Glutamate 170 of Human L -3-Hydroxyacyl-CoA Dehydrogenase Is Required for Proper Orientation of the Catalytic Histidine and Structural Integrity of the Enzyme*

L -3-Hydroxyacyl-CoA dehydrogenase (HAD), the pen- ultimate enzyme in the (cid:1) -oxidation spiral, reversibly catalyzes the conversion of L -3-hydroxyacyl-CoA to the corresponding 3-ketoacyl-CoA. Similar to other dehydrogenases, HAD contains a general acid/base, His 158 , which is within hydrogen bond distance of a carboxylate, Glu 170 . To investigate its function in this catalytic dyad, Glu 170 was replaced with glutamine (E170Q), and the mutant enzyme was characterized. Whereas substrate and cofactor binding were unaffected by the mutation, E170Q exhibited diminished catalytic activity. Protonation of the catalytic histidine did not restore wild-type activity, indicating that modulation of the p K a of His 158 is not the sole function of Glu 170 . The pH profile of charge transfer complex formation, an independent indicator of active site integrity, was unaltered by the amino acid substitution, but the intensity of the charge transfer band was diminished. This observation, cou-pled with significantly reduced enzymatic stability of the E170Q mutant, implicates Glu 170 in maintenance of active site architecture. Examination of the crystal structure of E170Q in complex with NAD (cid:2) and aceto-acetyl-CoA (R (cid:3) 21.9%, R free (cid:3) 27.6%, and are modeled such that potential hydrogen bond interactions are maximized. Coordinates and structure factors for the E170Q model have been deposited in the Protein Data Bank under the accession code 1IL0.

ring, making HAD a "B-side"-specific dehydrogenase (2). As illustrated, His 158 is thought to serve as a general acid/base in the catalytic mechanism of HAD, with catalysis facilitated by the presence of a conserved glutamate residue, Glu 170 (3)(4)(5)(6). However, the precise role of Glu 170 in catalysis has not been established.
Crystallographic studies of HAD have shown that Glu 170 is located adjacent to His 158 in the enzyme active site (5,6). The backbone carbonyl and amide groups of Glu 170 are positioned to form hydrogen bonds with the backbone amide and carbonyl groups of His 158 , respectively. In addition, a carboxylate oxygen of Glu 170 is within hydrogen bond distance of ND1 of His 158 as depicted in Scheme 1. However, Glu 170 also forms contacts that may be important for dynamic enzyme movements. HAD exhibits a two-domain topology (7), with the N-terminal domain (residues 12-200) adopting a ␤-␣-␤ fold similar to NAD(P) ϩbinding enzymes and the C-terminal domain (residues 201-302) consisting primarily of ␣-helices involved in subunit dimerization. Two distinct conformers of the enzyme structure have been identified (5,6). In the open conformation, as described for apo and cofactor-bound enzyme, a large cleft is observed between the NAD ϩ -binding and the C-terminal domains. The addition of substrate results in a conformational change in which the NAD ϩ -binding domain rotates inward toward the dimer interface, sequestering the enzyme active site. This domain shift appears necessary for high affinity substrate binding and critical for effective catalysis. The linker region, composed of residues 201-207, relates the two domains and contains a consensus Pro 203 -Gly 204 -Phe 205 sequence, which appears to be the pivot point for domain movement. Numerous interactions between Glu 170 and residues within this region are observed, including a hydrogen bond with a conserved water molecule.
Charge transfer complex formation by HAD provides a spectroscopic assay for structural integrity that can be used to complement binding studies and kinetic analysis. As described previously, the abortive ternary complex composed of HAD, NAD ϩ , and AACoA exhibits a broad absorbance band centered between 410 and 420 nm (6). AACoA, which is bound as an enolate in the abortive complex, acts as an electron donating species and the nicotinamide ring of NAD ϩ serves as the electron acceptor. The intensity of the charge transfer band is pH-dependent, with the protonation of a single group resulting in decreased enolate and charge transfer complex formation. The spectroscopic properties of the charge transfer complex are sensitive to perturbations in the protein structure and can be used to probe active site integrity.
To evaluate the contribution of Glu 170 to catalysis, this residue was substituted with glutamine by site-directed mutagenesis, and the resultant enzyme (E170Q) was analyzed by ki-netic, spectroscopic, and x-ray crystallographic techniques. Comparison of the mutant and native enzyme characteristics suggests that Glu 170 is required for enzyme stability and positioning of the imidazole ring of His 158 for efficient catalysis. In addition, an ionizable group with an apparent pK a of 5.3 has been identified as an important determinant in efficient substrate binding and charge transfer complex formation. These studies expand our understanding of the catalytic mechanism of L-3-hydroxyacyl-CoA dehydrogenase and provide additional insight into the potential nature of analogous Glu-His and Asp-His catalytic dyads.

Site-directed Mutagenesis, Expression, and Purification of E170Q-
The E170Q mutation was created in the previously reported HAD expression plasmid using the Stratagene QuikChange mutagenesis kit according to the manufacturer's protocol. The enzyme was expressed in Escherichia coli strain BL21(DE3)pLysS and purified by affinity chromatography using a nickel-chelating column (Novagen), as described previously (6). For protein crystallization, the N-terminal histidine tag was removed by thrombin cleavage. Purified protein was dialyzed against 50 mM potassium phosphate buffer, pH 7.2, containing 1 mM dithiothreitol and stored at 4°C.
Determination of Enzyme Activity-Enzyme activity was measured on a Cary-50 spectrophotometer at 25°C by monitoring the disappearance of NADH (⑀ 340 ϭ 6220 M Ϫ1 cm Ϫ1 ) upon reduction of acetoacetyl-CoA (AACoA), as described by Noyes & Bradshaw (2). For standard screening of enzyme activity, the assay was performed in citrate/phosphate buffer, pH 7.0, containing 100 M NADH and 50 M AACoA. For pH dependence studies, a citrate/phosphate buffer system of uniform ionic strength, maintained by the addition of KCl, was employed (8). Measurements were made at a saturating concentration of NADH (100 M) with acetoacetyl-CoA as the varied substrate. At an indicated pH, a substrate saturation curve was determined in triplicate, and values of V max and K m were obtained. pH profile data were plotted as the log of the kinetic parameter against pH and fitted using the appropriate rate equation, as described by Cleland (9). To assess enzyme stability, protein (1 mg/ml; citrate/phosphate buffer, pH 7.0) was incubated at 37°C. Aliquots were withdrawn at the indicated time, and activity was measured under standard conditions.
Characterization of the Charge Transfer Complex-L-3-Hydroxyacyl-CoA dehydrogenase in complex with NAD ϩ and AACoA has been shown to form a charge transfer complex, as indicated by a broad band centered between 410 and 420 nm (6). To determine the pH dependence of the charge transfer complex, absorbance measurements were made at 412 nm with either wild-type or mutant enzyme as a function of pH. Spectroscopic measurements were performed on a Cary-50 spectrophotometer at 25°C, using the citrate/phosphate/KCl buffer system described above. To 130 l of buffer of a given pH, 20 l of a preequilibrated stock solution of enzyme, NAD ϩ , and AACoA were added to yield final concentrations of 1.15 mg/ml enzyme, 2 mM NAD ϩ , and 2 mM AACoA. Measurements were made in triplicate and analyzed using the Origin 6.0 software package.
Crystallization, X-ray Diffraction Studies, and Structure Determination-Crystals of the abortive ternary complex composed of E170Q, NAD ϩ , and acetoacetyl-CoA were grown by the hanging drop method out of a solution of 50 mM N-[2-acetamido]-2-iminodiacetic acid, pH 6.5, within the precipitant range of 14 -19% polyethylene glycol 4000 at a protein concentration of 5 mg/ml as described for wild-type protein (6). Diffraction data were collected at the University of Minnesota using radiation produced by a Rigaku RU-200BH rotating anode fitted with an Osmic MaxFlux confocal multilayer fixed focus optical system and an R-AXIS IV ϩϩ image plate system. Crystals were maintained under cryogenic conditions by an X-stream cooling system, and data were analyzed with the Crystal Clear software package (10) (Molecular Structure Corp.). The structure of the E170Q abortive ternary complex was isomorphous with the corresponding wild-type complex (Protein Data Bank accession code 1F0Y).
Model Building and Refinement-The E170Q model was refined using the crystallography and NMR system CNS, version 0.3 (11) and rebuilt after each round of refinement using the program O (12), as described previously (5). Ambiguous regions of the electron density map were evaluated using 2ԽF o Խ Ϫ ԽF c Խ simulated annealing omit maps, in which designated regions of the structure were omitted and the remaining model was subjected to simulated annealing prior to map calculation. Water molecules obeying proper hydrogen bonding constraints with electron densities greater than 1.0 on a 2ԽF o Խ Ϫ ԽF c Խ map and 4.0 on an ԽF o Խ Ϫ ԽF c Խ map were included as model refinement neared completion. Bound substrate and cofactor were also added at this point using coordinates from the analogous wild-type structure. Note that the positions of O3 and C4 of acetoacetyl-CoA could not be assigned unambiguously in the electron density. The current models have O3 positioned within hydrogen bonding distance of N⑀2 of His 158 , the proposed catalytic base. Similarly, the orientations of His 158 , Gln 170 , and Asn 208 side chains cannot be determined unequivocally and are modeled such that potential hydrogen bond interactions are maximized. Coordinates and structure factors for the E170Q model have been deposited in the Protein Data Bank under the accession code 1IL0.

Protein Expression and Initial Characterization of E170Q-
The E170Q mutant enzyme was overexpressed in E. coli and purified to homogeneity. E170Q protein yields were ϳ4-fold lower than wild-type enzyme, with 15-20 mg produced per liter of cell culture. In addition, E170Q appeared to be less stable than native enzyme. Visible precipitate was observed in the concentrated E170Q stock solution (ϳ10 mg/ml) after 5-7 days when stored at 4°C; native enzyme was stable indefinitely under similar conditions. In addition, the specific activity of the E170Q protein decreased ϳ10% over this time period, while that of native enzyme remained unchanged. However, initial specific activity values for the E170Q enzyme could be restored by removal of the precipitated protein. As observed with wildtype protein, uncleaved and thrombin-cleaved E170Q exhibited similar levels of enzyme activity, cofactor affinity, and substrate binding (data not shown), indicating that the N-terminal histidine tag did not significantly alter enzyme structure or function.
A comparison of native and E170Q kinetic parameters is provided in Table I. The concentration of acetoacetyl-CoA was varied at a saturating NADH concentration (100 M), with initial rates determined in triplicate. Data were then fit to the Michaelis-Menten equation to obtain V max and K m . The sensitivity of the kinetic assay employed precluded determination of a K m for NADH by the reciprocal experiments. However, as judged by fluorescence quenching experiments (5), E170Q bound reduced cofactor with equal or slightly greater affinity than native enzyme; K d values were typically less than 1 M (data not shown). The E170Q mutation did not alter substrate binding as evidenced by comparable K m values for native and SCHEME 1  (13)(14)(15), a potential role of Glu 170 is to elevate the apparent pK a of His 158 , rendering the enzyme-catalyzed reaction more reversible at neutral pH. The E170Q mutation would therefore be predicted to alter the pH dependence of a given kinetic parameter. V max and K m values were determined as described above for native and mutant enzymes over the pH range of 5.0 -8.0 in 0.2 increments. Steady-state kinetic parameters are presented as log values versus pH (Fig. 1), with the data fit according to the methodology of Cleland (9) to obtain apparent pK values (Table II). Catalytic efficiency (V max /K m ) was pH-independent for native enzyme in the neutral pH range (Fig. 1A), with decreased enzyme efficiency in the acidic region of the curve indicative of a single ionizable group with an apparent pK a of 5.14 Ϯ 0.53 (Table II). In contrast, the pH dependence of E170Q enzyme activity fit to a bell-shaped curve described by two ionization events. The first (pK a1 ϭ 5.18 Ϯ 0.57) was comparable with wild-type enzyme, and the second (pK a2 ϭ 6.80 Ϯ 0.45) accounted for the basic limb of the curve.
To assign the observed pK a values to a binding event or to catalysis, individual kinetic parameters were examined. A plot of log V max versus pH (Fig. 1B) revealed that catalysis by E170Q was dependent on a single ionizable group with an apparent pK a ϭ 6.69 Ϯ 0.12, comparable with the value of 6.8 obtained from the V max /K m analysis (Table II). In contrast, native enzyme turnover appeared to be essentially pH-independent. A slight decrease in V max was seen in the basic pH range (pH Ͼ7.0), but this observation was attributable to underestimation of V max ; considerable product inhibition was detected in this pH range, reducing the calculated initial rates. Native and E170Q enzymes exhibited nearly identical K m values for AACoA over the pH range tested. Plots of log (1/K m ) versus pH (Fig. 1, C and D) indicated apparent pK a values of 5.24 Ϯ 0.38 and 5.43 Ϯ 0.29, which were associated with substrate binding to wild-type and mutant proteins, respectively (Table II). These values are in good agreement with those obtained from the V max /K m analysis.

Comparison of Native HAD and E170Q Charge Transfer
Complexes-A charge transfer complex composed of HAD, NAD ϩ , and AACoA has been described (6), and its spectroscopic properties were used to probe the integrity of the E170Q active site. Difference spectroscopy experiments were performed in which the sum of the individual spectra of the components was subtracted from the spectrum of the charge transfer complex. Both native and mutant enzymes displayed similar difference spectra with a sharp peak at 302 nm, corresponding to the enolate species of AACoA, and a broad peak centered between 410 and 420 nm, indicative of charge transfer (data not shown). The topologies of the two spectra were nearly identical, suggesting that key components involved in charge transfer complex formation were not dramatically perturbed by the E170Q mutation. However, the E170Q difference spectrum was uniformly reduced in magnitude ϳ3-fold relative to that of the native enzyme, indicating generally diminished enolate formation. This could result from lower affinity of the protein for AACoA or from altered positioning of key residues. The former was discounted by the observation that a 5-fold increase of AACoA concentration did not produce an increase in the intensity of the charge transfer band.
To further investigate the possible structural effects resulting from the E170Q mutation, the pH dependence of enolate formation by native and mutant HAD was assessed indirectly by measuring charge transfer complex formation as a function of pH (Fig. 2). As observed in the difference spectroscopy measurements, native and E170Q charge transfer bands exhibited an ϳ3-fold difference in magnitude at a given pH. However, a similar pH transition for charge transfer complex formation was observed for both native and mutant enzymes. Charge transfer band intensities were plotted as a function of pH and fit to a sigmoidal curve, yielding a single transition at pH 5.25 Ϯ 0.02 and 5.33 Ϯ 0.08 for native and E170Q enzymes, respectively. Values for this transition are, furthermore, in agreement with pK a values for AACoA binding obtained by kinetic measurements.
Assessment of Protein Stability-Decreased stability of the E170Q mutant protein was suggested by the formation of precipitate upon storage of the enzyme, as described above. This observation, along with previous structural characterizations, suggested that Glu 170 may function to maintain active site structure, prompting a detailed investigation of relative protein stability. An examination of tryptophan fluorescence as a function of guanidine hydrochloride concentration indicated a sharp transition at low denaturant concentration (ϳ0.75 M) for native enzyme (data not shown). Significant stability differences between native and E170Q HAD could not be detected by this method, possibly because of the location of the single tryptophan residue of the enzyme; Trp 265 is positioned at the interface of the NAD ϩ -binding and C-terminal domains near the hinge region of the protein. This region has been demonstrated to undergo significant conformational changes in response to substrate binding and may be more susceptible to FIG. 1. pH dependence of the kinetic parameters for the reduction of acetoacetyl-CoA by native and E170Q L-3-hydroxyacyl-CoA dehydrogenases. Initial rates were measured at a saturating concentration of NADH cofactor (100 M) with AACoA as the varied substrate. At an indicated pH, a substrate saturation curve was determined in triplicate, and the data were fit to the Michaelis-Menten equation to yield values for V max and K m . Apparent pK a values were obtained by fitting the appropriate equation to the observed data, as described under "Experimental Procedures," and are summarized in Table II. A, log(V max /K m ) versus pH for native (f) and E170Q (q) HAD. B, log (V max ) versus pH for native (f) and E170Q (q) HAD. C and D illustrate the pH dependence of AACoA binding (log 1/K m ) for native and E170Q enzyme, respectively.  Fig. 1 with the appropriate equation as described under "Experimental Procedures." pK a2 Native 5.14 Ϯ 0.53 -a b 5.24 Ϯ 0.38 E170Q 5.18 Ϯ 0.57 6.80 Ϯ 0.45 6.69 Ϯ 0.12 5.43 Ϯ 0.29 a Native enzyme does not display a basic limb (Fig. 1A). b The slight decrease in catalytic activity observed at pH values greater than 7 (Fig. 1B) is attributable to strong product inhibition and therefore underestimation of V max .
Importance of Glu 170 in L-3-Hydroxyacyl-CoA Dehydrogenase disruption by chemical denaturants. Hence, tryptophan quenching may not be a reliable probe for overall structural integrity. Therefore, an alternative measure of protein stability was developed. Native and E170Q proteins (1 mg/ml, pH 7.0) were incubated at 37°C, and enzyme activity was monitored over time. As shown in Fig. 3, native enzyme activity (squares) was completely retained over the time course, whereas the E170Q activity (circles) declined with a half-life of ϳ2 h. Furthermore, loss of enzyme activity was accompanied by precipitation of the E170Q protein, indicating reduced structural stability of the mutant relative to native HAD.
X-ray Diffraction Studies and Model Refinement-Structural studies of the E170Q mutant were undertaken to aid in the interpretation of the biochemical observations and provide a molecular model of catalysis by the mutant enzyme. X-ray diffraction data and model refinement statistics are summarized in Table III. E170Q enzyme crystals belonged to space group P2 1 2 1 2 1 and were isomorphous with those of the previously determined HAD abortive ternary complex. A nearly complete data set was collected with diffraction data extending beyond 2.2 Å. To minimize model bias, a simulated annealing omit map was calculated for the region about the E170Q mutation and used during model building. The final E170Q model has well defined geometry with nearly 90% of all residues found in the "most favorable" region of the Ramachandran plot (16). It is important to note that the resolution of the current model (2.2 Å) precluded discrimination between carbon, oxygen, and nitrogen atoms, and therefore rotamer assignments of side chains were based on optimizing potential hydrogen bond interactions.
Comparison of Native and E170Q L-3-Hydroxyacyl-CoA Dehydrogenase Structures-The E170Q enzyme structure was compared with the analogous wild-type HAD model to assess the effects of the mutation on protein structure. Overall, the E170Q and native models are quite similar with a root mean square deviation for ␣-carbons of 0.46 Å and overall B-factors of 29.9 Å 2 and 21.5 Å 2 , respectively. However, subtle differences between the two models are observed throughout the active site region. This region of the final E170Q model (dark gray) and its corresponding 2PF o PϪPF c P omit map density is superimposed upon the analogous wild-type structure (light gray) in Fig. 4. As illustrated, the electron density map supports an E170Q model in which the imidazole ring of His 158 has rotated ϳ90°relative to its placement in the native HAD structure. In addition, the loop containing the E170Q mutation has retreated slightly from the enzyme active site.
The position of the His 158 side chain was the most dramatic difference observed between the native and E170Q structures. Shown in Fig. 5 are the native (Fig. 5A) and E170Q (Fig. 5B) enzyme active sites, with relevant interatomic distances of Յ3.5 Å illustrated as dashed lines; actual distances are summarized in Table IV. Instead of hydrogen bonding with the side chain of Gln 170 in a fashion similar to wild-type protein, the imidazole ring of His 158 is rotated ϳ90°in the E170Q model.  New hydrogen bonds are suggested between ND1 of His 158 and the backbone carbonyl of Phe 159 as well as between NE2 of His 158 and the side chain amide of Asn 208 . As a result, NE2 of His 158 is removed more than 0.7 Å from the 3-keto oxygen of AACoA (O3) and no longer positioned for optimal hydrogen bond formation. This is reflected by an increase in the average B-factor of O3, from 13.4 Å 2 in native enzyme to 50.0 Å 2 in the E170Q model, and is consistent with reduced catalytic efficiency of the mutant enzyme. Other than at the C3 carbonyl of AACoA, substrate binding appears to be unaffected by the E170Q mutation, with comparable protein/ligand interactions observed in native and mutant enzymes.
An examination of potential hydrogen bond interactions in the E170Q model provided additional insight into its altered active site geometry. The center of each active site contains a conserved water molecule, which has been observed in every HAD structure determined to date. This conserved water molecule may have a structural role, since it forms numerous hydrogen bonds with residues located in the hinge region of the protein. The water molecule in the E170Q model (Wat 984) is less constrained than its counterpart in the native enzyme (Wat 809) as evidenced by its elevated B-factor. The average B-factor of this water molecule (subunits A and B) is 29.4 Å 2 for E170Q as compared with 9.7 Å 2 for native enzyme. Although the conserved water molecule exhibits similar hydrogen bond interactions in both models, the interatomic distance between the water molecule and the backbone amide of Asn 208 increases by over 0.6 Å in the E170Q model relative to wild-type protein (Table IV). This perturbation appears to result indirectly from the E170Q mutation.
The glutamate to glutamine substitution disrupts the interactions of residue 170 with Ile 206 and Val 207 . In the native enzyme, one of the carboxylate oxygens of Glu 170 (OE2) is within hydrogen bond distance to the backbone amides of both residues (Table IV, Fig. 5). In contrast, the 3.5 Å interatomic distance between the side chain carbonyl of Gln 170 (OE1) and the backbone amide of Ile 206 suggests only a weak hydrogen bond in the E170Q model. This relaxation of the hydrogen bond network adjacent to Asn 208 may account for the increased interatomic distance between its backbone amide and the conserved water molecule. Taken in sum, the seemingly minor disruptions to hydrogen bond contacts in the active site region may be responsible for the large decrease in E170Q enzyme stability.

DISCUSSION
Previous structural studies of HAD from this laboratory have suggested the importance of several enzyme active site residues in the reversible oxidation of 3-hydroxyacyl-CoA substrates (5, 6). Of particular interest is a histidine-glutamate dyad (His 158 -Glu 170 ) thought to be essential for catalysis (3,4). In the current report, Glu 170 was replaced with a glutamine residue by site-directed mutagenesis, and the mutant enzyme (E170Q) was characterized by kinetic, spectroscopic, and crystallographic analysis. These studies suggest that Glu 170 facilitates catalysis by orienting the imidazole ring of His 158 to optimize productive interaction with the substrate. Furthermore, Glu 170 appears to be critical for the structural stability of HAD.
Site-directed mutagenesis was employed to investigate the role of the His 158 -Glu 170 dyad in catalysis by human HAD. Mutant proteins in which His 158 was replaced by either a leucine (H158L) or glutamine (H158Q) residue had greatly diminished enzymatic activity (data not shown). Attempts at quantitative characterization of these mutants were unsuccessful due to protein instability; both H158L and H158Q formed detectable precipitates moments after clarification. Therefore, efforts focused on the characterization of a mutant protein in which Glu 170 was replaced with a glutamine residue (E170Q). This mutation was selected in an effort to minimize disruption of the active site architecture; glutamine is isosteric with glutamate and capable of forming multiple hydrogen bonds. However, the substitution removes the negative charge located adjacent to the active site histidine, His 158 . Kinetic values for the reduction of acetoacetyl-CoA (AACoA) by native HAD and E170Q are given in Table I. Whereas substrate affinity appears unaltered by the mutation, a Ͼ60-fold reduction in enzymatic activity is detected at pH 7.0.
E170Q consistently displays reduced enzyme activity as compared with wild-type protein over the pH range of 5.0 -8.0 (Fig.  1B), indicating the importance of Glu 170 in the catalytic dyad. A potential function of Glu 170 is to elevate the apparent pK a of FIG. 4. 2ԽF o Խ ؊ ԽF c Խ simulated annealing omit map of the E170Q mutant active site. Gln 170 , His 158 , NAD ϩ , and AACoA as well as residues and water molecules within 3.5 Å of these groups were omitted, and the remaining model was subjected to simulated annealing prior to map calculation. Shown are the resultant map contoured at 1.25 , the final E170Q model (dark gray), and, for reference, the corresponding native structure (light gray). The electron density map supports an E170Q model in which the imidazole ring of His 158 has rotated ϳ90°relative to its placement in the native HAD structure, while adjacent residues are essentially undisturbed. Relatively weak electron density, which has been omitted for clarity, is observed for the structural water molecule of the E170Q model, Wat 984.
the active site histidine, His 158 , which is proposed to serve as a general acid in the reduction of AACoA to L-3-hydroxybutyryl-CoA (4). As such, enzyme activity was predicted to exhibit a pH dependence reflecting the protonation state of His 158 . The enzymatic rate of the E170Q mutant displays a definitive pH dependence, indicating that catalysis is facilitated by the pro-tonation of a single ionizable group with an apparent pK a of 6.69 Ϯ 0.12 (Table II). It is not unreasonable to assign the observed pK a to His 158 , considering its proposed role in catalysis, but the inability to generate a stable His 158 mutant makes such an assignment difficult to confirm. Unfortunately, the apparent pK a of His 158 in wild-type HAD cannot be determined  Importance of Glu 170 in L-3-Hydroxyacyl-CoA Dehydrogenase using the current methodology. Native enzyme activity is virtually pH-independent over the range tested (pH 5-8), suggesting that the effect of His 158 protonation state on activity is probably masked by a rate-limiting step not associated with catalysis. Considerable product inhibition is observed for wildtype protein, indicating that product release may be the ratedetermining step, but further kinetic studies will be needed to confirm such an assertion. Nonetheless, the pH dependence of enzymatic turnover (Fig.  1B) does provide valuable information about the role of Glu 170 in catalysis. If the sole purpose of Glu 170 is to elevate the apparent pK a of His 158 , wild-type enzyme activity should be restored to the E170Q mutant by reduction of pH. Assuming a pK a value of 6.7 (Table II), His 158 would be fully protonated at pH 5.0 and capable of providing a proton to AACoA upon hydride transfer from NADH. However, E170Q HAD remains ϳ30-fold less active than native enzyme at this pH (Fig. 1B), suggesting that efficient catalysis requires more than just protonation of His 158. Clearly, Glu 170 has additional contacts within the active site that impact enzyme function.
Active site integrity can be further assessed by the formation of a charge transfer complex composed of HAD, NAD ϩ , and AACoA. Within this abortive ternary complex, the electron-rich enolate form of AACoA is juxtaposed to the electron-deficient nicotinamide ring of NAD ϩ , giving rise to the charge transfer band centered between 410 and 420 nm (6). Enolate formation, as judged by absorbance at 302 nm, is reduced dramatically in the E170Q mutant (data not shown), resulting in a comparably diminished charge transfer absorbance at 412 nm (Fig. 2). The disparity between native and E170Q charge transfer complex formation is nearly constant over the pH range tested with the intensity of the native charge transfer band ϳ3-fold greater than that of the E170Q mutant. Although E170Q exhibits a reduced ability to promote enolate formation, probably as a result of improper positioning of the active site histidine (Fig.  4), the precise role of His 158 in charge transfer complex formation is not known. It may facilitate preferential binding of the enolate species of AACoA by serving as a hydrogen bond donor or it may actively promote enolate formation by abstracting the proton from the enol form. The pK a of 6.7 attributed to His 158 by kinetic analysis of the E170Q mutant would suggest the former; His 158 would be primarily protonated over much of the pH range assayed for charge transfer complex formation and thus able to function as a hydrogen bond donor. However, that pK a assignment has not been confirmed, nor can it be assumed that the pK a of His 158 in the charge transfer complex is the same as in the Michaelis complex.
To facilitate interpretation of the biochemical data, x-ray crystallographic studies were undertaken. The E170Q abortive ternary complex structure is quite similar to the corresponding native model and exhibits an overall conservation of active site architecture (Figs. 4 and 5). The glutamate to glutamine substitution does result in perturbations of side chain orientations, most notably the rotation of the imidazole ring of His 158 by ϳ90°relative to native enzyme. Although glutamine can potentially form a hydrogen bond with NE2 of His 158 , there is no evidence for such an interaction within the E170Q crystal structure. Instead, the imidazole ring of His 158 forms hydrogen bonds between its ND1 and the backbone carbonyl of Phe 159 as well as between its NE2 and the side chain amide of Asn 208 (Table IV). Thus, it appears that His 158 requires Glu 170 to serve as a strong hydrogen bond acceptor and/or provide a negative charge for proper orientation of its imidazole ring.
Disruption of the hydrogen bond between the ND1 of His 158 and the carboxylate oxygen of Glu 170 alters the preferred orientation of the imidazole ring, resulting in an enzyme active site that is no longer optimized for efficient catalysis. The interatomic distance between NE2 of His 158 to O3 of AACoA increases from 2.62 Å in the native structure to 3.37 Å in the E170Q model (Table IV). In such a conformation, it is unlikely that His 158 would be able to effectively serve as a general acid/general base in the E170Q mutant. Instead, the imidazole ring of His 158 would have to adopt an orientation similar to wild-type enzyme to form a favorable interaction with the substrate. Accordingly, a greater than 3-fold increase in the Bfactor for O3 of AACoA is observed for the mutant structure relative to that of native enzyme. In addition to its contact with His 158 , the enolate of AACoA is stabilized by a hydrogen bond between its thioester carbonyl (O1) and the backbone amide of Asn 161 . However, comparable interatomic distances of 2.98 and 3.21 Å are observed for wild-type and mutant enzymes, respectively, indicating that this interaction is not significantly compromised by the E170Q mutation (Table IV). Thus, the reduction in E170Q enzyme activity, as well as the decrease in charge transfer complex formation, could be attributed, at least in part, to the rotation of the His 158 side chain.
Another consequence of the E170Q mutation is the disruption of the hydrogen bond network in the hinge region of the enzyme (residues 201-207), which relates the NAD ϩ -binding domain to the C-terminal domain. In native protein, a carboxylate oxygen of Glu 170 forms hydrogen bonds with the backbone amides of Ile 206 and Val 207 . In contrast, Gln 170 forms only a weak hydrogen bond (ϳ3.5 Å) with the amide of Ile 206 (Fig. 5, Table IV). In addition, the backbone amide of Asn 208 moves more than 0.6 Å away from the conserved water molecule located at the enzyme active site in the E170Q structure. The result is a 3-fold increase in the B-factor for the water molecule in the E170Q model as compared with native enzyme. A manifestation of these structural perturbations is reduced E170Q protein stability at 37°C (Fig. 3). Over the course of 4 h, wild-type HAD retains complete enzyme activity as opposed to the ϳ4-fold reduction in E170Q catalysis. Reduced protein stability could also account for the diminished protein yield from overexpression and purification of E170Q. Furthermore, disruption of the hinge region may contribute to reduced catalytic activity. Rotation of the NAD ϩ -binding domain inward toward the dimer interface is required for efficient substrate binding and catalysis. The structural rearrangements in this region, although subtle, may impair the dynamic movement of the enzyme.
Several examples of catalytic dyads composed of a histidine and an aspartate residue have been extensively characterized. As discussed by Levy and co-workers (15), the aspartate residue of such active site pairs has been postulated to serve various functions. In the case of glucose 6-phosphate dehydrogenase (15) and mandelate racemase (14), it appears that the aspartate residue facilitates catalysis by altering the pK a of the histidine. In both cases, the position of the histidine residue is relatively undisturbed by replacement of aspartate with an asparagine residue. Similarly, the position of the active site histidine is unaffected by analogous aspartate substitutions in either trypsin (17) or ribonuclease A (18). In these examples, aspartate is proposed to stabililize the productive tautomer of the active site histidine. Subsequent NMR studies with ribonuclease A have demonstrated that aspartate has only a modest effect on the pK a of the active site histidine (19). A third predicted function of aspartate is orientation of the imidazole ring of the histidine for optimal catalysis. However, significant perturbations in positioning of the imidazole ring as a result of aspartate substitution by asparagine have not been reported.
Comparable studies with histidine-glutamate pairs have not been as comprehensive. D-Isomer-specific 2-hydroxyacid dehy-drogenases (Pfam accession number: PF00389 (20)), including D-3-phosphoglycerate dehydrogenase, D-glycerate dehydrogenase, and D-lactate dehydrogenase, exhibit a conserved supersecondary structural motif related to, but distinguishable from, L-specific NAD ϩ -dependent dehydrogenases (21) and typically contain an active site histidine-glutamate pair. An exception is formate dehydrogenase that instead utilizes a histidine-glutamine pair (22). Replacement of the conserved glutamate (or glutamine) residue of a D-2-hydroxyacid dehydrogenase generally reduces catalytic activity and/or alters the pH profile of the enzyme (22)(23)(24). However, crystallographic studies of these mutant enzymes have not been described. The current study represents, to our knowledge, the first detailed report of a structural characterization of a histidine-glutamate catalytic dyad. It is worth noting that although HAD contains a histidine-glutamate active site pair, it is an L-specific dehydrogenase and is structurally more closely related to L-lactate and L-malate dehydrogenases (5). Thus, the results obtained using native and E170Q HAD, although consistent with previous studies, are not necessarily representative of the D-isomerspecific 2-hydroxyacid dehydrogenase enzyme family.
In addition to demonstrating the importance of the His 158 -Glu 170 dyad, comparison of the kinetic and spectroscopic properties of native and E170Q enzymes provides valuable mechanistic information regarding substrate binding and charge transfer complex formation. In particular, the importance of an unidentified residue with an apparent pK a of 5.3 has been demonstrated. Native and E170Q HAD exhibit comparable K m values for AACoA at each pH tested (Fig. 1, C and D), consistent with structural studies, which indicate that Glu 170 does not directly participate in substrate binding (Fig. 5). Examination of the pH profile of either protein suggests that an ionizable group with an apparent pK a of ϳ5.3 is associated with substrate binding (Table II). Similarly, the pH profiles of charge transfer complex formation by E170Q and HAD are nearly identical, each yielding an apparent pK a of ϳ5.3 for enolate formation. It seems likely that the similar pK a values associated with AACoA binding and enolate formation are related.
One possibility is that the pK a of 5.3 reflects a group on the substrate. AACoA is presumed to occur primarily in its enol form in solution and, as such, has three ionizable groups. The amine group on the adenine ring and the ribose phosphate of coenzyme A have reported pK a values of 4.0 and 6.4, respectively (25), although these values may shift under the current assay conditions. It is unlikely, however, that the protonation of either of these groups would significantly affect binding. The adenosine diphosphate portion of coenzyme A is only weakly associated with the enzyme as indicated by its relatively high B-factors (6) and is probably not a major determinant in substrate binding (26). The major contact sites between HAD and AACoA involve only the pantetheine and fatty acyl portions of the substrate. Thus, the only ionizable group in the solution form of AACoA likely to be relevant to binding is its enol group. However, the pK a of enolate formation in solution has been estimated to be greater than 8.5 (27), much larger than the pK a associated with substrate binding. Circumstantially, it appears that the relevant ionizable group resides within the enzymecofactor binary complex.
Decreased enolate formation could result from reduced affinity of the HAD-NAD ϩ binary complex for AACoA. The kinetic studies indicate that at low pH, an unidentified group becomes protonated and increases the K m for AACoA. This decrease in substrate affinity correlates well with decreased charge transfer formation. However, the charge transfer studies were carried out at an AACoA concentration of 2 mM, nearly 30-fold higher than the highest determined K m of AACoA (64.8 Ϯ 3.18 M). At pH 5.0, the intensity of the charge transfer band is already reduced by greater than 50%, although enzyme should be saturated with AACoA. Furthermore, a 5-fold increase in either AACoA or NAD ϩ concentration did not increase the intensity of the charge transfer band (data not shown), indicating that reduced substrate binding cannot account for the decrease in charge transfer complex formation.
Examination of the E170Q kinetic and charge transfer studies does provide a possible explanation for the observed pK a . The pH profile of the K m values suggests that substrate binding is reduced upon protonation of an ionizable group with an apparent pK a of 5.3 (Fig. 1, C and D). However, charge transfer complex formation, which exhibits the same apparent pK a as substrate binding (Fig. 2), cannot be increased by the addition of excess AACoA. HAD may still bind AACoA with reduced affinity when this group is protonated but may fail to adopt the conformation necessary for enolate and charge transfer complex formation. Structural studies of HAD have suggested that FIG. 6. Contribution of glutamate 110 to L-3-hydroxyacyl-CoA dehydrogenase structure and function. In both panels, nitrogen atoms are colored in dark gray, carbon in light gray, oxygen in white, and all other atom types in medium gray. A, the ribbon diagram illustrates a top view of an E170Q HAD monomer with residues Glu 110 and Ser 137 rendered as ball and stick models. Glu 110 resides within a loop linking the fourth ␤-strand and sixth ␣-helix of the NAD ϩ -binding domain of HAD; similarly, Ser 137 is located on a loop relating the fifth ␤-strand and seventh ␣-helix. Both loops protrude into the enzyme active site, which is located in the cleft between the NAD ϩ -binding and the C-terminal domains. B, hydrogen bonds formed by Glu 110 in the E170Q abortive ternary complex. In addition to its involvement in cofactor binding, Glu 110 also appears to position the loop containing Ser 137 . high affinity substrate binding and effective catalysis require a large conformation change in which the NAD ϩ -binding domain of the enzyme swings toward the C-terminal dimerization domain, sequestering the enzyme active site (6). The apparent pK a of 5.3 may therefore reflect the protonation state of a group involved in this domain shift.
The trigger responsible for the domain shift observed upon substrate binding has not yet been identified. Ser 137 undergoes the biggest conformational change of active site residues upon formation of the abortive ternary complex, as compared with the binary HAD-NAD ϩ complex. In the binary complex, the hydroxyl group of Ser 137 is found within hydrogen bond distance of NE2 of His 158 (5). Upon abortive ternary complex formation, the backbone carbon of Ser 137 shifts greater than 1 Å. Its hydroxyl group is found equidistant to O3 of AACoA and the 2Ј-hydroxyl group of the nicotinamide ribose of NAD ϩ (Fig.  5A) and is no longer close enough to hydrogen-bond to His 158 (6). The conservation of amino acid sequence in the Ser 137 loop is consistent with such a critical role in catalysis. However, a pK a shift to 5.3 would be quite dramatic for a serine residue.
A more suitable assignment for the pK a of 5.3 is the carboxylate of a highly conserved glutamate residue that appears to be essential for proper positioning of Ser 137 . Glu 110 is found within a loop linking the fourth ␤-strand and sixth ␣-helix of the NAD ϩ -binding domain of HAD (Fig. 6A) and has been implicated in cofactor binding (5). In particular, the side chain of Glu 110 forms hydrogen bonds with the hydroxyl groups of the nicotinamide ribose (Fig. 6B). Binding of NADH by native HAD does not exhibit a strong pH dependence in the pH range of 5-8, suggesting that deprotonation of Glu 110 is not required for effective cofactor binding. However, a carboxylate oxygen of Glu 110 is also within hydrogen bond distance of the backbone amides of Ser 137 and Ser 138 (Fig. 6B). Thus, Glu 110 may help orient the mobile loop containing the active site residue, Ser 137 , and reflect the unassigned pK a ; protonation of this carboxylate oxygen could disrupt the interactions with the mobile Ser 137 loop and affect substrate binding and enolate formation. Further studies are required to assign the pK a of 5.3, and as such, efforts are currently under way to establish the roles of Glu 110 and Ser 137 in binding and catalysis.
The characterization of E170Q HAD described is consistent with previously reported studies of the homologous E. coli fatty acid oxidation complex and provides insight into a clinical mutation found in human trifunctional protein (TFP). Analogous to the E170Q mutation, substitution of Glu 462 to a glutamine (E462Q) in the E. coli fatty acid oxidation complex results in decreased protein stability and reduced oxidative and reductive catalysis, whereas cofactor and substrate binding are unaffected (4). The fatty acid oxidation complex of E. coli is homologous to human TFP, a membrane-associated multifunctional enzyme that catalyzes three successive reactions of the ␤-oxidation spiral (28,29). An internal region of TFP ␣-subunit displays considerable sequence homology to HAD, suggesting that the E170Q mutant of HAD is an appropriate model system to investigate a comparable clinical mutation in human TFP. An E474Q mutation found within the ␣-subunit of TFP is often associated with acute fatty liver of pregnancy and hepatic dysfunction in heterozygous mothers carrying affected fetuses (30 -32) and is comparable with the E170Q mutation described. By analogy, the elevated levels of long chain 3-hydroxy fatty acid metabolites found in these patients may be attributable to reduced catalytic efficiency due to improper positioning of the active site histidine and possibly decreased protein stability.
A histidine-glutamate catalytic dyad is conserved in all known L-3-hydroxyacyl-CoA dehydrogenases, consistent with its critical role in catalysis. In this report, we have demonstrated that Glu 170 of human HAD is required for positioning of the active site histidine and the structural stability of the enzyme. Despite its similar size and potential for hydrogen bond formation, a glutamine residue cannot fulfill this requirement, suggesting that the negative charge of Glu 170 and/or its ability to serve as a strong hydrogen bond acceptor is needed for optimal catalysis and maintenance of active site integrity. Characterization of native and E170Q enzymes has also led to the identification of an ionizable group with an apparent pK a of 5.3 as an important determinant in substrate binding and charge transfer complex formation. The identity of this residue has been tentatively assigned to a second conserved glutamate residue, Glu 110 , but further study will be required to confirm this assertion.