Site-directed Mutagenesis of the Active Site Region in the Quinate/Shikimate 5-Dehydrogenase YdiB of Escherichia coli*

YdiB and its paralog AroE are members of the quinate/shikimate 5-dehdrogenase family. Enzymes from this family function in the shikimate pathway that is essential for survival of microorganisms and plants and represent potential drug targets. Recent YdiB and AroE crystal structures revealed the presence of a NAD(P)-binding and a catalytic domain. We carried out site-directed mutagenesis of 8 putative active site residues in YdiB from Escherichia coli and analyzed structural and kinetic properties of the mutant enzymes. Our data indicate critical roles for an invariant lysine and aspartate residue in substrate binding and allowed us to differentiate between two previously proposed models for the binding of the substrate in the active site. Comparison of several YdiB and AroE structures led us to conclude that, upon cofactor binding and domain closure, the 2 identified binding residues are repositioned to bind to the substrate. Although the lysine residue contributes to some extent to the stabilization of the transition state, we did not identify any residue as catalytically essential. This indicates that catalysis does not operate through a general acid-base mechanism, as thought originally. Our improved understanding of the medically and agriculturally important quinate/shikimate 5-dehydrogenase family at the molecular level may prove useful in the development of novel herbicides and antimicrobial agents.

The shikimate pathway of prokaryotes, fungi, plants, and apicomplexa is essential for survival. The main route of this pathway leads to the biosynthesis of chorismate, the precursor of essential aromatic compounds including vitamins and amino acids ( Fig. 1) (1, 2). The successful targeting of the shikimate pathway in crop plants by glyphosate has spurred further efforts in herbicide development (3). More recently, the occurrence of herbicide-resistant weeds (4) has led to the development of transgenic crops with increased glyphosate tolerance (5,6). In bacteria, the shikimate pathway has been subject to metabolic engineering, aimed at the possible industrial production of high value hydroaromatic compounds (7). Lastly, their absence in metazoans makes the enzymes of the shikimate pathway potential targets for novel antimicrobial agents (3, 8 -10). With a view to the design of inhibitors of the shikimate pathway, the understanding at the molecular level of enzymes involved in this pathway has received much attention over the last 25 years (3,11).
Step 4 of the pathway, the reversible NADPH-dependent reduction of 3-dehydroshikimate to shikimate (Fig. 2) is catalyzed by the enzyme shikimate dehydrogenase (EC 1.1.1.25), a member of the quinate/shikimate 5-dehdrogenase family. In addition to the widely distributed bacterial NADP-dependent shikimate dehydrogenase AroE, Escherichia coli, Salmonella typhimurium, Streptococcus pneumoniae, and Haemophilus influenzae also possess a paralogous enzyme, YdiB. YdiB from E. coli is a dual specificity quinate/shikimate NAD-dependent dehydrogenase, and its possible evolution and metabolic role have been discussed recently (12). Although the involvement of AroE in the shikimate pathway is well established for E. coli and S. typhimurium (13), the actual biological function of YdiB remains unclear, nor is it known whether 3-dehydroshikimate or quinate represents the natural substrate of YdiB. Nevertheless, the YdiB enzymes from S. pneumoniae and H. influenzae were recently shown to be essential for in vitro growth of these human pathogens (14). A similar occurrence of two family members is also found in filamentous fungi. In addition to a NADP-dependent shikimate dehydrogenase, they possess a related enzyme, quinate dehydrogenase (EC 1.1. 1.24), that catalyzes the NAD-dependent oxidation of quinate. This is the first step of the quinate pathway ( Fig. 1), which affords growth on quinate as a carbon source (15).
In fungi (16) and the apicomplexan parasite Toxoplasma gondii (17), the NADP-dependent shikimate dehydrogenase constitutes the C-terminal module of a penta-functional polypeptide, which combines the enzymatic activities for steps 2-6 of the shikimate pathway (Fig. 1). In plants (18) and bacteria from the genus Chlamydia, this enzyme forms the C-terminal module of a bifunctional polypeptide catalyzing steps 3 and 4. Most bacterial shikimate dehydrogenases, and the fungal quinate-oxidizing dehydrogenases, on the other hand, represent monofunctional enzymes of 29 -36 kDa.
Few studies have been performed on the kinetic or chemical mechanism of catalysis for this family of enzymes. For the shikimate dehydrogenase from Pisum sativum, it has been shown that the kinetic mechanism is ordered Bi Bi in both directions, with the cofactor adding first (19). As additionally demonstrated for the shikimate dehydrogenase from E. coli (20), the mechanism involves the stereoselective transfer of hydrogen between the A side of NADPH and the substrate (19).
Recently, the crystal structures of AroE from E. coli, Meth-anococcus jannaschii, and H. influenzae (12,21,22) and of YdiB from E. coli (12,23) were solved in complex with NADP and NAD, respectively. Structures of apoenzymes are also available for H. influenzae AroE (22) and YdiB. 1 All of these structures reveal a common fold comprising two domains separated by a cleft. Although AroE from E. coli and AroE and YdiB from H. influenzae exist as monomers (12,22), YdiB from E. coli (12,23) and AroE from M. jannaschii (21) both homodimerize via their N-terminal domains. This portion of the protein possesses a unique ␣-␤-␣ sandwich motif, which also includes an ␣-helical hairpin structure at the C terminus of the protein and is further referred to as domain 1. The intervening sequence, domain 2, forms a Rossmann fold, to which the dinucleotide cofactor is bound with the A side of the nicotinamide ring facing the interdomain cleft (12,21,23). A direct comparison of the cofactor complexes for the two E. coli enzymes AroE and YdiB by Michel et al. (12) revealed structural differences in their nucleotide-binding motifs, which likely account for the 10-fold higher affinity of YdiB for NAD than for NADP. In contrast to E. coli YdiB, AroE enzymes are generally NADP-specific. Despite the recent availability of several crystal structures for enzyme-cofactor complexes, as well as for apoenzymes of quinate/ shikimate 5-dehydrogenases, no structural data on substrate binding are yet available, and so the nature of the active site in The Active Site of E. coli YdiB this agriculturally and medically important enzyme family has remained ambiguous. The active site is identified by the position of the nicotinamide moiety and appears to be defined by a conserved cluster of residues from domain 1 (12,21,23). We carried out mutational analyses of 8 putative active site residues from domain 1 in YdiB from E. coli, exhausting all apparent candidate residues for catalysis. A conserved asparagine proved to be essential for proper folding. In addition to structural characterization and steady state kinetics, comparison of the mutated residues in different YdiB and AroE crystal structures supports roles for an invariant aspartate and lysine residue in substrate binding. Surprisingly, none of the mutated residues was essential for catalysis, suggesting that the primary catalytic mechanism does not involve general acid-base catalysis.

EXPERIMENTAL PROCEDURES
Preparation and Characterization of YdiB Mutants-Site-directed mutagenesis was performed on the plasmid containing the ydiB gene (12) using the QuikChange TM XL site-directed mutagenesis kit (Stratagene, La Jolla, CA). Forward versions of the mutagenic primers are listed in Table I. The sequences of the mutant ydiB genes were confirmed by DNA sequencing. YdiB enzymes were overexpressed and purified as described previously for wild-type protein (12). Protein concentrations were determined using the Bio-Rad protein assay (Bio-Rad Laboratories) based on the original Bradford assay (24) with bovine serum albumin as the standard. Mass spectra were recorded using an Agilent 1100 Series liquid chromatograph/mass selective detector instrument (Agilent Technologies, Mississauga, Ontario, Canada) in electrospray ionization mode and analyzed using Agilent ChemStation software (version A.09.01). A sample of purified YdiB protein (0.1-0.2 mg/ml) was diluted 1:100 (v/v) in 10% (v/v) acetonitrile, 0.1% (v/v) formic acid prior to injection. For native PAGE, proteins were analyzed using a 12% (w/v) polyacrylamide gel (375 mM Tris/HCl, pH 8.8) containing 8.7% (v/v) glycerol. After the addition of one equal volume of 2ϫ loading buffer (125 mM Tris/HCl, pH 6.8, 20% (v/v) glycerol, 10 g/ml bromphenol blue), 3 g of protein were loaded per lane. Electrophoresis was performed at 125 mA for 3 h with cooling on ice. Gels were stained with 0.1% (w/v) Coomassie Blue R-350. Dynamic light scattering and analytical gel filtration were performed as described previously (12). In brief, for dynamic light scattering measurements, 50 l of protein (0.3 mg/ml) were cleared by centrifugation and transferred to the 96-well plate. Analytical gel filtration was performed by applying 180 g of protein to the column at a flow rate of 0.5-0.6 ml/min. CD spectra were recorded using a model J-710 spectropolarimeter from Jasco Inc. (Easton, MD) with an N 2 flow rate of 5 liters/min and were analyzed using the Jasco J-700 software for Windows v1.10.00. Protein samples (200 l) at a concentration of ϳ0.6 -0.8 mg/ml in buffer (50 mM Tris/HCl, pH 7.5, 200 mM NaCl, 5% (v/v) glycerol) were loaded into a CD circular quartz cuvette (path length ϭ 0.05 cm). For each sample, five scans were acquired at a scan speed of 50 nm/min, with a bandwidth of 1.0 nm and a step resolution of 0.2 nm in the wavelength range of 200 -250 nm. The scans were subsequently averaged. Spectra of buffer alone were recorded accordingly and subtracted from the protein spectra.
Steady State Kinetics-Enzymatic activities were measured in 100 mM Tris/HCl, pH 9, in the reverse direction as described previously (12) with the following modifications. Determinations were set up in quadruplicate in 96-well plates using a multichannel pipette for synchronous mixing of the reagents. To determine the substrate-and cofactor-dependent kinetic properties of YdiB variants, we prepared 11 different concentrations each of shikimate, quinate, and NAD by one-in-one dilutions (v/v) of a 4 or 20 mM substrate stock solution and a 4 mM NAD stock solution, respectively. Per well, 180 l of substrate or cofactor were added to 10 l of enzyme, and the reactions were started by adding 20 l of a 40 mM solution of NAD or substrate (i.e. shikimate or quinate), respectively. Initial velocities of enzyme activity were determined by recording NADH absorbance at 340 nm in a 96-well plate reader. Each reading was calibrated by including a series of eight one-in-one dilutions (v/v) of reduced cofactor in a volume of 210 l/well, starting from 0.4 mM NADH. Kinetic data were evaluated by non-linear regression analysis with the Michaelis-Menten equation ), using the SigmaPlot software (SPSS Science, Chicago, IL). The catalytic constant, k cat , was calculated using the equation Structure Comparison-Structural models 2 for YdiB from E. coli and H. influenzae and for AroE from E. coli, H. influenzae, and M. jannaschii were superimposed, and root mean square deviation calculations were carried out using the Swiss-PDBviewer (26).

RESULTS AND DISCUSSION
Structural Characterization of YdiB Mutants-The structure of E. coli YdiB has been determined independently by two groups (12,23). Although one of the E. coli YdiB models (PDB 3 number 1NPD) includes all 288 residues of the protein (23), the other (PDB number 1O9B) includes residues 7-286 (12). In the current work, we made use of the same YdiB construct and YdiB purification protocol as described in the latter report (12). To determine whether the purified enzyme was truncated, we analyzed the wild-type protein by mass spectroscopy. Liquid chromatograph/mass selective detector analysis gave a mass for YdiB of 31,361 Da, as compared with a calculated mass of 31,371 Da, which agrees within the expected mass error of about 0.02%. This mass includes the additional Gly-Ser residues that remain at the N terminus after thrombin cleavage with protein expressed using the pGEX-4T1 vector (12). In the following, we will refer to residues according to E. coli YdiB numbering. Where specified otherwise, E. coli YdiB numbering will be referenced in parentheses.
Sequence alignments of quinate/shikimate 5-dehydrogenase family members from Pfam (27) and the NCBI Conserved Domain Database 4 (28) indicate that the residues Gly 63 , Ser/ Thr 67 , Pro 69 , Lys 71 , Asn 92 , Asp 107 , Gly 131 , Gly 133 , Gly 134 , Gly 255 , and Gln 262 are likely fully conserved. In addition, Asn 105 and Arg 156 are also conserved in nearly all sequences. The cluster of glycines (at positions 131, 133, and 134) is part of the P-loop that interacts with the phosphates of the NAD(P) cofactor, and the carbonyl group of Gly 255 hydrogen-bonds to the amide group of the nicotinamide ring. We have concentrated on the highly conserved residues Ser 67 , Lys 71 , Asn 92 , Asp 107 , and Gln 262 in the vicinity of the putative substratebinding site as the most likely candidates for residues essential for catalysis. Additionally, we selected the partially conserved residues Ser 22 , Tyr 39 , and Thr 106 as potential substrate-binding residues (Table I, Fig. 3). We have not attempted to mutate Arg 156 as it is remote from the catalytic site and stacks with the adenine ring of NAD. Similarly, Asn 105 appears to be shielded from the active site by Asp 107 and likely plays a structural role. The mutants S22A, Y39F, S67A, K71G, N92A, T106A, D107A, and Q262A were expressed and, with the exception of N92A, purified to apparent homogeneity as assessed by SDS-PAGE (not shown). The N92A mutant protein appeared more sensitive to proteolytic cleavage than wild-type YdiB or any of the other mutants and was not isolated in a pure form.
To assess the effects of the various mutations on YdiB structure, we characterized the conformational integrity of the purified proteins using several methods. Native PAGE showed well focused bands for the wild-type protein, as well as for the S22A, K71G, Q262A, Y39F, and S67A mutants, and less-focused bands for the D107A and T106A mutants (Fig. 4A) (25). Although 1NPY and 1P74 represent apoenzymes, all remaining models feature complexes with the NAD(P) cofactor. 3 The abbreviation used is: PDB, Protein Data Bank . 4 The respective sequence alignments of quinate/shikimate 5-dehydrogenase family members can be accessed through Pfam (http://www. sanger.ac.uk/Software/Pfam/) under accession number Pf01488 and through the NCBI Protein Database (http://www.ncbi.nlm.nih.gov) under NCBI accession number COG0169.
N92A mutant appeared as a smear. Dynamic light scattering measurements on the various purified proteins were mostly in agreement with the results from native PAGE, with the N92A mutant being polydisperse and yielding an apparent molecular mass of 300 kDa (not shown). Consistent with the results from native PAGE, the general shape of the CD spectra for wild-type enzyme and the S67A and K71G mutants were similar, whereas the alanine mutations of Thr 106 and Asp 107 caused apparent shifts (Fig. 4B). The N92A mutant exhibited a spectrum that deviated significantly from either wild-type YdiB or the other YdiB mutants. In analytical gel filtration experiments, wild-type YdiB exhibited an apparent mass of ϳ57 kDa (calculated mass for the recombinant dimer: 62.7 kDa). In addition, the wild-type contained a very small amount (0.2%) of a species with an apparent mass of ϳ33 kDa (likely monomeric form) and some protein that eluted in the void volume of the column. About 65% of the N92A mutant protein appeared in the void volume and was likely aggregated, with the remainder eluting at about 33 kDa. Although the T106A mutant also showed considerable amounts of apparent aggregate (12%) and a 33-kDa species (1.8%), it largely eluted as a 57-kDa species (likely dimeric form). The D107A profile was most similar to that of the wild type.
Taken together, these data suggested that the N92A protein was improperly folded. The protein was partly degraded, likely by trace amounts of proteolytic enzymes. We assume that the fragments aggregated into larger species. The Asn 92 residue is located at a ␤-bulge, at the N-terminal end of strand ␤4 within the 6-stranded mixed ␤-sheet of domain 1 (Fig. 3), and participates in a hydrogen-bonding network through its side chain and main chain atoms (12), thereby offering an explanation for the apparent effect of this mutation on the YdiB structure. Mutation of either Asn 92 or Thr 106 to alanine would disrupt the H-bond between Asn 92(OD1) and Thr 106(OG) . The side chain of Thr 106 also forms a hydrogen bond with Gln 262 , helping to orient both side chains in the active site. Thr 106 and Asp 107 are a Altered nucleotides, that introduced the desired mutations, are underlined. Additional silent mutations, introducing restriction enzyme cleavage sites for screening purposes, are printed in italics.
b The degree of conservation for the mutated residue in E. coli YdiB among ‫57ف‬ members of the quinate/shikimate 5-dehydrogenase family, as aligned in the NCBI Conserved Domain Database 4 (28), is indicated.
c Threonine, not serine is the prevailing amino acid at position 67, with a conservation level of 40%. The mutated residues are shown with carbon, nitrogen, and oxygen colored in green, purple, and red, respectively. The NC4 atom of the NAD cofactor is marked by a red asterisk. The nicotinamide and adenine nucleotide portions of the cofactor are shown in purple and magenta, respectively. The corresponding section of the protein is displayed in the background as a semitransparent ribbon drawing. The figure was made with PyMOL (40). located at the junction between the two domains, at the bottom of the interdomain cleft. Mutation of either of these residues to alanine likely results in some adjustment of the secondary or tertiary structure influencing the observed differences in the CD spectra as compared with wild-type YdiB. It may also influence the relative positioning and flexibility of the two domains of YdiB, which we discuss below.
Steady State Kinetics of YdiB Mutants-Kinetic measurements were performed using either quinate or shikimate as substrate to reduce the NAD cofactor. Table II summarizes the YdiB kinetic measurements. No activity at all could be detected for the N92A mutant, which is apparently incorrectly folded. Moreover, with the D107A mutant, activity could only be detected toward shikimate but not quinate. The kinetic parameters for the S22A and Y39F mutants remained largely unchanged as compared with the wild-type enzyme, indicating that neither Ser 22 nor Tyr 39 plays a role in the reaction. Some moderate changes were observed with the S67A and Q262A mutants. The Q262A mutant showed a general ϳ3-fold reduction in catalytic efficiency (k cat /K M ) for both substrates. Although the conservation of Gln 262 (Table I) suggested a critical role for this residue, this was not evident from the relatively mild effect on kinetic parameters by its mutation to alanine. The S67A mutation, on the other hand, reduced the catalytic efficiency ϳ6-fold with quinate as a substrate but had little effect on shikimate conversion. This may indicate that Ser 67 has additional interactions with quinate as compared with shikimate.
The mutants K71G, T106A, and D107A showed marked reductions in catalytic efficiencies in the substrate-dependent kinetics (Table II). As compared with the wild-type enzyme, the catalytic efficiency of NAD reduction by these mutants with shikimate was reduced by 3, 2, and nearly 4 orders of magnitude, respectively. For K71G and T106A, at least a 10-fold greater reduction in catalytic efficiency was observed with quinate than with shikimate. No activity toward quinate could be detected for the D107A mutant. The loss of activity in K71G, T106A, and D107A was predominantly due to significantly increased K M values for the substrate, with much smaller effects on k cat values. Indeed, the k cat values determined from substrate-dependent kinetics for the T106A and D107A mutants remained practically unchanged, whereas those for the K71G mutant were reduced ϳ9-fold. For the latter mutant, a significant decrease in K M for the NAD cofactor was observed (higher affinity), which offset the parallel decrease in k cat values, leading to a continued efficient catalytic reduction of NAD. By comparison, no pronounced changes in the measured cofactor kinetics were observed for the T106A and D107A mutants. We note that we could not saturate the K71G and T106A mutants with quinate and could not saturate the D107A mutant with shikimate during the cofactor kinetics due to high K M values for the respective substrates (Table II). Therefore, the catalytic constants for the cofactor are somewhat underestimated in these cases. We further suggest that the lack of detectable activity for the D107A mutant with quinate reflects the highest loss of substrate affinity.
We cannot rule out that steps other than the chemical transformation, i.e. the hydride transfer reaction, affected the apparent k cat values. The measured differences in kinetic parameters could also be affected by structural perturbation in some of the mutant enzymes, as discussed above. Nevertheless, our data (Table II) indicate important roles for Lys 71 and Asp 107 in substrate binding in the Michaelis complex. The general decrease in k cat for the structurally intact K71G mutant likely reflects a contribution of this residue to stabilization of the transition state. Importantly, none of the mutated residues was found to be essential for catalysis.
Active Site Structures of YdiB and AroE Enzymes-The sequence identity of E. coli YdiB with related proteins of known structure ranges from 23 (H. influenzae AroE) to 36% (M. jannaschii AroE). The overall structure of E. coli YdiB comprises two domains (23), which are each similar to the structures of the corresponding domains of H. influenzae YdiB 1 and of AroE from E. coli (12), H. influenzae (22), and M. jannaschii (21). However, as was shown previously based on the superposition of four independent molecules of AroE and two molecules of YdiB (12), the individual molecules display conformational variability, characterized by a ϳ25°rotation of one domain relative to the other. Domain movements of similar magnitude can also be observed in the YdiB structure determined by Benach et al. (23).
For spatial comparison of the substrate-binding site, we considered superposition of the individual chains using the putative active site residues of domain 1. Among the 16 individual chains compared (Table III) one molecule of these apoenzyme structures each, this reorientation is associated with the presence of a hydrogen bond between Gln 262 and Asp 107 (Table III). Furthermore, a salt bridge is formed between Lys 71 and Asp 107 in 5 out of 10 enzyme-cofactor complexes without major conformational changes as compared with the apo-enzyme structures.
When the NAD(P)-binding domains are superimposed independently, the cofactors in various structures cluster closely together. Relative to domain 1, however, the disposition of NAD(P) molecules shows a broader distribution that results primarily from domain flexibility (12). This is most pronounced for the adenine dinucleotide portion of NAD. In contrast, conserved hydrogen bonds between the cofactor amide group and the carbonyl groups of an invariant Gly 255 and of Cys 232 anchor the nicotinamide ring of the cofactor to domain 1. The distance between the NAD(P) NC4 hydrogen donor/acceptor and the closest residue from domain 1, the conserved Asp 107(OD1) , is within 3.4 and 3.9 Å in eight of nine enzyme-NAD complex molecules  a The identities of the NE1 and OE1 atoms of Gln 262 cannot be unequivocally assigned in the models for any of the enzymes. Therefore, atom identities were assigned by superimposing the active site in each molecule with the active site in chain A of 1O9B, and determining, which of the two side chain heteroatoms in the corresponding glutamine residue could be aligned with the designated NE1 and OE1 of Gln 262 in 1O9B with minimal side chain movements.
b Chain A of 1NYT contains a molecule of dithiothreitol, hydrogen-bonded to the residues equivalent to Ser 67 , Lys 71 , Asn 92 , Asp 107 , and Gln 262 . The interactions with the lysine and glutamine residues have been proposed to resemble substrate-enzyme interactions (12).
c The nicotinamide and ribose moieties in the nicotinamide mononucleotide portion of the cofactor were disordered and not included in 1P77 (22). d Chain A of 1NVT was omitted from the comparison because marked structural differences were seen in the model of the cofactor and the enzyme in the vicinity of the nicotinamide ring.
e In chain A of 1P74, the side chain of the aspartate residue equivalent to Asp 107 is rotated so that its OD2-atom is closer to both the lysine and the glutamine residue than its OD1-atom. (Table III). An increased distance of 4.4 Å for chain C of the E. coli AroE structure is indicative of a more open enzyme conformation as compared with the other three molecules in the same structure. The increased distance in this molecule is further associated with the presence of an Asp 102 -Gln 244 (Asp 107 -Gln 262 ) hydrogen bond, as observed in one molecule each from the apoenzyme structures of H. influenzae YdiB and AroE, as described above.
Within our comparison set of quinate/shikimate 5-dehydrogenase molecules (Table III), the Asp 107 -Gln 262 hydrogen bond and the Lys 71 -Asp 107 salt bridge are mutually exclusive. The former is observed predominantly in the structures of apoenzymes, whereas the latter is only found in enzyme-cofactor complexes, which were found to adopt more closed conformations. The breaking of the Asp 107 -Gln 262 hydrogen bond and the formation of Lys 71 -Asp 107 salt bridge, therefore, may be associated with active site closure after cofactor binding and the formation of a productive catalytic site, as proposed previously (12).
Substrate Binding and Catalysis-To assign functional roles to putative active site residues in the quinate/shikimate 5-dehydrogenase family, three models for ternary enzyme-cofactorsubstrate complexes have been proposed previously, representing two essentially different possible orientations of the substrate in the active site (Fig. 5) (22), shows 3-dehydroshikimate bound to H. influenzae AroE. As far we can deduce from the report (22), the orientation of the substrate is similar to the model of Benach et al. (23).
Our structure-function analysis included 5 of 7 putative substrate-interacting residues from models A and B (Fig. 5) together. We observed a clear decrease in substrate affinity only for the K71G and D107A mutants (Table II). Previous studies demonstrated a high importance of the C-4 hydroxyl over the C-3 hydroxyl (Fig. 2, shikimate numbering) in the binding of shikimate and 3-dehydoshkimate to the shikimate dehydrogenase from P. sativum (29) and of shikimate to E. coli AroE (30). It was concluded that a charged group is hydrogen-bonded to the C-4 hydroxyl (30). In this light, the substrate orientation in Model A (Fig. 5) is more consistent with important contributions of Lys 71 and Asp 107 , but not Gln 262 , to substrate binding.
Here, the C-4 hydroxyl interacts with Lys 65 (Lys 71 ) and Asp 102 (Asp 107 ), and the C-3 hydroxyl interacts with Gln 244 (Gln 262 ), whereas Model B places Lys 71 near the C-1 carboxylate group of the substrate and suggests an interaction between Gln 262 and the C-4 hydroxyl (22). None of the previously proposed models for substrate binding directly implicates Thr 106 , for which we also observed a clear loss in substrate affinity (Table  II). This is likely an indirect effect due to the structural significance of this residue, as discussed above. The much more pronounced loss in substrate affinity for quinate than for shikimate in the K71G, T106A, and probably also the D107A mutants (Table II) may reflect a more robust overall binding for shikimate.
Against the background of our structural comparison and mutational analysis, we propose that, upon cofactor binding and domain closure, the 2 fully conserved residues Lys 71 and Asp 107 form a salt bridge, which orients them for the initial binding of the substrate, i.e. 3-dehydroshikimate or shikimate, as proposed by Model A (Fig. 5) (12). This interaction may persist during the hydride transfer reaction, with only Lys 71 contributing appreciably to transition state stabilization as already mentioned. YdiB was previously suggested to operate by general acid-base catalysis (12,23). During substrate reduction (Fig. 2), the enzyme would stabilize the accumulating negative charge at the hydride-accepting C-3 carbonyl oxygen of the substrate by a concerted proton transfer reaction (23). Our kinetic analysis, however, did not identify any residue as catalytically essential, one that would represent a potential acid-base catalyst.
The retention of high catalytic rates for a NADP(H)-dependent reductase in the absence of general acid-base catalysis is not The approximate orientation of the substrate relative to neighboring protein side chains is projected. Proposed protein-substrate interactions in the ternary complex are highlighted by gray background shading. In model A, the C-4 and C-3 hydroxyl of the substrate are coordinated by Lys 71 /Asp 107 and Gln 262 , respectively, whereas Ser 20 , Ser 22 , and Tyr 234 hydrogen-bond to the C-1 carboxylate. In model B, the C-5, C-4, and C-3 hydroxyls hydrogen-bond to Asp 107 , Gln 262 , and Tyr 234 , respectively. The results from this study support model A.
unprecedented. After mutation of the primary acid-base catalyst Tyr 55 in 3␣-hydroxysteroid dehydrogenase, Penning and coworkers (32) reported wild-type-like k cat values for quinone reduction via a mechanism different from 3-ketosteroid reduction by the native enzyme and most likely facilitated by a proximity effect. In addition to the classical models of barrier crossing in enzyme reactions, the work of Klinam and co-workers (39) demonstrated the relevance of quantum mechanical tunneling over a wide range of physiological temperatures as another route (reviewed in Ref. 33). Domain flexibility in liver alcohol dehydrogenase (34) potentially lowers the hydrogen tunneling distance (35). A role for protein dynamics in liver alcohol dehydrogenase was recently substantiated by Plapp and colleagues (36,37), who implicated the nicotinamide-binding site in the energetics of the conformational change in addition to its proposed role in cofactor activation (38). The significance of such properties for catalysis in YdiB remains speculative.
In conclusion, the results from this study do not support a role for general acid-base catalysis for E. coli YdiB. To elucidate the catalytic pathway, structural information for the ternary enzyme-cofactor-substrate complex and more detailed kinetic analyses of enzyme mutants are needed. Our data suggest primary roles for the invariant residues Lys 71 and Asp 107 in substrate binding, which is most consistent with a specific substrate binding mode (Fig. 5, Model A). These insights into the structure-function relationship of YdiB as an enzyme of the quinate/shikimate 5-dehydrogenase family may prove useful in the design of novel shikimate pathway inhibitors as potential herbicides or antimicrobial agents.