The Active Site of Escherichia coliUDP-N-acetylglucosamine Acyltransferase

UDP-N-acetylglucosamine (UDP-GlcNAc) acyltransferase (LpxA) catalyzes the reversible transfer of anR-3-hydroxyacyl chain from R-3-hydroxyacyl-acyl carrier protein to the glucosamine 3-OH of UDP-GlcNAc in the first step of lipid A biosynthesis. Lipid A is required for the growth and virulence of most Gram-negative bacteria, making its biosynthetic enzymes intriguing targets for the development of new antibacterial agents. LpxA is a member of a large family of left-handed β-helical proteins, many of which are acyl- or acetyltransferases. We now demonstrate that histidine-, lysine-, and arginine-specific reagents effectively inhibit LpxA of Escherichia coli, whereas serine- and cysteine-specific reagents do not. Using this information in conjunction with multiple sequence alignments, we constructed site-directed alanine substitution mutations of conserved histidine, lysine, and arginine residues. Many of these mutant LpxA enzymes show severely decreased specific activities under standard assay conditions. The decrease in activity corresponds to decreasedk cat/K m ,UDP-GlcNAcvalues for all the mutants. With the exception of H125A, in which no activity is seen under any assay condition, the decrease ink cat/K m ,UDP-GlcNAcmainly reflects an increasedK m ,UDP-GlcNAc. His125 ofE. coli LpxA may therefore function as a catalytic residue, possibly as a general base. LpxA does not catalyze measurable UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc hydrolysis or UDP-GlcNAc/UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc exchange, arguing against a ping-pong mechanism with an acyl-enzyme intermediate.

Lipid A is required for growth of E. coli and most other Gram-negative bacteria (16,17). Lipid A is also necessary for maintaining the integrity of the outer membrane as a barrier to toxic chemicals (18,19). Furthermore, lipid A is a potent activator of innate immunity in animal systems (3,20,21). The study of the enzymes involved in lipid A biosynthesis should therefore prove useful for the development of new anti-infective drugs (22). All enzymes involved in E. coli lipid A biosynthesis have now been identified, and their structural genes have been cloned (3,(22)(23)(24).
The only enzyme of the pathway for which an x-ray structure is available is LpxA (25). LpxA is a trimer of identical subunits, and it represents the first example of a protein with a lefthanded, parallel ␤-helix as the predominating feature of its secondary structure (25). The 10 coils of the ␤-helix of E. coli LpxA are specified by 24 complete and six incomplete hexapeptide repeats, most of which are contiguous (25). Three repeats (18 amino acid residues) make up one turn of the ␤-helix. Many other bacterial and eucaryotic proteins contain similar contiguous hexapeptide repeats (26). Three additional hexapeptide repeat proteins have recently been crystallized and shown to contain the same left-handed ␤-helix structure as seen in LpxA. These are a carbonic anhydrase from Methanosarcina thermophila (27), a tetrahydrodipicolinate N-succinyltransferase from Mycobacterium bovis BCG (28), and a xenobiotic acetyltransferase from Pseudomonas aeruginosa (29). Like LpxA, these enzymes are trimers.
There are no structural clues to the location or mechanism of the LpxA active site, because the LpxA crystal structure was solved in the absence of substrates or inhibitors (25). However, the sequences of more than 15 LpxAs are now available. The amino acid residues conserved across all LpxAs cluster around a deep cleft located between adjacent subunits (Fig. 2), which contains multiple histidines and other basic residues. Because of its symmetry, LpxA has three such clefts (not all visible in Fig. 2). The ACP substrate of LpxA consists of 77 amino acid residues and is very acidic (30,31). Accordingly, it is plausible that the substrates R-3-hydroxymyristoyl-ACP and UDP-Glc-NAc might bind to this basic cleft. Further evidence for the importance of the cleft comes from recent studies of the acyl chain length specificity of E. coli LpxA (32). Substitution of glycine 173 (red residue in the shaded area of Fig. 2) with methionine switches the acyl chain length selectivity of E. coli LpxA from 14 carbons to 10 (32).
We now present chemical modification studies to demonstrate that histidine, lysine, and arginine residues are important in LpxA catalysis, whereas serine and cysteine residues are not. We also use site-directed mutagenesis to examine the effects of changing conserved histidine, lysine, and arginine residues. The combined findings support the view that His 125 of E. coli LpxA is directly involved in catalysis, whereas other conserved basic residues may participate in substrate binding. We also show that high levels of homogeneous LpxA do not catalyze UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc hydrolysis or UDP-GlcNAc/UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc exchange. The results argue against a ping-pong mechanism with an acyl-enzyme intermediate but are consistent with the direct transfer of the acyl chain from R-3-hydroxyacyl-ACP to the glucosamine 3-OH of UDP-GlcNAc.

EXPERIMENTAL PROCEDURES
Materials-[␣-32 P]UTP was purchased from NEN Life Science Products. Tryptone, yeast extract, brain heart infusion medium, and agar were from Difco. Antibiotics, glucosamine-1-phosphate, and ACP were products of Sigma. Chloroform, methanol, and acetic acid were from Mallinckrodt. All other chemicals were obtained from Sigma or Mallinckrodt. Silica gel-60 thin layer plates (0.25 mm) were purchased from Merck. Restriction enzymes, Klenow, and T4 DNA ligase were from New England BioLabs or Roche Molecular Biochemicals. Shrimp alkaline phosphatase was from U. S. Biochemical Corp. Primers for mutagenesis were custom-made by Life Technologies, Inc. The LpxC inhibitor L-573,655 (17) was provided by Dr. A. Patchett (Merck Research Laboratories, Rahway, NJ).
Standard Assay of LpxA Activity in the Forward Direction-The assay monitors the conversion of [␣-32 P]UDP-GlcNAc to [␣-32 P]UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc (7,8). The standard reaction mixture (10 -20 l) contains 40 mM HEPES, pH 8, 1 mg/ml BSA, 10 M R-3hydroxymyristoyl-ACP, and 10 M [␣-32 P]UDP-GlcNAc (2 ϫ 10 6 cpm/ nmol). Substrates were synthesized as described previously (33). The reaction is started by the addition of enzyme (either purified or in a cell extract). The reaction mixture is incubated at 30°C for 1-10 min. For measuring initial rates, the enzyme concentrations (monomer) are typically 1-10 nM. The reactions are terminated by spotting 2-2.5-l portions onto a silica thin layer chromatography plate. After the spots air dry, the plates are developed in the solvent chloroform/methanol/water/ acetic acid (25:15:4:2, v/v/v/v). The plates are dried and then exposed to imaging screens overnight at room temperature. The plates are visualized, and the extent of the reaction is quantified using a Molecular Dynamics PhosphorImager, operated with ImageQuant software.
For assays of E. coli crude extracts, 0.2 mg/ml L-573,655 (17) from a 10 mg/ml stock in dimethyl sulfoxide is added to the assay mixtures to inhibit further metabolism of the LpxA reaction product by LpxC pres-FIG. 1. Role of LpxA in E. coli lipid A biosynthesis. The first step in lipid A biosynthesis is catalyzed by LpxA (3). The transfer of the R-3-hydroxymyristoyl moiety from R-3-hydroxymyristoyl-ACP to UDP-GlcNAc is reversible and thermodynamically unfavorable (8). Therefore, the deacetylase encoded by lpxC is the first committed step of the pathway and is regulated in response to lipid A content (35). LpxD displays significant sequence homology to LpxA, including conservation of the ␤-helix and the equivalent of His 125 (46). Structures of the other biosynthetic intermediates and the genes encoding the enzymes of the rest of the pathway have been reviewed elsewhere (3,22).  (25). Basic residues that are conserved (46)  Gly 173 is buried deeply within the proposed active site cleft, as indicated by the shading. The only known exceptions to the conserved basic residues shown above are H160Y in N. meningitidis and K76S in C. vinosum. ent in these extracts (Fig. 1). This is unnecessary for assays of purified LpxA or of Corynebacterium glutamicum extracts expressing recombinant E. coli LpxA.
Preparation of Cell Extracts-For assays of cell free extracts, 50-ml cultures of BL21(DE3)/pLysE strains carrying plasmids containing wild type and mutant lpxA genes were grown at 37°C in LB (10 g/liter tryptone, 5 g/liter yeast extract, 10 g/liter NaCl) (36) with 100 g/ml ampicillin (225 rpm). At A 600 ϭ 0.6, they were induced with 1 mM IPTG and grown for 3 more hours. Cells were washed once with 5 ml of 10 mM potassium phosphate, pH 7, containing 0.2 M NaCl and 20% glycerol, resuspended in 2 ml of the same, and stored at Ϫ80°C. Cells were broken by one passage through a French pressure cell at 18,000 psi and centrifuged at 10,000 ϫ g for 20 min to remove cell debris. Membranefree supernatants were prepared by centrifugation at 150,000 ϫ g for 90 min.
For expression of LpxAs in the absence of chromosomally encoded residual activity, 500-ml cultures of C. glutamicum strains (37) carrying plasmids containing wild type and mutant lpxA genes were grown at 30°C in LB broth supplemented with 5 g/ml chloramphenicol and 50 g/ml rifampicin to A 600 ϭ 1. Cells were washed once with 40 ml of 10 mM potassium phosphate, pH 7, containing 0.2 M NaCl and 20% glycerol, resuspended in 2 ml of the same, and stored at Ϫ80°C. Cells were broken by three passages through a French pressure cell at 18,000 psi and centrifuged at 10,000 ϫ g for 20 min to remove cell debris.
Purification of LpxA-Wild type E. coli LpxA was purified from BL21(DE3)/pLysE/pTO1 (33). pTO1 is a pET23c (Novagen) vector containing wild type E. coli lpxA. Mutant LpxAs were purified from BL21(DE3)/pLysE strains carrying pET23c based plasmids containing mutant lpxA genes (see below). BL21(DE3)/pLysE cultures (500 ml) were grown in LB with 100 g/ml ampicillin at 37°C (225 rpm), induced with 1 mM IPTG at A 600 ϭ 0.6, and grown for 3 more hours (final A 600 , ϳ2). Cells were washed once with 50 ml of 10 mM potassium phosphate, pH 7, containing 0.2 M NaCl and 20% glycerol, resuspended in 20 ml of the same, and stored at Ϫ80°C. Cells were broken by one passage through a French pressure cell at 18,000 psi and centrifuged at 10,000 ϫ g for 20 min to remove cell debris. Membrane-free supernatants were prepared by centrifugation at 150,000 ϫ g for 90 min (final volume, 17 ml).
A 40-ml Gel Matrex Green A (Amicon) column (8) was prepared by washing with 5 column volumes 8 M urea/0.5 M NaOH and equilibrating with 5 column volumes of 10 mM potassium phosphate, pH 7, containing 0.2 M NaCl and 20% glycerol. The membrane-free supernatant was diluted to 10 mg/ml (final volume, 25 ml) and applied to this column at 1.5 ml/min. The column was washed with the equilibration buffer until no more protein emerged as judged by A 280 . The protein was eluted with a 400-ml 0.2-2.5 M NaCl gradient (in 10 mM potassium phosphate, pH 7, containing 20% glycerol). LpxA elutes at approximately 0.6 M NaCl (8). Fractions containing LpxA were pooled and dialyzed against 2 ϫ 2l portions of 20 mM Tris, pH 7.4, containing 20% glycerol. This pool was applied at 1 ml/min to an 8 ml Source 15Q (Amersham Pharmacia Biotech) column equilibrated in the same buffer. The column was washed in the equilibration buffer until no more protein eluted. The protein was then eluted with a 160 ml of 0 -0.6 M NaCl gradient (in 20 mM Tris, pH 8, containing 20% glycerol). LpxA elutes at approximately 0.2 M NaCl. Fractions containing LpxA were pooled, divided into aliquots, and then stored at Ϫ80°C.
Modification of LpxA with Diethylpyrocarbonate-In the following experiments, LpxA concentrations are expressed as the amount of monomer/unit volume. For inhibition studies, 20 M purified LpxA was incubated at room temperature in 40 mM potassium phosphate, pH 7, with various concentrations of DEPC (Sigma) for 10 min. DEPC solutions (in ethanol) were made each day just prior to an experiment. Exact DEPC concentrations were determined by reacting with an excess of imidazole (Sigma) and measuring the A 240 (⑀ ϭ 3200 M Ϫ1 ) (38). DEPC reactions with LpxA at room temperature were quenched by a 1:50 dilution into 70 M imidazole (on ice) (at least a 10-fold molar excess over the DEPC). DEPC, ethanol, and imidazole at the concentrations used have no effect on the substrates of the LpxA reaction, and ethanol and imidazole have no effect on LpxA activity at the levels at which they are carried over into the assay mixture (not shown). The LpxA/DEPC/ imidazole mixtures were further diluted with 1 mg/ml BSA and assayed by the standard method using 40 mM potassium phosphate, pH 7, instead of 40 mM HEPES, pH 8, but this modification has little effect on LpxA activity.
For substrate protection studies, 20 M purified LpxA was incubated at room temperature for 10 min in 40 mM potassium phosphate, pH 7, and 200 M DEPC in the presence of 20 mM UDP-GlcNAc, or 40 M R-3-hydroxymyristoyl-ACP. These concentrations correspond to about 20 times the K m (8) for each substrate. The reactions were quenched by a 1:50 dilution into 40 M imidazole (a 10-fold molar excess). The mixtures were then further diluted as appropriate into 1 mg/ml BSA and assayed by the standard method with 40 mM potassium phosphate, pH 7, taking into account the residual substrate concentrations carried over from the preincubation.
For NH 2 OH reactivation studies, 20 M purified LpxA was incubated with and without 100 M DEPC for 5 min at room temperature in 40 mM potassium phosphate, pH 7. The reactions were quenched by a 1:50 dilution into 20 M imidazole. These mixtures were then diluted 1:2 into 40 mM potassium phosphate, pH 7, and 1 mg/ml BSA or into the same containing a final concentration of 10 mM NH 2 OH (Sigma). A 100 mM stock of NH 2 OH in water was made fresh each day. At various times, portions were diluted further into 1 mg/ml BSA as appropriate and assayed by the standard method with 40 mM potassium phosphate, pH 7.
For spectrophotometric quantification of the extent of chemical modification, 20 M purified LpxA was incubated with 20 -90 M DEPC for 15 min at room temperature. During the reaction, the A 240 was followed to determine the concentration of modified histidines, and the A 280 was also measured to exclude major structural changes in protein folding. At various times throughout the reaction, 3-l portions were withdrawn and quenched with a 10-fold excess of imidazole in 1 mg/ml BSA. These sample were kept on ice until the end of the study, at which time all of the samples were assayed for LpxA activity by the standard method with 40 mM potassium phosphate, pH 7.
Modification of LpxA with Lysine-, Arginine-and Serine-specific Reagents-In all experiments, 20 M purified LpxA (monomer) was incubated at room temperature for 30 min with various concentrations of pyridoxal 5Ј-phosphate, phenylglyoxal, or phenylmethanesulfonyl fluoride. Pyridoxal 5Ј-phosphate reactions were quenched by a 1:2 dilution into 10 mM sodium borohydride (corresponding to at least a 10-fold molar excess over pyridoxal 5Ј-phosphate). All reactions were then diluted appropriately and assayed for LpxA activity. Pyridoxal 5Ј-phosphate (Sigma) was prepared in 40 mM HEPES, pH 8, whereas phenylglyoxal and sodium borohydride were dissolved in water, and the phenylmethanesulfonyl fluoride stock solution was prepared in isopropanol. None of these reagents has any effect on the substrates of the LpxA reaction, and isopropanol and sodium borohydride have no effect on LpxA activity at the levels carried over into the assay system. The pyridoxal 5Ј-phosphate and phenylglyoxal studies were done in 40 mM HEPES, pH 8 (39,40). The phenylmethanesulfonyl fluoride studies were done in 40 mM potassium phosphate, pH 7 (41).
Modification of LpxA with Cysteine-specific Reagents-In all experiments, 10 M purified LpxA (monomer) was incubated at room temperature for 30 min with various concentrations of methyl methane thiosulfonate or N-ethyl maleimide. Both reagents were from Sigma, and stock solutions were prepared in water. After 30 min, samples were diluted with 1 mg/ml BSA and assayed by the standard method with 40 mM potassium phosphate, pH 7.
General Recombinant DNA Techniques-Recombinant DNA techniques were carried out as described by Sambrook et al. (42). Plasmid DNA was prepared using the Qiagen Spin Miniprep kit or the Bigger Prep Plasmid DNA Preparation Kit (5 Prime 3 3 Prime, Inc., Boulder, CO). Restriction endonucleases, Klenow, T4 DNA ligase, and shrimp alkaline phosphatase were used according to the manufacturers' specifications. DNA was extracted from gels using the GeneClean kit (Bio 101, Inc.) according to the manufacturer's directions.
Site-directed Mutagenesis-LpxA mutants were made by the method of Kunkel et al. (43) using pTO1 (33) as template. Plasmid pTO1 contains wild type E. coli lpxA in a pET23c (Novagen) vector. Resulting plasmids are named for the mutated gene they carry, prefaced by pTO, e.g. pTO-H122A. Each mutated gene was sequenced to ensure that only the desired mutation was present. Plasmids were transformed into CaCl 2 competent BL21(DE3)/pLysE (Novagen) cells.
Cloning Mutant lpxA Genes into pGK1-pET23c-based plasmids containing wild type and mutant lpxA genes were digested with NdeI and BamHI to excise the lpxA gene. This fragment was treated with Klenow fragment and then ligated into a BamHI-digested and Klenow fragment-treated pGK1 Corynebacterium/E. coli shuttle vector (37). The plasmid with wild type lpxA is pTO8 (32). Other plasmids are named for the mutated gene that they carry prefaced by pGK, e.g. pGK-H122A. Plasmids were introduced into C. glutamicum R163 by electroporation (32,44). Competent cells were made by growing a 500-ml culture of C. glutamicum in LB broth at 30°C (225 rpm) to A 600 ϭ 0.2. Cells were washed twice with 15% glycerol in water and resuspended in 2.5 ml of the same. Electroporation was done under the following conditions: 50 l of cells and 1 l (1 g) of DNA in a 0.2-cm cuvette (Bio-Rad) at 2.5 kV, 25 microfarad, and 200 ⍀. Cells were allowed to recover for 2 h at 30°C in 1 ml of brain heart infusion (BHI) medium and then plated appropriately to yield approximately 100 colonies at 30°C on BHI plates (15 g/liter agar) supplemented with 50 g/ml rifampicin and 5 g/ml chloramphenicol.
Western Blotting-C. glutamicum extracts (1.5 g) and purified LpxA (5 ng) were analyzed by SDS-polyacrylamide gel electrophoresis (15%). Proteins were transferred to a polyvinylidene difluoride (Bio-Rad) membrane by the following procedure. A piece of extra thick filter paper (Bio-Rad) was wetted in 1ϫ transfer buffer (25 mM Tris base and 200 mM glycine) containing 20% methanol. On top of this was placed the polyvinylidene difluoride membrane that had been prewetted in methanol, followed by water and then by 1ϫ transfer buffer containing 20% methanol. The gel was placed on top of the membrane and covered with a piece of extra thick filter paper wetted in 1ϫ transfer buffer containing 0.05% SDS. The transfer was done at 20 V for 40 min with a Bio-Rad Trans-Blot SD semi-dry transfer apparatus. After transfer, the polyvinylidene difluoride membrane was blocked overnight at room temperature with blocking buffer (PBS containing 0.2% Tween-20 (Sigma) and 5% Kroger nonfat dry milk; PBS consists of 1.45 g of Na 2 HPO 4 , 0.2 g of KH 2 PO 4 , 0.2 g of KCl, and 8.0 g of NaCl/liter at pH 7.4). The next day the membrane was washed twice quickly with 25 ml of PBS containing 0.2% Tween-20 and then incubated for 1 h at room temperature with 20 ml of blocking buffer containing primary antibody (provided by Dr. Garry Dotson). The primary antibody is polyclonal rabbit antibody generated against purified His-tagged LpxA (45). This antibody was partially purified using a Ni 2ϩ column charged with His-tagged LpxA and then a protein A column (45). The primary antibody was used at a 1:10,000 dilution from a 1 mg/ml stock. The membrane was then washed three times for 5 min with 30 ml of blocking buffer. The membrane was washed twice quickly with 25 ml of PBS containing 0.2% Tween-20 and then incubated for 1 h at room temperature with 20 ml of blocking buffer containing secondary antibody. The secondary antibody is donkey anti-rabbit immunoglobulin conjugated with horseradish peroxidase (Amersham Pharmacia Biotech). The secondary antibody was used at a 1:20,000 dilution from a 0.5 mg/ml stock. The membrane was washed as described above and then washed three more times with 75 ml of PBS containing 0.2% Tween-20. Detection was done using enhanced chemiluminescence reagents (Amersham Pharmacia Biotech) according to the manufacturer's directions.
Determination of Kinetic Parameters-Initial velocities were measured with the standard assay in the forward direction using 100 M R-3-OH-myristoyl-ACP (wild type LpxA K m,acyl-ACP ϳ1 M (8)) and varying concentrations of UDP-GlcNAc. Purified wild type LpxA (3.6 nM monomer), as well as extracts of C. glutamicum cells (1-6 mg/ml) carrying plasmids that express wild type or mutant lpxA genes, were assayed at concentrations appropriate for following the initial rate. Because of the unfavorable equilibrium constant of the LpxA reaction (8), initial velocities could only be measured for Յ10 mM UDP-GlcNAc. For purified wild type LpxA and extracts containing wild type, K76A, or K76R LpxA, steady-state kinetic parameters were determined by fitting the initial velocities to the Michaelis-Menten equation by nonlinear regression using the Kaleidograph (Synergy Software) curve-fitting program. For extracts containing H122A, H144A, or H160A LpxAs, the initial velocities increased linearly with UDP-GlcNAc concentrations up to 10 mM. In these cases, k cat /K m values were determined by fitting initial velocities to a line using the Kaleidograph program, and lower limits for k cat and K m were then extrapolated.
The concentration of LpxA in C. glutamicum extracts was estimated in the following manner. The specific activity of purified wild type LpxA under standard conditions is 265,000 pmol/min/mg, and the specific activity under standard conditions in C. glutamicum extracts containing wild type LpxA is 347 pmol/min/mg. Dividing these two values gives the conversion factor of 764 mg extract/mg LpxA. Because expression levels are very similar in all the C. glutamicum extracts for wild type and mutant LpxAs, as judged by Western blotting (not shown), this conversion factor was used to estimate the k cat and k cat /K m values for all the LpxAs expressed in the C. glutamicum system.

RESULTS
Inactivation of LpxA by DEPC-Incubation of purified LpxA with DEPC, a histidine reagent, causes a rapid loss of enzymatic activity (Fig. 3). LpxA is stable under the preincubation conditions in the absence of DEPC. The loss of activity depends upon the time of preincubation and the concentration of DEPC, and inactivation is complete in 10 min with only a small (4-fold) molar excess of DEPC over LpxA monomer.
To determine whether the inactivation of LpxA results from modification of an active site residue, we looked for substrate protection against DEPC inactivation. UDP-GlcNAc has no effect, but R-3-hydroxymyristoyl-ACP largely protects LpxA from inactivation by DEPC (data not shown). ACP itself contains one histidine residue, and consequently care was taken to ensure that ACP was not simply scavenging the DEPC reagent. The incubation included a 2-fold molar excess of acyl-ACP over LpxA monomer, and a 10-fold molar excess of DEPC. Therefore, even if the histidine of ACP reacted with the DEPC, there would still be more than enough DEPC (an 8-fold excess) to cause full inactivation of LpxA. Protection by acyl-ACP suggests that the modified residue accounting for the inactivation is in the E. coli LpxA active site or at least in the cleft proposed to dock ACP (Fig. 2), which contains four conserved histidine residues.
Reactivation of DEPC-inactivated LpxA by NH 2 OH-In some cases, DEPC has been shown to react with side chains other than histidine (38). The small amount of DEPC needed for the inactivation of LpxA and the rapid rate of the inactivation (Fig.  3) argue that side chains other than histidine are not being modified. In addition, the inactivation by DEPC can be reversed by incubation with NH 2 OH (not shown). NH 2 OH reverses DEPC modification of histidine and tyrosine residues but not of lysine or cysteine side chains (38). The reaction of DEPC with tyrosine residues on LpxA can be ruled out, given that no change in the absorbance of LpxA at 280 nm occurs upon reaction with DEPC (not shown). Thus, the inactivation of LpxA by DEPC can be attributed solely to the modification of histidine residue(s).

Correlation between the Number of Histidine Residues Modified and the Loss of LpxA
Activity-When DEPC reacts with a histidine residue, a carbethoxyhistidine moiety is formed that absorbs light at 240 nm (⑀ ϭ 3200 M Ϫ1 ) (38). We used this phenomenon to calculate the number of histidine residues modified per LpxA monomer during the inactivation process. Throughout the incubation of LpxA with DEPC, portions of the reaction mixture were removed, quenched with imidazole, and assayed for LpxA activity. The correlation between the number of histidines modified and the loss of catalytic activity is shown in Fig. 4. When an average of one histidine residue/monomer is modified, only 20% of the initial activity remains. Because LpxA contains 12 histidines/monomer, it is reasonable that some monomers would have acquired more than one modified histidine residue, whereas 20% might have no modified histidines. These results suggest that one modified histidine/monomer leads to complete inactivation of LpxA. This would correspond to three modified histidines/trimer, consistent with the idea of three functional active sites/trimer.
Effects of Other Chemical Modification Reagents on LpxA Activity-We next examined LpxA inactivation by pyridoxal 5Ј-phosphate/sodium borohydride, phenylglyoxal, and phenylmethanesulfonyl fluoride, lysine-, arginine-, and serine-specific reagents, respectively (40). Low millimolar concentrations of both pyridoxal 5Ј-phosphate/sodium borohydride and phenylglyoxal are capable of inhibiting LpxA completely (not shown). Phenylmethanesulfonyl fluoride has little effect on LpxA activity (not shown). In addition, no time-dependent inactivation of LpxA is seen using the cysteine reagent, methyl methane thiosulfonate (not shown). These findings suggest that lysine and arginine residue(s) might play important roles in LpxA substrate binding and/or catalysis, whereas serine and cysteine residues probably do not, consistent with the fact that there are no conserved serine or cysteine residues in diverse LpxAs.
Site-directed Mutagenesis of E. coli LpxA-The sequences of LpxA proteins from 15 Gram-negative bacteria were found by BLAST searching of the nonredundant protein data base (46). All LpxAs have conserved, solvent-exposed histidine residues at positions corresponding to 122, 125, and 144 in the E. coli LpxA sequence (Fig. 2). The histidine at position 160 (Fig. 2) is conserved in all LpxAs except for the enzyme from Neisseria meningitidis (47), in which that position is a phenylalanine. Lysine 76 (Fig. 2) is conserved across all LpxAs except for the enzyme from Chromatium vinosum, in which it is a serine. Arginine 204 (Fig. 2) is conserved in all LpxA homologues identified to date. Accordingly, each of these residues was mutated to alanine, and the mutant LpxAs were overexpressed under T7 promoter control in the E. coli strain BL21(DE3)/ pLysE. All the mutants expressed comparable levels of LpxA protein, as judged by analysis of membrane-free supernatants by SDS-polyacrylamide gel electrophoresis and Coomassie Blue staining (not shown).
The relative activities of the membrane-free supernatants containing these LpxA mutants as compared with membranefree supernatant containing overexpressed wild type LpxA are as follows (Table I). H122A, H144A, and K76A LpxAs have 2-6% wild type specific activity under standard assay conditions in extracts made from IPTG induced cells, whereas H160A and H125A have less than 1% of the wild type activity ( Table I). Mutation of the conserved arginine at position 204 (Fig. 2) to alanine likewise displayed Ͻ1% wild type activity. 2 Each mutant shows increased protein levels upon induction with IPTG as judged by gel electrophoresis (not shown), and all but H125A and H160A show corresponding 5-10-fold induction of LpxA activity compared with extracts of uninduced cells ( Table I). The observation that increased levels of H125A protein do not lead to increased activity suggests that H125A LpxA is completely inactive. Indeed, the activity of H125A extracts is in the range of the pET23c vector control (Table I).
The next set of mutant LpxAs that we constructed included  H122N, H125N, H144N, H160F, and K76R (Table I). Asparagine can sometimes take the place of histidine in hydrogen bond formation. The H160F mutant was made to test whether the N. meningitidis residue could take the place of His 160 in E. coli LpxA, and the K76R mutant was made to determine whether a positive charge at position 76 was sufficient to support activity. K76R LpxA has 5-fold more activity than K76A LpxA, but none of the other mutants have significantly more activity than the original alanine mutants (Table I). Like H125A, the extracts of strains expressing H125N do not show increased LpxA activity upon induction of these mutant proteins with IPTG (Table I), nor was activity restored by the addition of imidazole to H125A LpxA, or a H125G mutant (not shown).
The mutant enzymes H122A, H125A, H144A, H160A, K76A, and K76R were purified as described under "Experimental Procedures." Each mutant LpxA eluted from each column at the same point in the NaCl gradients as did wild type LpxA. Analytical ultracentrifugation was used to demonstrate that each of these mutant enzymes formed trimers in aqueous solution, just like the wild type (data not shown). These results taken together suggest that the global protein folds of the mutant LpxAs are similar to that of the wild type. The loss of activity associated with each of the mutations (Table I) is therefore likely to be due to the loss of a residue that is important for catalysis and/or substrate binding rather than to a major structural change. However, the possibility of small local conformational changes associated with these point mutations cannot yet be excluded.
Expression of E. coli LpxA Mutants in C. glutamicum R163-When LpxA mutants are expressed in E. coli, the background of wild type LpxA activity made from the chromosomal copy of the essential lpxA gene complicates the kinetic analysis. Because the mutant LpxAs all purify exactly like the wild type enzyme, even the "purified" mutant enzymes are contaminated with 2 C. Sweet and C. R. H. Raetz, unpublished results. residual wild type LpxA. To avoid this problem, we expressed H122A, H125A, H144A, H160A, K76A, and K76R LpxAs in a Gram-positive host, C. glutamicum (37). Gram-positive bacteria do not contain any of the lpx genes, because they do not make lipid A (48). C. glutamicum has previously been shown to be a good system for expressing and assaying LpxA point mutants (32).
Mutated lpxA genes were removed from the pET23c vector and cloned into the C. glutamicum/E. coli shuttle vector, pGK1 (37). These plasmids were then used to express the mutant LpxAs in C. glutamicum. Overexpression of the enzymes in these extracts is 2 orders of magnitude lower than in the T7 promoter-driven E. coli system, and the LpxA protein band cannot be visualized by Coomassie staining after SDS-polyacrylamide gel electrophoresis of C. glutamicum crude extracts (not shown). Therefore, a Western blot with an anti-LpxA antibody was used to demonstrate that all mutant LpxAs were expressed to similar levels (data not shown).
The relative specific activities under standard assay conditions of the LpxA mutants compared with the wild type, when expressed in C. glutamicum extracts, are shown in Table II. With no LpxA activity background from the host cells, it is clear that H125A is completely inactive. The other mutants have similar relative specific activities as observed in the E. coli extracts (Table I).
Kinetic Analysis of LpxA Point Mutants-The initial kinetic analysis of the LpxA point mutants was carried out using the C. glutamicum extracts, because they have no wild type LpxA background activity, and the mutants all seem to behave similarly in the C. glutamicum system as in E. coli extracts (Tables  I and II). In addition, the wild type k cat and K m,UDP-GlcNAc at saturating levels of acyl-ACP (100 M) (8) are the same for purified LpxA and wild type LpxA expressed in C. glutamicum extracts (Table III).
The K m,UDP-GlcNAc , k cat , and k cat /K m,UDP-GlcNAc values for the LpxA point mutants are shown in Table III. The k cat /K m,UDP-GlcNAc values for the LpxA mutants are lower than wild type in all cases. This decrease is due to an increase in K m,UDP-GlcNAc for all of the mutants. In addition, H160A LpxA probably has a lower k cat than wild type. However, H125A shows no activity under any assay condition, and therefore no kinetic parameters can be measured. These results suggest that Lys 76 , His 122 , His 144 , and possibly His 160 play roles in substrate binding and that His 125 may be directly involved in catalysis.
Probing the Catalytic Mechanism of LpxA-Acetyl-and acyltransferases generally work by one of two mechanisms: either through an acyl-enzyme intermediate (41) or by direct transfer assisted by general base catalysis (49,50). If the LpxA mechanism involves an acyl-enzyme intermediate, it might function either by a ping-pong mechanism or by a sequential mechanism, the latter requiring the ternary complex for acyl-enzyme intermediate formation. To investigate some of these possibilities, we assayed LpxA in the reverse direction using [␣-32 P]UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc and ACP-SH as the substrates, because of the greater sensitivity of this assay and the possibility of testing other SH containing acceptors. If LpxA works by a ping-pong mechanism, then it might be able to hydrolyze the acyl chain of [␣- 32    completion in less than 1 min (Fig. 5A, closed circles). As predicted by the unfavorable equilibrium constant measured in the forward direction (8), O 3 S transfer under these conditions is highly favorable. Interestingly, acyl acceptors other than ACP show some slight activity under these conditions. Coenzyme A, pantetheine, and DTT (each 10 M) can all support the reverse LpxA reaction to a very small extent (Fig. 5A, closed squares, diamonds, and triangles). N-Acetylcysteamine and glutathione were also tested but do not support the reverse reaction above the DTT background (not shown). In the absence of LpxA, none of these reagents can deacylate the substrate (Fig. 5B) (Fig. 5A, ϫ). The small amount of background seen at time 0 in all cases is due to contaminating [␣-32 P]UDP-GlcNAc in the [␣-32 P]UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc substrate preparation, and this is not increased by the addition of LpxA alone (Fig. 5A). Thus, a thiol-containing acyl acceptor is required for LpxA activity in the reverse direction, suggesting that LpxA does not work by a ping-pong mechanism with an acyl-enzyme intermediate.

DISCUSSION
LpxA catalyzes the first step of lipid A biosynthesis in Gramnegative bacteria (3,7) and is the only enzyme of lipid A biosynthesis for which structural information is currently available (25). The secondary structure of LpxA contains a left-handed parallel ␤-helix (25) shared by only three other proteins of known structure (27)(28)(29). One turn of the ␤-helix corresponds to three contiguous hexapeptide repeats, a motif that is found in the primary sequence of LpxA and a number of acyl-and acetyltransferases (26).
Our goal in this paper was to do an initial characterization of the active site and mechanism of action of LpxA. This information is potentially useful for the development of LpxA inhibitors with antibacterial activity. Using chemical modification and site-directed mutagenesis, we identified six residues in E. coli LpxA that are important to catalysis. Four of these residues, His 122 , His 125 , His 144 , and His 160 , are all side chains of the ␤-helix of LpxA, whereas Lys 76 is located on a loop extending out from the middle of the ␤-helix domain, and Arg 204 is located in the C-terminal ␣-helical region of LpxA (25). All these residues cluster around a deep cleft situated between the monomers of the LpxA trimer (Fig. 2). Another residue in the same area, glycine 173 (Fig. 2, red), is involved in determining the acyl-chain length selectivity of LpxA (32). We presume that LpxA has three identical active sites, consistent with the locations of the active sites in other ␤-helix proteins (27)(28)(29). The active site of M. thermophila carbonic anhydrase is defined by the location of its catalytic zinc atom (27). The locations of the active sites of the P. aeruginosa xenobiotic acetyltransferase (29) and of the M. bovis tetrahydrodipicolinate N-succinyltransferase (28,51) are based on x-ray structures of co-crystals with substrate or inhibitor, respectively.
What roles might the conserved residues of the putative LpxA active site play in catalysis? To answer this question, we investigated LpxA mutants in which each residue was mutated to alanine (Tables I and II). The H125A LpxA mutant is completely inactive as shown by activity assays of the enzyme expressed in the Gram-positive host C. glutamicum (Table II). However, the other mutant LpxAs do possess low levels of activity. The absolute requirement for His 125 suggests a catalytic role, such as that of a general base in a direct transfer mechanism (Fig. 6), whereas the other residues may be involved in substrate binding.
We further investigated these LpxA mutants by measuring K m,UDP-GlcNAc , k cat , and k cat /K m,UDP-GlcNAc values for the enzymes expressed in C. glutamicum. We did not measure K m, acyl-ACP values because it is not technically feasible with the current assay. The wild type K m,acyl-ACP is ϳ2 M (8). The wild type K m,UDP-GlcNAc value is 3 orders of magnitude higher (1 mM), and the mutant LpxAs have K m,UDP-GlcNAc values that are another order of magnitude higher (ϳ10 mM) ( Table III). The equilibrium constant for the LpxA reaction is 0.01 (8). With these constraints, it is impossible to measure the conversion of [␣-32 P]UDP-GlcNAc to [␣-32 P]UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc at very high UDP-GlcNAc and very low acyl-ACP concentrations. Determination of accurate K m,acyl-ACP values for the LpxA mutants will have to await the development of a new LpxA assay involving radioactive acyl-ACP as the donor and unlabeled UDP-GlcNAc as the acceptor.
We measured the K m,UDP-GlcNAc of purified wild type LpxA to be 1 mM, an order of magnitude higher than that reported previously (8), and two-fold higher than that reported for wild type LpxA in cell extracts (7). The substrate preparations used in the present study are of higher quality than those used in the earlier investigations. In addition, the use of a Phosphor-Imager in our work has greatly facilitated the collection of a much larger number of accurate data points. For these reasons  (Table II). This reflects a higher K m,UDP-GlcNAc in all cases, suggesting that the mutated residues are involved in UDP-GlcNAc binding and/or formation of the ternary complex of UDP-GlcNAc, acyl-ACP, and LpxA. In particular, K76A has a K m,UDP-GlcNAc over an order of magnitude higher than wild type (Table III). The value is two-fold lower when Lys 76 is replaced with arginine. A positive charge at position 76 (Fig. 2) could be involved in binding the phosphates of UDP-GlcNAc or an acidic side chain of the acyl-ACP. In addition to a higher K m,UDP-GlcNAc , H160A LpxA has a lower k cat than wild type LpxA (Table III). His 160 may therefore play a more direct role in catalysis and/or substrate binding. More detailed kinetic analyses in conjunction with crystal structures of LpxA in complex with its substrates will be needed to define more precisely the roles of these histidine and lysine residues.
We propose that LpxA is not likely to utilize a ping-pong mechanism with an acyl enzyme intermediate, because LpxA does not remove the acyl chain from UDP-3-O-(R-3-hydroxymyristoyl)-GlcNAc in the absence of a thiol-containing acylacceptor (Fig. 5). The absence of either a conserved serine or cysteine residue and the inability of serine-or cysteine-specific reagents to inactivate LpxA render any mechanism involving an acyl-enzyme intermediate on serine or cysteine extremely unlikely. However, we cannot exclude the possibility of a mechanism in which binding of both the acyl-donor and the acylacceptor is required for the formation of an acyl-enzyme intermediate in the ternary complex on a residue other than serine or cysteine, perhaps on His 125 . Nevertheless, a more likely role for His 125 would be as a general base, abstracting the hydrogen from the 3-OH of UDP-GlcNAc and preparing that substrate for direct nucleophilic attack of the acyl chain carbonyl group of the acyl-ACP (Fig. 6). Interestingly, LpxD (Fig. 1) displays significant sequence homology to LpxA (11) in that it also contains an extensive ␤-helix domain, and it appears to have conserved the equivalent of the LpxA His 125 residue (46).
A better studied precedent for the mechanism shown in Fig.  6 is that proposed for the type III chloramphenicol acetyltransferase (49,50), in which co-crystals of the enzyme with chloramphenicol, affinity labeling, site-directed mutagenesis, and kinetic studies all strongly support the model. However, the chloramphenicol acetyltransferase does differ from LpxA in that it can catalyze two distinct acetylations of its substrate.
In the process of studying the reverse LpxA reaction, we discovered that, in addition to ACP, pantetheine, coenzyme A, and DTT function as weak acyl-acceptors (Fig. 5). Although this finding contradicts earlier reports that R-3-hydroxymyristoyl-CoA is not a substrate in the forward direction (7,8), LpxA activity with pantetheine, CoA, or DTT in the reverse direction is 5 orders of magnitude slower than with ACP and is therefore unlikely to have any physiological significance. Given the technical constraints of the LpxA assay in the forward direction, it would be extremely difficult to detect acyltransferase activity with such poor substrate analogs.  6. A hypothetical catalytic mechanism for LpxA. In this proposed mechanism, His 125 acts as a general base to increase the nucleophilicity of the glucosamine 3-OH of UDP-GlcNAc during its attack on the thioester carbonyl of the acyl-ACP donor. This model is based on the available data for LpxA and is similar to what has been proposed for chloramphenicol acetyl transferase (49,50). The positive charges involved in substrate binding may come from His 122 , His 144 , His 160 , Arg 204 , and/or Lys 76 . The LpxA hydrocarbon ruler appears to involve Gly 173 or its surroundings (see Fig. 2), as described previously (32).
The LpxA assay in the reverse direction (Fig. 5) offers a very sensitive probe for certain low activity substrates. The sensitivity of the reverse assay might therefore be useful for additional kinetic and thermodynamic studies of LpxA. For instance, a comparison of the LpxA equilibrium constants for coenzyme A versus ACP, approached from the product side (as in Fig. 5), might provide some insights into the physical basis for the unfavorable equilibrium constant observed for the LpxA reaction with its natural substrates. The acyl chain length specificity in the reverse direction, as well as the binding constants for each of the LpxA substrates and products, might also prove to be very informative.