Structural Insight into Bioremediation of Triphenylmethane Dyes by Citrobacter sp. Triphenylmethane Reductase*

Triphenylmethane dyes are aromatic xenobiotic compounds that are widely considered to be one of the main culprits of environmental pollution. Triphenylmethane reductase (TMR) from Citrobacter sp. strain KCTC 18061P was initially isolated and biochemically characterized as an enzyme that catalyzes the reduction of triphenylmethane dyes. Information from the primary amino acid sequence suggests that TMR is a dinucleotide-binding motif-containing enzyme; however, no other functional clues can be derived from sequence analysis. We present the crystal structure of TMR in complex with NADP+ at 2.0-Å resolution. Despite limited sequence similarity, the enzyme shows remarkable structural similarity to short-chain dehydrogenase/reductase (SDR) family proteins. Functional assignments revealed that TMR has features of both classic and extended SDR family members and does not contain a conserved active site. Thus, it constitutes a novel class of SDR family proteins. On the basis of simulated molecular docking using the substrate malachite green and the TMR/NADP+ crystal structure, together with site-directed mutagenesis, we have elucidated a potential molecular mechanism for triphenylmethane dye reduction.

The expanding use of synthetic dyes such as triphenylmethane, azo, and anthraquinone is alarming, given that the release of colored compounds into the environment may cause substantial ecological damage, not only due to their color, which may have an impact on photosynthesis in aquatic plants, but also because many dyes and their breakdown products may be toxic and/or mutagenic to living organisms (1). Biological decolorization and degradation is an environmentally friendly and cost-competitive alternative to chemical decomposition (2). One key to efficient dye degradation is to use broad-spectrum and highly efficient dye-decolorizing microorganisms. Several microorganisms, such as bacteria, yeast, and fungi have been investigated for their ability to biodecolorize triphenyl-methane dyes (3)(4)(5)(6)(7)(8)(9). Partial biochemical studies of the decolorization of triphenylmethane dyes have been carried out in fungi. These studies revealed that the laccase, peroxidase, and lignin peroxidase enzymes from fungi are involved in the decolorization of dyes (10,11). Recently, a novel dye-decolorizing strain of bacteria, Citrobacter sp., was isolated from effluenttreated plant soil at a textile and dyeing industrial site. The extracellular culture filtrate of this strain showed a broad-spectrum decolorization efficiency for azo and triphenylmethane dyes (12). The enzyme that was responsible for the decolorization of triphenylmethane dyes was purified and biochemically characterized, and the gene encoding the enzyme was cloned (13). Based on biochemical analysis, the enzyme, designated TMR, 3 catalyzes the NAD(P)H-dependent reduction of triphenylmethane dyes as a homodimer in solution, and has a substrate specificity that is dependent on the chemical structures of triphenylmethane dyes. The tmr gene encodes a 287-amino acid protein, and sequence analysis revealed the presence of a dinucleotide-binding motif (GXXGXXG) in the N terminus. Recently, using a transformation-based approach, a 60-kb plasmid, pGNB1, containing a triphenylmethane reductase gene and encoding a protein that is 99.7% identical to TMR, was isolated from a bacterial community in the activated sludge compartment of a wastewater treatment plant. Escherichia coli carrying pGNB1 showed a very similar dye-decolorization spectrum as TMR (14). The tmr gene has also been found in the food-borne pathogens Listeria monocytogenes (15) and Aeromonas hydrophila (7). Although efforts aimed at expanding the bioremediation of dye pollutants have made some progress, research into the development of bioremediation technologies have generally focused on the isolation and characterization of microorganisms able to decolorize synthetic dyes. Key information on the molecular mechanisms underlying the degradation of dye pollutants remains to be elucidated.
In this study, we attempted to extend our knowledge of bioremediation using a structural approach. We crystallized TMR using the introduction of surface entropy reduction method, and solved the TMR structure at a resolution of 2.0 Å. We describe the three-dimensional structure of TMR in complex with the cofactor NADP ϩ . Structural inspection of TMR reveals that, although it exhibits a typical SDR fold, the func-tional signatures of this family of proteins are not conserved. On the basis of the crystal structure and supporting mutagenesis data, we propose a catalytic mechanism for triphenylmethane dye-decolorization by TMR.

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-The tmr gene encoding TMR from Citrobacter sp. strain KCTC 18061P was amplified from pET-TMR (13) by PCR. The gene was subcloned into the EcoRI and HindIII sites of the expression vector pHis-Parallel1 (17), a hexahistidine (His 6 ) fusion protein expression vector containing a recombinant TEV protease (rTEV) cleavage site, to generate the overexpression plasmid pHisP-TMR. Mutations were introduced into pHisP-TMR using QuikChange (Stratagene), according to the manufacturer's protocol. E. coli Rosetta-gami (DE3) (Novagen) harboring the overexpression plasmid was grown in LB-ampicillin medium at 37°C until the cultures reached an A 600 between 0.6 and 0.8. The temperature was lowered to 25°C, and protein expression was induced by treatment with 0.5 mM isopropyl-␤-D-thiogalactopyranoside for 14 h. The cells were harvested by centrifugation at 5,000 ϫ g for 20 min at 4°C. The cell pellets were resuspended in buffer A (50 mM Tris-HCl, pH 8.0, and 300 mM NaCl), and the cell suspensions were ultrasonicated. The crude cell extracts were centrifuged at 11,000 ϫ g for 1 h at 4°C. Cell lysate containing His 6 -tagged protein was loaded onto a 10 ml nickel-nitrilotriacetic acid agarose (Qiagen) column that was pre-equilibrated with buffer A at 4°C. The resin was washed with buffer A, and bound protein was eluted with buffer A containing 200 mM imidazole. The His 6 tag was then released from the protein by incubation with rTEV (Amersham Biosciences), followed by nickel-nitrilotriacetic acid agarose and size exclusion chromatography. After purification, the recombinant protein contained a seven-residue cloning artifact (GAMDPEF) in its N terminus. The homogeneity of the protein was assessed by 10% SDS-PAGE and Coomassie Blue staining. Purified protein was dialyzed against 20 mM Tris-HCl, pH 7.5, concentrated to 15 mg/ml using a YM-10 membrane (Amicon), and stored at Ϫ80°C until use. Enzyme assays and kinetic analysis were performed with purified His 6 -tagged protein in 50 mM Tris-HCl, pH 7.0. Gel-filtration analysis was carried out on a Superdex-75 10/30 column (Amersham Biosciences). The column was calibrated using the following molecular mass standards: ribonuclease (13.7 kDa), chymotrypsinogen (25 kDa), ovalbumin (43 kDa), and bovine serum albumin (67 kDa).
Crystallization-A Leu-21 to methionine (L21M)/L235M double mutation was introduced into TMR to facilitate selenomethionine (SeMet) incorporation, because there are no naturally occurring methionines in the protein other than the first and last residues. However, we were unable to successfully crystallize this mutant as an x-ray diffraction quality crystal, despite extensive efforts. Therefore, surface entropy reduction (18) was carried out to improve the crystal quality of the mutant protein.
Lys-22 and Lys-23 were replaced with alanine, and the SeMetlabeled L21M/K22A/K23A/L235M mutant protein was expressed in the methionine auxotroph strain E. coli B834(DE3) (Novagen), which was grown in minimal medium supplemented with 50 mg/ml SeMet under the same conditions as the cells containing the native plasmid. The purification procedure for the SeMet-substituted protein was identical to that of the native protein, except for the addition of 5 mM methionine to all of the buffers. All of the crystallization trials were carried out using the sitting-drop technique at 21°C. The initial trials were performed using the commercially available sparse-matrix screening kits Crystal Screen I and II from Hampton Research, and Wizard I and II from Emerald Biostructures. Initial crystals of native TMR were produced using 0.05 M KH 2 PO 4 and 20% polyethylene glycol 8000, and x-ray diffraction-quality crystals appeared in the presence of 0.05 M KH 2 PO 4 and 18% polyethylene glycol 8000 in 1 week. Striking results were obtained with the surface entropy reduced L21M/K22A/K23A/L235M mutant protein. The mutant yielded crystal hits under four different screening conditions during the initial screening stage (data not shown). The best crystals of the SeMet-substituted mutant were obtained using 0.28 M NaH 2 PO 4 /1.12 M K 2 HPO 4 , 0.2 M NaCl, and 0.1 M imidazole, pH 8.0, in 1 day.
Kinetic Analysis-The enzyme assay was performed as described previously by Jang et al. (13), with slight modifications. Typical assay mixtures contained 30 mM HEPES, pH 7.5, 20 M basic fuchsin, 0.1 mM NADH, and a suitable amount of the enzyme in a total volume of 1 ml. The reaction was initiated by the addition of the enzyme at 30°C and spectrophotometrically monitored at 544 nm. One unit of enzyme activity was defined as the amount of enzyme required to reduce 1 mol of basic fuchsin per minute using a molar absorption coefficient of 116,000 M Ϫ1 cm Ϫ1 . Kinetic parameters for basic fuchsin were collected at a concentration range of 2 to 80 M with a constant concentration of NADH (0.1 mM). The kinetic constants for NADPH were determined in the concentration range of 40 to 200 M in the presence of a fixed concentration of basic fuchsin (40 M) at 340 nm using a molar absorption coefficient of 6,220 M Ϫ1 cm Ϫ1 . At least three independent determinations were performed for each kinetic constant. All of the kinetic results were analyzed using SigmaPlot Version 8.0 software (SPSS Inc.).
Data Collection, Structure Determination, and Refinement-The structure of the L21M/K22A/K23A/L235M mutant was determined using the phases obtained from a single-wavelength anomalous diffraction experiment, with data collected at the selenium peak wavelength. Single-wavelength anomalous diffraction data for the SeMet-substituted mutant was collected at Pohang Accelerator Laboratory (PAL) beam line 6C at 2.5-Å resolution at Ϫ173°C using crystals that were frozen in a cryosolution consisting of 0.32 M NaH 2 PO 4 /1.28 M K 2 HPO 4 , 0.2 M NaCl, 0.1 M imidazole, pH 8.0, and 35% glycerol. The data for native TMR was collected at beam line 4A at PAL at 2.0-Å resolution, using crystals that were frozen in a cryo-solution consisting of 0.05 M KH 2 PO 4 , 25% polyethylene glycol 8000 and 10% glycerol. All of the x-ray diffraction data were processed and scaled using the HKL2000 software package (19). The SeMet-mutant crystal belongs to the space group P2 1 2 1 2 1 . There are four molecules in the asymmetric unit with a packing density of 3.17 Å 3 /Da, corresponding to an estimated solvent content of 61%. Anomalous differences from the selenium peak data set were used to locate the 12 selenium atoms using the program SOLVE (20). Density modification and subsequent automated model building were performed using the program RESOLVE (21). The initial RESOLVE model was used as a guide to manually build the remainder of the protein into densitymodified maps with the programs COOT (22) and O (23). The final cycles of refinement included translation-liberation-screw refinement of the isotropic displacement parameters performed with REFMAC5 (24). The cycles of manual rebuilding and refinement resulted in R free and R cryst factors of 23.9% and 19.0%, respectively. The model contains 1140 amino acid residues, including a cloning artifact (phenylalanine) at the N terminus of the molecule, and 242 water molecules. There are two missing amino acids (Lys-286 and Met-287) in the C terminus of each chain. The native TMR crystal belongs to the space group C222 1 . There is a single molecule in the asymmetric unit with a packing density of 2.51 Å 3 /Da, corresponding to an estimated solvent content of 51%. The structure was solved by molecular replacement using the refined model of the SeMet crystal with the program MOLREP (25). The refinement included the translation-liberation-screw procedure. Refinement resulted in R free and R cryst values of 24.2% and 20.0%, respectively. No density was visible for the seven-residue cloning artifact at the N terminus and two amino acid residues at the C terminus. The model contains 284 amino acid residues, a molecule of NADP ϩ , and 136 water molecules. Both native and SeMet TMR models satisfied the quality criteria limits of the program PROCHECK (26). The crystallographic data statistics are summarized in Table 1. The atomic coordinates and structure factor amplitudes of the native and SeMet proteins have been deposited in the Protein Data Bank (27) under the accession codes 2VRB and 2VRC, respectively.
Molecular Modeling of the Ternary Protein/Cofactor/Substrate Complex-Molecular docking of the substrate malachite green was performed using the structure of the TMR/NADP ϩ complex. Malachite green was automatically docked into the putative active site of the TMR/NADP ϩ complex by the program DS Flexible Docking in Discovery Studio, and the docked model was energy-minimized using the program CDOCKER in Discovery Studio 2.0 (Accelrys, Inc., San Diego, CA). The active site was defined with a 6-Å radius sphere from the putative catalytic site of TMR.

RESULTS
Overall Structure of TMR-Analysis of the primary amino acid sequence of TMR using a publicly available program failed to reveal any structure-related functional information. Consequently, we generated crystals of a SeMet-incorporated variant of TMR (L21M/K22A/K23A/L235M) and used these crystals to solve the three-dimensional structure of the protein. Similarly, the structure of native TMR was determined, using the structure of SeMet-TMR as a model. The overall TMR structure consists of two domains. The large N-terminal domain is comprised of a Rossmann fold with a central core of parallel ␤-sheets (␤1-␤6 and ␤8) flanked by seven ␣ helices (␣1-␣5, ␣7, and ␣11) and two 3 10 helices on both sides. The Rossmann fold generates an NAD(P)H-binding domain with a dinucleotidebinding motif ( 7 GXXGXXG 13 ). Typically, the Rossmann fold is composed of three parallel ␤ strands linked by two ␣ helices in the topological order ␤␣␤␣␤, and binds one nucleotide. Thus, the binding domain for dinucleotides such as NAD(P)H consists of two paired Rossmann folds that each bind one nucleotide moiety of the cofactor molecule. The cofactor-binding domain of TMR has three additional ␣-helices (␣3, ␣7, and ␣11) and a ␤-strand (␤8) that disrupt the symmetry of the Rossman folds. The small C-terminal domain contains a parallel ␤-sheet (␤7 and ␤9), five ␣ helices (␣6, ␣8 -␣10, and ␣12) and a 3 10 helix. This domain is assumed to function in substrate binding. The ␣10 and 3 10 helices, together with the last ␣ helix (␣5) of the Rossman fold domain form a hydrophobic pocket that is immediately adjacent to the nicotinamide ring of the cofactor (Figs. 1a, 3a, and 3c, also see below).
A search conducted using DALI (28) revealed that TMR is structurally related to the SDR fold family of proteins (29), which includes the human NADP signaling protein HSCARG (PDB code 2EXX, Z ϭ 26.8) (30), the Aspergillus nidulans negative transcriptional regulator NmrA (PDB code 1K6I, Z ϭ 24.3) (31), basil eugenol synthase (PDB code 2QZZ, Z ϭ 23.8) (32), Thuja plicata phenylcoumaran benzylic ether reductase (PDB code 1QYC, Z ϭ 23.6) (33) and Arabidopsis thaliana SQD1, which is involved in the biosynthesis of the sulfoquinovosyl headgroup of plant sulfolipids (PDB code 1QRR, Z ϭ 19.9) (34). This structural similarity would not have been readily inferred from sequence data alone, because the pairwise sequence identities of these proteins with TMR are very low (Ͻ19%). Key topological features in all of these proteins, however, are conserved, although the secondary structure content in the small C-terminal domain varies among the proteins (Fig.  1b). These differences in the C-terminal regions might lead to different substrate specificities, indicating that the SDR family is a functionally heterogeneous family of proteins. The most closely related structure to TMR appeared to be the NADP ϩbinding restructured HSCARG protein, with a root mean square deviation of 2.6 Å for 268 equivalent residues.
Quaternary Structure of TMR-Most of the SDR enzymes are dimers or tetramers in their active conformation (35). Although size exclusion chromatography revealed that native TMR behaves as a dimer in solution (data not shown), the crystal structure of native TMR has a single molecule in the asymmetric unit with the dimensions 58.8 ϫ 141.7 ϫ 37.8 Å. The crystal packing, however, reveals how the molecule generates a dimer contact with one other protein monomer. In addition, SeMet-TMR (L21M/K22A/K23A/L235M) contains four molecules in the asymmetric unit, which are arranged as a dimer of homodimers. A dimer with a 2-fold pseudosymmetry axis binds to another dimer with a second 2-fold pseudosymmetric axis perpendicular to the first axis. The long helices ␣4 and ␣5 (residues 79 -97 and 116 -132) in the large domain of each monomer pack against one another to form an antiparallel four-helix bundle at the subunit interface (Fig. 1c). Hydrophobic interactions by nonpolar residues are prominent at the interface, and a few hydrogen bonds also appear to play a major role in dimer stability. Residues involved in hydrophobic interactions at the interface are Thr-81, Leu-82, Ile-84, Val-85, Ala-88, Lys-92, His-119, Leu-122, Ala-123, Ala-127, and Thr-130. Hydrogen bonding side chains at the interface include Asp-79, Thr-81, Arg-95, and Thr-130. Asp-79 and Arg-95 also form salt bridges with each other. This dimer arrangement is consistent with the results of PISA (36). The monomer surface area buried in the dimer interface is as large as 917 Å 2 , whereas the next largest interface buries only 555 Å 2 of surface area. This interface has the highest possible PISA complexation significance score (CSS ϭ 0.76), compared with a CSS of 0 for all other possible interfaces. These results suggested that the dimer structure arranged by the antiparallel four-helix bundle represents the active conformation of TMR (Fig. 1d).
Cofactor Binding to TMR-Although no additional cofactor was added during expression, purification, or crystallization, early in the refinement of native TMR, it became evident that the protein contained a molecule of NADP(H). The cofactor groups were well defined by their electron density, with the exception of the nicotinamide moiety, which appeared to have a somewhat poorly defined electron density, particularly at positions C2, C4, and C5, in comparison to the rest of the cofactors (Fig. 2a). The cofactor appeared to be in the oxidized form according to liquid chromatography-tandem mass spectrometry analysis (see supplemental data). We co-crystallized native TMR with 10 mM NADP ϩ under the same crystallization conditions as native TMR to improve the quality of the electron density map of the cofactor. Under these conditions, the crystal diffracted to a 1.96-Å resolution, and the refined model is essentially the same as that of native TMR. The phenylalanine cloning artifact in the N terminus and two amino acid residues (Lys-286 and Met-287) in the C terminus are included in the model. There are also five glycerols in the model. The overall quality of the electron density of the cofactor was improved, but the nicotinamide ring was still poorly defined (Fig. 2b). These results suggest that there is flexibility at the C4 position in the catalytic cycle. The crystallographic data statistics for the model (PDB code 2JL1) are summarized in Table 1. Liquid chromatography-tandem mass spectrometry analysis of the SeMet-substituted TMR mutant was also performed (see supplemental data), and indicated that the SeMet protein is also complexed with NADP ϩ . Low resolution data may result in a poorly defined NADP ϩ electron density map of the mutant protein. The NADP ϩ binding site of TMR is topologically equivalent to that of other members of the SDR family, and the structural features of the cofactor binding site are virtually identical to those of HSCARG (PDB code 2EXX) and basil eugenol synthase (PDB code 2QZZ) (Fig. 1b). The location of the cofactor is also equivalent to that of the negative transcriptional regulator NmrA (PDB code 1K6I), phenylcoumaran benzylic ether reductase (PDB code 1QYC) and SQD1 (PDB code 1QRR) proteins (data not shown).
The cofactor is bound through a large number of polar and non-polar interactions (Fig. 2c). The adenine ring adopts an anti conformation to its sugar moiety and is stacked between the ␦-guanido group of Arg-34 and the phenolic group of Tyr-54. The adenine ring (AN1) is further stabilized by hydrogen bonding with the Tyr-54 backbone amide. The adenine amino group is hydrogen-bonded with the Asp-53 side chain OD1. As is common for dinucleotide-binding proteins, the adenine is selectively recognized by the conserved motifs DXXX and DXXD in classic and extended SDRs, respectively (29). In TMR, this motif consists of 53 DYNQ 56 . The adenine-ribose adopts a C2Ј-endo pucker conformation, and its hydroxyl group is hydrogen-bonded with the backbone amide nitrogen of Gly-10. The 2Ј-phosphate groups AOP1 and AOP2 are hydrogenbonded with the side chains OG1 of Thr-9 and NH2 of Arg-34, respectively. It is worthwhile noting that Arg-34 is the determinant basic residue of NADPH-preferring enzymes of the extended P1 SDR subfamily (29). The central diphosphate group forms hydrogen bonds with backbone amide nitrogens of Gln-11 and Leu-12, and with the guanidinium group of Arg-175. The nicotinamide ribose has the C3Ј-endo pucker confor- mation, and its hydroxyl groups are involved in hydrogen bonds with the carbonyl oxygens of Gly-75 and Ser-74. The nicotinamide ring adopts an anti conformation with its B-face stacked against the side chain of Tyr-143 and the A-face directed toward the substrate binding pocket. This orientation of the nicotinamide ring appears to also be influenced by interactions between the nicotinamide carboxyl and amide groups with the backbone amide of Tyr-143 and the carbonyl of Ala-141, respectively.
Critical residues involved in cofactor binding were assessed using site-directed mutagenesis. The cofactor binding residues Arg-34, Asp-53, Tyr-54, Tyr-143, and Arg-175, were mutated, as indicated in Table 2. Kinetic analysis revealed that replacement of these residues has a relatively small effect on K m , but decreased the turnover number of TMR by 40-to 700-fold (k cat ). Thus, the catalytic efficiency (k cat /K m ) was diminished 20-to 100-fold. These results, together with the crystallographic data, confirmed a key role for Arg-34, Asp-53, Tyr-54, Tyr-143, and Arg-175 side chains in NADP(H) binding.
Modeled Ternary TMR/NADP ϩ /Malachite Green Complex Structure-To gain insight into the catalytic mechanism of TMR, we prepared co-crystals of TMR in complex with cofac-tor and triphenylmethane substrates, such as malachite green and crystal violet, or their reduced products. However, these efforts were unsuccessful. We then performed a simulated docking of malachite green to the putative substrate binding pocket of TMR (PDB code 2VRB), which is easily identifiable by comparison with homologous structures and is lined by Pro-76, Tyr-78, Tyr-107, Ile-115, Leu-117, His-121, Leu-142, Phe-146, Phe-147, Phe-244, Ile-247, Ala-250, and Ile-251. The calculated volume of the substrate binding pocket was 943.2 Å 3 (Fig. 3c) (37). Malachite green docked well into the putative substrate binding site of the TMR⅐NADP ϩ binary complex, located in the immediate vicinity of the nicotinamide ring, similar to other SDR family enzymes, making ideal interactions with the cofactor. The volume and hydrophobic environment of the pocket were consistent with the size and structural properties of the substrate. In the model, the distance between NC4 of the nicotinamide ring and C10 of malachite green is ϳ3.8 Å, which is compatible with a direct hydride transfer from NADPH to the substrate (Fig. 3, a and b). The aliphatic and aromatic residues in the substrate binding site play a crucial role in substrate interaction. The phenylmethyl moiety of malachite green is anchored to a pocket formed by Leu-142, Phe-146, Phe-147, FIGURE 2. Interactions between TMR and the NADP ؉ cofactor. a, the F o Ϫ F c electron density map for NADP ϩ . b, improved electron density of the cofactor derived by co-crystallization of native TMR with 10 mM NADP ϩ . The electron density maps were contoured at 2.5 . c, schematic diagram depicting the non-covalent interactions between NADP ϩ and the binding site residues of TMR (PDB code 2VRB). The interactions were identified with the program LIGPLOT (16).
and Ile-251. The phenyl ring of the substrate is stacked with the phenyl of Phe-146 through ainteraction and stabilizes enzyme-substrate interactions that facilitate a proper distance for the hydride transfer from NADPH to malachite green. The dimethylaminophenyl moiety of malachite green makes hydrophobic contacts with Tyr-107, Ile-115, Leu-117, and His-121. The other dimethylaminophenyl moiety forms hydrophobic contacts with Pro-76, Tyr-78, and Ile-247. The pseudo-phenyl ring of the moiety is stacked against the cofactor nicotinamide ring, which is a common feature of the substrate binding modes of SDR family enzymes. The hydroxyl group of Tyr-78 interacts with the amino group of the dimethylaminophenyl moiety and allows the dissipation of the delocalized positive charge of malachite green. Tyr-78 located in the loop between ␤4 and ␣4 is highly flexible in the SeMet-substituted mutant TMR structure and swings in toward the catalytic center and out into the solvent active site, exhibiting remarkably different side-chain conformations in each chain (Fig. 3d). It is also highly flexible in the structure of native TMR. Thus, it is likely that Tyr-78 plays a role in substrate recruitment into the catalytic site.
To further explore the roles of potential key residues involved in substrate recognition, we performed site-directed mutagenesis of the targeting residues ( Table 3). The mutations Y78A, I115A, L142D, and I247A resulted in a 2-to 50-fold decrease in k cat and a 4-to 7-fold decrease in catalytic efficiency (k cat /K m ). The mutations L117A, H121L, and I251A did not result in significant differences in catalytic efficiency as compared with wild-type TMR, whereas L117D, H121D, and I251S resulted in substantial decreases in catalytic efficiency. These results support the hypothesis that hydrophobic interactions between TMR and its substrates are key for catalysis. It is worthwhile noting that the positions of Leu-117 and His-121 correspond to the positions of the strictly conserved catalytic residues Tyr and Lys, respectively, in other SDR enzymes (see "Discussion").

TMR Comprises a Novel Class of SDR Family Proteins-The
SDR enzyme family has great functional diversity, with amino acid sequence identities that are only 15-30%. Their substrate spectrum ranges from alcohols, sugars, steroids, and aromatic compounds to xenobiotics (35). Despite low sequence identity, the three-dimensional structural folds can be superimposed on specific conserved sequence motifs that exhibit only small deviations. Variations in these motifs have been used to define SDR subfamilies. As expected, TMR contains an NAD(P)H-binding-   (29). The presence of the NADP(H)-preferring sig-nature basic residue Arg-34 located at the first loop position after the second ␤-strand suggests that TMR can be further classified into the extended P1 SDR subfamily (29). On the other hand, the adenine amino group recognition motif 53 DXXX 56 in TMR indicates that the enzyme might belong to the classic SDRs. Surprisingly, functional assignments revealed that the TMR active site is unique. In general, SDR family enzymes have two invariant catalytic residues, Tyr and Lys, which are present in a conserved YXXXK (or YXXMXXXK) motif located in ␣5 (29). In addition, a Ser residue in a conserved GXXXXXSS or SSXXXXG motif located in ␤5 is also partly involved catalysis. These conserved catalytic residues are missing in TMR. The 117 LXXXH 121 motif is present in the corresponding position, and although there is no equivalent Ser residue, Tyr-107 is present. These results suggest that TMR does not belong to any of the existing SDR subfamilies and may represent the first member of a new class of SDR family  enzymes. These differences might also explain why we were unable to extract structural or functional information from the primary amino acid sequence of TMR using the public data base. Sequence alignment using BLAST (38) revealed that the positions of the Leu and His residues of TMR are well conserved among a number of bacterial proteins of unknown function, showing similarity levels of 53-99% (Fig. 4). The putative substrate-binding residues in these proteins are similar to those of TMR, with a few exceptions. The NADPH determinant Arg and the adenine recognition Asp are also strictly conserved in this group of proteins. However, it is not clear whether these proteins possess similar functions to TMR. Nonetheless, the distinct active site and the presence of highly conserved sequence motifs that are a mix of classic and extended SDR family members suggest that these proteins define a novel class of SDR family enzymes. Implications for Substrate Binding and Electron Transfer-The proposed catalytic mechanism of the typical SDRs is governed by a catalytic triad of Ser(Thr)-Tyr-Lys that is involved in oxidoreductase activity. Previous studies support the concept that, in general, the Tyr residue functions as the catalytic base, whereas the Ser (or Thr) residue stabilizes the substrate and the lysine forms hydrogen bonds with the nicotinamide ribose hydroxyl group and assists in the reprotonation of the tyrosine residue during catalysis (39). The standard reduction of SDRs is considered to take place through two steps: first, the cofactor . Structure-related functional sequence conservation between TMR and homologous proteins. Shown above the alignments are elements of the secondary structure of TMR. The numbering shown is from TMR. Red triangles indicate the dinucleotide-binding motif. Blue triangles indicate critical cofactor binding residues. Closed green circles represent residues involved in the putative substrate binding hydrophobic pocket. Purple stars represent putative catalytic residues located at the corresponding positions of the conserved catalytic residues Tyr and Lys, respectively, in the SDR family proteins. Strictly conserved residues are highlighted with red boxes. Biological sources and accession codes for the sequences are as follows: EC, Erwinia carotovora subsp. atroseptica SCRI1043 (gi:49613893); SC, Streptomyces coelicolor A3(2) (gi:21222976); YE, Yersinia enterocolitica subsp. enterocolitica 8081 (gi:123442037); SF, Shigella flexneri 2a str. 301 (gi:24115476); BL, Bacillus licheniformis ATCC 14580 (gi:52784068); and EColi, E. coli 536 (gi:110644577). Sequence alignments were assembled using T-COFFEE software and visualized using ESPript software, both located on the ExPASy Proteomics Server (au.expasy.org/). hydride is transferred onto a polarized double bond in the substrate molecule, and second, the emerging negative charge on the substrate is neutralized in a subsequent protonation, probably using the catalytic triad. When TMR is superimposed with SQD1 (PDB code 1QRR) (34), one of the most structurally closed enzymes, the catalytic residues Thr, Tyr, and Lys are replaced with Tyr-107, Leu-117, and His-121, respectively, in TMR, suggesting that the catalytic chemistry of TMR is different than other SDRs (Fig. 5). In our mutagenesis analysis, L117D and H121D resulted in a 52-and 7-fold, respectively, decrease in catalytic efficiency, whereas L117A and H121L mutants had a similar catalytic efficiency as wild-type TMR (Table 3). These results imply that Leu-117 and His-121 function as catalytic residues in TMR through a hydrophobic interaction system with substrates, rather than the proton-relay reaction system that occurs in typical SDRs via hydrogen-bond interactions. Structural inspection and the docking and mutagenesis studies provide a basis for the following model of triphenylmethane dye reduction by TMR. The nicotinamide ring of the NADPH cofactor is situated in a cleft between two domains of TMR, with the nicotinamide ring in an anti conformation, thus ensuring the position of the C4 atom, which would permit the pro-R hydride transfer to be directly apposed to the substrate. The triphenylmethane dye (i.e. malachite green) is located in the proposed substrate binding pocket through hydrophobic interactions, as described in the docking model. In this case, the C10 atom of malachite green is positioned close to the C4 atom of the nicotinamide ring. Such positioning of the nicotinamide ring and the substrate would be ideal for direct hydride attack on C10 and reduction of the dye into colorless leucomalachite green (Fig. 6). However, we cannot rule out the possibility that the conserved catalytic mechanism of SDR family members is present in TMR in vivo in Citrobacter sp. If this is the case, then His-121 and Tyr-107 might be involved in a proton-relay catalytic reaction, and Leu-117 probably contrib-utes to hydrophobic interactions with the substrate. A detailed understanding of the catalytic mechanism of TMR with real substrates in Citrobacter sp. remains to be elucidated.
Previous results showed that TMR decolorization activity is dependent on the chemical structure of the dyes (13). Our model, in combination with a structural inspection of the substrate binding pocket, might help explain the substrate specificity of TMR. The most efficient TMR substrate appeared to be malachite green, because it exhibited favorable structural features when modeled with the ternary complex. Crystal violet was a less favorable TMR substrate than malachite green, perhaps because of the additional dimethylamino group. The corresponding phenylmethyl moiety of malachite green is anchored to a pocket formed by Leu-142, Phe-146, Phe-147, and Ile-251, and stabilizes the enzyme-substrate interaction through astacking system with the phenyl of Phe-146. The magenta dye basic fuchsin, which has a positively charged amino group in each phenylmethane moiety, may be a less favorable substrate in the hydrophobic environment of the substrate binding pocket as compared with malachite green. The structures of other dyes, such as brilliant green, bromphenol blue, methyl red, and Congo red, suggest that they would be incompatible with the size and hydrophobic restrictions of the pocket environment, resulting in no TMR activity in the presence of these dyes.
In summary, structural analysis of TMR enabled us to distinguish it as a unique member of the SDR family of proteins, in terms of functional motifs and its potential catalytic chemistry. Structure-based mechanisms of decolorization of triphenylmethane dyes will provide essential clues for the development of bioremediation systems.