Ϫ1 s Ϫ1 was found for product-bound reduced m-AldO. These are typical rate constants found for flavoprotein oxidases. For other oxidases, it has been shown that the bimolecular reoxidation rate depends to some extent on whether or not the product is bound to the reduced enzyme species (21).
The kinetic measurements were used to analyze the kinetic mechanism, which could be a ping-pong or ternary complex mechanism. Taking together the kinetic data of the reductive and oxidative half-reaction yields values for most kinetic parameters of the lower kinetic cycle shown in Fig. 5. From this cycle, which represents a ping-pong mechanism, and its associated kinetic constants, values for K m, xylitol , and k cat can be calculated using Equations 3 and 4. The calculated K m for xylitol yields a value of 0.034 mM, whereas the calculated k cat value is 2.6 s Ϫ1 . Both values are significantly lower than the K m and k cat values obtained with steady-state kinetic experiments (Table 2). Furthermore, it was found that the calculated value for the redox state of AldO during steady-state catalysis (2.6% in oxidized state) did not correspond to the measured redox state during steady-state catalysis (32% in oxidized state). So far, we assumed that the observed third phase in the reductive halfreaction (3.5 s Ϫ1 ) represents product release (k 4 ). However, we also simulated a scenario in which product release is not spectrally observable. For this, we varied the rate of product release from 3.5 s Ϫ1 up to 100 s Ϫ1 . The corresponding calculated kinetic parameters were not consistent with the measured kinetic parameters.
The incompatibility of the determined steady-state kinetic parameters and the calculated kinetic parameters based on a ping-pong kinetic mechanism suggests that another kinetic mechanism is operative. An alternative mechanism is represented in the upper cycle in Fig. 5. This upper cycle implies a ternary complex mechanism in which the reduced m-AldO, still having the product bound, is being oxidized and subsequently releases the product. Assuming a product release rate of 80 s Ϫ1 , we calculated a K m, xylitol , k cat , and redox state of 0.17 mM, 13 s Ϫ1 , and 29% in the oxidized state, respectively, from the ternary complex kinetic model (Equations 5 and 6). These values approach the measured steady-state kinetic parameters ( Table 2). The above mentioned calculations were performed by using the observed k 3 and k Ϫ3 values. However, it was shown by simulations that these observed rates are lower than the true rates, so the mechanistic calculations were also performed with k 3 and k Ϫ3 values of up to 80 s Ϫ1 and 21 s Ϫ1 , respectively. These variations did not result in significant changes of the calculated steady-state parameters and redox state. This suggests that m-AldO mainly follows a ternary complex mechanism during steady-state catalysis. This is further supported by the set of parallel Lineweaver-Burk plots that were obtained from O 2and xylitol-dependent steady-state kinetic measurements (see supplemental materials). Normally, this implies that a pingpong mechanism is operative. However, depending on some individual kinetic parameters, enzymes obeying a ternary complex mechanism can also yield parallel Lineweaver-Burk plots (25,26). By simulating these plots using the pre-steadystate kinetic parameters determined for m-AldO, a similar set of parallel lines was obtained (see supplemental materials). From Fig. 5 and all kinetic parameters, it is clear that the dominant type of kinetic mechanism depends on the concentration of O 2 . At higher O 2 concentrations, the ternary complex mechanism would be dominant, whereas at relatively low concentrations of O 2 , a ping-pong mechanism may partly be operative.
This study shows that m-AldO is an oxidase that can efficiently convert a number of alditols and, to a lesser extent, other polyols. It is known that carbohydrate oxidases are widely used in diagnostic applications, the food and drink industry, and carbohydrate synthesis. Based on the findings in this paper, m-AldO is a good candidate for further biocatalytic exploration in the field of biosynthesis and biosensors.

TABLE 2 Determined and calculated kinetic parameters for m-AldO
The calculated values were obtained from the kinetic constants in Fig. 5  a Numbers represent the percentage of m-AldO that is in the oxidized state.