Structural Insight into Substrate Differentiation of the Sugar-metabolizing Enzyme Galactitol Dehydrogenase from Rhodobacter sphaeroides D*

Galactitol 2-dehydrogenase (GatDH) belongs to the protein superfamily of short-chain dehydrogenases. As an enzyme capable of the stereo- and regioselective modification of carbohydrates, it exhibits a high potential for application in biotechnology as a biocatalyst. We have determined the crystal structure of the binary form of GatDH in complex with its cofactor NAD(H) and of the ternary form in complex with NAD(H) and three different substrates. The active form of GatDH constitutes a homo-tetramer with two magnesium-ion binding sites each formed by two opposing C termini. The catalytic tetrad is formed by Asn116, Ser144, Tyr159, and Lys163. GatDH structurally aligns well with related members of the short-chain dehydrogenase family. The substrate binding pocket can be divided into two parts of different size and polarity. In the smaller part, the side chains of amino acids Ser144, Ser146, and Asn151 are important determinants for the binding specificity and the orientation of (pro-) chiral compounds. The larger part of the pocket is elongated and flanked by polar and non-polar residues, enabling a rather broad substrate spectrum. The presented structures provide valuable information for a rational design of this enzyme to improve its stability against pH, temperature, or solvent concentration and to optimize product yield in bioreactors.

Dehydrogenases represent an important class of enzymes in biotechnological processes increasingly used in the chemical or pharmaceutical industry due to their enantio-and stereoselective oxidative and reductive catalytic properties (1)(2)(3)(4). A dehydrogenase with a promising catalytic potential is galactitol dehydrogenase (galactitol:NAD ϩ 5-oxidoreductase (GatDH) 3 ), an enzyme originally isolated from a galactitol-utilizing mutant of the bacterium Rhodobacter sphaeroides (5). GatDH is a homotetrameric protein that requires Mg 2ϩ for maintenance of its quaternary structure and enzymatic activity. It catalyzes the dehydrogenation of a variety of polyvalent aliphatic alcohols and polyols to the corresponding ketones and ketoses, respectively, and in the reverse reaction it reduces prochiral ketones with high stereoselectivity yielding the corresponding S-configured secondary alcohols (5)(6)(7). Based on these catalytic capabilities, GatDH was used with cofactor recycling (8,9) for several preparative conversions. Oxidation at C5 of galactitol gave the rare sugar L-tagatose in almost quantitative yields (6). L-Tagatose is a precursor for the synthesis of the glycosidase inhibitor 1-deoxygalactonojirimycin (10) and is a substrate of pyranose 2-oxidase yielding the interesting synthon 5-keto-psicose (11). Similarly, xylitol is oxidized at C4 to give L-xylulose (5). We also demonstrated the preparation of several (R)-1,2-diols by racemic resolution with GatDH as well as the synthesis of several S-configured aliphatic alcohols by reducing corresponding prochiral ketones (7). Furthermore, the suitability of GatDH for conversions with electrochemical cofactor regeneration was demonstrated with GatDH-bound covalently to the surface of a gold electrode (12). These examples indicate that GatDH has a considerable application potential, which is only limited because of intrinsic deficiencies of the enzyme such as low thermal and operational stability and restricted pH tolerance (5). A question that has not been addressed so far is its tolerance toward organic solvents in the case of alcohol conversions. There are also open questions concerning the region-and stereoselectivity of GatDH which, apparently, are influenced by the chain length and position of the carbonyl group of the substrate. With regard to polyol conversions, the starting materials for rare sugar syntheses, it seems that GatDH preferentially accepts polyvalent alcohols with D-threo-configuration adjacent to the primary alcohol group (Fig. 1). Thus, GatDH is not only an attractive biocatalyst for the synthesis of certain rare ketoses but also for asymmetric synthesis of enantiomerically pure diols (7). To provide a basis for a thorough understanding of the enzymatic properties of GatDH, structure determination of the substrate-free and substrate-bound enzyme in complex with its cofactor was performed. The presented structures provide insight into the properties of the active site of the enzyme and represent the prerequisite for optimization of this and similar dehydrogenases (13)(14)(15)(16)(17)(18)(19) that are available for biocatalytic processes, especially in electrochemical enzyme reactors.

EXPERIMENTAL PROCEDURES
Expression and Purification of GatDH-For production of recombinant GatDH from R. sphaeorides strain D, the respective gene was cloned into the vector pET24a (Novagen Inc.) with an N-terminal (His) 6 affinity tag. For removal of the affinity tag, a cassette coding for the tobacco etch virus-protease cleavage site was ligated into the plasmid. The new plasmid was transformed into Escherichia coli strain BL21(DE3)gold (Novagen). A 6-liter culture was grown in lysogeny broth (LB) medium supplemented with the antibiotic kanamycin (50 g ml Ϫ1 ) at 310 K until an A 600 of 1.5 was reached. The production of (His) 6 -tagged GatDH was induced by addition of isopropyl-␤-D-thiogalactopyranoside (final concentration, 0.5 mM) and expressed for 5 h at 303 K. The cells were harvested by centrifugation (10 min, 6,000 ϫ g, 277 K), washed in buffer (20 mM Bis-Tris, pH 6.5, 1 mM MgCl 2 , 0.1 mM phenylmethylsulfonyl fluoride), frozen in liquid nitrogen, and stored at 193 K. For purification, cells were thawed, resuspended in 120 ml of lysis buffer (20 mM phosphate buffer, pH 7.4, 500 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride), and disrupted at 16,000 p.s.i. by a cell homogenizer (Avestin Inc.). To digest the DNA, small amounts of DNase were added. After centrifugation (120 min, 75,000 ϫ g, 277 K) the supernatant was applied onto a nickel-nitrilotriacetic acid-superflow matrix (Qiagen), and the column was washed to baseline with 20 mM phosphate buffer, pH 7.4, 500 mM NaCl, 0.1 mM phenylmethylsulfonyl fluoride and 20 mM imidazole. His-tagged proteins were eluted in the same buffer in multiple gradient steps from 20 mM to 500 mM imidazole. Fractions containing (His) 6 -GatDH (identified by SDS-PAGE analysis, stained with Coomassie-Blue) were pooled and dialyzed against 50 mM Tris-HCl, pH 8.0, supplemented with 1 mM dithiothreitol and 10 mM EDTA. The affinity tag was removed with tobacco etch virus-protease S219V in a molar ratio of 1:10 (protease to protein). For the removal of uncleaved (His) 6 -GatDH and (His) 6 -tagged tobacco etch virus-protease, the cleavage reaction mixture was treated with nickel-nitrilotriacetic acid-matrix (Qiagen Inc.), equilibrated with 20 mM Bis-Tris, pH 6.5, and 1 mM imidazole. The supernatant, containing pure GatDH without the affinity tag (SDS-PAGE analysis, staining with Coomassie Blue) was dialyzed against 20 mM Bis-Tris, pH 6.5, and concentrated to 12 mg ml Ϫ1 (0.45 mM). The homogeneity of the purified protein was further verified by matrix-assisted laser desorption ionization time-of-flight. The activity of GatDH toward the used polyols was determined with an assay described by Schneider et al. (5).
Multiple Angle Light-scattering Measurements-For protein separation, the asymmetric flow field-flow fractionation technique (AFFF, Eclipse, Wyatt Technology) with a spacer of 490 nm and a cellulose membrane with 5-kDa cut-off was used. The system was connected to a UV-detector (Agilent), a multi-angle light-scattering detector (miniDAWN, ϭ 690 nm, Wyatt Technology) and a refractive index detector (Agilent). The mobile phase was 20 mM Bis-Tris buffer, pH 6.5, with and without 1 mM MgCl 2 , respectively. The analysis was carried out at a cross flow rate of 1.5-3 ml/min at room temperature (ϳ298 K). The protein concentration was 1 mg/ml in 20 mM Bis-Tris, pH 6.5, with or without 1 mM MgCl, and with or without 1 mM NAD ϩ or NADH, respectively. Finally, the molecular weight was calculated using the ASTRA software (Wyatt Technology).
Crystallization-All crystallization setups were performed at 291 K using the vapor-diffusion method with hanging drops. For crystallization of the holoenzyme, 1 mM NAD ϩ or NADH was added to the protein solution (0.45 mM) prior to crystallization. 1 l of protein solution (in 20 mM Bis-Tris, pH 6.5) was mixed with 1 l of reservoir solution and equilibrated against 1 ml of reservoir solution. Crystal screening was carried out using Crystal Screen TM , Crystal Screen 2 TM , and PEG/Ion Screen TM (Hampton Research). The best crystallization condition identified was used for further optimization using additive screens 1-3 TM (Hampton Research). The first crystallization condition contained 200 mM sodium acetate, 100 mM sodium cacodylate, pH 6.5, and 30% (w/v) PEG 8000. Under this condition crystals in space group P4 1 2 1 2 (with unit cell dimensions a ϭ b ϭ 109 Å, c ϭ 125 Å) grew within 1 week. Adding 4% (v/v) 2-propanol to the reservoir solution improved the crystal quality in respect to x-ray diffraction and resulted in crystals grown in the orthorhombic space group P2 1 2 1 2 1 (with unit cell dimensions a ϭ 98 Å, b ϭ 107 Å, and c ϭ 109 Å). These crystals were used for an initial crystal structure determination. For soaking and co-crystallization experiments, a second crystallization con-FIGURE 1. Some substrates and reactions of GatDH. A, GatDH performs the NAD ϩ -dependent regioselective dehydrogenation of the sugar alcohol galactitol at C5 producing L-tagatose and NADH. Similarly, xylitol is oxidized at C4 to L-xylulose (reaction not shown) (5). B, hydrogenation of short, aliphatic 1-hydroxy-ketones preferably yields the S-enantiomer of the 1,2-diol (5, 7). C, structures of substances co-crystallized with GatDH or used for soaking experiments of GatDH crystals. dition was optimized containing 100 mM MES, pH 5.5-5.9, 200 mM MgCl 2 , and 10 -20% (w/v) methoxy poly(ethylene glycol) 5000. Crystals produced under these conditions grew in the orthorhombic space group P2 1 2 1 2 1 (with unit cell dimensions a ϭ 65 Å, b ϭ 115 Å, and c ϭ 124 Å). Structures of GatDH with bound substrate were obtained utilizing two different strategies: co-crystallization with excess of substrate as well as soaking of protein crystals. For soaking of crystals, the reservoir solution from the second crystallization condition was supplemented with 1 mM NAD ϩ as well as with the substrate or putative inhibitor in concentrations in the range of 10 -100 mM. Crystals were soaked in increasing concentrations of MPEG 5000 (until 30 -35% (w/v)) within ϳ60 min. For co-crystallization the substrate was added to the protein solution in a final concentration of 5-50 mM prior to crystallization. To identify potential binding sites for bivalent metal ions, a protein solution containing 0.75 mM GatDH was dialyzed against a solution containing 10 mM EDTA to remove Mg 2ϩ . In the corresponding crystallization setups, MgCl 2 was replaced by 200 mM CoCl 2 in the protein and reservoir solution.
Data Collection, Processing, and Structure Refinement-Data collection was performed at cryogenic temperatures (100 K) after flash cooling of the crystals in liquid nitrogen. Depending on the crystallization condition either PEG 8000 or MPEG 5000 up to a final concentration of 30% (w/v) were used as cryoprotectants. X-ray diffraction data sets were collected on a sealed tube X-ray generator (IS, Incoatec Inc.) attached to a MARdtb goniostat and a MAR345 image plate detector (MAR Research, Norderstedt, Germany) or at various synchrotron beamlines (see Table 1). All data sets were indexed, integrated, and scaled using the program package XDS/XSCALE (20,21) or Mosflm/ Scala (22). The parameters for data collection and data processing are summarized in Table 1. For the phase determination of the first data set the molecular replacement method was used. Multiple sequence alignments with related proteins from the short-chain dehydrogenases/reductases family were performed with ClustalW (23). A homology model was generated as a starting search structure using the Swiss Model internet server (24). The modeling was based on the structure of gluconate 5-dehydrogenase from Thermotoga maritima (TM0441) (PDB entry 1VL8). The rotational and translational searches were performed with the program MOLREP (25). This first refined structure of GatDH was used as a search model to solve all subsequent structures of GatDH by molecular replacement. All structures were inspected and manually adjusted with the programs O (26) and COOT (27). Refinement was performed with REFMAC5 (28). Omit maps were generated by using the randomized omit map procedure (29). The coordinates of the questioned peptide regions were removed from the model, and each of the remaining coordinates was randomly translated up to 0.2 Å. This altered model was subjected to 10 rounds of restrained refinement with REFMAC5, and omit electron density maps with coefficients 2F obs Ϫ 1F calc were calculated. The quality of the refined structure was verified with the programs PROCHECK (30) and SFCHECK (31). The final refinement statistics for all structures are given in Table 1. Assignment of secondary structure elements was performed with DSSP (32). Identification of structurally related proteins in the Protein Data Bank (PDB) (33), was performed with the DALI server (34,35). Graphical representations were designed in PyMOL (36).

RESULTS
Overall Structure-The gene of GatDH extended by an Nterminal (His) 6 affinity tag followed by a tobacco etch virus protease cleavage site was heterologously expressed in E. coli, and the produced protein was purified to homogeneity via affinity chromatography and size-exclusion chromatography. The final amount of pure (His) 6 -GatDH was 80 mg/liter cell culture. The structure of GatDH with its cofactor NAD(H), but without substrate, was solved to 1.25-Å resolution ( Table 1). All 254 amino acid residues of the polypeptide chain and the cofactor NAD ϩ in the cofactor binding site were well defined in the electron-density maps.
Independent from the bound cofactor or the concentration of MgCl 2 , GatDH was found to be tetrameric in solution, as analyzed by size-exclusion chromatography as well as by static (dynamic light scattering) and dynamic (multiple-angle laser light scattering) light scattering (data not shown). In line with this finding, GatDH crystals displayed a homo-tetramer with point-group symmetry D 2 formed by a dimer of two homo-dimers (Fig. 2C). The interaction interfaces between monomer A1/A2 and B1/B2, coordinated through the helices ␣E and ␣F, are 1581 Å 2 , between monomer A1/B1 and A2/B2, coordinated through ␣F, ␣G, ␤F, the N terminus, and ␣B, are 1786 Å 2 . The contact area between A1/B2 and A2/B1 (452 Å 2 ) is formed through the last four amino acids of the C terminus and the loop preceding helix ␣F. One element of this interface is one magnesium ion binding site per dimer. Each magnesium ion is complexed by the two carboxyl groups of the C termini of two opposing peptide chains. Therefore, the homo-tetramer contains two magnesium binding sites not associated with the bound cofactor (Fig. 3).
NAD(H) Binding Site-The cofactor binding site in SDR enzymes is characterized by the conserved TGXXXGXG cofactor binding motif. In GatDH the corresponding TGAGSGIG segment (residues 18 -24) is located between the first ␤-strand (␤A) and the first ␣-helix (␣B) (Figs. 2 and 4A). This sequence is identical to the one in human estradiol 17-␤-dehydrogenase type 8 (PDB entry 2PD6) (40). Further fingerprints are the motifs NNAG (at position 86 -89) and PG (at position 189 -190), both involved in cofactor binding at the active site (reviewed in Refs. 41,42). In 70% of the cases, SDR enzymes with NAD(H) dependence comprise an aspartate residue at the C terminus of the second ␤-strand, and SDRs with NADP(H) dependence comprise an arginine residue (43). GatDH contains at this end of ␤B the sequence Asp 42 -Arg 43 . SDR enzymes con- Helices are represented by circles and ␤-strands by triangles. The nomenclature of secondary structure elements is according to 3␣/20␤-hydroxysteroid dehydrogenase (60). C, quaternary structure of GatDH. GatDH forms a homo-tetramer of point group symmetry D 2 . The individual protein chains are differently colored. Two magnesium ions (green and orange spheres) are coordinated each by two opposing C termini (A1-B2 and A2-B1, respectively). The active sites are positioned toward the surface of the protein as indicated by the NAD(H) molecules in ball-and-stick representation. The tetramer is oriented according to the PQR coordinate system (61) and displayed along the R-axis.

Structure of Substrate-bound Galactitol Dehydrogenase
taining this rare sequence are NAD-dependent (43), which is in accordance with the observed NAD(H) dependence of GatDH (5, 7). Asp 42 of GatDH discriminates against a phosphate group attached to the 2Ј-OH of the nicotinamide cofactor, whereas the side chain of Arg 43 interacts via -interactions with the adenine base NAD(H). The cofactor is bound in a deep cleft and is protected against the environment through the helix ␣FG1 (Figs. 2 and 4A). The cofactor binds via 22 H-bonds directly and via six well coordinated water molecules with the protein. One of these water molecules is hydrogen bonded to the first and to the last conserved glycine residues within the TGXXXGXG signature motif and to the side-chain hydroxyl group of Ser 92 at the end of ␤-strand ␤D. An equivalent water molecule is described as highly conserved in dinucleotide-binding proteins (44).
Mapping of the Substrate Binding Site-The active site in SDR enzymes is characterized by the sequence motif Tyr-X-X-X-Lys, with tyrosine as the most frequently conserved and the lysine second highly conserved residue in all members of the SDR family (45). In GatDH, Tyr 159 and Lys-163 form the catalytic tetrad together with Ser 144 and Asn 116 , which superimpose very well with the catalytic tetrad of human 3␤/17␤-hydroxysteroid dehydrogenase, suggesting a similar kind of reaction mechanism (41,46). The putative active site in GatDH is formed by a pocket at the end of the deep cleft in which the cofactor is bound. This cleft is predominantly shielded against the environment via helix ␣FG1 (Figs. 2 and 4A).
In GatDH without bound substrate, one large cavity next to the nicotinamide ring of the cofactor could be identified (Fig. 4 (B and C) and supplemental Fig. 3). This cavity has a volume of ϳ250 Å 3 , a length of ϳ14 Å, and a diameter of ϳ8 -10 Å with a small oval opening to the solvent environment of a diameter of ϳ2-5 Å. It is formed by residues 96 -98, 144 -146, 151, 154 -156, 159 -160, 189 -191, 196 -197, 200, 210, and the nicotinamide moiety of the cofactor. It displays mostly an apolar surface. Aliphatic molecules with up to eight carbon atoms and even cyclic compounds might fit within this cavity. The main-chain trace of the protein is displayed as a ribbon in gray. The carbon atoms of the cofactor NAD ϩ are colored in orange, and those of the amino acid residues are in green. One-letter codes are used to identify amino acids. A, electron density map around the cofactor and the substrate 1,2-(S)-pentanediol (PD, carbon atoms in magenta). The final A -weighted (2F obs Ϫ F calc ) electron density omit map drawn in blue and contoured at 2 was calculated after removal of the substrate from the model. Side chains in the surrounding neighborhood are displayed as sticks. The blue dashed lines mark the coordinating hydrogen bonds related to substrate binding and the catalyzed redox reaction. B, stick representation of the cofactor NAD ϩ and substrate 1,2-(S)-pentanediol (PD) superimposed with the mesh representation of the substrate binding pocket. The substrate binding pocket was analyzed with the program VOIDOO (62) and displayed with PyMOL (36). The cavity is displayed in a mesh representation and colored in orange for a probe radius of 1.4 Å (inner part of the binding pocket). The side chain of those amino acids that define the substrate binding pocket are displayed as sticks and are superimposed on the cofactor, the substrate, and the binding cavity. C, superposition of meso-erythritol (carbon atoms in cyan) together with its two coordinating water molecules (cyan spheres) onto the structure of GatDH with bound 1,2-(S)-pentanediol (carbon atoms in magenta) and the mesh representation of the substrate binding pocket in lime green for a probe radius of 1.1 Å (indicating the access path of the substrate). The two serine residues Ser 144 and Ser 146 are represented as sticks indicating their importance for the positioning of the substrate (in the case of 1,2-(S)-pentanediol) or the well coordinated water molecules within the substrate binding pocket. For clarity, the other amino acid residues of the binding pocket are not shown.
In crystals grown under crystallization condition I, a nearly tetragon-shaped electron density was found within the putative substrate binding pocket next to the nicotinamide ring. The interpretation of this electron density was ambiguous. One possible interpretation of this electron density is the two different binding modes of one acetate molecule averaged over the sampled crystal volume. Based on this interpretation, the carbonyl group of the acetate molecule is strongly bound via five H-bonds to side chains in the active site pocket (Ser 144 , Ser 146 , Asn 151 , and Tyr 159 ). To gain further insight into the mode of substrate binding we tried to map the substrate binding pocket with different substances. Because bound acetate might prevent potential substrates and/or inhibitors from binding, a second crystallization condition was optimized, which does not contain acetate or similar negatively charged substances. Based on the kinetic studies and identified substrates (5), we have performed soaking and co-crystallization experiments with 20 different compounds.
In co-crystallization experiments with a racemic (R,S)-mixture of 1,2-pentanediol (K m ϭ 1.2 mM) the electron density was of sufficient quality to verify that only the S-enantiomer bound to the active site (Fig. 4). Forming a hydrogen-bonding network, Ser 144 of the catalytic tetrad together with the side chains of Ser 146 and Asn 151 stabilize the primary, non-transformed hydroxyl-group of 1,2-(S)-pentanediol (Fig. 5).
The secondary, to be oxidized hydroxyl-group, forms two hydrogen bonds with Tyr 159 -OH and the amide group of the nicotinamide moiety of the cofactor. The aliphatic tail is positioned within the larger part of the binding pocket. A second binding site for 1,2-pentanediol could be identified within the crystal-packing interface between two GatDH tetramers. This binding induced a small relative rotation of the protein molecules causing a small change in the overall crystal packing and as a consequence a different unit cell compared with the other crystal structures of GatDH (Table 1).
Further complex structures were derived from crystals soaked with 10 mM meso-erythritol (displays 4% relative activity for oxidation compared with galactitol (5); K m could not be determined) or 10 mM xylitol (displays 410% relative activity for oxidation compared with galactitol (5); K m ϭ 22 mM). The electron density for these two polyols showed heterogeneity in the occupation of the cofactor binding pocket and the associated protein environment. This reflects the possibility of incomplete soaking. Therefore, the interpretation of the electron density within the substrate binding pocket was performed after the final refinement of the remaining protein, cofactor, and water structure. The final omit maps were of sufficient quality to unambiguously interpret the remaining electron density as one substrate (or product) molecule per binding pocket with additional water molecules in close proximity (Fig. 4).
Metal Dependence-The activity of GatDH was reported to be strictly dependent on the presence of divalent metal ions (5). Although the alcohol dehydrogenase members of the mediumchain dehydrogenases need Zn 2ϩ for their enzymatic activity (47), only a few members of the SDR enzymes are found to be dependent of divalent metals, e.g. the R-specific alcohol dehydrogenase of Lactobacillus brevis (48) and within the active site of gluconate 5-dehydrogenase of Streptococcus suis a Ca 2ϩ was identified (49). For identification of further binding sites for bivalent metals in GatDH, Mg 2ϩ was replaced by Co 2ϩ . Bound Co 2ϩ could be identified in electron density maps via their anomalous contribution to the x-ray scattering. A single-wavelength anomalous dispersion data set was collected, and an anomalous difference Patterson map was calculated. Although two strong electron density peaks at the C termini verified the cobalt ions bound to the magnesium binding sites in the interface of the tetramer, further binding sites could not be identified.

DISCUSSION
Based on typical sequence motifs (the NAD-binding TGXXXGXG motif, the active site YXXXK motif, and a conserved serine residue located in the active site) GatDH can be grouped into the subfamily of "classic" short-chain dehydrogenases (45, 50 -56).
Mapping the Substrate Binding Pocket-For the structural elucidation of the substrate binding site, we could solve the structures of GatDH in complex with its cofactor NAD ϩ and its substrates xylitol, meso-erythritol, and 1,2-S-pentanediol (Fig. 1C). Interestingly, these three compounds presented two different binding modes within the substrate binding pocket. Whereas in the case of 1,2-S-pentanediol, the substrate is bound in close proximity to the nicotinamide ring, in the case of meso-erythritol and xylitol the substrate molecules are positioned further away and well ordered water molecules are substrate-or product-induced inhibition such that a second bound molecule has to diffuse out of the binding pocket to clear the way for the converted molecule. However, such a mode of inhibition could not be detected in the case of meso-erythritol and xylitol, respectively. Further kinetic characterizations with a broader spectrum of substrates will clarify this point. Based on the elongated shape of the binding pocket, it seems to be more likely that acyclic sugar molecules are easier to be transformed than their bulky cyclic counterparts.
Conclusion-Kinetic characterization of the enzyme GatDH revealed a high degree of stereoselectivity within a widespread substrate spectrum covering sugars, sugar alcohols, secondary alcohols, or corresponding ketones. These characteristics make GatDH a very interesting enzyme in industrial biotechnology for the production of optically pure building blocks and the bioconversion of bioactive compounds. Here, we have presented different structures of GatDH in complex with its cofactor NAD(H). One high resolution structure represents the holoenzyme with cofactor but without bound substrate. Three structures could additionally be solved with different bound substrates. Together, these structures provide insight into the substrate binding pocket to rationalize the observed substrate spectrum and reaction selectivity. This information will help to further optimize the substrate specificity, the catalytic activity, and the stability toward pH, temperature, and solvents by directed evolution or rational design of enzyme variants. In addition it will help to rationalize the results of kinetic characterization of the enzyme and in silico docking experiments.