UDP-3-O-((R)-3-hydroxymyristoyl)-N-acetylglucosamine Deacetylase Functions through a General Acid-Base Catalyst Pair Mechanism*

UDP-3-O-((R)-3-hydroxymyristoyl)-N-acetylglucosamine deacetylase (LpxC) is a zinc-dependent enzyme that catalyzes the deacetylation of UDP-3-O-((R)-3-hydroxymyristoyl)-N-acetylglucosamine to form UDP-3-O-(R-hydroxymyristoyl)glucosamine and acetate. The structural similarity of the active site of LpxC to metalloproteases led to the proposal that LpxC functions via a metalloprotease-like mechanism. The pH dependence of kcat/Km catalyzed by Escherichia coli and Aquifex aeolicus LpxC displayed a bell-shaped curve (EcLpxC yields apparent pKa values of 6.4 ± 0.1 and 9.1 ± 0.1), demonstrating that at least two ionizations are important for maximal activity. Metal substitution and mutagenesis experiments suggest that the basic limb of the pH profile is because of deprotonation of a zinc-coordinated group such as the zinc-water molecule, whereas the acidic limb of the pH profile is caused by protonation of either Glu78 or His265. Furthermore, the magnitude of the activity decreases and synergy observed for the active site mutants suggest that Glu78 and His265 act as a general acid-base catalyst pair. Crystal structures of LpxC complexed with cacodylate or palmitate demonstrate that both Glu78 and His265 hydrogen-bond with the same oxygen atom of the tetrahedral intermediate and the product carboxylate. These structural features suggest that LpxC catalyzes deacetylation by using Glu78 and His265 as a general acid-base pair and the zinc-bound water as a nucleophile.

Lipopolysaccharide (LPS) 1 molecules form the outer membrane of Gram-negative bacteria and serve to exclude hydrophobic and negatively charged molecules. Lipid A is the hydrophobic portion of LPS that is responsible for anchoring LPS to the membrane and is essential for the viability of Gram-nega-tive bacteria (1). Lipid A is also known as endotoxin and is the immunomodulatory portion of LPS that triggers the immune system in septic shock. As a consequence, inhibition of lipid A biosynthesis is proposed as a strategy for both the development of novel antibiotics and anti-endotoxins in the treatment of septic shock (1)(2)(3)(4)(5).
In Escherichia coli lipid A is synthesized from UDP-N-acetylglucosamine in a 10-step pathway (Fig. 1). UDP-3-O-((R)-3hydroxymyristoyl)-N-acetylglucosamine deacetylase (LpxC) is a zinc-dependent enzyme that catalyzes the hydrolysis of UDP-3-O-((R)-3-hydroxymyristoyl)-N-acetylglucosamine (myr-UDP-GlcNAc) to form UDP-3-O-(R-hydroxymyristoyl)glucosamine and acetate (6). The deacetylation of myr-UDP-GlcNAc is the committed step in the biosynthesis of lipid A (7); therefore, this enzyme is a target for the development of inhibitors as antibiotics for the treatment of Gram-negative infections (2, 8 -10). A comprehensive understanding of the catalytic mechanism and structure of LpxC will facilitate the development of potent and specific inhibitors of this enzyme.
The enzyme LpxC belongs to a group of enzymes known as the zinc hydrolases. Mononuclear zinc hydrolases can be broadly categorized by two general catalytic mechanisms as follows: one that uses a single bifunctional general acid-general base catalyst (GABC) (i.e. metalloproteases), and a second mechanism that uses a GABC pair (i.e. histone deacetylases) to carry out catalysis (11)(12)(13)(14). The structure of Aquifex aeolicus LpxC (AaLpxC) has been solved using x-ray crystallography (15) and NMR spectroscopy (16). This structure (Fig. 2) reveals that LpxC contains a unique fold and a novel zinc-binding motif, both of which distinguish LpxC from other known zinc hydrolases (15). One bound zinc ion is an essential cofactor for LpxC catalytic activity (Zn A ), whereas excess zinc leads to enzyme inhibition (6) by Zn B (Fig. 2b) (15). The catalytic zinc ion (Zn A ) is coordinated by His 79 , His 238 , Asp 242 , and a solvent molecule, whereas the inhibitory Zn B ion is coordinated by Glu 78 , His 265 , a fatty acid, and a bridging solvent molecule. Previous mutagenesis experiments suggest that three active site residues (Glu 78 , Asp 246 , and His 265 ) are important for catalytic activity, as mutation of these residues to Ala decreases the catalytic activity Ͼ10 2 -fold (17). On the basis of these structural and mutagenesis data, LpxC was proposed to function via a metalloprotease-like mechanism using a single general acid-base catalytic side chain (Glu 78 ) and a zinc-water nucleophile to catalyze deacetylation of its substrate (Fig. 3A), as described in detail for metalloproteases (11)(12)(13)15). The Zn A ion and His 265 provide stabilization of the transition states and oxyanion intermediate prior to protonation of the amine leaving group and collapse of the oxyanion intermediate. In contrast, the deacetylation mechanism for histone deacetylases has been proposed to use a pair of His side chains, one func-tioning as a general acid and the other as a general base to activate the zinc-water nucleophile (12,18). The details of the LpxC catalytic mechanism have not yet been fully elucidated. The pH dependence of AaLpxC wild-type and mutants has been reported recently (19), suggesting that Glu 78 functions as a general base.
Here we investigate the catalytic mechanism of LpxC using metal substitution, mutagenesis, pH dependence, and x-ray crystallography. The k cat /K m for E. coli LpxC (EcLpxC) has a bell-shaped dependence on pH with two pK a values of ϳ6.4 and 9.1 similar to values recently reported for AaLpxC (19). Metal substitution and mutagenesis experiments suggest that the basic limb of the pH profile represents ionization of a metal-coordinated group, such as the zinc-water molecule. Kinetic evaluation of LpxC mutants and crystal structures of LpxC complexed with cacodylate or palmitate support a mechanism in which Glu 78 and His 265 function as a general acid-base catalyst pair, wherein the His 265 side chain functions as a general acid to protonate the amine leaving group, whereas Glu 78 functions as a general base to activate the zinc-water nucleophile.

EXPERIMENTAL PROCEDURES
Mutagenesis and Protein Expression-All mutant plasmids were prepared using the QuickChange site-directed mutagenesis kit (Stratagene). The LpxC variants were overexpressed and purified according to published procedures (6,8,20,21). For expression of less stable LpxC mutants, cells were incubated overnight at 25°C following addition of isopropyl D-thiogalactopyranoside. Additionally, 100 M isopropyl Dthiogalactopyranoside and 200 M ZnSO 4 were added at induction for expression of the D246A EcLpxC mutant.
All purification steps were carried out at room temperature and 4°C for AaLpxC and EcLpxC, respectively. Following lysis using a microfluidizer, the cell debris was pelleted by centrifugation, and the supernatant was loaded onto a DEAE-Sepharose column in 25 mM HEPES, pH 7, 2 mM dithiothreitol. Both AaLpxC and EcLpxC were eluted with a linear gradient (0 -250 mM KCl). The LpxC was further fractionated using a Reactive Red 120 column with a linear gradient (0 -250 mM KCl for EcLpxC and 150 -500 mM KCl for AaLpxC). The LpxC enzymes were Ͼ95% pure as assessed by SDS-PAGE and were stored at Ϫ80°C.
All metals were removed from the purified enzymes by incubation with 20 mM dipicolinic acid, as described previously (6). The apoLpxC was reconstituted with stoichiometric amounts of ZnSO 4 , CoSO 4 , or NiSO 4 prior to use in assays (6). The final metal-enzyme stoichiometry was determined using inductively coupled plasma emission mass spectrometry. For LpxC prepared using these methods, no fatty acid contaminants were detected by mass spectrometry following extraction of LpxC with methylene chloride.
LpxC Assay-The deacetylase activity was measured as described previously (6,23). Briefly, assay mixtures containing buffer, bovine serum albumin (fatty acid-free, 1 mg/ml), triscarboxyethylphosphine (0.5 mM), and [ 14 C]UDP-3-O-((R)-hydroxymyristoyl)-N-acetylglucosamine were pre-equilibrated at the assay temperature (EcLpxC at 30°C, AaLpxC at 60°C), and the reactions were initiated by the addition of enzyme. After incubation for various times, the reaction was quenched by the addition of sodium hydroxide, which also cleaves the myristate substituent for ease of separation. The substrate and product were separated on PEI-cellulose TLC plates (0.1 M guanidinium HCl) and quantified by scintillation counting. The initial rate of product formation (Ͻ20% reaction) was determined from these data. For the EcLpxC mutants the initial rate was measured in 20 mM bis-Tris propane, pH 7.5, at seven to nine different concentrations (25 nM Equation 2 was fit to the pH rate profiles for the E78A/H265A mutants, where k 1 is the k cat /K m at the pH optimum, and k 2 is the pH independent value of k cat /K m at low pH. LpxC is stable over this pH range under the assay conditions. The catalytic activity of LpxC decreases Յ2-fold after incubation at the altered pH for Յ5 min followed by measuring the activity at pH 7.5. For the solvent isotope effect experiments, the initial rates at subsaturating substrate concentrations in H 2 O were compared with the initial rates in ϳ95% D 2 O. The pD values obtained for the D 2 O buffers using the pH meter readings were corrected by adding 0.4 to these values (24). Crystallography-The C193A/⌬Asp 284 -Leu 294 variant of A. aeolicus LpxC was used for crystallography experiments as described (15). Crystallization was achieved by equilibrating a hanging drop containing 3 l of protein solution (3 mg/ml LpxC, 100 mM HEPES, pH 7.5, 180 mM NaCl, 9 -14% PEG3350, and 0.5 mM ZnSO 4 ) and 3 l of precipitant buffer (100 mM HEPES, pH 7.5, 180 mM NaCl, 9 -14% PEG3350, and 0.5 mM ZnSO 4 ) over a reservoir containing ϳ1 ml of precipitant buffer. Crystals with maximum dimensions of 0.3 ϫ 0.15 ϫ 0.15 mm 3 grew within 3 days and were gradually transferred to a stabilization buffer of 100 mM sodium cacodylate, pH 6.0, 180 mM NaCl, 11-16% PEG 3350, 0.5 mM ZnSO 4 , and 1% glycerol. Crystals were flash-cooled in liquid nitrogen following cryoprotection with 22% glycerol and diffracted x-rays to 2.1 Å at the Argonne National Laboratory (IMCA-CAT, Argonne, IL). Crystals were isomorphous with those prepared at pH 7.0 (15) and belong to space group P6 1 with unit cell dimensions a ϭ b ϭ

FIG. 3. The proposed mechanisms for LpxC using either a single bifunctional GABC (A) or a GABC pair (B).
101.3 Å, c ϭ 122.7 Å. Data were indexed and merged using the program HKL2000 (25). The structure was solved by molecular replacement using the zinc-inhibited enzyme, excluding all zinc ions, solvent, and fatty acid molecules, as a search probe for rotation and translation functions calculated with the program AmoRe (26). It was clear in initial electron density maps that the inhibitory metal ion, Zn B , had dissociated and that a tetrahedral cacodylate anion was coordinated to the catalytic metal ion Zn A . In native AaLpxC, a fatty acid interpretable as myristate or disordered palmitate occupied the hydrophobic tunnel and coordinated to Zn B ; because fatty acids were not included in the crystallization medium, this fatty acid must be a remnant of heterologous expression in E. coli. The fatty acid remained bound in the LpxCcacodylate complex, and its carboxylate group was displaced to two alternate conformations as a consequence of Zn B dissociation. Iterative cycles of refinement and model rebuilding were performed with the programs CNS (27) and O (28), respectively, to improve the structure as monitored by R free . Data collection and refinement statistics are reported in Table I. In a second experiment, crystals were gradually transferred to a stabilization buffer containing 100 mM bis-Tris, pH 6.0, 180 mM NaCl, 11-16% PEG 3350, 0.5 mM ZnSO 4 , and 1% glycerol. Following transfer to a 22% glycerol cryoprotectant, crystals were flash-cooled in liquid nitrogen and yielded x-ray diffraction data to 2.7 Å using an R axis IVϩϩ image plate detector mounted on a Rigaku-200HB rotating anode x-ray generator. Diffraction data were indexed and merged using the program d*Trek (29). The structure of zinc-inhibited LpxC, excluding all zinc ions, solvent, and fatty acid molecules, was used as a search probe in molecular replacement calculations with the program EPMR (30) to phase the initial electron density map. The atomic model was refined and rebuilt using the programs CNS (27) and O (28), respectively. Strict NCS constraints were used during the initial stages of refinement and then relaxed into appropriately weighted restraints as indicated by R free as refinement progressed. Data collection and refinement statistics are reported in Table I. Figs. 2a, 6a, and 7a were prepared using the program Bobscript (31). (17) and the crystal structure of the zincinhibited LpxC (15) have suggested that residues Glu 78 , His 265 , and Asp 246 are important for catalytic activity. Therefore, the steady-state parameters K m , k cat , and k cat /K m were determined for the E78A, H265A, D246A, and E78A/H265A EcLpxC mutants and compared with the values obtained for WT EcLpxC (Table II). The E78A, D246A, and H265A single mutations all decreased k cat /K m values by 400 -2200-fold, whereas the largest decrease in k cat /K m was measured for the E78A/H265A double mutant (ϳ1.5 ϫ 10 4 -fold). The decrease in k cat /K m values ob-served for the LpxC mutants is predominantly explained by a decrease in the k cat values for these enzymes. The smallest decrease in k cat was observed for the E78A mutant (18-fold), whereas the largest decrease was observed for the E78A/ H265A mutant (1.6 ϫ 10 3 -fold). These mutations have more modest effects on the K m values, with the E78A mutation causing the largest increase (23-fold). The observed changes in the K m values rule out the possibility that the activity in the mutants is because of WT contamination. To elucidate further the functional roles of these residues, we measured the pH dependence of catalysis.

Mutations Decrease Catalytic Activity-Previous mutagenesis experiments
pH Dependence of the LpxC-catalyzed Reaction-The pH dependence of the LpxC-catalyzed reaction for the EcLpxC and AaLpxC was determined under subsaturating substrate (k cat / K m ) conditions to identify ionizations in the free enzyme or substrate that are important for catalytic activity. The k cat /K m values included the rate constants for substrate binding through formation of acetate (first irreversible step) (32). In contrast, the kinetic parameter k cat included the rate constants after formation of the E⅐S complex through product dissociation. For enzymes where product release is rate-limiting, as is likely for wild-type EcLpxC, 2 k cat provided little information about the rate constant of the chemical step (k chem ). The pH dependence of AaLpxC was also determined because the structure of A. aeolicus LpxC is known, the protein stability is enhanced, and AaLpxC can be readily purified from the background WT E. coli LpxC. The pH dependence of k cat /K m for WT EcLpxC (Fig. 4) affords a bellshaped curve with two apparent pK a values of 6.4 Ϯ 0.1 and 9.1 Ϯ 0.1. The pK a values measured for the reaction catalyzed by wild-type AaLpxC are decreased by 0.5-1 pH units (Table  III) and are comparable with literature values (19). These data indicate that at least two ionizations in either LpxC or the substrate are important for catalytic activity.
Mutations Alter the pH Dependence of LpxC-The pH dependence of k cat /K m for LpxC mutants at Glu 78 , Asp 246 , and His 265 were examined to determine whether ionization of these groups was responsible for one or more of the ionizations observed in the pH profile ( Fig. 4 and Table III). In EcLpxC the mutation of Glu 78 , Asp 246 , or His 265 to Ala caused a 300 -3000-2 J. E. Jackman and C. A. Fierke, unpublished data.  fold decrease in k cat /K m at the pH maximum. The same mutations in AaLpxC decreased the value of k cat /K m at the pH maximum about 10-fold less due mainly to an ϳ6-fold decrease in the catalytic activity of the WT AaLpxC. All of these single mutations in both EcLpxC and AaLpxC retained a bell-shaped pH dependence consistent with two ionizations, and except for E78A AaLpxC, the pK a values remained within 0.5 unit of the wild-type values. In E78A AaLpxC the value of pK a1 increased by ϳ1 pH unit, whereas the value of pK a2 decreased, with both of these values approaching 7 (Table III). Taken together, these data suggested that ionization of none of these groups alone is essential for the bell-shaped pH dependence.
To examine further the origin of the ionizations that alter the activity of LpxC, the pH dependence of the E78A/H265A double mutant was measured in both EcLpxC and AaLpxC. These mutants displayed the largest decreases in k cat /K m values (1,700 -37,000-fold) (Table III), consistent with the loss of a GABC. Furthermore, the pH dependence of these mutant enzymes was significantly altered. The value of pK a2 was either increased modestly (EcLpxC) or unchanged (AaLpxC). However, at low pH the large decrease in catalytic activity described previously by pK a1 was not observed (Fig. 4 and Table III). The activity at low pH was relatively independent of pH. The small decrease in catalytic activity observed at low pH can be described by an ionization (pK a3 ϭ 7.4 Ϯ 0.1) that decreases the catalytic activity modestly (ϳ4-fold). These data demonstrated that an ionization that caused the acidic limb of the pH profile was observed in both the E78A and H265A single mutants but not in the E78A/H265A double mutant. One potential explanation of these data is that the catalytic mechanism uses Glu 78 and His 265 as a GABC pair instead of employing one of these side chains as a single GABC (Fig. 3B), and ionization of either residue could lead to the ioni-zation observed as pK a1 in wild-type and the single mutants.
pK a2 Reflects Ionization of the Metal-Water Ligand-To ascertain if one of the observed pK a values reflected ionization of the metal-bound water molecule, the pH dependence of LpxC substituted with Co(II) or Ni(II) was examined (Table IV). Previous experiments have demonstrated that these metal ions are capable of supporting LpxC activity in lieu of zinc (6). These data show that pK a2 shifts upward (Ni(II) Ͼ Co(II) Ͼ Zn(II)) upon substitution of Zn(II) with Co(II) or Ni(II) for the E. coli and A. aeolicus enzymes, paralleling the measured acidities of these metal-water ligands (11,33). Additionally, the value of pK a2 increased upon substitution of Zn(II) with Co(II) for the H265A AaLpxC mutant. The upward shift in the value of pK a2 as the metal-water acidity decreased suggested that pK a2 reflected ionization of a metal-coordinated group, such as water in both the WT and mutant LpxC enzymes.
Solvent Isotope Effect-In both proposed mechanisms (Fig.  3), Glu 78 functioned as a general base catalyst. Solvent isotope effects (K H 2 O /K D 2 O ) of 2-4 were typically observed for reactions that proceed via GBC (34). The pH dependence of the solvent isotope effects for WT, E78A, H265A, and E78A/H265A E. coli LpxC were measured ( Fig. 5 and Table V). A solvent isotope effect (V/K H 2 O /V/K D 2 O ) of ϳ2.1 was calculated for WT EcLpxC by comparing the values at the respective pH maximum (Fig.  5), consistent with the transfer of a proton from solvent in the transition state. In the H265A and E78A/H265A mutants, a decreased solvent isotope effect (V/K H 2 O /V/K D 2 O ϭ 1.4) was observed. Most unexpectedly, the E78A mutant had an inverse solvent isotope effect of ϳ0.2 determined from the ratios of the V/K values at the pH maxima. However, in the E78A mutant, the value of the solvent isotope effect was difficult to measure accurately because the values of pK a1 and pK a2 were comparable.
Structure of the LpxC-Cacodylate Complex-The gradual transfer of LpxC crystals from pH 7.0 to pH 6.0 resulted in the complete dissociation of the inhibitory zinc ion, Zn B , and the electron density map of the LpxC active site revealed that a buffer molecule, the tetrahedral cacodylate anion, coordinated to the catalytic zinc ion, Zn A (Fig. 6a). The cacodylate anion coordinated to Zn A in a monodentate fashion (O-2-Zn A separation ϭ 2.1 Å). The O-1-Zn A separation (2.9 Å) was too long for an inner sphere metal coordination interaction. The O-1 atom of cacodylate also accepted hydrogen bonds from the side chains of Glu 78 and His 265 , whereas the O-2 atom accepted a hydrogen bond from Thr 191 (Fig. 6b). Most importantly, the intermolecular interactions of the cacodylate anion may mimic those of the tetrahedral intermediate and its flanking transition states in LpxC catalysis (Fig. 6c).
A fatty acid molecule interpreted as myristate or disordered palmitate occupied the hydrophobic tunnel and coordinated to Zn B in the active site of the zinc-inhibited enzyme (15). This fatty acid remained in the hydrophobic tunnel upon cacodylate binding to Zn A and was clearly interpretable as palmitate. Its a The metal-substituted enzymes were prepared with a stoichiometry of 1:1 as described under "Experimental Procedures." b The initial rate for LpxC-catalyzed deacetylase activity was determined at 30°C (20 mM bis-Tris propane, pH 7.5, 1 mg/ml bovine serum albumin, 0.5 mM triscarboxyethylphosphine) with myr-UDP-GlcNAc as the substrate. The kinetic parameters were obtained from the initial velocities, as described under "Experimental Procedures."  Table III were determined by fitting an equation including two ionizations (Equation 1) or two ionizations where one pK a decreases the activity modestly (Equation 2) (dashed line) to the data. carboxylate group was displaced to two alternate conformations and did not contact the zinc-bound cacodylate anion. In one conformation the carboxylate hydrogen bonds with His 58 , and in the alternate conformation the carboxylate hydrogen bonds with a water molecule. Neither cacodylate binding nor the pH change triggered any significant conformational changes, and the root mean square deviation was 0.2 Å for 267 C-␣ atoms between the cacodylate-bound and the zinc-inhibited LpxC structures.
Structure of the LpxC-Palmitate Complex-Equilibration of LpxC crystals at pH 6.0 with a noncoordinating buffer, bis-Tris, resulted in the dissociation of Zn B and the coordination of the palmitate carboxylate to Zn A (Fig. 7a). Presumably, the palmitate molecule corresponded to the fatty acid that coordinated to Zn B in the zinc-inhibited enzyme (15). The electron density map of the LpxC-palmitate complex revealed that palmitate extended from the hydrophobic tunnel, and its carboxylate head group made a nearly symmetric bidentate coordination interaction with Zn A (Fig. 7a). Palmitate carboxylate oxygen O-1 accepted hydrogen bonds from Glu 78 and His 265 , and carboxylate oxygen O-2 accepted a hydrogen bond from Thr 191 (Fig. 7b). These hydrogen bond interactions were comparable with those accommodating the binding of the tetrahedral cacodylate anion (Fig. 6). Neither palmitate binding nor the pH change triggered any significant conformational changes, and the root mean square deviation is 0.2 Å for 267 C-␣ atoms between the palmitate-bound and the zinc-inhibited LpxC structures.

coli and A. aeolicus LpxC
The k cat /K m values for deacetylation of myr-UDP-GlcNAc catalyzed by EcLpxC and AaLpxC were determined at 30 and 60°C, respectively, as a function of pH. The pH of the buffers was measured at the reaction temperature. The pH dependence of the k cat /K m values was used to determine the pK a values as described under "Experimental Procedures."

LpxC mutant
LpxC source Active site metal ion a pK a1 pK a2 k cat /K m pH maximum a The single catalytic Zn(II) cofactor was removed by incubation with dipicolinic acid and replaced by incubation with Co(II) or Ni(II), as described under "Experimental Procedures." b Ec and Aa denote E. coli and A. aeolicus LpxC, respectively.

FIG. 5. Solvent isotope effect of E. coli WT LpxC in H 2 O (•) and 95% D 2 O (E) and E78A (f) in H 2 O and 95% D 2 O (Ⅺ).
The values for V/K were measured at 30°C using subsaturating concentrations of myr-UDP-GlcNAc (50 -200 nM), as described under "Experimental Procedures." The pK a values listed in Table V were determined by fitting an equation including two ionizations (Equation 1) to these data.

TABLE V Solvent isotope effects on E. coli LpxC
The LpxC activity was determined using myr-UDP-GlcNAc as the substrate. The pH dependence of k cat /K m was fit using Equations 1 or 2 to calculate the pK a values, as described under "Experimental Procedures."

DISCUSSION
Mutations Significantly Decrease k cat and k cat /K m -The large decreases in the k cat and k cat /K m values for the E78A, D246A, H265A, and E78A/H265A mutants further demonstrate the importance of these side chains in the catalytic mechanism. Preliminary transient kinetic data 2 suggest that for WT EcLpxC the kinetic parameter k cat is limited by a rate constant for a step that occurs after the rate constant for deacetylation (k chem ). However, in these LpxC mutants the deacetylation rate constant is decreased, so that k cat is presumed to measure k chem . Therefore, the decreases in k cat are a lower limit for the effects these mutations have on k chem . The large diminution in k cat (18 -950-fold) observed for mutations at Glu 78 , Asp 246 , and His 265 demonstrates the catalytic importance of these residues in LpxC. The E78A/H265A double mutation has partially additive effects on both k cat and k cat /K m compared with the E78A and H265A single mutations.
pH Dependence of the LpxC-catalyzed Reaction-The pH dependence of the LpxC-catalyzed reaction for the E. coli and A. aeolicus enzymes indicates that at least two ionizations are important for catalytic activity (Fig. 4). The pH dependence can be reasonably described by two ionizations (one protonation and one deprotonation) that each decrease the catalytic activity by more than 100-fold. The pH dependence under k cat /K m conditions reflects ionizations in both the unbound enzyme and substrate (35). However, mutagenesis and metal-substitution experiments alter both of the pK a values (Tables III and IV), suggesting that ionization of the substrate does not cause either of the primary pK a values. The simplest explanation of the pH dependence data is that the acidic limb represents an ionization whose deprotonation is important for activity, and the basic limb of the pH profile represents an ionization wherein protonation is important for activity. pK a2 Reflects Metal-Water Ionization-The molecular origins of the ionizations in the pH rate profile were investigated through metal substitution and mutagenesis studies. The correlation between the value of pK a2 for metal-substituted LpxC (Tables IV) with the Lewis acidity of the substituted metal ions (11,33) in WT and the AaH265A mutant suggests that pK a2 represents ionization of a metal-coordinated group, most likely the water molecule. We propose that one role of zinc in the reaction is to lower the pK a value of the ligated water molecule, which serves as the nucleophile in the LpxC-catalyzed reaction. However, ionization of the metal-water to metal-hydroxide would be predicted to increase the rate constant for nucleophilic attack of the metal-solvent ligand on the carbonyl carbon of the substrate, similar to the reaction catalyzed by carbonic anhydrase. The simplest explanation for the observed decrease in activity following ionization of the metal-water molecule would be that the proton on this water molecule is required for protonation of the amine leaving group to facilitate breakdown of the tetrahedral intermediate to form products (Fig. 3). pK a1 May Reflect Ionization of Glu 78 -Two ionizations are observed for wild-type LpxC and each of the H265A, D246A, and E78A mutants. However, the acidic limb of the pH profile essentially disappears in the E78A/H265A double mutant suggesting that ionization of either Glu 78 or His 265 is observed in pK a1 . Although the data do not clearly distinguish between these two groups, several pieces of evidence argue that ionization of Glu 78 is the origin of pK a1 in WT, D246A, and H265A LpxC enzymes. First, no changes are observed in the measured pK a values for the H265A and D246A mutations (Table III), indicating that ionization of neither of these side chains is observed in the pH rate profile. In contrast, the E78A mutation significantly increases the value of pK a1 in AaLpxC. Second, the crystal structures (Figs. 6 and 7) indicate that Glu 78 donates a hydrogen bond to the zinc-bound carboxylate oxygen of palmitate and to O-1 of cacodylate, suggesting that Glu 78 is protonated at pH 6, consistent with a pK a1 of 6.4. Third, the pK a of His 265 in AaLpxC has been measured by pH titration of the chemical shift in the NMR spectrum as 7.6, significantly higher than the measured values of pK a1 (36). However, the retention of two ionizations in the pH rate profile of the E78A mutant suggests that ionization of a different group, such as His 265 , is observed in the E78A mutant. The positioning of both His 265 and Glu 78 within hydrogen bond distance of O-1 of the bound cacodylate and palmitate suggests that either group could function as a GBC to activate the zinc-water nucleophile.
A third ionization that modulates activity 4-fold is unmasked in the E78A/H265A mutant. This pK a could reflect ionization of a number of different groups, including the diphosphate in the substrate (UDP-GlcNAc pK a values of ϳ2 and 7.5) 3 and could also alter the shape of the pH rate profile of WT LpxC.
The magnitude of the decreases in k cat /K m at the pH maximum for the E78A and H265A mutations (Table II) is on the low side for deletion of a side chain that functions as a GABC (10 2 -10 4 -fold); however, the 10 3 -10 4 -fold lower activity of the E78A/H265A double mutant is more consistent with the loss of a GABC (37,38). Furthermore, the magnitude of the decrease in the maximal k cat /K m for the E78A/H265A mutant is ϳ10-fold smaller than predicted if the effects of the single mutations are completely additive. In the absence of structural perturbations, partially additive effects can be explained by either a cooperative interaction of the mutated residues to facilitate the same rate-limiting step or independent interactions of the mutated residues to facilitate consecutive nonrate-limiting steps (39,40). The partially additive effects of the E78A and H265A mutants suggest that both groups might stabilize a common intermediate, for example by functioning as a general acid-base pair in catalysis (Fig. 3B).
Solvent Isotope Effect-The solvent isotope effect of 2 (k cat / 3 M. Hernick and C. A. Fierke, unpublished data. K m ) that is observed for WT LpxC is consistent with the ratelimiting transfer of a solvent proton in the transition state, as indicated by a general base catalysis mechanism (34). The most striking result from these experiments is the inverse solvent isotope effect observed for the E78A mutant. One possible explanation for this result is that the values of the pK a are inverted in E78A EcLpxC such that a small fraction of the total enzyme is active at any pH. The addition of D 2 O could alter the values of the pK a in the E78A mutant so that a higher fraction of the active enzyme is observed in D 2 O and therefore an inverse solvent isotope effect is obtained. However, the inverse isotope effect could also be explained by a change in mechanism for the E78A mutant to a reaction that proceeds via general acid catalyst, such as rate-limiting acid-catalyzed breakdown of the tetrahedral intermediate or a reaction that involves use of a metal-chelated water (34,41).
LpxC-Cacodylate and -Palmitate Complexes Support a General Acid-base Pair Mechanism-Both Glu 78 and His 265 hydrogen bond with the O-1 atoms of zinc-bound cacodylate (Fig. 6) and palmitate (Fig. 7), suggesting that both residues could simultaneously interact with a zinc-bound solvent nucleophile. Notably, palmitate can be considered as a "bi-product analogue," i.e. it contains structural features present in both products of LpxC catalysis: a carboxylate group like that of acetate, and a fatty acid side chain like that of myr-UDP-GlcNAc. Therefore, the intermolecular interactions of palmitate, and especially those of the palmitate carboxylate, are relevant for understanding product binding in the LpxC active site. The close contact between Glu 78 and the zinc-bound carboxylate oxygen of palmitate likely implies that Glu 78 is predominantly protonated at pH 6.0 to accommodate this interaction, consistent with the pK a value of 6.4 in the acidic limb of the pH rate profile assigned to Glu 78 . It is possible that a similar interaction accommodates the binding of the acetate product in the final step of the hydrolysis reaction. However, the ionization of Glu 78 to the negatively charged carboxylate is required to regenerate the active form of the general base, so the resultant electrostatic repulsion between the negatively charged carboxylate group of Glu 78 and zinc-bound palmitate may facilitate product release. Combining these structural data with the pH rate profiles leads to the proposal that pK a1 reflects ionization of Glu 78 in WT and H265A LpxC, which functions as a general base to activate the zinc-water nucleophile. However, in E78A LpxC the side chain of His 265 replaces Glu 78 as the GBC, albeit with reduced efficiency, and ionization of this group is observed in the pH rate profile as pK a1 . Although it is possible for the apparent pK a values for the acidic and basic groups to be switched (35), the observed values for pK a1 and pK a2 are reasonable for ionization of an active site Glu and zinc-bound water, respectively.
Overall Mechanism-On the basis of these data we propose that Glu 78 and His 265 function as a GABC pair and that the ionizations observed for WT LpxC represent the ionization of Glu 78 (pK a1 ) and the zinc-bound water molecule (pK a2 ). The assignment of these pK a values to Glu 78 and the zinc-bound water molecule is consistent with those reported in a recent study; however, the mechanisms arrived at based on these data are different (19). We propose that Glu 78 and His 265 are poised to function as a GABC pair in the LpxC-catalyzed reaction via the mechanism shown in Fig. 3B, with zinc and Thr 191 providing stabilization of the transition states and oxyanion intermediate. This mechanism is consistent with the structure of the bound tetrahedral cacodylate anion (Fig. 6), which mimics the structure of the oxyanionic intermediate and its flanking transition states; the hydrogen bond interactions of cacodylate with both Glu 78 and His 265 demonstrate that both of these side chains are positioned to function as general acid-base catalysts. The finding that the E78A and H265A mutations have partially additive effects can also be explained by this mechanism, because these residues are functioning cooperatively in the chemical step. One of the major differences between the mechanisms in Fig. 3 is the role of His 265 . This residue is structurally/ electronically precluded from fulfilling both of the proposed roles as an electrostatic catalyst (Fig. 3A) and general acid catalyst (Fig. 3B). We propose that the mechanism wherein His 265 functions as a general acid catalyst is preferable to the metalloprotease-like mechanism where His 265 provides stabilization of the transition states and oxyanion intermediate for the following reasons. The His 265 is part of a His/Asp chargerelay, and normally these charge-relays mediate proton-transfer reactions (42,43). Although the distances from both Glu 78 and His 265 to the CH 3 of the cacodylate complex (equivalent to leaving group amine in the product) are longer than expected for a proton donor (Ͼ4 Å), His 265 is closer. Moreover, His 265 can make a closer approach to the leaving group by rotating about side chain torsion angle 1 , which would weaken or break the hydrogen bond with Asp 246 . This would also lower the pK a of His 265 and facilitate proton transfer to the leaving amino group and collapse of the tetrahedral intermediate. Participation of Thr 191 in the reaction via electrostatic stabilization of the oxyanion intermediate and flanking transition states is supported by the structures of the LpxC-cacodylate and -palmitate complexes (Figs. 6 and 7) where hydrogen bonds between Thr 191 and ligand O-2 (corresponding to the carbonyl oxygen atom of the substrate) atoms are observed. In summary, these data support the proposal that the LpxC-catalyzed reaction proceeds via a general acid-base pair mechanism.