Critical residues for structure and catalysis in short-chain dehydrogenases/reductases.

Short-chain dehydrogenases/reductases form a large, evolutionarily old family of NAD(P)(H)-dependent enzymes with over 60 genes found in the human genome. Despite low levels of sequence identity (often 10-30%), the three-dimensional structures display a highly similar alpha/beta folding pattern. We have analyzed the role of several conserved residues regarding folding, stability, steady-state kinetics, and coenzyme binding using bacterial 3beta/17beta-hydroxysteroid dehydrogenase and selected mutants. Structure determination of the wild-type enzyme at 1.2-A resolution by x-ray crystallography and docking analysis was used to interpret the biochemical data. Enzyme kinetic data from mutagenetic replacements emphasize the critical role of residues Thr-12, Asp-60, Asn-86, Asn-87, and Ala-88 in coenzyme binding and catalysis. The data also demonstrate essential interactions of Asn-111 with active site residues. A general role of its side chain interactions for maintenance of the active site configuration to build up a proton relay system is proposed. This extends the previously recognized catalytic triad of Ser-Tyr-Lys residues to form a tetrad of Asn-Ser-Tyr-Lys in the majority of characterized short-chain dehydrogenases/reductase enzymes.

Since the discovery of fundamental differences between insect-type and liver-type alcohol dehydrogenases (1), corresponding to the protein families of "short-chain" dehydrogenases/reductases (SDR) 1 and "medium-chain" dehydrogenases/ reductases (MDR), respectively, SDR enzymes have received much attention. They constitute a large protein family with well over 2000 annotated enzyme and species variant sequences in databases and are represented in all life forms with minimally 60 genes found in the human genome (2)(3)(4). The SDR enzymes span several EC classes, from oxidoreductases and lyases to isomerases, with NAD(P)(H)-dependent oxidoreductases constituting the majority of forms. In this class, many enzymes with different specificities act on steroids, prostaglandins, aliphatic alcohols, and xenobiotics.
The pairwise sequence identity between different enzymes is low, typically 10 -30%, but all available three-dimensional structures (ϳ20) display a highly similar ␣/␤ folding pattern (5)(6)(7)(8)(9)(10)(11). Most SDR enzymes have a 250 -350-residue core structure, frequently with additional N-or C-terminal transmembrane domains or signal peptides. Conserved sequence regions cover a variable N-terminal Gly-X 3 -Gly-X-Gly motif as part of the nucleotide binding region and the active site with a triad of catalytically important Ser, Tyr, and Lys residues, of which Tyr is the most conserved residue within the whole family (2,8,12) (Fig. 1). The functions of the residues at these particular sites have been elucidated by a combination of chemical modifications, sequence comparisons, structure analyses, and site-directed replacements (2,13). However, several other conserved but still variable residues have not been analyzed in detail. To understand their role, we now carried through a mutagenesis study using bacterial 3␤/17␤ hydroxysteroid dehydrogenase (3␤/17␤-HSD) analyzing effects on enzymatic function and the stability of replacements, assisted by x-ray crystallography and docking analysis. Other residues have been studied before (13), but novel segments investigated include Asp-60 (between ␤C and ␣D), an NNAG motif (Asn-86 -Gly-89 in ␤D), and Asn-111 in ␣E (Fig. 1).
SDS-PAGE and Protein Determination-Purity of protein samples was assessed by SDS-PAGE on 10% gels. Protein concentration was determined by amino acid analysis of pure samples after hydrolysis in 6 M HCl followed by ninhydrin-based quantification on an LKB Alpha plus analyzer.
Biophysical Characterization of 3␤/17␤-HSD and Mutants-Analytical ultracentrifugation was performed using a Beckman XL-1 ultracentrifuge. For sedimentation equilibrium and velocity studies, both absorbance at 280 nm and interference were recorded at different rotor speeds and protein concentrations at 20°C. Sodium phosphate buffer (30 mM, 150 mM NaCl, pH 7.5) was used. Data evaluation was performed using the Origin 4.0 software package. Partial specific volumes of wild-type and mutant proteins and buffer densities were calculated using the program Sednterp. Double sector cells were used for sedimentation velocity studies. Interference scans were recorded in intervals of 2 min.
Assessment of wild-type and mutant conformations was achieved by circular dichroism spectroscopy by recording the ellipticity as a function of wavelength between 260 and 195 nm using an AVIV Model62 DS spectropolarimeter. Conformational stability was determined by titration of the individual proteins with guanidine-HCl using a titration robot and by monitoring CD at 222 nm as described earlier (12,13).

RESULTS AND DISCUSSION
Overexpression, Purification, Folding, and Stability Analysis of Wild Type and Mutants of 3␤/17␤-HSD-Wild-type and mutant 3␤/17␤-HSD forms were overexpressed in Escherichia coli strain BL21 and purified by metal-chelate chromatography (Fig. 2). Proteins were analyzed regarding folding, stability, coenzyme binding, and steady-state kinetics. Judged from CD spectroscopy, wild-type and mutant forms displayed essentially identical spectral characteristics, indicating similar secondary structure properties. Stability measurements were performed by titration with guanidinium hydrochloride, monitored by CD spectroscopy and by determination of the transition temperatures using differential scanning calorimetry. Judging from these experiments, no significant differences between wild-type and mutant proteins could be detected, indicating similar fold and conformational stability (data not shown).
Determination of Kinetic and Binding Constants-Steadystate kinetics were determined with different substrates for oxo-reductase and ␤-dehydrogenase reactions at C3 and C17 of the steroid. The kinetic and coenzyme binding data obtained are summarized in Tables I and II. Bacterial 3␤/17␤-HSD is able to catalyze specific dehydrogenations/reductions at positions 3 and 17 of steroids with different conformations, i.e. being trans-or cis-configured between ring A and B. First order constants k cat /K m for dehydrogenase reactions are an order of magnitude higher (1.7 ϫ 10 6 s Ϫ1 M Ϫ1 for 17␤-HSD activity, 1.0 ϫ 10 6 s Ϫ1 M Ϫ1 for 3␤-HSD activity with isoUDCA) as com-     The mutants compared for dehydrogenase and reductase activities have residue exchanges at positions Thr-12, Asp-60, Asn-86, Asn-87, Ala-88, Asn-111, Ser-138, and Tyr-151, surrounding the coenzyme binding region or located at the active site. Depending on the amino acid substitution, differential effects on enzymatic constants were observed. Completely inactive enzymes were obtained with the N111L, S138A, and Y151F mutants, whereas partially active enzymes with significant changes were observed for the other substitutions ( Table I).
Replacement of Thr-12 by Ala results in an enzyme that is only able to catalyze the reductive reaction at position C3 and the corresponding dehydrogenase activity with DHEA as substrate. All other activities are not detectable. Exchange to Ser results in an enzyme with enzymatic properties largely similar to those of the wild-type form (Table I). A profound change in cofactor binding (Table II)    Taken together, the drastic reduction in k cat /K m values for the oxidative reaction of the Asp-60, Asn-86, Asn-87, and Ala-88 forms is mainly due to increased K m for NAD ϩ without significant decrease in binding affinities, indicating inhibition of substrate formation without affecting association of the enzyme-substrate complex. The k cat /K m values for the reductive reactions are slightly lowered, accompanied by a decrease in NADH binding. Steroid structure is a further contributing factor to the differential decrease in oxidation since differences between isoUDCA and DHEA are observed.
Determination of Kinetic and Binding Constants of Active Site Mutants-Mutations performed within or close to the active site were residues Ser-138 and Tyr-151, recognized previously to be part of a catalytic triad with Lys-155, and the conserved Asn-111, located within helix ␣E, which forms the main subunit interaction surface. Mutants N111L, S138A, and Y151F are enzymatically inactive. Coenzyme binding constants of these mutants are changed but within the range observed with the mutants described above (from 0.2 M for NAD ϩ for Y151F to 12 M for NADH for S138A). As described earlier (13), substitution of Ser for Thr at position 138 yields an active enzyme. Loss of activity observed for S138A and Y151F mutants thus underscores the critical role of these residues in catalysis.
Crystallographic Analysis of 3␤/17␤-HSD and Comparison with High Resolution SDR Structures-The apo structure of wild-type 3␤/17␤-HSD was determined by x-ray crystallography to a resolution of 1.2 Å. 2 The enzyme forms a tetramer with subunit interactions similar to those of tetrameric SDR structures (9,10,17). The geometry of coenzyme binding and active sites was sufficiently similar to allow docking analysis of a 3␤/17␤-HSD ternary complex with NAD ϩ and a steroid substrate (3␤-OH-5-androsten-17-one) (Figs. 3 and 4).
The coenzyme is bound through few specific contacts, performed through other residues than Thr-12, Asn-86, and with indicated exceptions, Asn-87. Among the residues investigated, Asp-60 and Ala-88 contribute directly to coenzyme binding, a view supported by the significantly altered coenzyme constants. In the modeled structure, the carboxyl side chain of Asp-60 is in weak H-bonding distance to the adenine ring (OD1 Asp-60-N6A 3.99 Å) (Fig. 2), and similar interactions between Asp-60 and coenzyme are observed in, for example, MLCR, 7␣-HSD, or CR (9,10,20). Side chain to backbone interactions are also observed (OD2Asp-60-OGSer-62: 2.7 Å), and the main role of Asp-60 appears to be in the stabilization of the turn between ␤C and ␣D as part of the adenine ring binding pocket (10). Ala-88 can make hydrophobic contacts to the ade- (5␣-androstane, 3-one, 17ol). Catalysis is initiated by proton transfer from Tyr-151 hydroxyl to the substrate carbonyl followed by hydrid transfer to C3 of the steroid. A proton relay is formed and involves the 2ЈOH of the ribose, the Lys-155 side chain, and a water molecule bound to the backbone carbonyl of Asn-111. ARPP, the adenosine ribose pyrophosphate moiety of NADH. nine ring and thus contributes to binding of the coenzyme. These hydrophobic interactions of Ala-88 appear to be an important characteristic since mutation to Ser significantly changes activities.

FIG. 5. Postulated reductive reaction mechanism of 3␤/17␤-HSD involving NADH and steroid substrate
Reaction Mechanism of 3␤/17␤-HSD-The previously determined triad of Ser-Tyr-Lys residues (positions 138, 151, and 155, respectively in 3␤/17␤-HSD) constitutes the active site (2) (Fig. 4). Our data extend this concept by addition of an essential Asn-111 to this triad to form an active site tetrad. Previous studies support the concept that Tyr-151 functions as the catalytic base (2), whereas Ser-138 stabilizes the substrate, and Lys-155 forms hydrogen bonds with the nicotinamide ribose moiety and lowers the pK a of the Tyr-OH to promote proton transfer (Fig. 5). In the 3␤/17␤-HSD apo structure, water molecules are bound to the Tyr-OH and Lys side chain, thus mimicking substrate and ribose hydroxyl group positions. Determination of the Drosophila alcohol dehydrogenase structure (21) revealed interaction of the conserved Asn-111 via a water molecule, binding to the active-site Lys-155, and this interaction is also observed in the present structure (OAsn-111-H 2 O: 2.76 Å). In the Drosophila alcohol dehydrogenase structure, a large hydrogen-bonded solvent network including the water molecule bound by Asn-111 and Lys-155 was found (21), thereby substantiating our assumption of a proton relay with access to bulk solvent molecules (cf. below). Inspection of available three-dimensional SDR structures reveals interactions of Asn-111 (Table III) similar to those in 3␤/17␤-HSD and Drosophila alcohol dehydrogenase, indicating a homologous role of Asn-111 in all these cases. Out of 20 SDR structures retrieved from the Protein Data Bank, 16 contain Asn at the position homologous to the one in 3␤/17␤-HSD, show similar side chain/ backbone interactions, and display the feature of having a connecting water molecule to the active site lysine. Moreover, the four structures without a homologous Asn-111 contain a Ser residue, which is connected through a water molecule to the active site Lys. An extended network is created in mouse sepiapterin reductase built (Protein Data Bank accession number 1nas) with an additional Arg residue involved. Based on this general configuration, we conclude that Asn-111 is important to stabilize the position of Lys-155, and furthermore, that a proton relay is formed in most if not all SDR structures at the active site, including coenzyme, substrate, Tyr-151, ribose 2ЈOH, Lys-155, water, and Asn-111 or a corresponding Ser (Figs. 4 -6). A proton relay system involving water and essential ribose contacts to the catalytic base similar to that found in horse liver alcohol dehydrogenase (22) has been postulated earlier also for SDR (7); however, here we provide direct evidence for a critical involvement of Asn-111 in this process. The stabilization of the active site geometry is thus achieved through maintaining the Lys-155 position and furthermore through Asn side chain interactions with main chain atoms of residues located within the segment preceding helix ␣E (Table III). Asn-111 is located within ␣E, the main dimerization interface in oligomeric SDRs. Notably, at this position, the helix forms a sharp kink. This motif created by the side chain of Asn-111 presumably forces its backbone carbonyl group to bind a water molecule instead of the amide group (Val-115) that would have been expected in an ␣-helical structure. Moreover, we found other water molecules in the same cavity, forming a small water-rich enclosure inside the protein fold. In 3␤/17␤-HSD, we observed four water molecules, and in other SDR structures, the number ranges from two to five. This waterfilled hydrophilic cavity is lined by other well conserved amino acids: the side chains of Thr-12 (for Protein Data Bank accession number 1a4u, Thr-114; for 1dhr, Thr-110; and for 1a27, Thr-118), Asn-86, and Ser-114 and the main chain carbonyl group of Ala-88. This hydrophilic pocket serves as a proton relay system by apparently acting as a proton bridge between Lys-155 and the bulk solvent, similar to a temporary proton reservoir during enzymatic catalysis (Fig. 6), and it can presumably stabilize a positively charged Lys-155. The proton transfer in 3␤/17␤-HSD between W1 and W2 can be mediated via a carbonyl group (Fig. 6), proceeding through an enolic intermediate (Fig. 7). This mechanism explains the necessity for a small amino acid (Gly-89) within the NNAG motif since the rotational freedom of a small residue at that position will more easily allow a positive charge to be temporarily localized on the amide group (Fig. 7, step 2). The amino acids (Thr-12, Asn-86, Ala-88, and Ser-114) lining the hydrophilic pocket appear to have a secondary role in stabilizing these water molecules and in allowing proton transfer among them, inside an otherwise hydrophobic environment (Fig. 6).
To exclude involvement of Asn-111 in oligomerization processes, we performed sedimentation analysis and found no changes between wild type and N111L. Thus Asn-111 interactions are not essential to subunit associations. We postulate that the Asn-111 side chain interactions to Ile-90 and Ser-154 (OD1Asn-111-NI90: 2.80 Å; ND2Asn-111-OIle-90 3.03 Å; ND2Asn-111-OGSer-154: 3.06 Å) and homologous interactions (Table III) are necessary to position the Asn-111 backbone carbonyl for binding of the water molecule, which participates in the postulated proton relay and stabilizes the geometry of active site residues. This view is supported by mutational analysis of a homologous Ser (Ser-154 in 3␤/17␤-HSD) in type 1 11␤-hydroxysteroid dehydrogenase, showing a critical involvement of this residue in catalysis (23).
Role of Conserved Residues in SDR Enzymes-With over 2000 sequences annotated in databases and about 20 crystal structures determined, a picture of the general SDR architecture and mechanism emerges. Considering all sequences, no strict positional conservation is noted. However, multiple sequence alignments revealed several consensus motifs, the most conserved being the N-terminal TGX 3 GXG around Thr-12, as part of the nucleotide binding fold, and the active site SYK triad, now shown to form a tetrad with the conserved Asn-111. The NNAG motif around residue 86, the conserved Asn-111, defined in this study, and further motifs (comprising a conserved Asn-179 in strand ␤F, the PG motif, and the conserved Thr-188; Fig. 1), identified through structural alignments and functional analyses (6,12,24) reveal the critical involvement of conserved elements for coenzyme binding, maintenance of the SDR scaffold, and catalysis. Notably, the recent structure determination of sequence-unrelated proteins displaying the SDR fold considerably extend structure-activity relationships (25). Thus the SDR domain structure appears to be a generic scaffold not only including dehydrogenases/reductase, lyase, epimerase, and hydratase activities but also comprising RNA binding proteins, kinases, and transcription factors (25). Greater understanding of the mechanistic and structural principles governing the SDR architecture will reveal novel substrate and protein-protein interactions and will facilitate the development of inhibitors directed against biologically relevant SDR targets. These efforts constitute avenues currently pursued at several pharmaceutical sites.